Ticket #37223: 37223-2-py-numpy.patch

File 37223-2-py-numpy.patch, 773.8 KB (added by seanfarley (Sean Farley), 12 years ago)

py-numpy: add python33 support

  • dports/python/py-numpy/Portfile 

    # HG changeset patch
# User Sean Farley <sean.michael.farley@gmail.com>
# Date 1354844785 21600
# Node ID b3dc6497233b7ca2fc9728e731ea5f0a5d207532
# Parent  4d41d9b6d8d1e92ce93e4c26493a9681e285b11e
py-numpy: add python33 support

diff --git a/dports/python/py-numpy/Portfile b/dports/python/py-numpy/Portfile
    a b  
    1616long_description        ${description}
    1717
    1818checksums               rmd160  9643c04a2e8fbb99cdb047281eedbbfb99423553 \
    1919                        sha256  0992d326147d0ed83bd059519897e7a8ee52dea5ee66bbe04c0ea1c502cd8618
    2020
    21 python.versions         24 25 26 27 31 32
     21python.versions         24 25 26 27 31 32 33
    2222
    2323if {${subport} != ${name}} {
    2424
    2525patchfiles              patch-f2py_setup.py.diff \
    2626                        patch-fcompiler_g95.diff
    2727
     28if {${python.version} == 33} {
     29    patchfiles-append   patch-python33-unicode.diff \
     30                        patch-python33-mtrand.diff \
     31                        patch-python33-shape.diff \
     32                        patch-python33-methods.diff
     33}
     34
    2835depends_lib-append      port:fftw-3 \
    2936                        port:py${python.version}-nose
    3037
    3138if {[variant_isset universal]} {
    3239    build.env-append    ARCHFLAGS="${configure.universal_ldflags}"

  • new file dports/python/py-numpy/files/patch-python33-methods.diff 

    diff --git a/dports/python/py-numpy/files/patch-python33-methods.diff b/dports/python/py-numpy/files/patch-python33-methods.diff
new file mode 100644
    - +  
     1--- numpy/core/src/multiarray/methods.old.c     2012-12-06 16:49:23.000000000 -0600
     2+++ numpy/core/src/multiarray/methods.c 2012-12-06 16:48:07.000000000 -0600
     3@@ -1476,7 +1476,7 @@
     4     if (!PyDataType_FLAGCHK(typecode, NPY_LIST_PICKLE)) {
     5         int swap=!PyArray_ISNOTSWAPPED(self);
     6         self->data = datastr;
     7-        if (!_IsAligned(self) || swap) {
     8+        if (!_IsAligned(self) || swap || (len <= 1000)) {
     9             intp num = PyArray_NBYTES(self);
     10             self->data = PyDataMem_NEW(num);
     11             if (self->data == NULL) {

  • new file dports/python/py-numpy/files/patch-python33-mtrand.diff 

    diff --git a/dports/python/py-numpy/files/patch-python33-mtrand.diff b/dports/python/py-numpy/files/patch-python33-mtrand.diff
new file mode 100644
    - +  
     1--- numpy/random/mtrand/mtrand.old.c    2012-05-19 08:41:41.000000000 -0500
     2+++ numpy/random/mtrand/mtrand.c        2012-12-06 15:02:43.000000000 -0600
     3@@ -1,16 +1,16 @@
     4-/* Generated by Cython 0.15.1 on Tue May  1 15:44:50 2012 */
     5+/* Generated by Cython 0.17.1 on Thu Dec  6 15:02:42 2012 */
     6 
     7 #define PY_SSIZE_T_CLEAN
     8 #include "Python.h"
     9 #ifndef Py_PYTHON_H
     10     #error Python headers needed to compile C extensions, please install development version of Python.
     11+#elif PY_VERSION_HEX < 0x02040000
     12+    #error Cython requires Python 2.4+.
     13 #else
     14-
     15 #include <stddef.h> /* For offsetof */
     16 #ifndef offsetof
     17 #define offsetof(type, member) ( (size_t) & ((type*)0) -> member )
     18 #endif
     19-
     20 #if !defined(WIN32) && !defined(MS_WINDOWS)
     21   #ifndef __stdcall
     22     #define __stdcall
     23@@ -22,36 +22,44 @@
     24     #define __fastcall
     25   #endif
     26 #endif
     27-
     28 #ifndef DL_IMPORT
     29   #define DL_IMPORT(t) t
     30 #endif
     31 #ifndef DL_EXPORT
     32   #define DL_EXPORT(t) t
     33 #endif
     34-
     35 #ifndef PY_LONG_LONG
     36   #define PY_LONG_LONG LONG_LONG
     37 #endif
     38-
     39-#if PY_VERSION_HEX < 0x02040000
     40-  #define METH_COEXIST 0
     41-  #define PyDict_CheckExact(op) (Py_TYPE(op) == &PyDict_Type)
     42-  #define PyDict_Contains(d,o)   PySequence_Contains(d,o)
     43+#ifndef Py_HUGE_VAL
     44+  #define Py_HUGE_VAL HUGE_VAL
     45+#endif
     46+#ifdef PYPY_VERSION
     47+#define CYTHON_COMPILING_IN_PYPY 1
     48+#define CYTHON_COMPILING_IN_CPYTHON 0
     49+#else
     50+#define CYTHON_COMPILING_IN_PYPY 0
     51+#define CYTHON_COMPILING_IN_CPYTHON 1
     52 #endif
     53-
     54 #if PY_VERSION_HEX < 0x02050000
     55   typedef int Py_ssize_t;
     56   #define PY_SSIZE_T_MAX INT_MAX
     57   #define PY_SSIZE_T_MIN INT_MIN
     58   #define PY_FORMAT_SIZE_T ""
     59+  #define CYTHON_FORMAT_SSIZE_T ""
     60   #define PyInt_FromSsize_t(z) PyInt_FromLong(z)
     61   #define PyInt_AsSsize_t(o)   __Pyx_PyInt_AsInt(o)
     62-  #define PyNumber_Index(o)    PyNumber_Int(o)
     63-  #define PyIndex_Check(o)     PyNumber_Check(o)
     64+  #define PyNumber_Index(o)    ((PyNumber_Check(o) && !PyFloat_Check(o)) ? PyNumber_Int(o) : \
     65+                                (PyErr_Format(PyExc_TypeError, \
     66+                                              "expected index value, got %.200s", Py_TYPE(o)->tp_name), \
     67+                                 (PyObject*)0))
     68+  #define PyIndex_Check(o)     (PyNumber_Check(o) && !PyFloat_Check(o) && !PyComplex_Check(o))
     69   #define PyErr_WarnEx(category, message, stacklevel) PyErr_Warn(category, message)
     70+  #define __PYX_BUILD_PY_SSIZE_T "i"
     71+#else
     72+  #define __PYX_BUILD_PY_SSIZE_T "n"
     73+  #define CYTHON_FORMAT_SSIZE_T "z"
     74 #endif
     75-
     76 #if PY_VERSION_HEX < 0x02060000
     77   #define Py_REFCNT(ob) (((PyObject*)(ob))->ob_refcnt)
     78   #define Py_TYPE(ob)   (((PyObject*)(ob))->ob_type)
     79@@ -59,7 +67,6 @@
     80   #define PyVarObject_HEAD_INIT(type, size) \
     81           PyObject_HEAD_INIT(type) size,
     82   #define PyType_Modified(t)
     83-
     84   typedef struct {
     85      void *buf;
     86      PyObject *obj;
     87@@ -73,7 +80,6 @@
     88      Py_ssize_t *suboffsets;
     89      void *internal;
     90   } Py_buffer;
     91-
     92   #define PyBUF_SIMPLE 0
     93   #define PyBUF_WRITABLE 0x0001
     94   #define PyBUF_FORMAT 0x0004
     95@@ -83,24 +89,44 @@
     96   #define PyBUF_F_CONTIGUOUS (0x0040 | PyBUF_STRIDES)
     97   #define PyBUF_ANY_CONTIGUOUS (0x0080 | PyBUF_STRIDES)
     98   #define PyBUF_INDIRECT (0x0100 | PyBUF_STRIDES)
     99-
     100+  #define PyBUF_RECORDS (PyBUF_STRIDES | PyBUF_FORMAT | PyBUF_WRITABLE)
     101+  #define PyBUF_FULL (PyBUF_INDIRECT | PyBUF_FORMAT | PyBUF_WRITABLE)
     102+  typedef int (*getbufferproc)(PyObject *, Py_buffer *, int);
     103+  typedef void (*releasebufferproc)(PyObject *, Py_buffer *);
     104 #endif
     105-
     106 #if PY_MAJOR_VERSION < 3
     107   #define __Pyx_BUILTIN_MODULE_NAME "__builtin__"
     108+  #define __Pyx_PyCode_New(a, k, l, s, f, code, c, n, v, fv, cell, fn, name, fline, lnos) \
     109+          PyCode_New(a, l, s, f, code, c, n, v, fv, cell, fn, name, fline, lnos)
     110 #else
     111   #define __Pyx_BUILTIN_MODULE_NAME "builtins"
     112+  #define __Pyx_PyCode_New(a, k, l, s, f, code, c, n, v, fv, cell, fn, name, fline, lnos) \
     113+          PyCode_New(a, k, l, s, f, code, c, n, v, fv, cell, fn, name, fline, lnos)
     114+#endif
     115+#if PY_MAJOR_VERSION < 3 && PY_MINOR_VERSION < 6
     116+  #define PyUnicode_FromString(s) PyUnicode_Decode(s, strlen(s), "UTF-8", "strict")
     117 #endif
     118-
     119 #if PY_MAJOR_VERSION >= 3
     120   #define Py_TPFLAGS_CHECKTYPES 0
     121   #define Py_TPFLAGS_HAVE_INDEX 0
     122 #endif
     123-
     124 #if (PY_VERSION_HEX < 0x02060000) || (PY_MAJOR_VERSION >= 3)
     125   #define Py_TPFLAGS_HAVE_NEWBUFFER 0
     126 #endif
     127-
     128+#if PY_VERSION_HEX > 0x03030000 && defined(PyUnicode_KIND)
     129+  #define CYTHON_PEP393_ENABLED 1
     130+  #define __Pyx_PyUnicode_READY(op)       (likely(PyUnicode_IS_READY(op)) ? \
     131+                                              0 : _PyUnicode_Ready((PyObject *)(op)))
     132+  #define __Pyx_PyUnicode_GET_LENGTH(u)   PyUnicode_GET_LENGTH(u)
     133+  #define __Pyx_PyUnicode_READ_CHAR(u, i) PyUnicode_READ_CHAR(u, i)
     134+  #define __Pyx_PyUnicode_READ(k, d, i)   PyUnicode_READ(k, d, i)
     135+#else
     136+  #define CYTHON_PEP393_ENABLED 0
     137+  #define __Pyx_PyUnicode_READY(op)       (0)
     138+  #define __Pyx_PyUnicode_GET_LENGTH(u)   PyUnicode_GET_SIZE(u)
     139+  #define __Pyx_PyUnicode_READ_CHAR(u, i) ((Py_UCS4)(PyUnicode_AS_UNICODE(u)[i]))
     140+  #define __Pyx_PyUnicode_READ(k, d, i)   ((k=k), (Py_UCS4)(((Py_UNICODE*)d)[i]))
     141+#endif
     142 #if PY_MAJOR_VERSION >= 3
     143   #define PyBaseString_Type            PyUnicode_Type
     144   #define PyStringObject               PyUnicodeObject
     145@@ -108,7 +134,6 @@
     146   #define PyString_Check               PyUnicode_Check
     147   #define PyString_CheckExact          PyUnicode_CheckExact
     148 #endif
     149-
     150 #if PY_VERSION_HEX < 0x02060000
     151   #define PyBytesObject                PyStringObject
     152   #define PyBytes_Type                 PyString_Type
     153@@ -127,7 +152,6 @@
     154   #define PyBytes_Concat               PyString_Concat
     155   #define PyBytes_ConcatAndDel         PyString_ConcatAndDel
     156 #endif
     157-
     158 #if PY_VERSION_HEX < 0x02060000
     159   #define PySet_Check(obj)             PyObject_TypeCheck(obj, &PySet_Type)
     160   #define PyFrozenSet_Check(obj)       PyObject_TypeCheck(obj, &PyFrozenSet_Type)
     161@@ -135,9 +159,7 @@
     162 #ifndef PySet_CheckExact
     163   #define PySet_CheckExact(obj)        (Py_TYPE(obj) == &PySet_Type)
     164 #endif
     165-
     166 #define __Pyx_TypeCheck(obj, type) PyObject_TypeCheck(obj, (PyTypeObject *)type)
     167-
     168 #if PY_MAJOR_VERSION >= 3
     169   #define PyIntObject                  PyLongObject
     170   #define PyInt_Type                   PyLong_Type
     171@@ -154,11 +176,9 @@
     172   #define PyInt_AsUnsignedLongMask     PyLong_AsUnsignedLongMask
     173   #define PyInt_AsUnsignedLongLongMask PyLong_AsUnsignedLongLongMask
     174 #endif
     175-
     176 #if PY_MAJOR_VERSION >= 3
     177   #define PyBoolObject                 PyLongObject
     178 #endif
     179-
     180 #if PY_VERSION_HEX < 0x03020000
     181   typedef long Py_hash_t;
     182   #define __Pyx_PyInt_FromHash_t PyInt_FromLong
     183@@ -167,16 +187,6 @@
     184   #define __Pyx_PyInt_FromHash_t PyInt_FromSsize_t
     185   #define __Pyx_PyInt_AsHash_t   PyInt_AsSsize_t
     186 #endif
     187-
     188-
     189-#if PY_MAJOR_VERSION >= 3
     190-  #define __Pyx_PyNumber_Divide(x,y)         PyNumber_TrueDivide(x,y)
     191-  #define __Pyx_PyNumber_InPlaceDivide(x,y)  PyNumber_InPlaceTrueDivide(x,y)
     192-#else
     193-  #define __Pyx_PyNumber_Divide(x,y)         PyNumber_Divide(x,y)
     194-  #define __Pyx_PyNumber_InPlaceDivide(x,y)  PyNumber_InPlaceDivide(x,y)
     195-#endif
     196-
     197 #if (PY_MAJOR_VERSION < 3) || (PY_VERSION_HEX >= 0x03010300)
     198   #define __Pyx_PySequence_GetSlice(obj, a, b) PySequence_GetSlice(obj, a, b)
     199   #define __Pyx_PySequence_SetSlice(obj, a, b, value) PySequence_SetSlice(obj, a, b, value)
     200@@ -195,11 +205,9 @@
     201         (likely((obj)->ob_type->tp_as_mapping) ? (PySequence_DelSlice(obj, a, b)) : \
     202             (PyErr_Format(PyExc_TypeError, "'%.200s' object doesn't support slice deletion", (obj)->ob_type->tp_name), -1)))
     203 #endif
     204-
     205 #if PY_MAJOR_VERSION >= 3
     206   #define PyMethod_New(func, self, klass) ((self) ? PyMethod_New(func, self) : PyInstanceMethod_New(func))
     207 #endif
     208-
     209 #if PY_VERSION_HEX < 0x02050000
     210   #define __Pyx_GetAttrString(o,n)   PyObject_GetAttrString((o),((char *)(n)))
     211   #define __Pyx_SetAttrString(o,n,a) PyObject_SetAttrString((o),((char *)(n)),(a))
     212@@ -209,7 +217,6 @@
     213   #define __Pyx_SetAttrString(o,n,a) PyObject_SetAttrString((o),(n),(a))
     214   #define __Pyx_DelAttrString(o,n)   PyObject_DelAttrString((o),(n))
     215 #endif
     216-
     217 #if PY_VERSION_HEX < 0x02050000
     218   #define __Pyx_NAMESTR(n) ((char *)(n))
     219   #define __Pyx_DOCSTR(n)  ((char *)(n))
     220@@ -218,6 +225,15 @@
     221   #define __Pyx_DOCSTR(n)  (n)
     222 #endif
     223 
     224+
     225+#if PY_MAJOR_VERSION >= 3
     226+  #define __Pyx_PyNumber_Divide(x,y)         PyNumber_TrueDivide(x,y)
     227+  #define __Pyx_PyNumber_InPlaceDivide(x,y)  PyNumber_InPlaceTrueDivide(x,y)
     228+#else
     229+  #define __Pyx_PyNumber_Divide(x,y)         PyNumber_Divide(x,y)
     230+  #define __Pyx_PyNumber_InPlaceDivide(x,y)  PyNumber_InPlaceDivide(x,y)
     231+#endif
     232+
     233 #ifndef __PYX_EXTERN_C
     234   #ifdef __cplusplus
     235     #define __PYX_EXTERN_C extern "C"
     236@@ -269,7 +285,7 @@
     237 #   else
     238 #     define CYTHON_UNUSED
     239 #   endif
     240-# elif defined(__ICC) || defined(__INTEL_COMPILER)
     241+# elif defined(__ICC) || (defined(__INTEL_COMPILER) && !defined(_MSC_VER))
     242 #   define CYTHON_UNUSED __attribute__ ((__unused__))
     243 # else
     244 #   define CYTHON_UNUSED
     245@@ -293,8 +309,12 @@
     246 static CYTHON_INLINE PyObject * __Pyx_PyInt_FromSize_t(size_t);
     247 static CYTHON_INLINE size_t __Pyx_PyInt_AsSize_t(PyObject*);
     248 
     249+#if CYTHON_COMPILING_IN_CPYTHON
     250 #define __pyx_PyFloat_AsDouble(x) (PyFloat_CheckExact(x) ? PyFloat_AS_DOUBLE(x) : PyFloat_AsDouble(x))
     251-
     252+#else
     253+#define __pyx_PyFloat_AsDouble(x) PyFloat_AsDouble(x)
     254+#endif
     255+#define __pyx_PyFloat_AsFloat(x) ((float) __pyx_PyFloat_AsDouble(x))
     256 
     257 #ifdef __GNUC__
     258   /* Test for GCC > 2.95 */
     259@@ -421,11 +441,9 @@
     260   rk_state *internal_state;
     261 };
     262 
     263-
     264 #ifndef CYTHON_REFNANNY
     265   #define CYTHON_REFNANNY 0
     266 #endif
     267-
     268 #if CYTHON_REFNANNY
     269   typedef struct {
     270     void (*INCREF)(void*, PyObject*, int);
     271@@ -438,8 +456,21 @@
     272   static __Pyx_RefNannyAPIStruct *__Pyx_RefNanny = NULL;
     273   static __Pyx_RefNannyAPIStruct *__Pyx_RefNannyImportAPI(const char *modname); /*proto*/
     274   #define __Pyx_RefNannyDeclarations void *__pyx_refnanny = NULL;
     275-  #define __Pyx_RefNannySetupContext(name)           __pyx_refnanny = __Pyx_RefNanny->SetupContext((name), __LINE__, __FILE__)
     276-  #define __Pyx_RefNannyFinishContext()           __Pyx_RefNanny->FinishContext(&__pyx_refnanny)
     277+#ifdef WITH_THREAD
     278+  #define __Pyx_RefNannySetupContext(name, acquire_gil) \
     279+          if (acquire_gil) { \
     280+              PyGILState_STATE __pyx_gilstate_save = PyGILState_Ensure(); \
     281+              __pyx_refnanny = __Pyx_RefNanny->SetupContext((name), __LINE__, __FILE__); \
     282+              PyGILState_Release(__pyx_gilstate_save); \
     283+          } else { \
     284+              __pyx_refnanny = __Pyx_RefNanny->SetupContext((name), __LINE__, __FILE__); \
     285+          }
     286+#else
     287+  #define __Pyx_RefNannySetupContext(name, acquire_gil) \
     288+          __pyx_refnanny = __Pyx_RefNanny->SetupContext((name), __LINE__, __FILE__)
     289+#endif
     290+  #define __Pyx_RefNannyFinishContext() \
     291+          __Pyx_RefNanny->FinishContext(&__pyx_refnanny)
     292   #define __Pyx_INCREF(r)  __Pyx_RefNanny->INCREF(__pyx_refnanny, (PyObject *)(r), __LINE__)
     293   #define __Pyx_DECREF(r)  __Pyx_RefNanny->DECREF(__pyx_refnanny, (PyObject *)(r), __LINE__)
     294   #define __Pyx_GOTREF(r)  __Pyx_RefNanny->GOTREF(__pyx_refnanny, (PyObject *)(r), __LINE__)
     295@@ -450,7 +481,7 @@
     296   #define __Pyx_XGIVEREF(r) do { if((r) != NULL) {__Pyx_GIVEREF(r);}} while(0)
     297 #else
     298   #define __Pyx_RefNannyDeclarations
     299-  #define __Pyx_RefNannySetupContext(name)
     300+  #define __Pyx_RefNannySetupContext(name, acquire_gil)
     301   #define __Pyx_RefNannyFinishContext()
     302   #define __Pyx_INCREF(r) Py_INCREF(r)
     303   #define __Pyx_DECREF(r) Py_DECREF(r)
     304@@ -461,6 +492,8 @@
     305   #define __Pyx_XGOTREF(r)
     306   #define __Pyx_XGIVEREF(r)
     307 #endif /* CYTHON_REFNANNY */
     308+#define __Pyx_CLEAR(r)    do { PyObject* tmp = ((PyObject*)(r)); r = NULL; __Pyx_DECREF(tmp);} while(0)
     309+#define __Pyx_XCLEAR(r)   do { if((r) != NULL) {PyObject* tmp = ((PyObject*)(r)); r = NULL; __Pyx_DECREF(tmp);}} while(0)
     310 
     311 static PyObject *__Pyx_GetName(PyObject *dict, PyObject *name); /*proto*/
     312 
     313@@ -469,15 +502,15 @@
     314 
     315 static void __Pyx_Raise(PyObject *type, PyObject *value, PyObject *tb, PyObject *cause); /*proto*/
     316 
     317-static void __Pyx_RaiseDoubleKeywordsError(
     318-    const char* func_name, PyObject* kw_name); /*proto*/
     319+static void __Pyx_RaiseDoubleKeywordsError(const char* func_name, PyObject* kw_name); /*proto*/
     320 
     321-static int __Pyx_ParseOptionalKeywords(PyObject *kwds, PyObject **argnames[],     PyObject *kwds2, PyObject *values[], Py_ssize_t num_pos_args,     const char* function_name); /*proto*/
     322+static int __Pyx_ParseOptionalKeywords(PyObject *kwds, PyObject **argnames[], \
     323+    PyObject *kwds2, PyObject *values[], Py_ssize_t num_pos_args, \
     324+    const char* function_name); /*proto*/
     325 
     326 static void __Pyx_RaiseArgtupleInvalid(const char* func_name, int exact,
     327     Py_ssize_t num_min, Py_ssize_t num_max, Py_ssize_t num_found); /*proto*/
     328 
     329-
     330 static CYTHON_INLINE PyObject *__Pyx_GetItemInt_Generic(PyObject *o, PyObject* j) {
     331     PyObject *r;
     332     if (!j) return NULL;
     333@@ -485,89 +518,102 @@
     334     Py_DECREF(j);
     335     return r;
     336 }
     337-
     338-
     339 #define __Pyx_GetItemInt_List(o, i, size, to_py_func) (((size) <= sizeof(Py_ssize_t)) ? \
     340                                                     __Pyx_GetItemInt_List_Fast(o, i) : \
     341                                                     __Pyx_GetItemInt_Generic(o, to_py_func(i)))
     342-
     343 static CYTHON_INLINE PyObject *__Pyx_GetItemInt_List_Fast(PyObject *o, Py_ssize_t i) {
     344-    if (likely(o != Py_None)) {
     345-        if (likely((0 <= i) & (i < PyList_GET_SIZE(o)))) {
     346-            PyObject *r = PyList_GET_ITEM(o, i);
     347-            Py_INCREF(r);
     348-            return r;
     349-        }
     350-        else if ((-PyList_GET_SIZE(o) <= i) & (i < 0)) {
     351-            PyObject *r = PyList_GET_ITEM(o, PyList_GET_SIZE(o) + i);
     352-            Py_INCREF(r);
     353-            return r;
     354-        }
     355+#if CYTHON_COMPILING_IN_CPYTHON
     356+    if (likely((0 <= i) & (i < PyList_GET_SIZE(o)))) {
     357+        PyObject *r = PyList_GET_ITEM(o, i);
     358+        Py_INCREF(r);
     359+        return r;
     360+    }
     361+    else if ((-PyList_GET_SIZE(o) <= i) & (i < 0)) {
     362+        PyObject *r = PyList_GET_ITEM(o, PyList_GET_SIZE(o) + i);
     363+        Py_INCREF(r);
     364+        return r;
     365     }
     366     return __Pyx_GetItemInt_Generic(o, PyInt_FromSsize_t(i));
     367+#else
     368+    return PySequence_GetItem(o, i);
     369+#endif
     370 }
     371-
     372 #define __Pyx_GetItemInt_Tuple(o, i, size, to_py_func) (((size) <= sizeof(Py_ssize_t)) ? \
     373                                                     __Pyx_GetItemInt_Tuple_Fast(o, i) : \
     374                                                     __Pyx_GetItemInt_Generic(o, to_py_func(i)))
     375-
     376 static CYTHON_INLINE PyObject *__Pyx_GetItemInt_Tuple_Fast(PyObject *o, Py_ssize_t i) {
     377-    if (likely(o != Py_None)) {
     378-        if (likely((0 <= i) & (i < PyTuple_GET_SIZE(o)))) {
     379-            PyObject *r = PyTuple_GET_ITEM(o, i);
     380-            Py_INCREF(r);
     381-            return r;
     382-        }
     383-        else if ((-PyTuple_GET_SIZE(o) <= i) & (i < 0)) {
     384-            PyObject *r = PyTuple_GET_ITEM(o, PyTuple_GET_SIZE(o) + i);
     385-            Py_INCREF(r);
     386-            return r;
     387-        }
     388+#if CYTHON_COMPILING_IN_CPYTHON
     389+    if (likely((0 <= i) & (i < PyTuple_GET_SIZE(o)))) {
     390+        PyObject *r = PyTuple_GET_ITEM(o, i);
     391+        Py_INCREF(r);
     392+        return r;
     393+    }
     394+    else if ((-PyTuple_GET_SIZE(o) <= i) & (i < 0)) {
     395+        PyObject *r = PyTuple_GET_ITEM(o, PyTuple_GET_SIZE(o) + i);
     396+        Py_INCREF(r);
     397+        return r;
     398     }
     399     return __Pyx_GetItemInt_Generic(o, PyInt_FromSsize_t(i));
     400+#else
     401+    return PySequence_GetItem(o, i);
     402+#endif
     403 }
     404-
     405-
     406 #define __Pyx_GetItemInt(o, i, size, to_py_func) (((size) <= sizeof(Py_ssize_t)) ? \
     407                                                     __Pyx_GetItemInt_Fast(o, i) : \
     408                                                     __Pyx_GetItemInt_Generic(o, to_py_func(i)))
     409-
     410 static CYTHON_INLINE PyObject *__Pyx_GetItemInt_Fast(PyObject *o, Py_ssize_t i) {
     411-    PyObject *r;
     412-    if (PyList_CheckExact(o) && ((0 <= i) & (i < PyList_GET_SIZE(o)))) {
     413-        r = PyList_GET_ITEM(o, i);
     414-        Py_INCREF(r);
     415-    }
     416-    else if (PyTuple_CheckExact(o) && ((0 <= i) & (i < PyTuple_GET_SIZE(o)))) {
     417-        r = PyTuple_GET_ITEM(o, i);
     418-        Py_INCREF(r);
     419+#if CYTHON_COMPILING_IN_CPYTHON
     420+    if (PyList_CheckExact(o)) {
     421+        Py_ssize_t n = (likely(i >= 0)) ? i : i + PyList_GET_SIZE(o);
     422+        if (likely((n >= 0) & (n < PyList_GET_SIZE(o)))) {
     423+            PyObject *r = PyList_GET_ITEM(o, n);
     424+            Py_INCREF(r);
     425+            return r;
     426+        }
     427     }
     428-    else if (Py_TYPE(o)->tp_as_sequence && Py_TYPE(o)->tp_as_sequence->sq_item && (likely(i >= 0))) {
     429-        r = PySequence_GetItem(o, i);
     430+    else if (PyTuple_CheckExact(o)) {
     431+        Py_ssize_t n = (likely(i >= 0)) ? i : i + PyTuple_GET_SIZE(o);
     432+        if (likely((n >= 0) & (n < PyTuple_GET_SIZE(o)))) {
     433+            PyObject *r = PyTuple_GET_ITEM(o, n);
     434+            Py_INCREF(r);
     435+            return r;
     436+        }
     437+    } else {  /* inlined PySequence_GetItem() */
     438+        PySequenceMethods *m = Py_TYPE(o)->tp_as_sequence;
     439+        if (likely(m && m->sq_item)) {
     440+            if (unlikely(i < 0) && likely(m->sq_length)) {
     441+                Py_ssize_t l = m->sq_length(o);
     442+                if (unlikely(l < 0)) return NULL;
     443+                i += l;
     444+            }
     445+            return m->sq_item(o, i);
     446+        }
     447     }
     448-    else {
     449-        r = __Pyx_GetItemInt_Generic(o, PyInt_FromSsize_t(i));
     450+#else
     451+    if (PySequence_Check(o)) {
     452+        return PySequence_GetItem(o, i);
     453     }
     454-    return r;
     455+#endif
     456+    return __Pyx_GetItemInt_Generic(o, PyInt_FromSsize_t(i));
     457 }
     458 
     459+static CYTHON_INLINE void __Pyx_RaiseTooManyValuesError(Py_ssize_t expected);
     460+
     461 static CYTHON_INLINE void __Pyx_RaiseNeedMoreValuesError(Py_ssize_t index);
     462 
     463-static CYTHON_INLINE void __Pyx_RaiseTooManyValuesError(Py_ssize_t expected);
     464+static CYTHON_INLINE int __Pyx_IterFinish(void); /*proto*/
     465 
     466 static int __Pyx_IternextUnpackEndCheck(PyObject *retval, Py_ssize_t expected); /*proto*/
     467 
     468 static int __Pyx_GetException(PyObject **type, PyObject **value, PyObject **tb); /*proto*/
     469 
     470-static CYTHON_INLINE int __Pyx_CheckKeywordStrings(PyObject *kwdict,
     471-    const char* function_name, int kw_allowed); /*proto*/
     472+static CYTHON_INLINE int __Pyx_CheckKeywordStrings(PyObject *kwdict, const char* function_name, int kw_allowed); /*proto*/
     473 
     474 static CYTHON_INLINE int __Pyx_TypeTest(PyObject *obj, PyTypeObject *type); /*proto*/
     475 
     476 #define __Pyx_SetItemInt(o, i, v, size, to_py_func) (((size) <= sizeof(Py_ssize_t)) ? \
     477                                                     __Pyx_SetItemInt_Fast(o, i, v) : \
     478                                                     __Pyx_SetItemInt_Generic(o, to_py_func(i), v))
     479-
     480 static CYTHON_INLINE int __Pyx_SetItemInt_Generic(PyObject *o, PyObject *j, PyObject *v) {
     481     int r;
     482     if (!j) return -1;
     483@@ -575,20 +621,38 @@
     484     Py_DECREF(j);
     485     return r;
     486 }
     487-
     488 static CYTHON_INLINE int __Pyx_SetItemInt_Fast(PyObject *o, Py_ssize_t i, PyObject *v) {
     489-    if (PyList_CheckExact(o) && ((0 <= i) & (i < PyList_GET_SIZE(o)))) {
     490-        Py_INCREF(v);
     491-        Py_DECREF(PyList_GET_ITEM(o, i));
     492-        PyList_SET_ITEM(o, i, v);
     493-        return 1;
     494+#if CYTHON_COMPILING_IN_CPYTHON
     495+    if (PyList_CheckExact(o)) {
     496+        Py_ssize_t n = (likely(i >= 0)) ? i : i + PyList_GET_SIZE(o);
     497+        if (likely((n >= 0) & (n < PyList_GET_SIZE(o)))) {
     498+            PyObject* old = PyList_GET_ITEM(o, n);
     499+            Py_INCREF(v);
     500+            PyList_SET_ITEM(o, n, v);
     501+            Py_DECREF(old);
     502+            return 1;
     503+        }
     504+    } else {  /* inlined PySequence_SetItem() */
     505+        PySequenceMethods *m = Py_TYPE(o)->tp_as_sequence;
     506+        if (likely(m && m->sq_ass_item)) {
     507+            if (unlikely(i < 0) && likely(m->sq_length)) {
     508+                Py_ssize_t l = m->sq_length(o);
     509+                if (unlikely(l < 0)) return -1;
     510+                i += l;
     511+            }
     512+            return m->sq_ass_item(o, i, v);
     513+        }
     514     }
     515-    else if (Py_TYPE(o)->tp_as_sequence && Py_TYPE(o)->tp_as_sequence->sq_ass_item && (likely(i >= 0)))
     516+#else
     517+#if CYTHON_COMPILING_IN_PYPY
     518+    if (PySequence_Check(o) && !PyDict_Check(o)) {
     519+#else
     520+    if (PySequence_Check(o)) {
     521+#endif
     522         return PySequence_SetItem(o, i, v);
     523-    else {
     524-        PyObject *j = PyInt_FromSsize_t(i);
     525-        return __Pyx_SetItemInt_Generic(o, j, v);
     526     }
     527+#endif
     528+    return __Pyx_SetItemInt_Generic(o, PyInt_FromSsize_t(i), v);
     529 }
     530 
     531 static CYTHON_INLINE void __Pyx_ExceptionSave(PyObject **type, PyObject **value, PyObject **tb); /*proto*/
     532@@ -596,17 +660,9 @@
     533 
     534 static PyObject *__Pyx_Import(PyObject *name, PyObject *from_list, long level); /*proto*/
     535 
     536-#include <string.h>
     537-
     538-static CYTHON_INLINE int __Pyx_PyBytes_Equals(PyObject* s1, PyObject* s2, int equals); /*proto*/
     539-
     540-static CYTHON_INLINE int __Pyx_PyUnicode_Equals(PyObject* s1, PyObject* s2, int equals); /*proto*/
     541+static CYTHON_INLINE npy_intp __Pyx_PyInt_from_py_npy_intp(PyObject *);
     542 
     543-#if PY_MAJOR_VERSION >= 3
     544-#define __Pyx_PyString_Equals __Pyx_PyUnicode_Equals
     545-#else
     546-#define __Pyx_PyString_Equals __Pyx_PyBytes_Equals
     547-#endif
     548+static CYTHON_INLINE void __Pyx_RaiseImportError(PyObject *name);
     549 
     550 static CYTHON_INLINE PyObject *__Pyx_PyInt_to_py_npy_intp(npy_intp);
     551 
     552@@ -642,19 +698,40 @@
     553 
     554 static CYTHON_INLINE signed PY_LONG_LONG __Pyx_PyInt_AsSignedLongLong(PyObject *);
     555 
     556-static CYTHON_INLINE npy_intp __Pyx_PyInt_from_py_npy_intp(PyObject *);
     557-
     558 static int __Pyx_check_binary_version(void);
     559 
     560-static PyTypeObject *__Pyx_ImportType(const char *module_name, const char *class_name, size_t size, int strict);  /*proto*/
     561+#if !defined(__Pyx_PyIdentifier_FromString)
     562+#if PY_MAJOR_VERSION < 3
     563+  #define __Pyx_PyIdentifier_FromString(s) PyString_FromString(s)
     564+#else
     565+  #define __Pyx_PyIdentifier_FromString(s) PyUnicode_FromString(s)
     566+#endif
     567+#endif
     568 
     569 static PyObject *__Pyx_ImportModule(const char *name); /*proto*/
     570 
     571-static void __Pyx_AddTraceback(const char *funcname, int __pyx_clineno,
     572-                               int __pyx_lineno, const char *__pyx_filename); /*proto*/
     573+static PyTypeObject *__Pyx_ImportType(const char *module_name, const char *class_name, size_t size, int strict);  /*proto*/
     574+
     575+typedef struct {
     576+    int code_line;
     577+    PyCodeObject* code_object;
     578+} __Pyx_CodeObjectCacheEntry;
     579+struct __Pyx_CodeObjectCache {
     580+    int count;
     581+    int max_count;
     582+    __Pyx_CodeObjectCacheEntry* entries;
     583+};
     584+static struct __Pyx_CodeObjectCache __pyx_code_cache = {0,0,NULL};
     585+static int __pyx_bisect_code_objects(__Pyx_CodeObjectCacheEntry* entries, int count, int code_line);
     586+static PyCodeObject *__pyx_find_code_object(int code_line);
     587+static void __pyx_insert_code_object(int code_line, PyCodeObject* code_object);
     588+
     589+static void __Pyx_AddTraceback(const char *funcname, int c_line,
     590+                               int py_line, const char *filename); /*proto*/
     591 
     592 static int __Pyx_InitStrings(__Pyx_StringTabEntry *t); /*proto*/
     593 
     594+
     595 /* Module declarations from 'numpy' */
     596 
     597 /* Module declarations from 'mtrand' */
     598@@ -686,6 +763,58 @@
     599 /* Implementation of 'mtrand' */
     600 static PyObject *__pyx_builtin_ValueError;
     601 static PyObject *__pyx_builtin_TypeError;
     602+static int __pyx_pf_6mtrand_11RandomState___init__(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_seed); /* proto */
     603+static void __pyx_pf_6mtrand_11RandomState_2__dealloc__(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self); /* proto */
     604+static PyObject *__pyx_pf_6mtrand_11RandomState_4seed(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_seed); /* proto */
     605+static PyObject *__pyx_pf_6mtrand_11RandomState_6get_state(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self); /* proto */
     606+static PyObject *__pyx_pf_6mtrand_11RandomState_8set_state(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_state); /* proto */
     607+static PyObject *__pyx_pf_6mtrand_11RandomState_10__getstate__(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self); /* proto */
     608+static PyObject *__pyx_pf_6mtrand_11RandomState_12__setstate__(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_state); /* proto */
     609+static PyObject *__pyx_pf_6mtrand_11RandomState_14__reduce__(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self); /* proto */
     610+static PyObject *__pyx_pf_6mtrand_11RandomState_16random_sample(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_size); /* proto */
     611+static PyObject *__pyx_pf_6mtrand_11RandomState_18tomaxint(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_size); /* proto */
     612+static PyObject *__pyx_pf_6mtrand_11RandomState_20randint(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_low, PyObject *__pyx_v_high, PyObject *__pyx_v_size); /* proto */
     613+static PyObject *__pyx_pf_6mtrand_11RandomState_22bytes(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, npy_intp __pyx_v_length); /* proto */
     614+static PyObject *__pyx_pf_6mtrand_11RandomState_24uniform(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_low, PyObject *__pyx_v_high, PyObject *__pyx_v_size); /* proto */
     615+static PyObject *__pyx_pf_6mtrand_11RandomState_26rand(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_args); /* proto */
     616+static PyObject *__pyx_pf_6mtrand_11RandomState_28randn(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_args); /* proto */
     617+static PyObject *__pyx_pf_6mtrand_11RandomState_30random_integers(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_low, PyObject *__pyx_v_high, PyObject *__pyx_v_size); /* proto */
     618+static PyObject *__pyx_pf_6mtrand_11RandomState_32standard_normal(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_size); /* proto */
     619+static PyObject *__pyx_pf_6mtrand_11RandomState_34normal(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_loc, PyObject *__pyx_v_scale, PyObject *__pyx_v_size); /* proto */
     620+static PyObject *__pyx_pf_6mtrand_11RandomState_36beta(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_a, PyObject *__pyx_v_b, PyObject *__pyx_v_size); /* proto */
     621+static PyObject *__pyx_pf_6mtrand_11RandomState_38exponential(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_scale, PyObject *__pyx_v_size); /* proto */
     622+static PyObject *__pyx_pf_6mtrand_11RandomState_40standard_exponential(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_size); /* proto */
     623+static PyObject *__pyx_pf_6mtrand_11RandomState_42standard_gamma(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_shape, PyObject *__pyx_v_size); /* proto */
     624+static PyObject *__pyx_pf_6mtrand_11RandomState_44gamma(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_shape, PyObject *__pyx_v_scale, PyObject *__pyx_v_size); /* proto */
     625+static PyObject *__pyx_pf_6mtrand_11RandomState_46f(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_dfnum, PyObject *__pyx_v_dfden, PyObject *__pyx_v_size); /* proto */
     626+static PyObject *__pyx_pf_6mtrand_11RandomState_48noncentral_f(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_dfnum, PyObject *__pyx_v_dfden, PyObject *__pyx_v_nonc, PyObject *__pyx_v_size); /* proto */
     627+static PyObject *__pyx_pf_6mtrand_11RandomState_50chisquare(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_df, PyObject *__pyx_v_size); /* proto */
     628+static PyObject *__pyx_pf_6mtrand_11RandomState_52noncentral_chisquare(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_df, PyObject *__pyx_v_nonc, PyObject *__pyx_v_size); /* proto */
     629+static PyObject *__pyx_pf_6mtrand_11RandomState_54standard_cauchy(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_size); /* proto */
     630+static PyObject *__pyx_pf_6mtrand_11RandomState_56standard_t(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_df, PyObject *__pyx_v_size); /* proto */
     631+static PyObject *__pyx_pf_6mtrand_11RandomState_58vonmises(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_mu, PyObject *__pyx_v_kappa, PyObject *__pyx_v_size); /* proto */
     632+static PyObject *__pyx_pf_6mtrand_11RandomState_60pareto(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_a, PyObject *__pyx_v_size); /* proto */
     633+static PyObject *__pyx_pf_6mtrand_11RandomState_62weibull(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_a, PyObject *__pyx_v_size); /* proto */
     634+static PyObject *__pyx_pf_6mtrand_11RandomState_64power(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_a, PyObject *__pyx_v_size); /* proto */
     635+static PyObject *__pyx_pf_6mtrand_11RandomState_66laplace(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_loc, PyObject *__pyx_v_scale, PyObject *__pyx_v_size); /* proto */
     636+static PyObject *__pyx_pf_6mtrand_11RandomState_68gumbel(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_loc, PyObject *__pyx_v_scale, PyObject *__pyx_v_size); /* proto */
     637+static PyObject *__pyx_pf_6mtrand_11RandomState_70logistic(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_loc, PyObject *__pyx_v_scale, PyObject *__pyx_v_size); /* proto */
     638+static PyObject *__pyx_pf_6mtrand_11RandomState_72lognormal(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_mean, PyObject *__pyx_v_sigma, PyObject *__pyx_v_size); /* proto */
     639+static PyObject *__pyx_pf_6mtrand_11RandomState_74rayleigh(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_scale, PyObject *__pyx_v_size); /* proto */
     640+static PyObject *__pyx_pf_6mtrand_11RandomState_76wald(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_mean, PyObject *__pyx_v_scale, PyObject *__pyx_v_size); /* proto */
     641+static PyObject *__pyx_pf_6mtrand_11RandomState_78triangular(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_left, PyObject *__pyx_v_mode, PyObject *__pyx_v_right, PyObject *__pyx_v_size); /* proto */
     642+static PyObject *__pyx_pf_6mtrand_11RandomState_80binomial(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_n, PyObject *__pyx_v_p, PyObject *__pyx_v_size); /* proto */
     643+static PyObject *__pyx_pf_6mtrand_11RandomState_82negative_binomial(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_n, PyObject *__pyx_v_p, PyObject *__pyx_v_size); /* proto */
     644+static PyObject *__pyx_pf_6mtrand_11RandomState_84poisson(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_lam, PyObject *__pyx_v_size); /* proto */
     645+static PyObject *__pyx_pf_6mtrand_11RandomState_86zipf(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_a, PyObject *__pyx_v_size); /* proto */
     646+static PyObject *__pyx_pf_6mtrand_11RandomState_88geometric(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_p, PyObject *__pyx_v_size); /* proto */
     647+static PyObject *__pyx_pf_6mtrand_11RandomState_90hypergeometric(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_ngood, PyObject *__pyx_v_nbad, PyObject *__pyx_v_nsample, PyObject *__pyx_v_size); /* proto */
     648+static PyObject *__pyx_pf_6mtrand_11RandomState_92logseries(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_p, PyObject *__pyx_v_size); /* proto */
     649+static PyObject *__pyx_pf_6mtrand_11RandomState_94multivariate_normal(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_mean, PyObject *__pyx_v_cov, PyObject *__pyx_v_size); /* proto */
     650+static PyObject *__pyx_pf_6mtrand_11RandomState_96multinomial(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, npy_intp __pyx_v_n, PyObject *__pyx_v_pvals, PyObject *__pyx_v_size); /* proto */
     651+static PyObject *__pyx_pf_6mtrand_11RandomState_98dirichlet(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_alpha, PyObject *__pyx_v_size); /* proto */
     652+static PyObject *__pyx_pf_6mtrand_11RandomState_100shuffle(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_x); /* proto */
     653+static PyObject *__pyx_pf_6mtrand_11RandomState_102permutation(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_x); /* proto */
     654 static char __pyx_k_1[] = "size is not compatible with inputs";
     655 static char __pyx_k_9[] = "algorithm must be 'MT19937'";
     656 static char __pyx_k_11[] = "state must be 624 longs";
     657@@ -1318,7 +1447,7 @@
     658   int __pyx_lineno = 0;
     659   const char *__pyx_filename = NULL;
     660   int __pyx_clineno = 0;
     661-  __Pyx_RefNannySetupContext("cont0_array");
     662+  __Pyx_RefNannySetupContext("cont0_array", 0);
     663 
     664   /* "mtrand.pyx":134
     665  *     cdef npy_intp i
     666@@ -1365,7 +1494,7 @@
     667     __Pyx_GOTREF(__pyx_t_4);
     668     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     669     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 137; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     670-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     671+    __Pyx_GOTREF(__pyx_t_2);
     672     __Pyx_INCREF(__pyx_v_size);
     673     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     674     __Pyx_GIVEREF(__pyx_v_size);
     675@@ -1470,7 +1599,7 @@
     676   int __pyx_lineno = 0;
     677   const char *__pyx_filename = NULL;
     678   int __pyx_clineno = 0;
     679-  __Pyx_RefNannySetupContext("cont1_array_sc");
     680+  __Pyx_RefNannySetupContext("cont1_array_sc", 0);
     681 
     682   /* "mtrand.pyx":151
     683  *     cdef npy_intp i
     684@@ -1517,7 +1646,7 @@
     685     __Pyx_GOTREF(__pyx_t_4);
     686     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     687     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 154; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     688-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     689+    __Pyx_GOTREF(__pyx_t_2);
     690     __Pyx_INCREF(__pyx_v_size);
     691     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     692     __Pyx_GIVEREF(__pyx_v_size);
     693@@ -1625,7 +1754,7 @@
     694   int __pyx_lineno = 0;
     695   const char *__pyx_filename = NULL;
     696   int __pyx_clineno = 0;
     697-  __Pyx_RefNannySetupContext("cont1_array");
     698+  __Pyx_RefNannySetupContext("cont1_array", 0);
     699 
     700   /* "mtrand.pyx":170
     701  *     cdef broadcast multi
     702@@ -1731,7 +1860,7 @@
     703     __Pyx_GOTREF(__pyx_t_5);
     704     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     705     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 179; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     706-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     707+    __Pyx_GOTREF(__pyx_t_2);
     708     __Pyx_INCREF(__pyx_v_size);
     709     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     710     __Pyx_GIVEREF(__pyx_v_size);
     711@@ -1886,7 +2015,7 @@
     712   int __pyx_lineno = 0;
     713   const char *__pyx_filename = NULL;
     714   int __pyx_clineno = 0;
     715-  __Pyx_RefNannySetupContext("cont2_array_sc");
     716+  __Pyx_RefNannySetupContext("cont2_array_sc", 0);
     717 
     718   /* "mtrand.pyx":198
     719  *     cdef npy_intp i
     720@@ -1933,7 +2062,7 @@
     721     __Pyx_GOTREF(__pyx_t_4);
     722     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     723     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 201; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     724-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     725+    __Pyx_GOTREF(__pyx_t_2);
     726     __Pyx_INCREF(__pyx_v_size);
     727     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     728     __Pyx_GIVEREF(__pyx_v_size);
     729@@ -2040,7 +2169,7 @@
     730   int __pyx_lineno = 0;
     731   const char *__pyx_filename = NULL;
     732   int __pyx_clineno = 0;
     733-  __Pyx_RefNannySetupContext("cont2_array");
     734+  __Pyx_RefNannySetupContext("cont2_array", 0);
     735 
     736   /* "mtrand.pyx":219
     737  *     cdef broadcast multi
     738@@ -2155,7 +2284,7 @@
     739     __Pyx_GOTREF(__pyx_t_5);
     740     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     741     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 229; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     742-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     743+    __Pyx_GOTREF(__pyx_t_2);
     744     __Pyx_INCREF(__pyx_v_size);
     745     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     746     __Pyx_GIVEREF(__pyx_v_size);
     747@@ -2327,7 +2456,7 @@
     748   int __pyx_lineno = 0;
     749   const char *__pyx_filename = NULL;
     750   int __pyx_clineno = 0;
     751-  __Pyx_RefNannySetupContext("cont3_array_sc");
     752+  __Pyx_RefNannySetupContext("cont3_array_sc", 0);
     753 
     754   /* "mtrand.pyx":250
     755  *     cdef npy_intp i
     756@@ -2374,7 +2503,7 @@
     757     __Pyx_GOTREF(__pyx_t_4);
     758     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     759     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 253; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     760-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     761+    __Pyx_GOTREF(__pyx_t_2);
     762     __Pyx_INCREF(__pyx_v_size);
     763     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     764     __Pyx_GIVEREF(__pyx_v_size);
     765@@ -2482,7 +2611,7 @@
     766   int __pyx_lineno = 0;
     767   const char *__pyx_filename = NULL;
     768   int __pyx_clineno = 0;
     769-  __Pyx_RefNannySetupContext("cont3_array");
     770+  __Pyx_RefNannySetupContext("cont3_array", 0);
     771 
     772   /* "mtrand.pyx":272
     773  *     cdef broadcast multi
     774@@ -2606,7 +2735,7 @@
     775     __Pyx_GOTREF(__pyx_t_5);
     776     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     777     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 283; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     778-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     779+    __Pyx_GOTREF(__pyx_t_2);
     780     __Pyx_INCREF(__pyx_v_size);
     781     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     782     __Pyx_GIVEREF(__pyx_v_size);
     783@@ -2778,7 +2907,7 @@
     784   int __pyx_lineno = 0;
     785   const char *__pyx_filename = NULL;
     786   int __pyx_clineno = 0;
     787-  __Pyx_RefNannySetupContext("disc0_array");
     788+  __Pyx_RefNannySetupContext("disc0_array", 0);
     789 
     790   /* "mtrand.pyx":303
     791  *     cdef npy_intp i
     792@@ -2820,7 +2949,7 @@
     793     __Pyx_GOTREF(__pyx_t_3);
     794     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     795     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 306; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     796-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     797+    __Pyx_GOTREF(__pyx_t_2);
     798     __Pyx_INCREF(__pyx_v_size);
     799     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     800     __Pyx_GIVEREF(__pyx_v_size);
     801@@ -2925,7 +3054,7 @@
     802   int __pyx_lineno = 0;
     803   const char *__pyx_filename = NULL;
     804   int __pyx_clineno = 0;
     805-  __Pyx_RefNannySetupContext("discnp_array_sc");
     806+  __Pyx_RefNannySetupContext("discnp_array_sc", 0);
     807 
     808   /* "mtrand.pyx":319
     809  *     cdef npy_intp i
     810@@ -2967,7 +3096,7 @@
     811     __Pyx_GOTREF(__pyx_t_3);
     812     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     813     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 322; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     814-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     815+    __Pyx_GOTREF(__pyx_t_2);
     816     __Pyx_INCREF(__pyx_v_size);
     817     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     818     __Pyx_GIVEREF(__pyx_v_size);
     819@@ -3074,7 +3203,7 @@
     820   int __pyx_lineno = 0;
     821   const char *__pyx_filename = NULL;
     822   int __pyx_clineno = 0;
     823-  __Pyx_RefNannySetupContext("discnp_array");
     824+  __Pyx_RefNannySetupContext("discnp_array", 0);
     825 
     826   /* "mtrand.pyx":338
     827  *     cdef broadcast multi
     828@@ -3184,7 +3313,7 @@
     829     __Pyx_GOTREF(__pyx_t_4);
     830     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     831     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 348; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     832-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     833+    __Pyx_GOTREF(__pyx_t_2);
     834     __Pyx_INCREF(__pyx_v_size);
     835     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     836     __Pyx_GIVEREF(__pyx_v_size);
     837@@ -3356,7 +3485,7 @@
     838   int __pyx_lineno = 0;
     839   const char *__pyx_filename = NULL;
     840   int __pyx_clineno = 0;
     841-  __Pyx_RefNannySetupContext("discdd_array_sc");
     842+  __Pyx_RefNannySetupContext("discdd_array_sc", 0);
     843 
     844   /* "mtrand.pyx":368
     845  *     cdef npy_intp i
     846@@ -3398,7 +3527,7 @@
     847     __Pyx_GOTREF(__pyx_t_3);
     848     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     849     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 371; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     850-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     851+    __Pyx_GOTREF(__pyx_t_2);
     852     __Pyx_INCREF(__pyx_v_size);
     853     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     854     __Pyx_GIVEREF(__pyx_v_size);
     855@@ -3505,7 +3634,7 @@
     856   int __pyx_lineno = 0;
     857   const char *__pyx_filename = NULL;
     858   int __pyx_clineno = 0;
     859-  __Pyx_RefNannySetupContext("discdd_array");
     860+  __Pyx_RefNannySetupContext("discdd_array", 0);
     861 
     862   /* "mtrand.pyx":387
     863  *     cdef broadcast multi
     864@@ -3615,7 +3744,7 @@
     865     __Pyx_GOTREF(__pyx_t_4);
     866     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     867     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 397; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     868-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     869+    __Pyx_GOTREF(__pyx_t_2);
     870     __Pyx_INCREF(__pyx_v_size);
     871     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     872     __Pyx_GIVEREF(__pyx_v_size);
     873@@ -3787,7 +3916,7 @@
     874   int __pyx_lineno = 0;
     875   const char *__pyx_filename = NULL;
     876   int __pyx_clineno = 0;
     877-  __Pyx_RefNannySetupContext("discnmN_array_sc");
     878+  __Pyx_RefNannySetupContext("discnmN_array_sc", 0);
     879 
     880   /* "mtrand.pyx":418
     881  *     cdef npy_intp i
     882@@ -3829,7 +3958,7 @@
     883     __Pyx_GOTREF(__pyx_t_3);
     884     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     885     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 421; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     886-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     887+    __Pyx_GOTREF(__pyx_t_2);
     888     __Pyx_INCREF(__pyx_v_size);
     889     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     890     __Pyx_GIVEREF(__pyx_v_size);
     891@@ -3937,7 +4066,7 @@
     892   int __pyx_lineno = 0;
     893   const char *__pyx_filename = NULL;
     894   int __pyx_clineno = 0;
     895-  __Pyx_RefNannySetupContext("discnmN_array");
     896+  __Pyx_RefNannySetupContext("discnmN_array", 0);
     897 
     898   /* "mtrand.pyx":439
     899  *     cdef broadcast multi
     900@@ -4056,7 +4185,7 @@
     901     __Pyx_GOTREF(__pyx_t_4);
     902     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     903     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 450; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     904-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     905+    __Pyx_GOTREF(__pyx_t_2);
     906     __Pyx_INCREF(__pyx_v_size);
     907     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     908     __Pyx_GIVEREF(__pyx_v_size);
     909@@ -4228,7 +4357,7 @@
     910   int __pyx_lineno = 0;
     911   const char *__pyx_filename = NULL;
     912   int __pyx_clineno = 0;
     913-  __Pyx_RefNannySetupContext("discd_array_sc");
     914+  __Pyx_RefNannySetupContext("discd_array_sc", 0);
     915 
     916   /* "mtrand.pyx":471
     917  *     cdef npy_intp i
     918@@ -4270,7 +4399,7 @@
     919     __Pyx_GOTREF(__pyx_t_3);
     920     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     921     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 474; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     922-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     923+    __Pyx_GOTREF(__pyx_t_2);
     924     __Pyx_INCREF(__pyx_v_size);
     925     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     926     __Pyx_GIVEREF(__pyx_v_size);
     927@@ -4378,7 +4507,7 @@
     928   int __pyx_lineno = 0;
     929   const char *__pyx_filename = NULL;
     930   int __pyx_clineno = 0;
     931-  __Pyx_RefNannySetupContext("discd_array");
     932+  __Pyx_RefNannySetupContext("discd_array", 0);
     933 
     934   /* "mtrand.pyx":490
     935  *     cdef flatiter itera
     936@@ -4479,7 +4608,7 @@
     937     __Pyx_GOTREF(__pyx_t_4);
     938     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     939     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 499; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     940-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     941+    __Pyx_GOTREF(__pyx_t_2);
     942     __Pyx_INCREF(__pyx_v_size);
     943     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     944     __Pyx_GIVEREF(__pyx_v_size);
     945@@ -4628,7 +4757,7 @@
     946   double __pyx_r;
     947   __Pyx_RefNannyDeclarations
     948   npy_intp __pyx_t_1;
     949-  __Pyx_RefNannySetupContext("kahan_sum");
     950+  __Pyx_RefNannySetupContext("kahan_sum", 0);
     951 
     952   /* "mtrand.pyx":513
     953  *     cdef double c, y, t, sum
     954@@ -4711,39 +4840,35 @@
     955   return __pyx_r;
     956 }
     957 
     958-/* "mtrand.pyx":558
     959+/* Python wrapper */
     960+static int __pyx_pw_6mtrand_11RandomState_1__init__(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     961+static int __pyx_pw_6mtrand_11RandomState_1__init__(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     962+  PyObject *__pyx_v_seed = 0;
     963+  int __pyx_r;
     964+  __Pyx_RefNannyDeclarations
     965+  __Pyx_RefNannySetupContext("__init__ (wrapper)", 0);
     966+  {
     967+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__seed,0};
     968+    PyObject* values[1] = {0};
     969+
     970+    /* "mtrand.pyx":558
     971  *     poisson_lam_max = np.iinfo('l').max - np.sqrt(np.iinfo('l').max)*10
     972  *
     973  *     def __init__(self, seed=None):             # <<<<<<<<<<<<<<
     974  *         self.internal_state = <rk_state*>PyMem_Malloc(sizeof(rk_state))
     975  *
     976  */
     977-
     978-static int __pyx_pf_6mtrand_11RandomState___init__(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     979-static int __pyx_pf_6mtrand_11RandomState___init__(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     980-  PyObject *__pyx_v_seed = 0;
     981-  int __pyx_r;
     982-  __Pyx_RefNannyDeclarations
     983-  PyObject *__pyx_t_1 = NULL;
     984-  PyObject *__pyx_t_2 = NULL;
     985-  PyObject *__pyx_t_3 = NULL;
     986-  int __pyx_lineno = 0;
     987-  const char *__pyx_filename = NULL;
     988-  int __pyx_clineno = 0;
     989-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__seed,0};
     990-  __Pyx_RefNannySetupContext("__init__");
     991-  {
     992-    PyObject* values[1] = {0};
     993     values[0] = ((PyObject *)Py_None);
     994     if (unlikely(__pyx_kwds)) {
     995       Py_ssize_t kw_args;
     996-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     997+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     998+      switch (pos_args) {
     999         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     1000         case  0: break;
     1001         default: goto __pyx_L5_argtuple_error;
     1002       }
     1003       kw_args = PyDict_Size(__pyx_kwds);
     1004-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     1005+      switch (pos_args) {
     1006         case  0:
     1007         if (kw_args > 0) {
     1008           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__seed);
     1009@@ -4751,7 +4876,7 @@
     1010         }
     1011       }
     1012       if (unlikely(kw_args > 0)) {
     1013-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "__init__") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 558; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     1014+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 558; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     1015       }
     1016     } else {
     1017       switch (PyTuple_GET_SIZE(__pyx_args)) {
     1018@@ -4770,6 +4895,21 @@
     1019   __Pyx_RefNannyFinishContext();
     1020   return -1;
     1021   __pyx_L4_argument_unpacking_done:;
     1022+  __pyx_r = __pyx_pf_6mtrand_11RandomState___init__(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_seed);
     1023+  __Pyx_RefNannyFinishContext();
     1024+  return __pyx_r;
     1025+}
     1026+
     1027+static int __pyx_pf_6mtrand_11RandomState___init__(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_seed) {
     1028+  int __pyx_r;
     1029+  __Pyx_RefNannyDeclarations
     1030+  PyObject *__pyx_t_1 = NULL;
     1031+  PyObject *__pyx_t_2 = NULL;
     1032+  PyObject *__pyx_t_3 = NULL;
     1033+  int __pyx_lineno = 0;
     1034+  const char *__pyx_filename = NULL;
     1035+  int __pyx_clineno = 0;
     1036+  __Pyx_RefNannySetupContext("__init__", 0);
     1037 
     1038   /* "mtrand.pyx":559
     1039  *
     1040@@ -4778,7 +4918,7 @@
     1041  *
     1042  *         self.seed(seed)
     1043  */
     1044-  ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state = ((rk_state *)PyMem_Malloc((sizeof(rk_state))));
     1045+  __pyx_v_self->internal_state = ((rk_state *)PyMem_Malloc((sizeof(rk_state))));
     1046 
     1047   /* "mtrand.pyx":561
     1048  *         self.internal_state = <rk_state*>PyMem_Malloc(sizeof(rk_state))
     1049@@ -4787,10 +4927,10 @@
     1050  *
     1051  *     def __dealloc__(self):
     1052  */
     1053-  __pyx_t_1 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__seed); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 561; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1054+  __pyx_t_1 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__seed); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 561; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1055   __Pyx_GOTREF(__pyx_t_1);
     1056   __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 561; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1057-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     1058+  __Pyx_GOTREF(__pyx_t_2);
     1059   __Pyx_INCREF(__pyx_v_seed);
     1060   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_seed);
     1061   __Pyx_GIVEREF(__pyx_v_seed);
     1062@@ -4813,6 +4953,15 @@
     1063   return __pyx_r;
     1064 }
     1065 
     1066+/* Python wrapper */
     1067+static void __pyx_pw_6mtrand_11RandomState_3__dealloc__(PyObject *__pyx_v_self); /*proto*/
     1068+static void __pyx_pw_6mtrand_11RandomState_3__dealloc__(PyObject *__pyx_v_self) {
     1069+  __Pyx_RefNannyDeclarations
     1070+  __Pyx_RefNannySetupContext("__dealloc__ (wrapper)", 0);
     1071+  __pyx_pf_6mtrand_11RandomState_2__dealloc__(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self));
     1072+  __Pyx_RefNannyFinishContext();
     1073+}
     1074+
     1075 /* "mtrand.pyx":563
     1076  *         self.seed(seed)
     1077  *
     1078@@ -4821,11 +4970,10 @@
     1079  *             PyMem_Free(self.internal_state)
     1080  */
     1081 
     1082-static void __pyx_pf_6mtrand_11RandomState_1__dealloc__(PyObject *__pyx_v_self); /*proto*/
     1083-static void __pyx_pf_6mtrand_11RandomState_1__dealloc__(PyObject *__pyx_v_self) {
     1084+static void __pyx_pf_6mtrand_11RandomState_2__dealloc__(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self) {
     1085   __Pyx_RefNannyDeclarations
     1086   int __pyx_t_1;
     1087-  __Pyx_RefNannySetupContext("__dealloc__");
     1088+  __Pyx_RefNannySetupContext("__dealloc__", 0);
     1089 
     1090   /* "mtrand.pyx":564
     1091  *
     1092@@ -4834,7 +4982,7 @@
     1093  *             PyMem_Free(self.internal_state)
     1094  *             self.internal_state = NULL
     1095  */
     1096-  __pyx_t_1 = (((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state != NULL);
     1097+  __pyx_t_1 = (__pyx_v_self->internal_state != NULL);
     1098   if (__pyx_t_1) {
     1099 
     1100     /* "mtrand.pyx":565
     1101@@ -4844,7 +4992,7 @@
     1102  *             self.internal_state = NULL
     1103  *
     1104  */
     1105-    PyMem_Free(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state);
     1106+    PyMem_Free(__pyx_v_self->internal_state);
     1107 
     1108     /* "mtrand.pyx":566
     1109  *         if self.internal_state != NULL:
     1110@@ -4853,52 +5001,44 @@
     1111  *
     1112  *     def seed(self, seed=None):
     1113  */
     1114-    ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state = NULL;
     1115-    goto __pyx_L5;
     1116+    __pyx_v_self->internal_state = NULL;
     1117+    goto __pyx_L3;
     1118   }
     1119-  __pyx_L5:;
     1120+  __pyx_L3:;
     1121 
     1122   __Pyx_RefNannyFinishContext();
     1123 }
     1124 
     1125-/* "mtrand.pyx":568
     1126+/* Python wrapper */
     1127+static PyObject *__pyx_pw_6mtrand_11RandomState_5seed(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     1128+static char __pyx_doc_6mtrand_11RandomState_4seed[] = "\n        seed(seed=None)\n\n        Seed the generator.\n\n        This method is called when `RandomState` is initialized. It can be\n        called again to re-seed the generator. For details, see `RandomState`.\n\n        Parameters\n        ----------\n        seed : int or array_like, optional\n            Seed for `RandomState`.\n\n        See Also\n        --------\n        RandomState\n\n        ";
     1129+static PyObject *__pyx_pw_6mtrand_11RandomState_5seed(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     1130+  PyObject *__pyx_v_seed = 0;
     1131+  PyObject *__pyx_r = 0;
     1132+  __Pyx_RefNannyDeclarations
     1133+  __Pyx_RefNannySetupContext("seed (wrapper)", 0);
     1134+  {
     1135+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__seed,0};
     1136+    PyObject* values[1] = {0};
     1137+
     1138+    /* "mtrand.pyx":568
     1139  *             self.internal_state = NULL
     1140  *
     1141  *     def seed(self, seed=None):             # <<<<<<<<<<<<<<
     1142  *         """
     1143  *         seed(seed=None)
     1144  */
     1145-
     1146-static PyObject *__pyx_pf_6mtrand_11RandomState_2seed(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     1147-static char __pyx_doc_6mtrand_11RandomState_2seed[] = "\n        seed(seed=None)\n\n        Seed the generator.\n\n        This method is called when `RandomState` is initialized. It can be\n        called again to re-seed the generator. For details, see `RandomState`.\n\n        Parameters\n        ----------\n        seed : int or array_like, optional\n            Seed for `RandomState`.\n\n        See Also\n        --------\n        RandomState\n\n        ";
     1148-static PyObject *__pyx_pf_6mtrand_11RandomState_2seed(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     1149-  PyObject *__pyx_v_seed = 0;
     1150-  rk_error __pyx_v_errcode;
     1151-  PyArrayObject *arrayObject_obj = 0;
     1152-  PyObject *__pyx_v_iseed = NULL;
     1153-  PyObject *__pyx_r = NULL;
     1154-  __Pyx_RefNannyDeclarations
     1155-  int __pyx_t_1;
     1156-  unsigned long __pyx_t_2;
     1157-  PyObject *__pyx_t_3 = NULL;
     1158-  PyObject *__pyx_t_4 = NULL;
     1159-  int __pyx_lineno = 0;
     1160-  const char *__pyx_filename = NULL;
     1161-  int __pyx_clineno = 0;
     1162-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__seed,0};
     1163-  __Pyx_RefNannySetupContext("seed");
     1164-  {
     1165-    PyObject* values[1] = {0};
     1166     values[0] = ((PyObject *)Py_None);
     1167     if (unlikely(__pyx_kwds)) {
     1168       Py_ssize_t kw_args;
     1169-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     1170+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     1171+      switch (pos_args) {
     1172         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     1173         case  0: break;
     1174         default: goto __pyx_L5_argtuple_error;
     1175       }
     1176       kw_args = PyDict_Size(__pyx_kwds);
     1177-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     1178+      switch (pos_args) {
     1179         case  0:
     1180         if (kw_args > 0) {
     1181           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__seed);
     1182@@ -4906,7 +5046,7 @@
     1183         }
     1184       }
     1185       if (unlikely(kw_args > 0)) {
     1186-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "seed") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 568; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     1187+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "seed") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 568; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     1188       }
     1189     } else {
     1190       switch (PyTuple_GET_SIZE(__pyx_args)) {
     1191@@ -4925,6 +5065,25 @@
     1192   __Pyx_RefNannyFinishContext();
     1193   return NULL;
     1194   __pyx_L4_argument_unpacking_done:;
     1195+  __pyx_r = __pyx_pf_6mtrand_11RandomState_4seed(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_seed);
     1196+  __Pyx_RefNannyFinishContext();
     1197+  return __pyx_r;
     1198+}
     1199+
     1200+static PyObject *__pyx_pf_6mtrand_11RandomState_4seed(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_seed) {
     1201+  CYTHON_UNUSED rk_error __pyx_v_errcode;
     1202+  PyArrayObject *arrayObject_obj = 0;
     1203+  PyObject *__pyx_v_iseed = NULL;
     1204+  PyObject *__pyx_r = NULL;
     1205+  __Pyx_RefNannyDeclarations
     1206+  int __pyx_t_1;
     1207+  unsigned long __pyx_t_2;
     1208+  PyObject *__pyx_t_3 = NULL;
     1209+  PyObject *__pyx_t_4 = NULL;
     1210+  int __pyx_lineno = 0;
     1211+  const char *__pyx_filename = NULL;
     1212+  int __pyx_clineno = 0;
     1213+  __Pyx_RefNannySetupContext("seed", 0);
     1214 
     1215   /* "mtrand.pyx":589
     1216  *         cdef rk_error errcode
     1217@@ -4943,8 +5102,8 @@
     1218  *         elif type(seed) is int:
     1219  *             rk_seed(seed, self.internal_state)
     1220  */
     1221-    __pyx_v_errcode = rk_randomseed(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state);
     1222-    goto __pyx_L6;
     1223+    __pyx_v_errcode = rk_randomseed(__pyx_v_self->internal_state);
     1224+    goto __pyx_L3;
     1225   }
     1226 
     1227   /* "mtrand.pyx":591
     1228@@ -4965,8 +5124,8 @@
     1229  *             iseed = int(seed)
     1230  */
     1231     __pyx_t_2 = __Pyx_PyInt_AsUnsignedLong(__pyx_v_seed); if (unlikely((__pyx_t_2 == (unsigned long)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 592; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1232-    rk_seed(__pyx_t_2, ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state);
     1233-    goto __pyx_L6;
     1234+    rk_seed(__pyx_t_2, __pyx_v_self->internal_state);
     1235+    goto __pyx_L3;
     1236   }
     1237 
     1238   /* "mtrand.pyx":593
     1239@@ -4993,7 +5152,7 @@
     1240  *         else:
     1241  */
     1242     __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 594; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1243-    __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     1244+    __Pyx_GOTREF(__pyx_t_4);
     1245     __Pyx_INCREF(__pyx_v_seed);
     1246     PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_seed);
     1247     __Pyx_GIVEREF(__pyx_v_seed);
     1248@@ -5011,8 +5170,8 @@
     1249  *             obj = <ndarray>PyArray_ContiguousFromObject(seed, NPY_LONG, 1, 1)
     1250  */
     1251     __pyx_t_2 = __Pyx_PyInt_AsUnsignedLong(__pyx_v_iseed); if (unlikely((__pyx_t_2 == (unsigned long)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 595; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1252-    rk_seed(__pyx_t_2, ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state);
     1253-    goto __pyx_L6;
     1254+    rk_seed(__pyx_t_2, __pyx_v_self->internal_state);
     1255+    goto __pyx_L3;
     1256   }
     1257   /*else*/ {
     1258 
     1259@@ -5036,9 +5195,9 @@
     1260  *
     1261  *     def get_state(self):
     1262  */
     1263-    init_by_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, ((unsigned long *)arrayObject_obj->data), (arrayObject_obj->dimensions[0]));
     1264+    init_by_array(__pyx_v_self->internal_state, ((unsigned long *)arrayObject_obj->data), (arrayObject_obj->dimensions[0]));
     1265   }
     1266-  __pyx_L6:;
     1267+  __pyx_L3:;
     1268 
     1269   __pyx_r = Py_None; __Pyx_INCREF(Py_None);
     1270   goto __pyx_L0;
     1271@@ -5055,6 +5214,18 @@
     1272   return __pyx_r;
     1273 }
     1274 
     1275+/* Python wrapper */
     1276+static PyObject *__pyx_pw_6mtrand_11RandomState_7get_state(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused); /*proto*/
     1277+static char __pyx_doc_6mtrand_11RandomState_6get_state[] = "\n        get_state()\n\n        Return a tuple representing the internal state of the generator.\n\n        For more details, see `set_state`.\n\n        Returns\n        -------\n        out : tuple(str, ndarray of 624 uints, int, int, float)\n            The returned tuple has the following items:\n\n            1. the string 'MT19937'.\n            2. a 1-D array of 624 unsigned integer keys.\n            3. an integer ``pos``.\n            4. an integer ``has_gauss``.\n            5. a float ``cached_gaussian``.\n\n        See Also\n        --------\n        set_state\n\n        Notes\n        -----\n        `set_state` and `get_state` are not needed to work with any of the\n        random distributions in NumPy. If the internal state is manually altered,\n        the user should know exactly what he/she is doing.\n\n        ";
     1278+static PyObject *__pyx_pw_6mtrand_11RandomState_7get_state(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused) {
     1279+  PyObject *__pyx_r = 0;
     1280+  __Pyx_RefNannyDeclarations
     1281+  __Pyx_RefNannySetupContext("get_state (wrapper)", 0);
     1282+  __pyx_r = __pyx_pf_6mtrand_11RandomState_6get_state(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self));
     1283+  __Pyx_RefNannyFinishContext();
     1284+  return __pyx_r;
     1285+}
     1286+
     1287 /* "mtrand.pyx":601
     1288  *                 obj.dimensions[0])
     1289  *
     1290@@ -5063,9 +5234,7 @@
     1291  *         get_state()
     1292  */
     1293 
     1294-static PyObject *__pyx_pf_6mtrand_11RandomState_3get_state(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused); /*proto*/
     1295-static char __pyx_doc_6mtrand_11RandomState_3get_state[] = "\n        get_state()\n\n        Return a tuple representing the internal state of the generator.\n\n        For more details, see `set_state`.\n\n        Returns\n        -------\n        out : tuple(str, ndarray of 624 uints, int, int, float)\n            The returned tuple has the following items:\n\n            1. the string 'MT19937'.\n            2. a 1-D array of 624 unsigned integer keys.\n            3. an integer ``pos``.\n            4. an integer ``has_gauss``.\n            5. a float ``cached_gaussian``.\n\n        See Also\n        --------\n        set_state\n\n        Notes\n        -----\n        `set_state` and `get_state` are not needed to work with any of the\n        random distributions in NumPy. If the internal state is manually altered,\n        the user should know exactly what he/she is doing.\n\n        ";
     1296-static PyObject *__pyx_pf_6mtrand_11RandomState_3get_state(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused) {
     1297+static PyObject *__pyx_pf_6mtrand_11RandomState_6get_state(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self) {
     1298   PyArrayObject *arrayObject_state = 0;
     1299   PyObject *__pyx_r = NULL;
     1300   __Pyx_RefNannyDeclarations
     1301@@ -5076,7 +5245,7 @@
     1302   int __pyx_lineno = 0;
     1303   const char *__pyx_filename = NULL;
     1304   int __pyx_clineno = 0;
     1305-  __Pyx_RefNannySetupContext("get_state");
     1306+  __Pyx_RefNannySetupContext("get_state", 0);
     1307 
     1308   /* "mtrand.pyx":632
     1309  *         """
     1310@@ -5096,7 +5265,7 @@
     1311   __Pyx_GOTREF(__pyx_t_3);
     1312   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     1313   __pyx_t_1 = PyTuple_New(2); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 632; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1314-  __Pyx_GOTREF(((PyObject *)__pyx_t_1));
     1315+  __Pyx_GOTREF(__pyx_t_1);
     1316   __Pyx_INCREF(__pyx_int_624);
     1317   PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_int_624);
     1318   __Pyx_GIVEREF(__pyx_int_624);
     1319@@ -5118,7 +5287,7 @@
     1320  *         state = <ndarray>np.asarray(state, np.uint32)
     1321  *         return ('MT19937', state, self.internal_state.pos,
     1322  */
     1323-  memcpy(((void *)arrayObject_state->data), ((void *)((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state->key), (624 * (sizeof(long))));
     1324+  memcpy(((void *)arrayObject_state->data), ((void *)__pyx_v_self->internal_state->key), (624 * (sizeof(long))));
     1325 
     1326   /* "mtrand.pyx":634
     1327  *         state = <ndarray>np.empty(624, np.uint)
     1328@@ -5138,7 +5307,7 @@
     1329   __Pyx_GOTREF(__pyx_t_2);
     1330   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     1331   __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 634; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1332-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     1333+  __Pyx_GOTREF(__pyx_t_3);
     1334   __Pyx_INCREF(((PyObject *)arrayObject_state));
     1335   PyTuple_SET_ITEM(__pyx_t_3, 0, ((PyObject *)arrayObject_state));
     1336   __Pyx_GIVEREF(((PyObject *)arrayObject_state));
     1337@@ -5162,7 +5331,7 @@
     1338  *
     1339  */
     1340   __Pyx_XDECREF(__pyx_r);
     1341-  __pyx_t_2 = PyInt_FromLong(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state->pos); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 635; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1342+  __pyx_t_2 = PyInt_FromLong(__pyx_v_self->internal_state->pos); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 635; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1343   __Pyx_GOTREF(__pyx_t_2);
     1344 
     1345   /* "mtrand.pyx":636
     1346@@ -5172,12 +5341,12 @@
     1347  *
     1348  *     def set_state(self, state):
     1349  */
     1350-  __pyx_t_3 = PyInt_FromLong(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state->has_gauss); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 636; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1351+  __pyx_t_3 = PyInt_FromLong(__pyx_v_self->internal_state->has_gauss); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 636; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1352   __Pyx_GOTREF(__pyx_t_3);
     1353-  __pyx_t_1 = PyFloat_FromDouble(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state->gauss); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 636; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1354+  __pyx_t_1 = PyFloat_FromDouble(__pyx_v_self->internal_state->gauss); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 636; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1355   __Pyx_GOTREF(__pyx_t_1);
     1356   __pyx_t_4 = PyTuple_New(5); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 635; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1357-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     1358+  __Pyx_GOTREF(__pyx_t_4);
     1359   __Pyx_INCREF(((PyObject *)__pyx_n_s__MT19937));
     1360   PyTuple_SET_ITEM(__pyx_t_4, 0, ((PyObject *)__pyx_n_s__MT19937));
     1361   __Pyx_GIVEREF(((PyObject *)__pyx_n_s__MT19937));
     1362@@ -5213,6 +5382,18 @@
     1363   return __pyx_r;
     1364 }
     1365 
     1366+/* Python wrapper */
     1367+static PyObject *__pyx_pw_6mtrand_11RandomState_9set_state(PyObject *__pyx_v_self, PyObject *__pyx_v_state); /*proto*/
     1368+static char __pyx_doc_6mtrand_11RandomState_8set_state[] = "\n        set_state(state)\n\n        Set the internal state of the generator from a tuple.\n\n        For use if one has reason to manually (re-)set the internal state of the\n        \"Mersenne Twister\"[1]_ pseudo-random number generating algorithm.\n\n        Parameters\n        ----------\n        state : tuple(str, ndarray of 624 uints, int, int, float)\n            The `state` tuple has the following items:\n\n            1. the string 'MT19937', specifying the Mersenne Twister algorithm.\n            2. a 1-D array of 624 unsigned integers ``keys``.\n            3. an integer ``pos``.\n            4. an integer ``has_gauss``.\n            5. a float ``cached_gaussian``.\n\n        Returns\n        -------\n        out : None\n            Returns 'None' on success.\n\n        See Also\n        --------\n        get_state\n\n        Notes\n        -----\n        `set_state` and `get_state` are not needed to work with any of the\n        random distributions in NumPy. If the internal state is manually altered,\n        the user should know exactly what he/she is doing.\n\n        For backwards compatibility, the form (str, array of 624 uints, int) is\n        also accepted although it is missing some information about the cached\n        Gaussian value: ``state = ('MT19937', keys, pos)``.\n\n        References\n        ----------\n        .. [1] M. Matsumoto and T. Nishimura, \"Mersenne Twister: A\n           623-dimensionally equidistributed uniform pseudorandom number\n           generator,\" *ACM Trans. on Modeling and Computer Simulation*,\n           Vol. 8, No. 1, pp. 3-30, Jan. 1998.\n\n        ";
     1369+static PyObject *__pyx_pw_6mtrand_11RandomState_9set_state(PyObject *__pyx_v_self, PyObject *__pyx_v_state) {
     1370+  PyObject *__pyx_r = 0;
     1371+  __Pyx_RefNannyDeclarations
     1372+  __Pyx_RefNannySetupContext("set_state (wrapper)", 0);
     1373+  __pyx_r = __pyx_pf_6mtrand_11RandomState_8set_state(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), ((PyObject *)__pyx_v_state));
     1374+  __Pyx_RefNannyFinishContext();
     1375+  return __pyx_r;
     1376+}
     1377+
     1378 /* "mtrand.pyx":638
     1379  *             self.internal_state.has_gauss, self.internal_state.gauss)
     1380  *
     1381@@ -5221,9 +5402,7 @@
     1382  *         set_state(state)
     1383  */
     1384 
     1385-static PyObject *__pyx_pf_6mtrand_11RandomState_4set_state(PyObject *__pyx_v_self, PyObject *__pyx_v_state); /*proto*/
     1386-static char __pyx_doc_6mtrand_11RandomState_4set_state[] = "\n        set_state(state)\n\n        Set the internal state of the generator from a tuple.\n\n        For use if one has reason to manually (re-)set the internal state of the\n        \"Mersenne Twister\"[1]_ pseudo-random number generating algorithm.\n\n        Parameters\n        ----------\n        state : tuple(str, ndarray of 624 uints, int, int, float)\n            The `state` tuple has the following items:\n\n            1. the string 'MT19937', specifying the Mersenne Twister algorithm.\n            2. a 1-D array of 624 unsigned integers ``keys``.\n            3. an integer ``pos``.\n            4. an integer ``has_gauss``.\n            5. a float ``cached_gaussian``.\n\n        Returns\n        -------\n        out : None\n            Returns 'None' on success.\n\n        See Also\n        --------\n        get_state\n\n        Notes\n        -----\n        `set_state` and `get_state` are not needed to work with any of the\n        random distributions in NumPy. If the internal state is manually altered,\n        the user should know exactly what he/she is doing.\n\n        For backwards compatibility, the form (str, array of 624 uints, int) is\n        also accepted although it is missing some information about the cached\n        Gaussian value: ``state = ('MT19937', keys, pos)``.\n\n        References\n        ----------\n        .. [1] M. Matsumoto and T. Nishimura, \"Mersenne Twister: A\n           623-dimensionally equidistributed uniform pseudorandom number\n           generator,\" *ACM Trans. on Modeling and Computer Simulation*,\n           Vol. 8, No. 1, pp. 3-30, Jan. 1998.\n\n        ";
     1387-static PyObject *__pyx_pf_6mtrand_11RandomState_4set_state(PyObject *__pyx_v_self, PyObject *__pyx_v_state) {
     1388+static PyObject *__pyx_pf_6mtrand_11RandomState_8set_state(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_state) {
     1389   PyArrayObject *arrayObject_obj = 0;
     1390   int __pyx_v_pos;
     1391   PyObject *__pyx_v_algorithm_name = NULL;
     1392@@ -5247,7 +5426,7 @@
     1393   int __pyx_lineno = 0;
     1394   const char *__pyx_filename = NULL;
     1395   int __pyx_clineno = 0;
     1396-  __Pyx_RefNannySetupContext("set_state");
     1397+  __Pyx_RefNannySetupContext("set_state", 0);
     1398 
     1399   /* "mtrand.pyx":687
     1400  *         cdef ndarray obj "arrayObject_obj"
     1401@@ -5268,7 +5447,9 @@
     1402  *             raise ValueError("algorithm must be 'MT19937'")
     1403  *         key, pos = state[1:3]
     1404  */
     1405-  __pyx_t_2 = __Pyx_PyString_Equals(__pyx_v_algorithm_name, ((PyObject *)__pyx_n_s__MT19937), Py_NE); if (unlikely(__pyx_t_2 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 688; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1406+  __pyx_t_1 = PyObject_RichCompare(__pyx_v_algorithm_name, ((PyObject *)__pyx_n_s__MT19937), Py_NE); __Pyx_XGOTREF(__pyx_t_1); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 688; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1407+  __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_2 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 688; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1408+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     1409   if (__pyx_t_2) {
     1410 
     1411     /* "mtrand.pyx":689
     1412@@ -5283,9 +5464,9 @@
     1413     __Pyx_Raise(__pyx_t_1, 0, 0, 0);
     1414     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     1415     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 689; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1416-    goto __pyx_L5;
     1417+    goto __pyx_L3;
     1418   }
     1419-  __pyx_L5:;
     1420+  __pyx_L3:;
     1421 
     1422   /* "mtrand.pyx":690
     1423  *         if algorithm_name != 'MT19937':
     1424@@ -5298,45 +5479,52 @@
     1425   __Pyx_GOTREF(__pyx_t_1);
     1426   if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
     1427     PyObject* sequence = __pyx_t_1;
     1428+    #if CYTHON_COMPILING_IN_CPYTHON
     1429+    Py_ssize_t size = Py_SIZE(sequence);
     1430+    #else
     1431+    Py_ssize_t size = PySequence_Size(sequence);
     1432+    #endif
     1433+    if (unlikely(size != 2)) {
     1434+      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
     1435+      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
     1436+      {__pyx_filename = __pyx_f[0]; __pyx_lineno = 690; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1437+    }
     1438+    #if CYTHON_COMPILING_IN_CPYTHON
     1439     if (likely(PyTuple_CheckExact(sequence))) {
     1440-      if (unlikely(PyTuple_GET_SIZE(sequence) != 2)) {
     1441-        if (PyTuple_GET_SIZE(sequence) > 2) __Pyx_RaiseTooManyValuesError(2);
     1442-        else __Pyx_RaiseNeedMoreValuesError(PyTuple_GET_SIZE(sequence));
     1443-        {__pyx_filename = __pyx_f[0]; __pyx_lineno = 690; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1444-      }
     1445       __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0);
     1446       __pyx_t_4 = PyTuple_GET_ITEM(sequence, 1);
     1447     } else {
     1448-      if (unlikely(PyList_GET_SIZE(sequence) != 2)) {
     1449-        if (PyList_GET_SIZE(sequence) > 2) __Pyx_RaiseTooManyValuesError(2);
     1450-        else __Pyx_RaiseNeedMoreValuesError(PyList_GET_SIZE(sequence));
     1451-        {__pyx_filename = __pyx_f[0]; __pyx_lineno = 690; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1452-      }
     1453       __pyx_t_3 = PyList_GET_ITEM(sequence, 0);
     1454       __pyx_t_4 = PyList_GET_ITEM(sequence, 1);
     1455     }
     1456     __Pyx_INCREF(__pyx_t_3);
     1457     __Pyx_INCREF(__pyx_t_4);
     1458+    #else
     1459+    __pyx_t_3 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 690; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1460+    __pyx_t_4 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 690; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1461+    #endif
     1462     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     1463-  } else {
     1464+  } else
     1465+  {
     1466     Py_ssize_t index = -1;
     1467     __pyx_t_5 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 690; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1468     __Pyx_GOTREF(__pyx_t_5);
     1469     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     1470     __pyx_t_6 = Py_TYPE(__pyx_t_5)->tp_iternext;
     1471-    index = 0; __pyx_t_3 = __pyx_t_6(__pyx_t_5); if (unlikely(!__pyx_t_3)) goto __pyx_L6_unpacking_failed;
     1472+    index = 0; __pyx_t_3 = __pyx_t_6(__pyx_t_5); if (unlikely(!__pyx_t_3)) goto __pyx_L4_unpacking_failed;
     1473     __Pyx_GOTREF(__pyx_t_3);
     1474-    index = 1; __pyx_t_4 = __pyx_t_6(__pyx_t_5); if (unlikely(!__pyx_t_4)) goto __pyx_L6_unpacking_failed;
     1475+    index = 1; __pyx_t_4 = __pyx_t_6(__pyx_t_5); if (unlikely(!__pyx_t_4)) goto __pyx_L4_unpacking_failed;
     1476     __Pyx_GOTREF(__pyx_t_4);
     1477     if (__Pyx_IternextUnpackEndCheck(__pyx_t_6(__pyx_t_5), 2) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 690; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1478+    __pyx_t_6 = NULL;
     1479     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     1480-    goto __pyx_L7_unpacking_done;
     1481-    __pyx_L6_unpacking_failed:;
     1482+    goto __pyx_L5_unpacking_done;
     1483+    __pyx_L4_unpacking_failed:;
     1484     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     1485-    if (PyErr_Occurred() && PyErr_ExceptionMatches(PyExc_StopIteration)) PyErr_Clear();
     1486-    if (!PyErr_Occurred()) __Pyx_RaiseNeedMoreValuesError(index);
     1487+    __pyx_t_6 = NULL;
     1488+    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
     1489     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 690; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1490-    __pyx_L7_unpacking_done:;
     1491+    __pyx_L5_unpacking_done:;
     1492   }
     1493   __pyx_t_7 = __Pyx_PyInt_AsInt(__pyx_t_4); if (unlikely((__pyx_t_7 == (int)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 690; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1494   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     1495@@ -5376,7 +5564,7 @@
     1496     __Pyx_GOTREF(__pyx_t_1);
     1497     __pyx_v_cached_gaussian = __pyx_t_1;
     1498     __pyx_t_1 = 0;
     1499-    goto __pyx_L8;
     1500+    goto __pyx_L6;
     1501   }
     1502   /*else*/ {
     1503 
     1504@@ -5391,52 +5579,59 @@
     1505     __Pyx_GOTREF(__pyx_t_1);
     1506     if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
     1507       PyObject* sequence = __pyx_t_1;
     1508+      #if CYTHON_COMPILING_IN_CPYTHON
     1509+      Py_ssize_t size = Py_SIZE(sequence);
     1510+      #else
     1511+      Py_ssize_t size = PySequence_Size(sequence);
     1512+      #endif
     1513+      if (unlikely(size != 2)) {
     1514+        if (size > 2) __Pyx_RaiseTooManyValuesError(2);
     1515+        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
     1516+        {__pyx_filename = __pyx_f[0]; __pyx_lineno = 695; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1517+      }
     1518+      #if CYTHON_COMPILING_IN_CPYTHON
     1519       if (likely(PyTuple_CheckExact(sequence))) {
     1520-        if (unlikely(PyTuple_GET_SIZE(sequence) != 2)) {
     1521-          if (PyTuple_GET_SIZE(sequence) > 2) __Pyx_RaiseTooManyValuesError(2);
     1522-          else __Pyx_RaiseNeedMoreValuesError(PyTuple_GET_SIZE(sequence));
     1523-          {__pyx_filename = __pyx_f[0]; __pyx_lineno = 695; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1524-        }
     1525         __pyx_t_4 = PyTuple_GET_ITEM(sequence, 0);
     1526         __pyx_t_3 = PyTuple_GET_ITEM(sequence, 1);
     1527       } else {
     1528-        if (unlikely(PyList_GET_SIZE(sequence) != 2)) {
     1529-          if (PyList_GET_SIZE(sequence) > 2) __Pyx_RaiseTooManyValuesError(2);
     1530-          else __Pyx_RaiseNeedMoreValuesError(PyList_GET_SIZE(sequence));
     1531-          {__pyx_filename = __pyx_f[0]; __pyx_lineno = 695; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1532-        }
     1533         __pyx_t_4 = PyList_GET_ITEM(sequence, 0);
     1534         __pyx_t_3 = PyList_GET_ITEM(sequence, 1);
     1535       }
     1536       __Pyx_INCREF(__pyx_t_4);
     1537       __Pyx_INCREF(__pyx_t_3);
     1538+      #else
     1539+      __pyx_t_4 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 695; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1540+      __pyx_t_3 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 695; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1541+      #endif
     1542       __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     1543-    } else {
     1544+    } else
     1545+    {
     1546       Py_ssize_t index = -1;
     1547       __pyx_t_5 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 695; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1548       __Pyx_GOTREF(__pyx_t_5);
     1549       __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     1550       __pyx_t_6 = Py_TYPE(__pyx_t_5)->tp_iternext;
     1551-      index = 0; __pyx_t_4 = __pyx_t_6(__pyx_t_5); if (unlikely(!__pyx_t_4)) goto __pyx_L9_unpacking_failed;
     1552+      index = 0; __pyx_t_4 = __pyx_t_6(__pyx_t_5); if (unlikely(!__pyx_t_4)) goto __pyx_L7_unpacking_failed;
     1553       __Pyx_GOTREF(__pyx_t_4);
     1554-      index = 1; __pyx_t_3 = __pyx_t_6(__pyx_t_5); if (unlikely(!__pyx_t_3)) goto __pyx_L9_unpacking_failed;
     1555+      index = 1; __pyx_t_3 = __pyx_t_6(__pyx_t_5); if (unlikely(!__pyx_t_3)) goto __pyx_L7_unpacking_failed;
     1556       __Pyx_GOTREF(__pyx_t_3);
     1557       if (__Pyx_IternextUnpackEndCheck(__pyx_t_6(__pyx_t_5), 2) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 695; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1558+      __pyx_t_6 = NULL;
     1559       __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     1560-      goto __pyx_L10_unpacking_done;
     1561-      __pyx_L9_unpacking_failed:;
     1562+      goto __pyx_L8_unpacking_done;
     1563+      __pyx_L7_unpacking_failed:;
     1564       __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     1565-      if (PyErr_Occurred() && PyErr_ExceptionMatches(PyExc_StopIteration)) PyErr_Clear();
     1566-      if (!PyErr_Occurred()) __Pyx_RaiseNeedMoreValuesError(index);
     1567+      __pyx_t_6 = NULL;
     1568+      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
     1569       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 695; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1570-      __pyx_L10_unpacking_done:;
     1571+      __pyx_L8_unpacking_done:;
     1572     }
     1573     __pyx_v_has_gauss = __pyx_t_4;
     1574     __pyx_t_4 = 0;
     1575     __pyx_v_cached_gaussian = __pyx_t_3;
     1576     __pyx_t_3 = 0;
     1577   }
     1578-  __pyx_L8:;
     1579+  __pyx_L6:;
     1580 
     1581   /* "mtrand.pyx":696
     1582  *         else:
     1583@@ -5459,7 +5654,7 @@
     1584  *         except TypeError:
     1585  *             # compatibility -- could be an older pickle
     1586  */
     1587-      __pyx_t_1 = PyArray_ContiguousFromObject(__pyx_v_key, NPY_ULONG, 1, 1); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 697; __pyx_clineno = __LINE__; goto __pyx_L11_error;}
     1588+      __pyx_t_1 = PyArray_ContiguousFromObject(__pyx_v_key, NPY_ULONG, 1, 1); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 697; __pyx_clineno = __LINE__; goto __pyx_L9_error;}
     1589       __Pyx_GOTREF(__pyx_t_1);
     1590       __Pyx_INCREF(((PyObject *)((PyArrayObject *)__pyx_t_1)));
     1591       arrayObject_obj = ((PyArrayObject *)__pyx_t_1);
     1592@@ -5468,8 +5663,8 @@
     1593     __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
     1594     __Pyx_XDECREF(__pyx_t_10); __pyx_t_10 = 0;
     1595     __Pyx_XDECREF(__pyx_t_11); __pyx_t_11 = 0;
     1596-    goto __pyx_L18_try_end;
     1597-    __pyx_L11_error:;
     1598+    goto __pyx_L16_try_end;
     1599+    __pyx_L9_error:;
     1600     __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
     1601     __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
     1602     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     1603@@ -5485,7 +5680,7 @@
     1604     __pyx_t_7 = PyErr_ExceptionMatches(__pyx_builtin_TypeError);
     1605     if (__pyx_t_7) {
     1606       __Pyx_AddTraceback("mtrand.RandomState.set_state", __pyx_clineno, __pyx_lineno, __pyx_filename);
     1607-      if (__Pyx_GetException(&__pyx_t_1, &__pyx_t_3, &__pyx_t_4) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 698; __pyx_clineno = __LINE__; goto __pyx_L13_except_error;}
     1608+      if (__Pyx_GetException(&__pyx_t_1, &__pyx_t_3, &__pyx_t_4) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 698; __pyx_clineno = __LINE__; goto __pyx_L11_except_error;}
     1609       __Pyx_GOTREF(__pyx_t_1);
     1610       __Pyx_GOTREF(__pyx_t_3);
     1611       __Pyx_GOTREF(__pyx_t_4);
     1612@@ -5497,7 +5692,7 @@
     1613  *         if obj.dimensions[0] != 624:
     1614  *             raise ValueError("state must be 624 longs")
     1615  */
     1616-      __pyx_t_5 = PyArray_ContiguousFromObject(__pyx_v_key, NPY_LONG, 1, 1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 700; __pyx_clineno = __LINE__; goto __pyx_L13_except_error;}
     1617+      __pyx_t_5 = PyArray_ContiguousFromObject(__pyx_v_key, NPY_LONG, 1, 1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 700; __pyx_clineno = __LINE__; goto __pyx_L11_except_error;}
     1618       __Pyx_GOTREF(__pyx_t_5);
     1619       __Pyx_INCREF(((PyObject *)((PyArrayObject *)__pyx_t_5)));
     1620       __Pyx_XDECREF(((PyObject *)arrayObject_obj));
     1621@@ -5506,20 +5701,20 @@
     1622       __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     1623       __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     1624       __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     1625-      goto __pyx_L12_exception_handled;
     1626+      goto __pyx_L10_exception_handled;
     1627     }
     1628-    __pyx_L13_except_error:;
     1629+    __pyx_L11_except_error:;
     1630     __Pyx_XGIVEREF(__pyx_t_9);
     1631     __Pyx_XGIVEREF(__pyx_t_10);
     1632     __Pyx_XGIVEREF(__pyx_t_11);
     1633     __Pyx_ExceptionReset(__pyx_t_9, __pyx_t_10, __pyx_t_11);
     1634     goto __pyx_L1_error;
     1635-    __pyx_L12_exception_handled:;
     1636+    __pyx_L10_exception_handled:;
     1637     __Pyx_XGIVEREF(__pyx_t_9);
     1638     __Pyx_XGIVEREF(__pyx_t_10);
     1639     __Pyx_XGIVEREF(__pyx_t_11);
     1640     __Pyx_ExceptionReset(__pyx_t_9, __pyx_t_10, __pyx_t_11);
     1641-    __pyx_L18_try_end:;
     1642+    __pyx_L16_try_end:;
     1643   }
     1644 
     1645   /* "mtrand.pyx":701
     1646@@ -5544,9 +5739,9 @@
     1647     __Pyx_Raise(__pyx_t_4, 0, 0, 0);
     1648     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     1649     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 702; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1650-    goto __pyx_L21;
     1651+    goto __pyx_L19;
     1652   }
     1653-  __pyx_L21:;
     1654+  __pyx_L19:;
     1655 
     1656   /* "mtrand.pyx":703
     1657  *         if obj.dimensions[0] != 624:
     1658@@ -5555,7 +5750,7 @@
     1659  *         self.internal_state.pos = pos
     1660  *         self.internal_state.has_gauss = has_gauss
     1661  */
     1662-  memcpy(((void *)((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state->key), ((void *)arrayObject_obj->data), (624 * (sizeof(long))));
     1663+  memcpy(((void *)__pyx_v_self->internal_state->key), ((void *)arrayObject_obj->data), (624 * (sizeof(long))));
     1664 
     1665   /* "mtrand.pyx":704
     1666  *             raise ValueError("state must be 624 longs")
     1667@@ -5564,7 +5759,7 @@
     1668  *         self.internal_state.has_gauss = has_gauss
     1669  *         self.internal_state.gauss = cached_gaussian
     1670  */
     1671-  ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state->pos = __pyx_v_pos;
     1672+  __pyx_v_self->internal_state->pos = __pyx_v_pos;
     1673 
     1674   /* "mtrand.pyx":705
     1675  *         memcpy(<void*>(self.internal_state.key), <void*>(obj.data), 624*sizeof(long))
     1676@@ -5574,7 +5769,7 @@
     1677  *
     1678  */
     1679   __pyx_t_7 = __Pyx_PyInt_AsInt(__pyx_v_has_gauss); if (unlikely((__pyx_t_7 == (int)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 705; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1680-  ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state->has_gauss = __pyx_t_7;
     1681+  __pyx_v_self->internal_state->has_gauss = __pyx_t_7;
     1682 
     1683   /* "mtrand.pyx":706
     1684  *         self.internal_state.pos = pos
     1685@@ -5584,7 +5779,7 @@
     1686  *     # Pickling support:
     1687  */
     1688   __pyx_t_12 = __pyx_PyFloat_AsDouble(__pyx_v_cached_gaussian); if (unlikely((__pyx_t_12 == (double)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 706; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1689-  ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state->gauss = __pyx_t_12;
     1690+  __pyx_v_self->internal_state->gauss = __pyx_t_12;
     1691 
     1692   __pyx_r = Py_None; __Pyx_INCREF(Py_None);
     1693   goto __pyx_L0;
     1694@@ -5606,6 +5801,17 @@
     1695   return __pyx_r;
     1696 }
     1697 
     1698+/* Python wrapper */
     1699+static PyObject *__pyx_pw_6mtrand_11RandomState_11__getstate__(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused); /*proto*/
     1700+static PyObject *__pyx_pw_6mtrand_11RandomState_11__getstate__(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused) {
     1701+  PyObject *__pyx_r = 0;
     1702+  __Pyx_RefNannyDeclarations
     1703+  __Pyx_RefNannySetupContext("__getstate__ (wrapper)", 0);
     1704+  __pyx_r = __pyx_pf_6mtrand_11RandomState_10__getstate__(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self));
     1705+  __Pyx_RefNannyFinishContext();
     1706+  return __pyx_r;
     1707+}
     1708+
     1709 /* "mtrand.pyx":709
     1710  *
     1711  *     # Pickling support:
     1712@@ -5614,8 +5820,7 @@
     1713  *
     1714  */
     1715 
     1716-static PyObject *__pyx_pf_6mtrand_11RandomState_5__getstate__(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused); /*proto*/
     1717-static PyObject *__pyx_pf_6mtrand_11RandomState_5__getstate__(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused) {
     1718+static PyObject *__pyx_pf_6mtrand_11RandomState_10__getstate__(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self) {
     1719   PyObject *__pyx_r = NULL;
     1720   __Pyx_RefNannyDeclarations
     1721   PyObject *__pyx_t_1 = NULL;
     1722@@ -5623,7 +5828,7 @@
     1723   int __pyx_lineno = 0;
     1724   const char *__pyx_filename = NULL;
     1725   int __pyx_clineno = 0;
     1726-  __Pyx_RefNannySetupContext("__getstate__");
     1727+  __Pyx_RefNannySetupContext("__getstate__", 0);
     1728 
     1729   /* "mtrand.pyx":710
     1730  *     # Pickling support:
     1731@@ -5633,7 +5838,7 @@
     1732  *     def __setstate__(self, state):
     1733  */
     1734   __Pyx_XDECREF(__pyx_r);
     1735-  __pyx_t_1 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__get_state); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 710; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1736+  __pyx_t_1 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__get_state); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 710; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1737   __Pyx_GOTREF(__pyx_t_1);
     1738   __pyx_t_2 = PyObject_Call(__pyx_t_1, ((PyObject *)__pyx_empty_tuple), NULL); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 710; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1739   __Pyx_GOTREF(__pyx_t_2);
     1740@@ -5655,6 +5860,17 @@
     1741   return __pyx_r;
     1742 }
     1743 
     1744+/* Python wrapper */
     1745+static PyObject *__pyx_pw_6mtrand_11RandomState_13__setstate__(PyObject *__pyx_v_self, PyObject *__pyx_v_state); /*proto*/
     1746+static PyObject *__pyx_pw_6mtrand_11RandomState_13__setstate__(PyObject *__pyx_v_self, PyObject *__pyx_v_state) {
     1747+  PyObject *__pyx_r = 0;
     1748+  __Pyx_RefNannyDeclarations
     1749+  __Pyx_RefNannySetupContext("__setstate__ (wrapper)", 0);
     1750+  __pyx_r = __pyx_pf_6mtrand_11RandomState_12__setstate__(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), ((PyObject *)__pyx_v_state));
     1751+  __Pyx_RefNannyFinishContext();
     1752+  return __pyx_r;
     1753+}
     1754+
     1755 /* "mtrand.pyx":712
     1756  *         return self.get_state()
     1757  *
     1758@@ -5663,8 +5879,7 @@
     1759  *
     1760  */
     1761 
     1762-static PyObject *__pyx_pf_6mtrand_11RandomState_6__setstate__(PyObject *__pyx_v_self, PyObject *__pyx_v_state); /*proto*/
     1763-static PyObject *__pyx_pf_6mtrand_11RandomState_6__setstate__(PyObject *__pyx_v_self, PyObject *__pyx_v_state) {
     1764+static PyObject *__pyx_pf_6mtrand_11RandomState_12__setstate__(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_state) {
     1765   PyObject *__pyx_r = NULL;
     1766   __Pyx_RefNannyDeclarations
     1767   PyObject *__pyx_t_1 = NULL;
     1768@@ -5673,7 +5888,7 @@
     1769   int __pyx_lineno = 0;
     1770   const char *__pyx_filename = NULL;
     1771   int __pyx_clineno = 0;
     1772-  __Pyx_RefNannySetupContext("__setstate__");
     1773+  __Pyx_RefNannySetupContext("__setstate__", 0);
     1774 
     1775   /* "mtrand.pyx":713
     1776  *
     1777@@ -5682,10 +5897,10 @@
     1778  *
     1779  *     def __reduce__(self):
     1780  */
     1781-  __pyx_t_1 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__set_state); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 713; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1782+  __pyx_t_1 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__set_state); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 713; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1783   __Pyx_GOTREF(__pyx_t_1);
     1784   __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 713; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1785-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     1786+  __Pyx_GOTREF(__pyx_t_2);
     1787   __Pyx_INCREF(__pyx_v_state);
     1788   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_state);
     1789   __Pyx_GIVEREF(__pyx_v_state);
     1790@@ -5709,6 +5924,17 @@
     1791   return __pyx_r;
     1792 }
     1793 
     1794+/* Python wrapper */
     1795+static PyObject *__pyx_pw_6mtrand_11RandomState_15__reduce__(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused); /*proto*/
     1796+static PyObject *__pyx_pw_6mtrand_11RandomState_15__reduce__(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused) {
     1797+  PyObject *__pyx_r = 0;
     1798+  __Pyx_RefNannyDeclarations
     1799+  __Pyx_RefNannySetupContext("__reduce__ (wrapper)", 0);
     1800+  __pyx_r = __pyx_pf_6mtrand_11RandomState_14__reduce__(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self));
     1801+  __Pyx_RefNannyFinishContext();
     1802+  return __pyx_r;
     1803+}
     1804+
     1805 /* "mtrand.pyx":715
     1806  *         self.set_state(state)
     1807  *
     1808@@ -5717,8 +5943,7 @@
     1809  *
     1810  */
     1811 
     1812-static PyObject *__pyx_pf_6mtrand_11RandomState_7__reduce__(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused); /*proto*/
     1813-static PyObject *__pyx_pf_6mtrand_11RandomState_7__reduce__(PyObject *__pyx_v_self, CYTHON_UNUSED PyObject *unused) {
     1814+static PyObject *__pyx_pf_6mtrand_11RandomState_14__reduce__(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self) {
     1815   PyObject *__pyx_r = NULL;
     1816   __Pyx_RefNannyDeclarations
     1817   PyObject *__pyx_t_1 = NULL;
     1818@@ -5727,7 +5952,7 @@
     1819   int __pyx_lineno = 0;
     1820   const char *__pyx_filename = NULL;
     1821   int __pyx_clineno = 0;
     1822-  __Pyx_RefNannySetupContext("__reduce__");
     1823+  __Pyx_RefNannySetupContext("__reduce__", 0);
     1824 
     1825   /* "mtrand.pyx":716
     1826  *
     1827@@ -5745,13 +5970,13 @@
     1828   __pyx_t_1 = PyObject_GetAttr(__pyx_t_2, __pyx_n_s____RandomState_ctor); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 716; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1829   __Pyx_GOTREF(__pyx_t_1);
     1830   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     1831-  __pyx_t_2 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__get_state); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 716; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1832+  __pyx_t_2 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__get_state); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 716; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1833   __Pyx_GOTREF(__pyx_t_2);
     1834   __pyx_t_3 = PyObject_Call(__pyx_t_2, ((PyObject *)__pyx_empty_tuple), NULL); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 716; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1835   __Pyx_GOTREF(__pyx_t_3);
     1836   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     1837   __pyx_t_2 = PyTuple_New(3); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 716; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1838-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     1839+  __Pyx_GOTREF(__pyx_t_2);
     1840   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_1);
     1841   __Pyx_GIVEREF(__pyx_t_1);
     1842   __Pyx_INCREF(((PyObject *)__pyx_empty_tuple));
     1843@@ -5779,38 +6004,36 @@
     1844   return __pyx_r;
     1845 }
     1846 
     1847-/* "mtrand.pyx":719
     1848+/* Python wrapper */
     1849+static PyObject *__pyx_pw_6mtrand_11RandomState_17random_sample(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     1850+static char __pyx_doc_6mtrand_11RandomState_16random_sample[] = "\n        random_sample(size=None)\n\n        Return random floats in the half-open interval [0.0, 1.0).\n\n        Results are from the \"continuous uniform\" distribution over the\n        stated interval.  To sample :math:`Unif[a, b), b > a` multiply\n        the output of `random_sample` by `(b-a)` and add `a`::\n\n          (b - a) * random_sample() + a\n\n        Parameters\n        ----------\n        size : int or tuple of ints, optional\n            Defines the shape of the returned array of random floats. If None\n            (the default), returns a single float.\n\n        Returns\n        -------\n        out : float or ndarray of floats\n            Array of random floats of shape `size` (unless ``size=None``, in which\n            case a single float is returned).\n\n        Examples\n        --------\n        >>> np.random.random_sample()\n        0.47108547995356098\n        >>> type(np.random.random_sample())\n        <type 'float'>\n        >>> np.random.random_sample((5,))\n        array([ 0.30220482,  0.86820401,  0.1654503 ,  0.11659149,  0.54323428])\n\n        Three-by-two array of random numbers from [-5, 0):\n\n        >>> 5 * np.random.random_sample((3, 2)) - 5\n        array([[-3.99149989, -0.52338984],\n               [-2.99091858, -0.79479508],\n               [-1.23204345, -1.75224494]])\n\n        ";
     1851+static PyObject *__pyx_pw_6mtrand_11RandomState_17random_sample(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     1852+  PyObject *__pyx_v_size = 0;
     1853+  PyObject *__pyx_r = 0;
     1854+  __Pyx_RefNannyDeclarations
     1855+  __Pyx_RefNannySetupContext("random_sample (wrapper)", 0);
     1856+  {
     1857+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__size,0};
     1858+    PyObject* values[1] = {0};
     1859+
     1860+    /* "mtrand.pyx":719
     1861  *
     1862  *     # Basic distributions:
     1863  *     def random_sample(self, size=None):             # <<<<<<<<<<<<<<
     1864  *         """
     1865  *         random_sample(size=None)
     1866  */
     1867-
     1868-static PyObject *__pyx_pf_6mtrand_11RandomState_8random_sample(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     1869-static char __pyx_doc_6mtrand_11RandomState_8random_sample[] = "\n        random_sample(size=None)\n\n        Return random floats in the half-open interval [0.0, 1.0).\n\n        Results are from the \"continuous uniform\" distribution over the\n        stated interval.  To sample :math:`Unif[a, b), b > a` multiply\n        the output of `random_sample` by `(b-a)` and add `a`::\n\n          (b - a) * random_sample() + a\n\n        Parameters\n        ----------\n        size : int or tuple of ints, optional\n            Defines the shape of the returned array of random floats. If None\n            (the default), returns a single float.\n\n        Returns\n        -------\n        out : float or ndarray of floats\n            Array of random floats of shape `size` (unless ``size=None``, in which\n            case a single float is returned).\n\n        Examples\n        --------\n        >>> np.random.random_sample()\n        0.47108547995356098\n        >>> type(np.random.random_sample())\n        <type 'float'>\n        >>> np.random.random_sample((5,))\n        array([ 0.30220482,  0.86820401,  0.1654503 ,  0.11659149,  0.54323428])\n\n        Three-by-two array of random numbers from [-5, 0):\n\n        >>> 5 * np.random.random_sample((3, 2)) - 5\n        array([[-3.99149989, -0.52338984],\n               [-2.99091858, -0.79479508],\n               [-1.23204345, -1.75224494]])\n\n        ";
     1870-static PyObject *__pyx_pf_6mtrand_11RandomState_8random_sample(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     1871-  PyObject *__pyx_v_size = 0;
     1872-  PyObject *__pyx_r = NULL;
     1873-  __Pyx_RefNannyDeclarations
     1874-  PyObject *__pyx_t_1 = NULL;
     1875-  int __pyx_lineno = 0;
     1876-  const char *__pyx_filename = NULL;
     1877-  int __pyx_clineno = 0;
     1878-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__size,0};
     1879-  __Pyx_RefNannySetupContext("random_sample");
     1880-  {
     1881-    PyObject* values[1] = {0};
     1882     values[0] = ((PyObject *)Py_None);
     1883     if (unlikely(__pyx_kwds)) {
     1884       Py_ssize_t kw_args;
     1885-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     1886+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     1887+      switch (pos_args) {
     1888         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     1889         case  0: break;
     1890         default: goto __pyx_L5_argtuple_error;
     1891       }
     1892       kw_args = PyDict_Size(__pyx_kwds);
     1893-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     1894+      switch (pos_args) {
     1895         case  0:
     1896         if (kw_args > 0) {
     1897           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__size);
     1898@@ -5818,7 +6041,7 @@
     1899         }
     1900       }
     1901       if (unlikely(kw_args > 0)) {
     1902-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "random_sample") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 719; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     1903+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "random_sample") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 719; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     1904       }
     1905     } else {
     1906       switch (PyTuple_GET_SIZE(__pyx_args)) {
     1907@@ -5837,6 +6060,19 @@
     1908   __Pyx_RefNannyFinishContext();
     1909   return NULL;
     1910   __pyx_L4_argument_unpacking_done:;
     1911+  __pyx_r = __pyx_pf_6mtrand_11RandomState_16random_sample(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_size);
     1912+  __Pyx_RefNannyFinishContext();
     1913+  return __pyx_r;
     1914+}
     1915+
     1916+static PyObject *__pyx_pf_6mtrand_11RandomState_16random_sample(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_size) {
     1917+  PyObject *__pyx_r = NULL;
     1918+  __Pyx_RefNannyDeclarations
     1919+  PyObject *__pyx_t_1 = NULL;
     1920+  int __pyx_lineno = 0;
     1921+  const char *__pyx_filename = NULL;
     1922+  int __pyx_clineno = 0;
     1923+  __Pyx_RefNannySetupContext("random_sample", 0);
     1924 
     1925   /* "mtrand.pyx":760
     1926  *
     1927@@ -5846,7 +6082,7 @@
     1928  *     def tomaxint(self, size=None):
     1929  */
     1930   __Pyx_XDECREF(__pyx_r);
     1931-  __pyx_t_1 = __pyx_f_6mtrand_cont0_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_double, __pyx_v_size); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 760; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1932+  __pyx_t_1 = __pyx_f_6mtrand_cont0_array(__pyx_v_self->internal_state, rk_double, __pyx_v_size); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 760; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     1933   __Pyx_GOTREF(__pyx_t_1);
     1934   __pyx_r = __pyx_t_1;
     1935   __pyx_t_1 = 0;
     1936@@ -5864,38 +6100,36 @@
     1937   return __pyx_r;
     1938 }
     1939 
     1940-/* "mtrand.pyx":762
     1941+/* Python wrapper */
     1942+static PyObject *__pyx_pw_6mtrand_11RandomState_19tomaxint(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     1943+static char __pyx_doc_6mtrand_11RandomState_18tomaxint[] = "\n        tomaxint(size=None)\n\n        Random integers between 0 and ``sys.maxint``, inclusive.\n\n        Return a sample of uniformly distributed random integers in the interval\n        [0, ``sys.maxint``].\n\n        Parameters\n        ----------\n        size : tuple of ints, int, optional\n            Shape of output.  If this is, for example, (m,n,k), m*n*k samples\n            are generated.  If no shape is specified, a single sample is\n            returned.\n\n        Returns\n        -------\n        out : ndarray\n            Drawn samples, with shape `size`.\n\n        See Also\n        --------\n        randint : Uniform sampling over a given half-open interval of integers.\n        random_integers : Uniform sampling over a given closed interval of\n            integers.\n\n        Examples\n        --------\n        >>> RS = np.random.mtrand.RandomState() # need a RandomState object\n        >>> RS.tomaxint((2,2,2))\n        array([[[1170048599, 1600360186],\n                [ 739731006, 1947757578]],\n               [[1871712945,  752307660],\n                [1601631370, 1479324245]]])\n        >>> import sys\n        >>> sys.maxint\n        2147483647\n        >>> RS.tomaxint((2,2,2)) < sys.maxint\n        array([[[ True,  True],\n                [ True,  True]],\n               [[ True,  True],\n                [ True,  True]]], dtype=bool)\n\n        ";
     1944+static PyObject *__pyx_pw_6mtrand_11RandomState_19tomaxint(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     1945+  PyObject *__pyx_v_size = 0;
     1946+  PyObject *__pyx_r = 0;
     1947+  __Pyx_RefNannyDeclarations
     1948+  __Pyx_RefNannySetupContext("tomaxint (wrapper)", 0);
     1949+  {
     1950+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__size,0};
     1951+    PyObject* values[1] = {0};
     1952+
     1953+    /* "mtrand.pyx":762
     1954  *         return cont0_array(self.internal_state, rk_double, size)
     1955  *
     1956  *     def tomaxint(self, size=None):             # <<<<<<<<<<<<<<
     1957  *         """
     1958  *         tomaxint(size=None)
     1959  */
     1960-
     1961-static PyObject *__pyx_pf_6mtrand_11RandomState_9tomaxint(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     1962-static char __pyx_doc_6mtrand_11RandomState_9tomaxint[] = "\n        tomaxint(size=None)\n\n        Random integers between 0 and ``sys.maxint``, inclusive.\n\n        Return a sample of uniformly distributed random integers in the interval\n        [0, ``sys.maxint``].\n\n        Parameters\n        ----------\n        size : tuple of ints, int, optional\n            Shape of output.  If this is, for example, (m,n,k), m*n*k samples\n            are generated.  If no shape is specified, a single sample is\n            returned.\n\n        Returns\n        -------\n        out : ndarray\n            Drawn samples, with shape `size`.\n\n        See Also\n        --------\n        randint : Uniform sampling over a given half-open interval of integers.\n        random_integers : Uniform sampling over a given closed interval of\n            integers.\n\n        Examples\n        --------\n        >>> RS = np.random.mtrand.RandomState() # need a RandomState object\n        >>> RS.tomaxint((2,2,2))\n        array([[[1170048599, 1600360186],\n                [ 739731006, 1947757578]],\n               [[1871712945,  752307660],\n                [1601631370, 1479324245]]])\n        >>> import sys\n        >>> sys.maxint\n        2147483647\n        >>> RS.tomaxint((2,2,2)) < sys.maxint\n        array([[[ True,  True],\n                [ True,  True]],\n               [[ True,  True],\n                [ True,  True]]], dtype=bool)\n\n        ";
     1963-static PyObject *__pyx_pf_6mtrand_11RandomState_9tomaxint(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     1964-  PyObject *__pyx_v_size = 0;
     1965-  PyObject *__pyx_r = NULL;
     1966-  __Pyx_RefNannyDeclarations
     1967-  PyObject *__pyx_t_1 = NULL;
     1968-  int __pyx_lineno = 0;
     1969-  const char *__pyx_filename = NULL;
     1970-  int __pyx_clineno = 0;
     1971-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__size,0};
     1972-  __Pyx_RefNannySetupContext("tomaxint");
     1973-  {
     1974-    PyObject* values[1] = {0};
     1975     values[0] = ((PyObject *)Py_None);
     1976     if (unlikely(__pyx_kwds)) {
     1977       Py_ssize_t kw_args;
     1978-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     1979+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     1980+      switch (pos_args) {
     1981         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     1982         case  0: break;
     1983         default: goto __pyx_L5_argtuple_error;
     1984       }
     1985       kw_args = PyDict_Size(__pyx_kwds);
     1986-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     1987+      switch (pos_args) {
     1988         case  0:
     1989         if (kw_args > 0) {
     1990           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__size);
     1991@@ -5903,7 +6137,7 @@
     1992         }
     1993       }
     1994       if (unlikely(kw_args > 0)) {
     1995-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "tomaxint") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 762; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     1996+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "tomaxint") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 762; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     1997       }
     1998     } else {
     1999       switch (PyTuple_GET_SIZE(__pyx_args)) {
     2000@@ -5922,6 +6156,19 @@
     2001   __Pyx_RefNannyFinishContext();
     2002   return NULL;
     2003   __pyx_L4_argument_unpacking_done:;
     2004+  __pyx_r = __pyx_pf_6mtrand_11RandomState_18tomaxint(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_size);
     2005+  __Pyx_RefNannyFinishContext();
     2006+  return __pyx_r;
     2007+}
     2008+
     2009+static PyObject *__pyx_pf_6mtrand_11RandomState_18tomaxint(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_size) {
     2010+  PyObject *__pyx_r = NULL;
     2011+  __Pyx_RefNannyDeclarations
     2012+  PyObject *__pyx_t_1 = NULL;
     2013+  int __pyx_lineno = 0;
     2014+  const char *__pyx_filename = NULL;
     2015+  int __pyx_clineno = 0;
     2016+  __Pyx_RefNannySetupContext("tomaxint", 0);
     2017 
     2018   /* "mtrand.pyx":807
     2019  *
     2020@@ -5931,7 +6178,7 @@
     2021  *     def randint(self, low, high=None, size=None):
     2022  */
     2023   __Pyx_XDECREF(__pyx_r);
     2024-  __pyx_t_1 = __pyx_f_6mtrand_disc0_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_long, __pyx_v_size); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 807; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2025+  __pyx_t_1 = __pyx_f_6mtrand_disc0_array(__pyx_v_self->internal_state, rk_long, __pyx_v_size); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 807; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2026   __Pyx_GOTREF(__pyx_t_1);
     2027   __pyx_r = __pyx_t_1;
     2028   __pyx_t_1 = 0;
     2029@@ -5949,48 +6196,33 @@
     2030   return __pyx_r;
     2031 }
     2032 
     2033-/* "mtrand.pyx":809
     2034- *         return disc0_array(self.internal_state, rk_long, size)
     2035- *
     2036- *     def randint(self, low, high=None, size=None):             # <<<<<<<<<<<<<<
     2037- *         """
     2038- *         randint(low, high=None, size=None)
     2039- */
     2040-
     2041-static PyObject *__pyx_pf_6mtrand_11RandomState_10randint(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2042-static char __pyx_doc_6mtrand_11RandomState_10randint[] = "\n        randint(low, high=None, size=None)\n\n        Return random integers from `low` (inclusive) to `high` (exclusive).\n\n        Return random integers from the \"discrete uniform\" distribution in the\n        \"half-open\" interval [`low`, `high`). If `high` is None (the default),\n        then results are from [0, `low`).\n\n        Parameters\n        ----------\n        low : int\n            Lowest (signed) integer to be drawn from the distribution (unless\n            ``high=None``, in which case this parameter is the *highest* such\n            integer).\n        high : int, optional\n            If provided, one above the largest (signed) integer to be drawn\n            from the distribution (see above for behavior if ``high=None``).\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single int is\n            returned.\n\n        Returns\n        -------\n        out : int or ndarray of ints\n            `size`-shaped array of random integers from the appropriate\n            distribution, or a single such random int if `size` not provided.\n\n        See Also\n        --------\n        random.random_integers : similar to `randint`, only for the closed\n            interval [`low`, `high`], and 1 is the lowest value if `high` is\n            omitted. In particular, this other one is the one to use to generate\n            uniformly distributed discrete non-integers.\n\n        Examples\n        --------\n        >>> np.random.randint(2, size=10)\n        array([1, 0, 0, 0, 1, 1, 0, 0, 1, 0])\n        >>> np.random.randint(1, size=10)\n        array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0])\n\n        Generate a 2 x 4 array of ints between 0 and 4, inclusive:\n\n        >>> np.random.randint(5, size=(2, 4))\n        array([[4, 0, 2, 1],\n               [3, 2, 2, 0]])\n\n        ";
     2043-static PyObject *__pyx_pf_6mtrand_11RandomState_10randint(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2044+/* Python wrapper */
     2045+static PyObject *__pyx_pw_6mtrand_11RandomState_21randint(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2046+static char __pyx_doc_6mtrand_11RandomState_20randint[] = "\n        randint(low, high=None, size=None)\n\n        Return random integers from `low` (inclusive) to `high` (exclusive).\n\n        Return random integers from the \"discrete uniform\" distribution in the\n        \"half-open\" interval [`low`, `high`). If `high` is None (the default),\n        then results are from [0, `low`).\n\n        Parameters\n        ----------\n        low : int\n            Lowest (signed) integer to be drawn from the distribution (unless\n            ``high=None``, in which case this parameter is the *highest* such\n            integer).\n        high : int, optional\n            If provided, one above the largest (signed) integer to be drawn\n            from the distribution (see above for behavior if ``high=None``).\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single int is\n            returned.\n\n        Returns\n        -------\n        out : int or ndarray of ints\n            `size`-shaped array of random integers from the appropriate\n            distribution, or a single such random int if `size` not provided.\n\n        See Also\n        --------\n        random.random_integers : similar to `randint`, only for the closed\n            interval [`low`, `high`], and 1 is the lowest value if `high` is\n            omitted. In particular, this other one is the one to use to generate\n            uniformly distributed discrete non-integers.\n\n        Examples\n        --------\n        >>> np.random.randint(2, size=10)\n        array([1, 0, 0, 0, 1, 1, 0, 0, 1, 0])\n        >>> np.random.randint(1, size=10)\n        array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0])\n\n        Generate a 2 x 4 array of ints between 0 and 4, inclusive:\n\n        >>> np.random.randint(5, size=(2, 4))\n        array([[4, 0, 2, 1],\n               [3, 2, 2, 0]])\n\n        ";
     2047+static PyObject *__pyx_pw_6mtrand_11RandomState_21randint(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2048   PyObject *__pyx_v_low = 0;
     2049   PyObject *__pyx_v_high = 0;
     2050   PyObject *__pyx_v_size = 0;
     2051-  long __pyx_v_lo;
     2052-  long __pyx_v_hi;
     2053-  long __pyx_v_rv;
     2054-  unsigned long __pyx_v_diff;
     2055-  long *__pyx_v_array_data;
     2056-  PyArrayObject *arrayObject = 0;
     2057-  npy_intp __pyx_v_length;
     2058-  npy_intp __pyx_v_i;
     2059-  PyObject *__pyx_r = NULL;
     2060+  PyObject *__pyx_r = 0;
     2061   __Pyx_RefNannyDeclarations
     2062-  int __pyx_t_1;
     2063-  long __pyx_t_2;
     2064-  PyObject *__pyx_t_3 = NULL;
     2065-  PyObject *__pyx_t_4 = NULL;
     2066-  PyObject *__pyx_t_5 = NULL;
     2067-  npy_intp __pyx_t_6;
     2068-  int __pyx_lineno = 0;
     2069-  const char *__pyx_filename = NULL;
     2070-  int __pyx_clineno = 0;
     2071-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__low,&__pyx_n_s__high,&__pyx_n_s__size,0};
     2072-  __Pyx_RefNannySetupContext("randint");
     2073+  __Pyx_RefNannySetupContext("randint (wrapper)", 0);
     2074   {
     2075+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__low,&__pyx_n_s__high,&__pyx_n_s__size,0};
     2076     PyObject* values[3] = {0,0,0};
     2077+
     2078+    /* "mtrand.pyx":809
     2079+ *         return disc0_array(self.internal_state, rk_long, size)
     2080+ *
     2081+ *     def randint(self, low, high=None, size=None):             # <<<<<<<<<<<<<<
     2082+ *         """
     2083+ *         randint(low, high=None, size=None)
     2084+ */
     2085     values[1] = ((PyObject *)Py_None);
     2086     values[2] = ((PyObject *)Py_None);
     2087     if (unlikely(__pyx_kwds)) {
     2088       Py_ssize_t kw_args;
     2089-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     2090+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     2091+      switch (pos_args) {
     2092         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     2093         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     2094         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     2095@@ -5998,10 +6230,9 @@
     2096         default: goto __pyx_L5_argtuple_error;
     2097       }
     2098       kw_args = PyDict_Size(__pyx_kwds);
     2099-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     2100+      switch (pos_args) {
     2101         case  0:
     2102-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__low);
     2103-        if (likely(values[0])) kw_args--;
     2104+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__low)) != 0)) kw_args--;
     2105         else goto __pyx_L5_argtuple_error;
     2106         case  1:
     2107         if (kw_args > 0) {
     2108@@ -6015,7 +6246,7 @@
     2109         }
     2110       }
     2111       if (unlikely(kw_args > 0)) {
     2112-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "randint") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 809; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2113+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "randint") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 809; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2114       }
     2115     } else {
     2116       switch (PyTuple_GET_SIZE(__pyx_args)) {
     2117@@ -6038,6 +6269,32 @@
     2118   __Pyx_RefNannyFinishContext();
     2119   return NULL;
     2120   __pyx_L4_argument_unpacking_done:;
     2121+  __pyx_r = __pyx_pf_6mtrand_11RandomState_20randint(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_low, __pyx_v_high, __pyx_v_size);
     2122+  __Pyx_RefNannyFinishContext();
     2123+  return __pyx_r;
     2124+}
     2125+
     2126+static PyObject *__pyx_pf_6mtrand_11RandomState_20randint(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_low, PyObject *__pyx_v_high, PyObject *__pyx_v_size) {
     2127+  long __pyx_v_lo;
     2128+  long __pyx_v_hi;
     2129+  long __pyx_v_rv;
     2130+  unsigned long __pyx_v_diff;
     2131+  long *__pyx_v_array_data;
     2132+  PyArrayObject *arrayObject = 0;
     2133+  npy_intp __pyx_v_length;
     2134+  npy_intp __pyx_v_i;
     2135+  PyObject *__pyx_r = NULL;
     2136+  __Pyx_RefNannyDeclarations
     2137+  int __pyx_t_1;
     2138+  long __pyx_t_2;
     2139+  PyObject *__pyx_t_3 = NULL;
     2140+  PyObject *__pyx_t_4 = NULL;
     2141+  PyObject *__pyx_t_5 = NULL;
     2142+  npy_intp __pyx_t_6;
     2143+  int __pyx_lineno = 0;
     2144+  const char *__pyx_filename = NULL;
     2145+  int __pyx_clineno = 0;
     2146+  __Pyx_RefNannySetupContext("randint", 0);
     2147 
     2148   /* "mtrand.pyx":866
     2149  *         cdef npy_intp i
     2150@@ -6067,7 +6324,7 @@
     2151  */
     2152     __pyx_t_2 = __Pyx_PyInt_AsLong(__pyx_v_low); if (unlikely((__pyx_t_2 == (long)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 868; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2153     __pyx_v_hi = __pyx_t_2;
     2154-    goto __pyx_L6;
     2155+    goto __pyx_L3;
     2156   }
     2157   /*else*/ {
     2158 
     2159@@ -6091,7 +6348,7 @@
     2160     __pyx_t_2 = __Pyx_PyInt_AsLong(__pyx_v_high); if (unlikely((__pyx_t_2 == (long)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 871; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2161     __pyx_v_hi = __pyx_t_2;
     2162   }
     2163-  __pyx_L6:;
     2164+  __pyx_L3:;
     2165 
     2166   /* "mtrand.pyx":873
     2167  *             hi = high
     2168@@ -6115,9 +6372,9 @@
     2169     __Pyx_Raise(__pyx_t_3, 0, 0, 0);
     2170     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     2171     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 874; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2172-    goto __pyx_L7;
     2173+    goto __pyx_L4;
     2174   }
     2175-  __pyx_L7:;
     2176+  __pyx_L4:;
     2177 
     2178   /* "mtrand.pyx":876
     2179  *             raise ValueError("low >= high")
     2180@@ -6145,7 +6402,7 @@
     2181  *             return rv
     2182  *         else:
     2183  */
     2184-    __pyx_v_rv = (__pyx_v_lo + ((long)rk_interval(__pyx_v_diff, ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state)));
     2185+    __pyx_v_rv = (__pyx_v_lo + ((long)rk_interval(__pyx_v_diff, __pyx_v_self->internal_state)));
     2186 
     2187     /* "mtrand.pyx":879
     2188  *         if size is None:
     2189@@ -6160,7 +6417,7 @@
     2190     __pyx_r = __pyx_t_3;
     2191     __pyx_t_3 = 0;
     2192     goto __pyx_L0;
     2193-    goto __pyx_L8;
     2194+    goto __pyx_L5;
     2195   }
     2196   /*else*/ {
     2197 
     2198@@ -6177,7 +6434,7 @@
     2199     __Pyx_GOTREF(__pyx_t_4);
     2200     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     2201     __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 881; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2202-    __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     2203+    __Pyx_GOTREF(__pyx_t_3);
     2204     __Pyx_INCREF(__pyx_v_size);
     2205     PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_v_size);
     2206     __Pyx_GIVEREF(__pyx_v_size);
     2207@@ -6227,7 +6484,7 @@
     2208  *                 array_data[i] = rv
     2209  *             return array
     2210  */
     2211-      __pyx_v_rv = (__pyx_v_lo + ((long)rk_interval(__pyx_v_diff, ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state)));
     2212+      __pyx_v_rv = (__pyx_v_lo + ((long)rk_interval(__pyx_v_diff, __pyx_v_self->internal_state)));
     2213 
     2214       /* "mtrand.pyx":886
     2215  *             for i from 0 <= i < length:
     2216@@ -6251,7 +6508,7 @@
     2217     __pyx_r = ((PyObject *)arrayObject);
     2218     goto __pyx_L0;
     2219   }
     2220-  __pyx_L8:;
     2221+  __pyx_L5:;
     2222 
     2223   __pyx_r = Py_None; __Pyx_INCREF(Py_None);
     2224   goto __pyx_L0;
     2225@@ -6268,6 +6525,28 @@
     2226   return __pyx_r;
     2227 }
     2228 
     2229+/* Python wrapper */
     2230+static PyObject *__pyx_pw_6mtrand_11RandomState_23bytes(PyObject *__pyx_v_self, PyObject *__pyx_arg_length); /*proto*/
     2231+static char __pyx_doc_6mtrand_11RandomState_22bytes[] = "\n        bytes(length)\n\n        Return random bytes.\n\n        Parameters\n        ----------\n        length : int\n            Number of random bytes.\n\n        Returns\n        -------\n        out : str\n            String of length `length`.\n\n        Examples\n        --------\n        >>> np.random.bytes(10)\n        ' eh\\x85\\x022SZ\\xbf\\xa4' #random\n\n        ";
     2232+static PyObject *__pyx_pw_6mtrand_11RandomState_23bytes(PyObject *__pyx_v_self, PyObject *__pyx_arg_length) {
     2233+  npy_intp __pyx_v_length;
     2234+  PyObject *__pyx_r = 0;
     2235+  __Pyx_RefNannyDeclarations
     2236+  __Pyx_RefNannySetupContext("bytes (wrapper)", 0);
     2237+  assert(__pyx_arg_length); {
     2238+    __pyx_v_length = __Pyx_PyInt_from_py_npy_intp(__pyx_arg_length); if (unlikely((__pyx_v_length == (npy_intp)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 889; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2239+  }
     2240+  goto __pyx_L4_argument_unpacking_done;
     2241+  __pyx_L3_error:;
     2242+  __Pyx_AddTraceback("mtrand.RandomState.bytes", __pyx_clineno, __pyx_lineno, __pyx_filename);
     2243+  __Pyx_RefNannyFinishContext();
     2244+  return NULL;
     2245+  __pyx_L4_argument_unpacking_done:;
     2246+  __pyx_r = __pyx_pf_6mtrand_11RandomState_22bytes(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), ((npy_intp)__pyx_v_length));
     2247+  __Pyx_RefNannyFinishContext();
     2248+  return __pyx_r;
     2249+}
     2250+
     2251 /* "mtrand.pyx":889
     2252  *             return array
     2253  *
     2254@@ -6276,10 +6555,7 @@
     2255  *         bytes(length)
     2256  */
     2257 
     2258-static PyObject *__pyx_pf_6mtrand_11RandomState_11bytes(PyObject *__pyx_v_self, PyObject *__pyx_arg_length); /*proto*/
     2259-static char __pyx_doc_6mtrand_11RandomState_11bytes[] = "\n        bytes(length)\n\n        Return random bytes.\n\n        Parameters\n        ----------\n        length : int\n            Number of random bytes.\n\n        Returns\n        -------\n        out : str\n            String of length `length`.\n\n        Examples\n        --------\n        >>> np.random.bytes(10)\n        ' eh\\x85\\x022SZ\\xbf\\xa4' #random\n\n        ";
     2260-static PyObject *__pyx_pf_6mtrand_11RandomState_11bytes(PyObject *__pyx_v_self, PyObject *__pyx_arg_length) {
     2261-  npy_intp __pyx_v_length;
     2262+static PyObject *__pyx_pf_6mtrand_11RandomState_22bytes(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, npy_intp __pyx_v_length) {
     2263   void *__pyx_v_bytes;
     2264   PyObject *__pyx_v_bytestring = NULL;
     2265   PyObject *__pyx_r = NULL;
     2266@@ -6288,16 +6564,7 @@
     2267   int __pyx_lineno = 0;
     2268   const char *__pyx_filename = NULL;
     2269   int __pyx_clineno = 0;
     2270-  __Pyx_RefNannySetupContext("bytes");
     2271-  assert(__pyx_arg_length); {
     2272-    __pyx_v_length = __Pyx_PyInt_from_py_npy_intp(__pyx_arg_length); if (unlikely((__pyx_v_length == (npy_intp)-1) && PyErr_Occurred())) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 889; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2273-  }
     2274-  goto __pyx_L4_argument_unpacking_done;
     2275-  __pyx_L3_error:;
     2276-  __Pyx_AddTraceback("mtrand.RandomState.bytes", __pyx_clineno, __pyx_lineno, __pyx_filename);
     2277-  __Pyx_RefNannyFinishContext();
     2278-  return NULL;
     2279-  __pyx_L4_argument_unpacking_done:;
     2280+  __Pyx_RefNannySetupContext("bytes", 0);
     2281 
     2282   /* "mtrand.pyx":912
     2283  *         """
     2284@@ -6318,7 +6585,7 @@
     2285  *         return bytestring
     2286  *
     2287  */
     2288-  rk_fill(__pyx_v_bytes, __pyx_v_length, ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state);
     2289+  rk_fill(__pyx_v_bytes, __pyx_v_length, __pyx_v_self->internal_state);
     2290 
     2291   /* "mtrand.pyx":914
     2292  *         bytestring = empty_py_bytes(length, &bytes)
     2293@@ -6345,45 +6612,34 @@
     2294   return __pyx_r;
     2295 }
     2296 
     2297-/* "mtrand.pyx":916
     2298- *         return bytestring
     2299- *
     2300- *     def uniform(self, low=0.0, high=1.0, size=None):             # <<<<<<<<<<<<<<
     2301- *         """
     2302- *         uniform(low=0.0, high=1.0, size=1)
     2303- */
     2304-
     2305-static PyObject *__pyx_pf_6mtrand_11RandomState_12uniform(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2306-static char __pyx_doc_6mtrand_11RandomState_12uniform[] = "\n        uniform(low=0.0, high=1.0, size=1)\n\n        Draw samples from a uniform distribution.\n\n        Samples are uniformly distributed over the half-open interval\n        ``[low, high)`` (includes low, but excludes high).  In other words,\n        any value within the given interval is equally likely to be drawn\n        by `uniform`.\n\n        Parameters\n        ----------\n        low : float, optional\n            Lower boundary of the output interval.  All values generated will be\n            greater than or equal to low.  The default value is 0.\n        high : float\n            Upper boundary of the output interval.  All values generated will be\n            less than high.  The default value is 1.0.\n        size : int or tuple of ints, optional\n            Shape of output.  If the given size is, for example, (m,n,k),\n            m*n*k samples are generated.  If no shape is specified, a single sample\n            is returned.\n\n        Returns\n        -------\n        out : ndarray\n            Drawn samples, with shape `size`.\n\n        See Also\n        --------\n        randint : Discrete uniform distribution, yielding integers.\n        random_integers : Discrete uniform distribution over the closed\n                          interval ``[low, high]``.\n        random_sample : Floats uniformly distributed over ``[0, 1)``.\n        random : Alias for `random_sample`.\n        rand : Convenience function that accepts dimensions as input, e.g.,\n               ``rand(2,2)`` would generate a 2-by-2 array of floats,\n               uniformly distributed over ``[0, 1)``.\n\n        Notes\n        -----\n        The probability density function of the uniform distribution is\n\n        .. math:: p(x) = \\frac{1}{b - a}\n\n        anywhere within the interval ``[a, b)``, and zero elsewhere.\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> s = np.random.uniform(-1,0,1000)\n\n        All values are w""ithin the given interval:\n\n        >>> np.all(s >= -1)\n        True\n        >>> np.all(s < 0)\n        True\n\n        Display the histogram of the samples, along with the\n        probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 15, normed=True)\n        >>> plt.plot(bins, np.ones_like(bins), linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     2307-static PyObject *__pyx_pf_6mtrand_11RandomState_12uniform(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2308+/* Python wrapper */
     2309+static PyObject *__pyx_pw_6mtrand_11RandomState_25uniform(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2310+static char __pyx_doc_6mtrand_11RandomState_24uniform[] = "\n        uniform(low=0.0, high=1.0, size=1)\n\n        Draw samples from a uniform distribution.\n\n        Samples are uniformly distributed over the half-open interval\n        ``[low, high)`` (includes low, but excludes high).  In other words,\n        any value within the given interval is equally likely to be drawn\n        by `uniform`.\n\n        Parameters\n        ----------\n        low : float, optional\n            Lower boundary of the output interval.  All values generated will be\n            greater than or equal to low.  The default value is 0.\n        high : float\n            Upper boundary of the output interval.  All values generated will be\n            less than high.  The default value is 1.0.\n        size : int or tuple of ints, optional\n            Shape of output.  If the given size is, for example, (m,n,k),\n            m*n*k samples are generated.  If no shape is specified, a single sample\n            is returned.\n\n        Returns\n        -------\n        out : ndarray\n            Drawn samples, with shape `size`.\n\n        See Also\n        --------\n        randint : Discrete uniform distribution, yielding integers.\n        random_integers : Discrete uniform distribution over the closed\n                          interval ``[low, high]``.\n        random_sample : Floats uniformly distributed over ``[0, 1)``.\n        random : Alias for `random_sample`.\n        rand : Convenience function that accepts dimensions as input, e.g.,\n               ``rand(2,2)`` would generate a 2-by-2 array of floats,\n               uniformly distributed over ``[0, 1)``.\n\n        Notes\n        -----\n        The probability density function of the uniform distribution is\n\n        .. math:: p(x) = \\frac{1}{b - a}\n\n        anywhere within the interval ``[a, b)``, and zero elsewhere.\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> s = np.random.uniform(-1,0,1000)\n\n        All values are w""ithin the given interval:\n\n        >>> np.all(s >= -1)\n        True\n        >>> np.all(s < 0)\n        True\n\n        Display the histogram of the samples, along with the\n        probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 15, normed=True)\n        >>> plt.plot(bins, np.ones_like(bins), linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     2311+static PyObject *__pyx_pw_6mtrand_11RandomState_25uniform(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2312   PyObject *__pyx_v_low = 0;
     2313   PyObject *__pyx_v_high = 0;
     2314   PyObject *__pyx_v_size = 0;
     2315-  PyArrayObject *__pyx_v_olow = 0;
     2316-  PyArrayObject *__pyx_v_ohigh = 0;
     2317-  PyArrayObject *__pyx_v_odiff = 0;
     2318-  double __pyx_v_flow;
     2319-  double __pyx_v_fhigh;
     2320-  PyObject *__pyx_v_temp = 0;
     2321-  PyObject *__pyx_r = NULL;
     2322+  PyObject *__pyx_r = 0;
     2323   __Pyx_RefNannyDeclarations
     2324-  int __pyx_t_1;
     2325-  PyObject *__pyx_t_2 = NULL;
     2326-  PyObject *__pyx_t_3 = NULL;
     2327-  PyObject *__pyx_t_4 = NULL;
     2328-  int __pyx_lineno = 0;
     2329-  const char *__pyx_filename = NULL;
     2330-  int __pyx_clineno = 0;
     2331-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__low,&__pyx_n_s__high,&__pyx_n_s__size,0};
     2332-  __Pyx_RefNannySetupContext("uniform");
     2333+  __Pyx_RefNannySetupContext("uniform (wrapper)", 0);
     2334   {
     2335+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__low,&__pyx_n_s__high,&__pyx_n_s__size,0};
     2336     PyObject* values[3] = {0,0,0};
     2337     values[0] = __pyx_k_15;
     2338     values[1] = __pyx_k_16;
     2339+
     2340+    /* "mtrand.pyx":916
     2341+ *         return bytestring
     2342+ *
     2343+ *     def uniform(self, low=0.0, high=1.0, size=None):             # <<<<<<<<<<<<<<
     2344+ *         """
     2345+ *         uniform(low=0.0, high=1.0, size=1)
     2346+ */
     2347     values[2] = ((PyObject *)Py_None);
     2348     if (unlikely(__pyx_kwds)) {
     2349       Py_ssize_t kw_args;
     2350-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     2351+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     2352+      switch (pos_args) {
     2353         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     2354         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     2355         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     2356@@ -6391,7 +6647,7 @@
     2357         default: goto __pyx_L5_argtuple_error;
     2358       }
     2359       kw_args = PyDict_Size(__pyx_kwds);
     2360-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     2361+      switch (pos_args) {
     2362         case  0:
     2363         if (kw_args > 0) {
     2364           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__low);
     2365@@ -6409,7 +6665,7 @@
     2366         }
     2367       }
     2368       if (unlikely(kw_args > 0)) {
     2369-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "uniform") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 916; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2370+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "uniform") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 916; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2371       }
     2372     } else {
     2373       switch (PyTuple_GET_SIZE(__pyx_args)) {
     2374@@ -6432,6 +6688,28 @@
     2375   __Pyx_RefNannyFinishContext();
     2376   return NULL;
     2377   __pyx_L4_argument_unpacking_done:;
     2378+  __pyx_r = __pyx_pf_6mtrand_11RandomState_24uniform(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_low, __pyx_v_high, __pyx_v_size);
     2379+  __Pyx_RefNannyFinishContext();
     2380+  return __pyx_r;
     2381+}
     2382+
     2383+static PyObject *__pyx_pf_6mtrand_11RandomState_24uniform(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_low, PyObject *__pyx_v_high, PyObject *__pyx_v_size) {
     2384+  PyArrayObject *__pyx_v_olow = 0;
     2385+  PyArrayObject *__pyx_v_ohigh = 0;
     2386+  PyArrayObject *__pyx_v_odiff = 0;
     2387+  double __pyx_v_flow;
     2388+  double __pyx_v_fhigh;
     2389+  PyObject *__pyx_v_temp = 0;
     2390+  PyObject *__pyx_r = NULL;
     2391+  __Pyx_RefNannyDeclarations
     2392+  int __pyx_t_1;
     2393+  PyObject *__pyx_t_2 = NULL;
     2394+  PyObject *__pyx_t_3 = NULL;
     2395+  PyObject *__pyx_t_4 = NULL;
     2396+  int __pyx_lineno = 0;
     2397+  const char *__pyx_filename = NULL;
     2398+  int __pyx_clineno = 0;
     2399+  __Pyx_RefNannySetupContext("uniform", 0);
     2400 
     2401   /* "mtrand.pyx":990
     2402  *         cdef object temp
     2403@@ -6469,14 +6747,14 @@
     2404  *         olow = <ndarray>PyArray_FROM_OTF(low, NPY_DOUBLE, NPY_ALIGNED)
     2405  */
     2406     __Pyx_XDECREF(__pyx_r);
     2407-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_uniform, __pyx_v_size, __pyx_v_flow, (__pyx_v_fhigh - __pyx_v_flow)); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 993; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2408+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_uniform, __pyx_v_size, __pyx_v_flow, (__pyx_v_fhigh - __pyx_v_flow)); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 993; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2409     __Pyx_GOTREF(__pyx_t_2);
     2410     __pyx_r = __pyx_t_2;
     2411     __pyx_t_2 = 0;
     2412     goto __pyx_L0;
     2413-    goto __pyx_L6;
     2414+    goto __pyx_L3;
     2415   }
     2416-  __pyx_L6:;
     2417+  __pyx_L3:;
     2418 
     2419   /* "mtrand.pyx":994
     2420  *         if not PyErr_Occurred():
     2421@@ -6526,7 +6804,7 @@
     2422   __Pyx_GOTREF(__pyx_t_3);
     2423   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     2424   __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 997; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2425-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     2426+  __Pyx_GOTREF(__pyx_t_2);
     2427   __Pyx_INCREF(((PyObject *)__pyx_v_ohigh));
     2428   PyTuple_SET_ITEM(__pyx_t_2, 0, ((PyObject *)__pyx_v_ohigh));
     2429   __Pyx_GIVEREF(((PyObject *)__pyx_v_ohigh));
     2430@@ -6570,7 +6848,7 @@
     2431  *     def rand(self, *args):
     2432  */
     2433   __Pyx_XDECREF(__pyx_r);
     2434-  __pyx_t_4 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_uniform, __pyx_v_size, __pyx_v_olow, __pyx_v_odiff); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1001; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2435+  __pyx_t_4 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_uniform, __pyx_v_size, __pyx_v_olow, __pyx_v_odiff); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1001; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2436   __Pyx_GOTREF(__pyx_t_4);
     2437   __pyx_r = __pyx_t_4;
     2438   __pyx_t_4 = 0;
     2439@@ -6594,6 +6872,23 @@
     2440   return __pyx_r;
     2441 }
     2442 
     2443+/* Python wrapper */
     2444+static PyObject *__pyx_pw_6mtrand_11RandomState_27rand(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2445+static char __pyx_doc_6mtrand_11RandomState_26rand[] = "\n        rand(d0, d1, ..., dn)\n\n        Random values in a given shape.\n\n        Create an array of the given shape and propagate it with\n        random samples from a uniform distribution\n        over ``[0, 1)``.\n\n        Parameters\n        ----------\n        d0, d1, ..., dn : int\n            Shape of the output.\n\n        Returns\n        -------\n        out : ndarray, shape ``(d0, d1, ..., dn)``\n            Random values.\n\n        See Also\n        --------\n        random\n\n        Notes\n        -----\n        This is a convenience function. If you want an interface that\n        takes a shape-tuple as the first argument, refer to\n        `random`.\n\n        Examples\n        --------\n        >>> np.random.rand(3,2)\n        array([[ 0.14022471,  0.96360618],  #random\n               [ 0.37601032,  0.25528411],  #random\n               [ 0.49313049,  0.94909878]]) #random\n\n        ";
     2446+static PyObject *__pyx_pw_6mtrand_11RandomState_27rand(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2447+  PyObject *__pyx_v_args = 0;
     2448+  PyObject *__pyx_r = 0;
     2449+  __Pyx_RefNannyDeclarations
     2450+  __Pyx_RefNannySetupContext("rand (wrapper)", 0);
     2451+  if (unlikely(__pyx_kwds) && unlikely(PyDict_Size(__pyx_kwds) > 0) && unlikely(!__Pyx_CheckKeywordStrings(__pyx_kwds, "rand", 0))) return NULL;
     2452+  __Pyx_INCREF(__pyx_args);
     2453+  __pyx_v_args = __pyx_args;
     2454+  __pyx_r = __pyx_pf_6mtrand_11RandomState_26rand(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_args);
     2455+  __Pyx_XDECREF(__pyx_v_args);
     2456+  __Pyx_RefNannyFinishContext();
     2457+  return __pyx_r;
     2458+}
     2459+
     2460 /* "mtrand.pyx":1003
     2461  *         return cont2_array(self.internal_state, rk_uniform, size, olow, odiff)
     2462  *
     2463@@ -6602,10 +6897,7 @@
     2464  *         rand(d0, d1, ..., dn)
     2465  */
     2466 
     2467-static PyObject *__pyx_pf_6mtrand_11RandomState_13rand(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2468-static char __pyx_doc_6mtrand_11RandomState_13rand[] = "\n        rand(d0, d1, ..., dn)\n\n        Random values in a given shape.\n\n        Create an array of the given shape and propagate it with\n        random samples from a uniform distribution\n        over ``[0, 1)``.\n\n        Parameters\n        ----------\n        d0, d1, ..., dn : int\n            Shape of the output.\n\n        Returns\n        -------\n        out : ndarray, shape ``(d0, d1, ..., dn)``\n            Random values.\n\n        See Also\n        --------\n        random\n\n        Notes\n        -----\n        This is a convenience function. If you want an interface that\n        takes a shape-tuple as the first argument, refer to\n        `random`.\n\n        Examples\n        --------\n        >>> np.random.rand(3,2)\n        array([[ 0.14022471,  0.96360618],  #random\n               [ 0.37601032,  0.25528411],  #random\n               [ 0.49313049,  0.94909878]]) #random\n\n        ";
     2469-static PyObject *__pyx_pf_6mtrand_11RandomState_13rand(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2470-  PyObject *__pyx_v_args = 0;
     2471+static PyObject *__pyx_pf_6mtrand_11RandomState_26rand(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_args) {
     2472   PyObject *__pyx_r = NULL;
     2473   __Pyx_RefNannyDeclarations
     2474   Py_ssize_t __pyx_t_1;
     2475@@ -6616,10 +6908,7 @@
     2476   int __pyx_lineno = 0;
     2477   const char *__pyx_filename = NULL;
     2478   int __pyx_clineno = 0;
     2479-  __Pyx_RefNannySetupContext("rand");
     2480-  if (unlikely(__pyx_kwds) && unlikely(PyDict_Size(__pyx_kwds) > 0) && unlikely(!__Pyx_CheckKeywordStrings(__pyx_kwds, "rand", 0))) return NULL;
     2481-  __Pyx_INCREF(__pyx_args);
     2482-  __pyx_v_args = __pyx_args;
     2483+  __Pyx_RefNannySetupContext("rand", 0);
     2484 
     2485   /* "mtrand.pyx":1041
     2486  *
     2487@@ -6628,10 +6917,7 @@
     2488  *             return self.random_sample()
     2489  *         else:
     2490  */
     2491-  if (unlikely(((PyObject *)__pyx_v_args) == Py_None)) {
     2492-    PyErr_SetString(PyExc_TypeError, "object of type 'NoneType' has no len()"); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1041; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2493-  }
     2494-  __pyx_t_1 = PyTuple_GET_SIZE(((PyObject *)__pyx_v_args));
     2495+  __pyx_t_1 = PyTuple_GET_SIZE(((PyObject *)__pyx_v_args)); if (unlikely(__pyx_t_1 == -1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1041; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2496   __pyx_t_2 = (__pyx_t_1 == 0);
     2497   if (__pyx_t_2) {
     2498 
     2499@@ -6643,7 +6929,7 @@
     2500  *             return self.random_sample(size=args)
     2501  */
     2502     __Pyx_XDECREF(__pyx_r);
     2503-    __pyx_t_3 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__random_sample); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1042; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2504+    __pyx_t_3 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__random_sample); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1042; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2505     __Pyx_GOTREF(__pyx_t_3);
     2506     __pyx_t_4 = PyObject_Call(__pyx_t_3, ((PyObject *)__pyx_empty_tuple), NULL); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1042; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2507     __Pyx_GOTREF(__pyx_t_4);
     2508@@ -6651,7 +6937,7 @@
     2509     __pyx_r = __pyx_t_4;
     2510     __pyx_t_4 = 0;
     2511     goto __pyx_L0;
     2512-    goto __pyx_L5;
     2513+    goto __pyx_L3;
     2514   }
     2515   /*else*/ {
     2516 
     2517@@ -6663,12 +6949,12 @@
     2518  *     def randn(self, *args):
     2519  */
     2520     __Pyx_XDECREF(__pyx_r);
     2521-    __pyx_t_4 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__random_sample); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1044; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2522+    __pyx_t_4 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__random_sample); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1044; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2523     __Pyx_GOTREF(__pyx_t_4);
     2524     __pyx_t_3 = PyDict_New(); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1044; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2525     __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     2526     if (PyDict_SetItem(__pyx_t_3, ((PyObject *)__pyx_n_s__size), ((PyObject *)__pyx_v_args)) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1044; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2527-    __pyx_t_5 = PyEval_CallObjectWithKeywords(__pyx_t_4, ((PyObject *)__pyx_empty_tuple), ((PyObject *)__pyx_t_3)); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1044; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2528+    __pyx_t_5 = PyObject_Call(__pyx_t_4, ((PyObject *)__pyx_empty_tuple), ((PyObject *)__pyx_t_3)); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1044; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2529     __Pyx_GOTREF(__pyx_t_5);
     2530     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     2531     __Pyx_DECREF(((PyObject *)__pyx_t_3)); __pyx_t_3 = 0;
     2532@@ -6676,7 +6962,7 @@
     2533     __pyx_t_5 = 0;
     2534     goto __pyx_L0;
     2535   }
     2536-  __pyx_L5:;
     2537+  __pyx_L3:;
     2538 
     2539   __pyx_r = Py_None; __Pyx_INCREF(Py_None);
     2540   goto __pyx_L0;
     2541@@ -6687,12 +6973,28 @@
     2542   __Pyx_AddTraceback("mtrand.RandomState.rand", __pyx_clineno, __pyx_lineno, __pyx_filename);
     2543   __pyx_r = NULL;
     2544   __pyx_L0:;
     2545-  __Pyx_XDECREF(__pyx_v_args);
     2546   __Pyx_XGIVEREF(__pyx_r);
     2547   __Pyx_RefNannyFinishContext();
     2548   return __pyx_r;
     2549 }
     2550 
     2551+/* Python wrapper */
     2552+static PyObject *__pyx_pw_6mtrand_11RandomState_29randn(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2553+static char __pyx_doc_6mtrand_11RandomState_28randn[] = "\n        randn([d1, ..., dn])\n\n        Return a sample (or samples) from the \"standard normal\" distribution.\n\n        If positive, int_like or int-convertible arguments are provided,\n        `randn` generates an array of shape ``(d1, ..., dn)``, filled\n        with random floats sampled from a univariate \"normal\" (Gaussian)\n        distribution of mean 0 and variance 1 (if any of the :math:`d_i` are\n        floats, they are first converted to integers by truncation). A single\n        float randomly sampled from the distribution is returned if no\n        argument is provided.\n\n        This is a convenience function.  If you want an interface that takes a\n        tuple as the first argument, use `numpy.random.standard_normal` instead.\n\n        Parameters\n        ----------\n        d1, ..., dn : `n` ints, optional\n            The dimensions of the returned array, should be all positive.\n\n        Returns\n        -------\n        Z : ndarray or float\n            A ``(d1, ..., dn)``-shaped array of floating-point samples from\n            the standard normal distribution, or a single such float if\n            no parameters were supplied.\n\n        See Also\n        --------\n        random.standard_normal : Similar, but takes a tuple as its argument.\n\n        Notes\n        -----\n        For random samples from :math:`N(\\mu, \\sigma^2)`, use:\n\n        ``sigma * np.random.randn(...) + mu``\n\n        Examples\n        --------\n        >>> np.random.randn()\n        2.1923875335537315 #random\n\n        Two-by-four array of samples from N(3, 6.25):\n\n        >>> 2.5 * np.random.randn(2, 4) + 3\n        array([[-4.49401501,  4.00950034, -1.81814867,  7.29718677],  #random\n               [ 0.39924804,  4.68456316,  4.99394529,  4.84057254]]) #random\n\n        ";
     2554+static PyObject *__pyx_pw_6mtrand_11RandomState_29randn(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2555+  PyObject *__pyx_v_args = 0;
     2556+  PyObject *__pyx_r = 0;
     2557+  __Pyx_RefNannyDeclarations
     2558+  __Pyx_RefNannySetupContext("randn (wrapper)", 0);
     2559+  if (unlikely(__pyx_kwds) && unlikely(PyDict_Size(__pyx_kwds) > 0) && unlikely(!__Pyx_CheckKeywordStrings(__pyx_kwds, "randn", 0))) return NULL;
     2560+  __Pyx_INCREF(__pyx_args);
     2561+  __pyx_v_args = __pyx_args;
     2562+  __pyx_r = __pyx_pf_6mtrand_11RandomState_28randn(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_args);
     2563+  __Pyx_XDECREF(__pyx_v_args);
     2564+  __Pyx_RefNannyFinishContext();
     2565+  return __pyx_r;
     2566+}
     2567+
     2568 /* "mtrand.pyx":1046
     2569  *             return self.random_sample(size=args)
     2570  *
     2571@@ -6701,10 +7003,7 @@
     2572  *         randn([d1, ..., dn])
     2573  */
     2574 
     2575-static PyObject *__pyx_pf_6mtrand_11RandomState_14randn(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2576-static char __pyx_doc_6mtrand_11RandomState_14randn[] = "\n        randn([d1, ..., dn])\n\n        Return a sample (or samples) from the \"standard normal\" distribution.\n\n        If positive, int_like or int-convertible arguments are provided,\n        `randn` generates an array of shape ``(d1, ..., dn)``, filled\n        with random floats sampled from a univariate \"normal\" (Gaussian)\n        distribution of mean 0 and variance 1 (if any of the :math:`d_i` are\n        floats, they are first converted to integers by truncation). A single\n        float randomly sampled from the distribution is returned if no\n        argument is provided.\n\n        This is a convenience function.  If you want an interface that takes a\n        tuple as the first argument, use `numpy.random.standard_normal` instead.\n\n        Parameters\n        ----------\n        d1, ..., dn : `n` ints, optional\n            The dimensions of the returned array, should be all positive.\n\n        Returns\n        -------\n        Z : ndarray or float\n            A ``(d1, ..., dn)``-shaped array of floating-point samples from\n            the standard normal distribution, or a single such float if\n            no parameters were supplied.\n\n        See Also\n        --------\n        random.standard_normal : Similar, but takes a tuple as its argument.\n\n        Notes\n        -----\n        For random samples from :math:`N(\\mu, \\sigma^2)`, use:\n\n        ``sigma * np.random.randn(...) + mu``\n\n        Examples\n        --------\n        >>> np.random.randn()\n        2.1923875335537315 #random\n\n        Two-by-four array of samples from N(3, 6.25):\n\n        >>> 2.5 * np.random.randn(2, 4) + 3\n        array([[-4.49401501,  4.00950034, -1.81814867,  7.29718677],  #random\n               [ 0.39924804,  4.68456316,  4.99394529,  4.84057254]]) #random\n\n        ";
     2577-static PyObject *__pyx_pf_6mtrand_11RandomState_14randn(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2578-  PyObject *__pyx_v_args = 0;
     2579+static PyObject *__pyx_pf_6mtrand_11RandomState_28randn(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_args) {
     2580   PyObject *__pyx_r = NULL;
     2581   __Pyx_RefNannyDeclarations
     2582   Py_ssize_t __pyx_t_1;
     2583@@ -6715,10 +7014,7 @@
     2584   int __pyx_lineno = 0;
     2585   const char *__pyx_filename = NULL;
     2586   int __pyx_clineno = 0;
     2587-  __Pyx_RefNannySetupContext("randn");
     2588-  if (unlikely(__pyx_kwds) && unlikely(PyDict_Size(__pyx_kwds) > 0) && unlikely(!__Pyx_CheckKeywordStrings(__pyx_kwds, "randn", 0))) return NULL;
     2589-  __Pyx_INCREF(__pyx_args);
     2590-  __pyx_v_args = __pyx_args;
     2591+  __Pyx_RefNannySetupContext("randn", 0);
     2592 
     2593   /* "mtrand.pyx":1097
     2594  *
     2595@@ -6727,10 +7023,7 @@
     2596  *             return self.standard_normal()
     2597  *         else:
     2598  */
     2599-  if (unlikely(((PyObject *)__pyx_v_args) == Py_None)) {
     2600-    PyErr_SetString(PyExc_TypeError, "object of type 'NoneType' has no len()"); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1097; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2601-  }
     2602-  __pyx_t_1 = PyTuple_GET_SIZE(((PyObject *)__pyx_v_args));
     2603+  __pyx_t_1 = PyTuple_GET_SIZE(((PyObject *)__pyx_v_args)); if (unlikely(__pyx_t_1 == -1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1097; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2604   __pyx_t_2 = (__pyx_t_1 == 0);
     2605   if (__pyx_t_2) {
     2606 
     2607@@ -6742,7 +7035,7 @@
     2608  *             return self.standard_normal(args)
     2609  */
     2610     __Pyx_XDECREF(__pyx_r);
     2611-    __pyx_t_3 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__standard_normal); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1098; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2612+    __pyx_t_3 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__standard_normal); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1098; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2613     __Pyx_GOTREF(__pyx_t_3);
     2614     __pyx_t_4 = PyObject_Call(__pyx_t_3, ((PyObject *)__pyx_empty_tuple), NULL); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1098; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2615     __Pyx_GOTREF(__pyx_t_4);
     2616@@ -6750,7 +7043,7 @@
     2617     __pyx_r = __pyx_t_4;
     2618     __pyx_t_4 = 0;
     2619     goto __pyx_L0;
     2620-    goto __pyx_L5;
     2621+    goto __pyx_L3;
     2622   }
     2623   /*else*/ {
     2624 
     2625@@ -6762,10 +7055,10 @@
     2626  *     def random_integers(self, low, high=None, size=None):
     2627  */
     2628     __Pyx_XDECREF(__pyx_r);
     2629-    __pyx_t_4 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__standard_normal); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1100; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2630+    __pyx_t_4 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__standard_normal); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1100; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2631     __Pyx_GOTREF(__pyx_t_4);
     2632     __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1100; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2633-    __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     2634+    __Pyx_GOTREF(__pyx_t_3);
     2635     __Pyx_INCREF(((PyObject *)__pyx_v_args));
     2636     PyTuple_SET_ITEM(__pyx_t_3, 0, ((PyObject *)__pyx_v_args));
     2637     __Pyx_GIVEREF(((PyObject *)__pyx_v_args));
     2638@@ -6777,7 +7070,7 @@
     2639     __pyx_t_5 = 0;
     2640     goto __pyx_L0;
     2641   }
     2642-  __pyx_L5:;
     2643+  __pyx_L3:;
     2644 
     2645   __pyx_r = Py_None; __Pyx_INCREF(Py_None);
     2646   goto __pyx_L0;
     2647@@ -6788,44 +7081,38 @@
     2648   __Pyx_AddTraceback("mtrand.RandomState.randn", __pyx_clineno, __pyx_lineno, __pyx_filename);
     2649   __pyx_r = NULL;
     2650   __pyx_L0:;
     2651-  __Pyx_XDECREF(__pyx_v_args);
     2652   __Pyx_XGIVEREF(__pyx_r);
     2653   __Pyx_RefNannyFinishContext();
     2654   return __pyx_r;
     2655 }
     2656 
     2657-/* "mtrand.pyx":1102
     2658- *             return self.standard_normal(args)
     2659- *
     2660- *     def random_integers(self, low, high=None, size=None):             # <<<<<<<<<<<<<<
     2661- *         """
     2662- *         random_integers(low, high=None, size=None)
     2663- */
     2664-
     2665-static PyObject *__pyx_pf_6mtrand_11RandomState_15random_integers(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2666-static char __pyx_doc_6mtrand_11RandomState_15random_integers[] = "\n        random_integers(low, high=None, size=None)\n\n        Return random integers between `low` and `high`, inclusive.\n\n        Return random integers from the \"discrete uniform\" distribution in the\n        closed interval [`low`, `high`].  If `high` is None (the default),\n        then results are from [1, `low`].\n\n        Parameters\n        ----------\n        low : int\n            Lowest (signed) integer to be drawn from the distribution (unless\n            ``high=None``, in which case this parameter is the *highest* such\n            integer).\n        high : int, optional\n            If provided, the largest (signed) integer to be drawn from the\n            distribution (see above for behavior if ``high=None``).\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single int is returned.\n\n        Returns\n        -------\n        out : int or ndarray of ints\n            `size`-shaped array of random integers from the appropriate\n            distribution, or a single such random int if `size` not provided.\n\n        See Also\n        --------\n        random.randint : Similar to `random_integers`, only for the half-open\n            interval [`low`, `high`), and 0 is the lowest value if `high` is\n            omitted.\n\n        Notes\n        -----\n        To sample from N evenly spaced floating-point numbers between a and b,\n        use::\n\n          a + (b - a) * (np.random.random_integers(N) - 1) / (N - 1.)\n\n        Examples\n        --------\n        >>> np.random.random_integers(5)\n        4\n        >>> type(np.random.random_integers(5))\n        <type 'int'>\n        >>> np.random.random_integers(5, size=(3.,2.))\n        array([[5, 4],\n               [3, 3],\n               [4, 5]])\n\n        Choose five random numbers from the set of five evenly-spaced\n        numbers between 0 and 2.5, inclusive (*i.e.*, from the set\n        :math:`{0, 5/8, 10/8, 15/8, 20/8}`):\n""\n        >>> 2.5 * (np.random.random_integers(5, size=(5,)) - 1) / 4.\n        array([ 0.625,  1.25 ,  0.625,  0.625,  2.5  ])\n\n        Roll two six sided dice 1000 times and sum the results:\n\n        >>> d1 = np.random.random_integers(1, 6, 1000)\n        >>> d2 = np.random.random_integers(1, 6, 1000)\n        >>> dsums = d1 + d2\n\n        Display results as a histogram:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(dsums, 11, normed=True)\n        >>> plt.show()\n\n        ";
     2667-static PyObject *__pyx_pf_6mtrand_11RandomState_15random_integers(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2668+/* Python wrapper */
     2669+static PyObject *__pyx_pw_6mtrand_11RandomState_31random_integers(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2670+static char __pyx_doc_6mtrand_11RandomState_30random_integers[] = "\n        random_integers(low, high=None, size=None)\n\n        Return random integers between `low` and `high`, inclusive.\n\n        Return random integers from the \"discrete uniform\" distribution in the\n        closed interval [`low`, `high`].  If `high` is None (the default),\n        then results are from [1, `low`].\n\n        Parameters\n        ----------\n        low : int\n            Lowest (signed) integer to be drawn from the distribution (unless\n            ``high=None``, in which case this parameter is the *highest* such\n            integer).\n        high : int, optional\n            If provided, the largest (signed) integer to be drawn from the\n            distribution (see above for behavior if ``high=None``).\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single int is returned.\n\n        Returns\n        -------\n        out : int or ndarray of ints\n            `size`-shaped array of random integers from the appropriate\n            distribution, or a single such random int if `size` not provided.\n\n        See Also\n        --------\n        random.randint : Similar to `random_integers`, only for the half-open\n            interval [`low`, `high`), and 0 is the lowest value if `high` is\n            omitted.\n\n        Notes\n        -----\n        To sample from N evenly spaced floating-point numbers between a and b,\n        use::\n\n          a + (b - a) * (np.random.random_integers(N) - 1) / (N - 1.)\n\n        Examples\n        --------\n        >>> np.random.random_integers(5)\n        4\n        >>> type(np.random.random_integers(5))\n        <type 'int'>\n        >>> np.random.random_integers(5, size=(3.,2.))\n        array([[5, 4],\n               [3, 3],\n               [4, 5]])\n\n        Choose five random numbers from the set of five evenly-spaced\n        numbers between 0 and 2.5, inclusive (*i.e.*, from the set\n        :math:`{0, 5/8, 10/8, 15/8, 20/8}`):\n""\n        >>> 2.5 * (np.random.random_integers(5, size=(5,)) - 1) / 4.\n        array([ 0.625,  1.25 ,  0.625,  0.625,  2.5  ])\n\n        Roll two six sided dice 1000 times and sum the results:\n\n        >>> d1 = np.random.random_integers(1, 6, 1000)\n        >>> d2 = np.random.random_integers(1, 6, 1000)\n        >>> dsums = d1 + d2\n\n        Display results as a histogram:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(dsums, 11, normed=True)\n        >>> plt.show()\n\n        ";
     2671+static PyObject *__pyx_pw_6mtrand_11RandomState_31random_integers(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2672   PyObject *__pyx_v_low = 0;
     2673   PyObject *__pyx_v_high = 0;
     2674   PyObject *__pyx_v_size = 0;
     2675-  PyObject *__pyx_r = NULL;
     2676+  PyObject *__pyx_r = 0;
     2677   __Pyx_RefNannyDeclarations
     2678-  int __pyx_t_1;
     2679-  PyObject *__pyx_t_2 = NULL;
     2680-  PyObject *__pyx_t_3 = NULL;
     2681-  PyObject *__pyx_t_4 = NULL;
     2682-  int __pyx_lineno = 0;
     2683-  const char *__pyx_filename = NULL;
     2684-  int __pyx_clineno = 0;
     2685-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__low,&__pyx_n_s__high,&__pyx_n_s__size,0};
     2686-  __Pyx_RefNannySetupContext("random_integers");
     2687+  __Pyx_RefNannySetupContext("random_integers (wrapper)", 0);
     2688   {
     2689+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__low,&__pyx_n_s__high,&__pyx_n_s__size,0};
     2690     PyObject* values[3] = {0,0,0};
     2691+
     2692+    /* "mtrand.pyx":1102
     2693+ *             return self.standard_normal(args)
     2694+ *
     2695+ *     def random_integers(self, low, high=None, size=None):             # <<<<<<<<<<<<<<
     2696+ *         """
     2697+ *         random_integers(low, high=None, size=None)
     2698+ */
     2699     values[1] = ((PyObject *)Py_None);
     2700     values[2] = ((PyObject *)Py_None);
     2701     if (unlikely(__pyx_kwds)) {
     2702       Py_ssize_t kw_args;
     2703-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     2704+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     2705+      switch (pos_args) {
     2706         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     2707         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     2708         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     2709@@ -6833,10 +7120,9 @@
     2710         default: goto __pyx_L5_argtuple_error;
     2711       }
     2712       kw_args = PyDict_Size(__pyx_kwds);
     2713-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     2714+      switch (pos_args) {
     2715         case  0:
     2716-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__low);
     2717-        if (likely(values[0])) kw_args--;
     2718+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__low)) != 0)) kw_args--;
     2719         else goto __pyx_L5_argtuple_error;
     2720         case  1:
     2721         if (kw_args > 0) {
     2722@@ -6850,7 +7136,7 @@
     2723         }
     2724       }
     2725       if (unlikely(kw_args > 0)) {
     2726-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "random_integers") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1102; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2727+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "random_integers") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1102; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2728       }
     2729     } else {
     2730       switch (PyTuple_GET_SIZE(__pyx_args)) {
     2731@@ -6873,6 +7159,22 @@
     2732   __Pyx_RefNannyFinishContext();
     2733   return NULL;
     2734   __pyx_L4_argument_unpacking_done:;
     2735+  __pyx_r = __pyx_pf_6mtrand_11RandomState_30random_integers(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_low, __pyx_v_high, __pyx_v_size);
     2736+  __Pyx_RefNannyFinishContext();
     2737+  return __pyx_r;
     2738+}
     2739+
     2740+static PyObject *__pyx_pf_6mtrand_11RandomState_30random_integers(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_low, PyObject *__pyx_v_high, PyObject *__pyx_v_size) {
     2741+  PyObject *__pyx_r = NULL;
     2742+  __Pyx_RefNannyDeclarations
     2743+  int __pyx_t_1;
     2744+  PyObject *__pyx_t_2 = NULL;
     2745+  PyObject *__pyx_t_3 = NULL;
     2746+  PyObject *__pyx_t_4 = NULL;
     2747+  int __pyx_lineno = 0;
     2748+  const char *__pyx_filename = NULL;
     2749+  int __pyx_clineno = 0;
     2750+  __Pyx_RefNannySetupContext("random_integers", 0);
     2751   __Pyx_INCREF(__pyx_v_low);
     2752   __Pyx_INCREF(__pyx_v_high);
     2753 
     2754@@ -6907,9 +7209,9 @@
     2755     __Pyx_INCREF(__pyx_int_1);
     2756     __Pyx_DECREF(__pyx_v_low);
     2757     __pyx_v_low = __pyx_int_1;
     2758-    goto __pyx_L6;
     2759+    goto __pyx_L3;
     2760   }
     2761-  __pyx_L6:;
     2762+  __pyx_L3:;
     2763 
     2764   /* "mtrand.pyx":1177
     2765  *             high = low
     2766@@ -6919,12 +7221,12 @@
     2767  *     # Complicated, continuous distributions:
     2768  */
     2769   __Pyx_XDECREF(__pyx_r);
     2770-  __pyx_t_2 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__randint); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1177; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2771+  __pyx_t_2 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__randint); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1177; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2772   __Pyx_GOTREF(__pyx_t_2);
     2773   __pyx_t_3 = PyNumber_Add(__pyx_v_high, __pyx_int_1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1177; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2774   __Pyx_GOTREF(__pyx_t_3);
     2775   __pyx_t_4 = PyTuple_New(3); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1177; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2776-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     2777+  __Pyx_GOTREF(__pyx_t_4);
     2778   __Pyx_INCREF(__pyx_v_low);
     2779   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_low);
     2780   __Pyx_GIVEREF(__pyx_v_low);
     2781@@ -6958,38 +7260,36 @@
     2782   return __pyx_r;
     2783 }
     2784 
     2785-/* "mtrand.pyx":1180
     2786+/* Python wrapper */
     2787+static PyObject *__pyx_pw_6mtrand_11RandomState_33standard_normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2788+static char __pyx_doc_6mtrand_11RandomState_32standard_normal[] = "\n        standard_normal(size=None)\n\n        Returns samples from a Standard Normal distribution (mean=0, stdev=1).\n\n        Parameters\n        ----------\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single value is\n            returned.\n\n        Returns\n        -------\n        out : float or ndarray\n            Drawn samples.\n\n        Examples\n        --------\n        >>> s = np.random.standard_normal(8000)\n        >>> s\n        array([ 0.6888893 ,  0.78096262, -0.89086505, ...,  0.49876311, #random\n               -0.38672696, -0.4685006 ])                               #random\n        >>> s.shape\n        (8000,)\n        >>> s = np.random.standard_normal(size=(3, 4, 2))\n        >>> s.shape\n        (3, 4, 2)\n\n        ";
     2789+static PyObject *__pyx_pw_6mtrand_11RandomState_33standard_normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2790+  PyObject *__pyx_v_size = 0;
     2791+  PyObject *__pyx_r = 0;
     2792+  __Pyx_RefNannyDeclarations
     2793+  __Pyx_RefNannySetupContext("standard_normal (wrapper)", 0);
     2794+  {
     2795+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__size,0};
     2796+    PyObject* values[1] = {0};
     2797+
     2798+    /* "mtrand.pyx":1180
     2799  *
     2800  *     # Complicated, continuous distributions:
     2801  *     def standard_normal(self, size=None):             # <<<<<<<<<<<<<<
     2802  *         """
     2803  *         standard_normal(size=None)
     2804  */
     2805-
     2806-static PyObject *__pyx_pf_6mtrand_11RandomState_16standard_normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2807-static char __pyx_doc_6mtrand_11RandomState_16standard_normal[] = "\n        standard_normal(size=None)\n\n        Returns samples from a Standard Normal distribution (mean=0, stdev=1).\n\n        Parameters\n        ----------\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single value is\n            returned.\n\n        Returns\n        -------\n        out : float or ndarray\n            Drawn samples.\n\n        Examples\n        --------\n        >>> s = np.random.standard_normal(8000)\n        >>> s\n        array([ 0.6888893 ,  0.78096262, -0.89086505, ...,  0.49876311, #random\n               -0.38672696, -0.4685006 ])                               #random\n        >>> s.shape\n        (8000,)\n        >>> s = np.random.standard_normal(size=(3, 4, 2))\n        >>> s.shape\n        (3, 4, 2)\n\n        ";
     2808-static PyObject *__pyx_pf_6mtrand_11RandomState_16standard_normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2809-  PyObject *__pyx_v_size = 0;
     2810-  PyObject *__pyx_r = NULL;
     2811-  __Pyx_RefNannyDeclarations
     2812-  PyObject *__pyx_t_1 = NULL;
     2813-  int __pyx_lineno = 0;
     2814-  const char *__pyx_filename = NULL;
     2815-  int __pyx_clineno = 0;
     2816-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__size,0};
     2817-  __Pyx_RefNannySetupContext("standard_normal");
     2818-  {
     2819-    PyObject* values[1] = {0};
     2820     values[0] = ((PyObject *)Py_None);
     2821     if (unlikely(__pyx_kwds)) {
     2822       Py_ssize_t kw_args;
     2823-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     2824+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     2825+      switch (pos_args) {
     2826         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     2827         case  0: break;
     2828         default: goto __pyx_L5_argtuple_error;
     2829       }
     2830       kw_args = PyDict_Size(__pyx_kwds);
     2831-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     2832+      switch (pos_args) {
     2833         case  0:
     2834         if (kw_args > 0) {
     2835           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__size);
     2836@@ -6997,7 +7297,7 @@
     2837         }
     2838       }
     2839       if (unlikely(kw_args > 0)) {
     2840-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "standard_normal") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1180; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2841+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "standard_normal") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1180; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2842       }
     2843     } else {
     2844       switch (PyTuple_GET_SIZE(__pyx_args)) {
     2845@@ -7016,6 +7316,19 @@
     2846   __Pyx_RefNannyFinishContext();
     2847   return NULL;
     2848   __pyx_L4_argument_unpacking_done:;
     2849+  __pyx_r = __pyx_pf_6mtrand_11RandomState_32standard_normal(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_size);
     2850+  __Pyx_RefNannyFinishContext();
     2851+  return __pyx_r;
     2852+}
     2853+
     2854+static PyObject *__pyx_pf_6mtrand_11RandomState_32standard_normal(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_size) {
     2855+  PyObject *__pyx_r = NULL;
     2856+  __Pyx_RefNannyDeclarations
     2857+  PyObject *__pyx_t_1 = NULL;
     2858+  int __pyx_lineno = 0;
     2859+  const char *__pyx_filename = NULL;
     2860+  int __pyx_clineno = 0;
     2861+  __Pyx_RefNannySetupContext("standard_normal", 0);
     2862 
     2863   /* "mtrand.pyx":1210
     2864  *
     2865@@ -7025,7 +7338,7 @@
     2866  *     def normal(self, loc=0.0, scale=1.0, size=None):
     2867  */
     2868   __Pyx_XDECREF(__pyx_r);
     2869-  __pyx_t_1 = __pyx_f_6mtrand_cont0_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_gauss, __pyx_v_size); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1210; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2870+  __pyx_t_1 = __pyx_f_6mtrand_cont0_array(__pyx_v_self->internal_state, rk_gauss, __pyx_v_size); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1210; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2871   __Pyx_GOTREF(__pyx_t_1);
     2872   __pyx_r = __pyx_t_1;
     2873   __pyx_t_1 = 0;
     2874@@ -7043,44 +7356,34 @@
     2875   return __pyx_r;
     2876 }
     2877 
     2878-/* "mtrand.pyx":1212
     2879- *         return cont0_array(self.internal_state, rk_gauss, size)
     2880- *
     2881- *     def normal(self, loc=0.0, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     2882- *         """
     2883- *         normal(loc=0.0, scale=1.0, size=None)
     2884- */
     2885-
     2886-static PyObject *__pyx_pf_6mtrand_11RandomState_17normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2887-static char __pyx_doc_6mtrand_11RandomState_17normal[] = "\n        normal(loc=0.0, scale=1.0, size=None)\n\n        Draw random samples from a normal (Gaussian) distribution.\n\n        The probability density function of the normal distribution, first\n        derived by De Moivre and 200 years later by both Gauss and Laplace\n        independently [2]_, is often called the bell curve because of\n        its characteristic shape (see the example below).\n\n        The normal distributions occurs often in nature.  For example, it\n        describes the commonly occurring distribution of samples influenced\n        by a large number of tiny, random disturbances, each with its own\n        unique distribution [2]_.\n\n        Parameters\n        ----------\n        loc : float\n            Mean (\"centre\") of the distribution.\n        scale : float\n            Standard deviation (spread or \"width\") of the distribution.\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        See Also\n        --------\n        scipy.stats.distributions.norm : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Gaussian distribution is\n\n        .. math:: p(x) = \\frac{1}{\\sqrt{ 2 \\pi \\sigma^2 }}\n                         e^{ - \\frac{ (x - \\mu)^2 } {2 \\sigma^2} },\n\n        where :math:`\\mu` is the mean and :math:`\\sigma` the standard deviation.\n        The square of the standard deviation, :math:`\\sigma^2`, is called the\n        variance.\n\n        The function has its peak at the mean, and its \"spread\" increases with\n        the standard deviation (the function reaches 0.607 times its maximum at\n        :math:`x + \\sigma` and :math:`x - \\sigma` [2]_).  This implies that\n        `numpy.random.normal` is more likely to return samples lying close to the\n        mean, rather than those far away.\n""\n        References\n        ----------\n        .. [1] Wikipedia, \"Normal distribution\",\n               http://en.wikipedia.org/wiki/Normal_distribution\n        .. [2] P. R. Peebles Jr., \"Central Limit Theorem\" in \"Probability, Random\n               Variables and Random Signal Principles\", 4th ed., 2001,\n               pp. 51, 51, 125.\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> mu, sigma = 0, 0.1 # mean and standard deviation\n        >>> s = np.random.normal(mu, sigma, 1000)\n\n        Verify the mean and the variance:\n\n        >>> abs(mu - np.mean(s)) < 0.01\n        True\n\n        >>> abs(sigma - np.std(s, ddof=1)) < 0.01\n        True\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 30, normed=True)\n        >>> plt.plot(bins, 1/(sigma * np.sqrt(2 * np.pi)) *\n        ...                np.exp( - (bins - mu)**2 / (2 * sigma**2) ),\n        ...          linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     2888-static PyObject *__pyx_pf_6mtrand_11RandomState_17normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2889+/* Python wrapper */
     2890+static PyObject *__pyx_pw_6mtrand_11RandomState_35normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     2891+static char __pyx_doc_6mtrand_11RandomState_34normal[] = "\n        normal(loc=0.0, scale=1.0, size=None)\n\n        Draw random samples from a normal (Gaussian) distribution.\n\n        The probability density function of the normal distribution, first\n        derived by De Moivre and 200 years later by both Gauss and Laplace\n        independently [2]_, is often called the bell curve because of\n        its characteristic shape (see the example below).\n\n        The normal distributions occurs often in nature.  For example, it\n        describes the commonly occurring distribution of samples influenced\n        by a large number of tiny, random disturbances, each with its own\n        unique distribution [2]_.\n\n        Parameters\n        ----------\n        loc : float\n            Mean (\"centre\") of the distribution.\n        scale : float\n            Standard deviation (spread or \"width\") of the distribution.\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        See Also\n        --------\n        scipy.stats.distributions.norm : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Gaussian distribution is\n\n        .. math:: p(x) = \\frac{1}{\\sqrt{ 2 \\pi \\sigma^2 }}\n                         e^{ - \\frac{ (x - \\mu)^2 } {2 \\sigma^2} },\n\n        where :math:`\\mu` is the mean and :math:`\\sigma` the standard deviation.\n        The square of the standard deviation, :math:`\\sigma^2`, is called the\n        variance.\n\n        The function has its peak at the mean, and its \"spread\" increases with\n        the standard deviation (the function reaches 0.607 times its maximum at\n        :math:`x + \\sigma` and :math:`x - \\sigma` [2]_).  This implies that\n        `numpy.random.normal` is more likely to return samples lying close to the\n        mean, rather than those far away.\n""\n        References\n        ----------\n        .. [1] Wikipedia, \"Normal distribution\",\n               http://en.wikipedia.org/wiki/Normal_distribution\n        .. [2] P. R. Peebles Jr., \"Central Limit Theorem\" in \"Probability, Random\n               Variables and Random Signal Principles\", 4th ed., 2001,\n               pp. 51, 51, 125.\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> mu, sigma = 0, 0.1 # mean and standard deviation\n        >>> s = np.random.normal(mu, sigma, 1000)\n\n        Verify the mean and the variance:\n\n        >>> abs(mu - np.mean(s)) < 0.01\n        True\n\n        >>> abs(sigma - np.std(s, ddof=1)) < 0.01\n        True\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 30, normed=True)\n        >>> plt.plot(bins, 1/(sigma * np.sqrt(2 * np.pi)) *\n        ...                np.exp( - (bins - mu)**2 / (2 * sigma**2) ),\n        ...          linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     2892+static PyObject *__pyx_pw_6mtrand_11RandomState_35normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     2893   PyObject *__pyx_v_loc = 0;
     2894   PyObject *__pyx_v_scale = 0;
     2895   PyObject *__pyx_v_size = 0;
     2896-  PyArrayObject *__pyx_v_oloc = 0;
     2897-  PyArrayObject *__pyx_v_oscale = 0;
     2898-  double __pyx_v_floc;
     2899-  double __pyx_v_fscale;
     2900-  PyObject *__pyx_r = NULL;
     2901+  PyObject *__pyx_r = 0;
     2902   __Pyx_RefNannyDeclarations
     2903-  int __pyx_t_1;
     2904-  PyObject *__pyx_t_2 = NULL;
     2905-  PyObject *__pyx_t_3 = NULL;
     2906-  PyObject *__pyx_t_4 = NULL;
     2907-  PyObject *__pyx_t_5 = NULL;
     2908-  int __pyx_lineno = 0;
     2909-  const char *__pyx_filename = NULL;
     2910-  int __pyx_clineno = 0;
     2911-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__loc,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     2912-  __Pyx_RefNannySetupContext("normal");
     2913+  __Pyx_RefNannySetupContext("normal (wrapper)", 0);
     2914   {
     2915+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__loc,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     2916     PyObject* values[3] = {0,0,0};
     2917     values[0] = __pyx_k_17;
     2918     values[1] = __pyx_k_18;
     2919+
     2920+    /* "mtrand.pyx":1212
     2921+ *         return cont0_array(self.internal_state, rk_gauss, size)
     2922+ *
     2923+ *     def normal(self, loc=0.0, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     2924+ *         """
     2925+ *         normal(loc=0.0, scale=1.0, size=None)
     2926+ */
     2927     values[2] = ((PyObject *)Py_None);
     2928     if (unlikely(__pyx_kwds)) {
     2929       Py_ssize_t kw_args;
     2930-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     2931+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     2932+      switch (pos_args) {
     2933         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     2934         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     2935         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     2936@@ -7088,7 +7391,7 @@
     2937         default: goto __pyx_L5_argtuple_error;
     2938       }
     2939       kw_args = PyDict_Size(__pyx_kwds);
     2940-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     2941+      switch (pos_args) {
     2942         case  0:
     2943         if (kw_args > 0) {
     2944           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__loc);
     2945@@ -7106,7 +7409,7 @@
     2946         }
     2947       }
     2948       if (unlikely(kw_args > 0)) {
     2949-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "normal") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1212; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2950+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "normal") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1212; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     2951       }
     2952     } else {
     2953       switch (PyTuple_GET_SIZE(__pyx_args)) {
     2954@@ -7129,6 +7432,27 @@
     2955   __Pyx_RefNannyFinishContext();
     2956   return NULL;
     2957   __pyx_L4_argument_unpacking_done:;
     2958+  __pyx_r = __pyx_pf_6mtrand_11RandomState_34normal(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_loc, __pyx_v_scale, __pyx_v_size);
     2959+  __Pyx_RefNannyFinishContext();
     2960+  return __pyx_r;
     2961+}
     2962+
     2963+static PyObject *__pyx_pf_6mtrand_11RandomState_34normal(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_loc, PyObject *__pyx_v_scale, PyObject *__pyx_v_size) {
     2964+  PyArrayObject *__pyx_v_oloc = 0;
     2965+  PyArrayObject *__pyx_v_oscale = 0;
     2966+  double __pyx_v_floc;
     2967+  double __pyx_v_fscale;
     2968+  PyObject *__pyx_r = NULL;
     2969+  __Pyx_RefNannyDeclarations
     2970+  int __pyx_t_1;
     2971+  PyObject *__pyx_t_2 = NULL;
     2972+  PyObject *__pyx_t_3 = NULL;
     2973+  PyObject *__pyx_t_4 = NULL;
     2974+  PyObject *__pyx_t_5 = NULL;
     2975+  int __pyx_lineno = 0;
     2976+  const char *__pyx_filename = NULL;
     2977+  int __pyx_clineno = 0;
     2978+  __Pyx_RefNannySetupContext("normal", 0);
     2979 
     2980   /* "mtrand.pyx":1297
     2981  *         cdef double floc, fscale
     2982@@ -7180,9 +7504,9 @@
     2983       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     2984       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     2985       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1301; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2986-      goto __pyx_L7;
     2987+      goto __pyx_L4;
     2988     }
     2989-    __pyx_L7:;
     2990+    __pyx_L4:;
     2991 
     2992     /* "mtrand.pyx":1302
     2993  *             if fscale <= 0:
     2994@@ -7192,14 +7516,14 @@
     2995  *         PyErr_Clear()
     2996  */
     2997     __Pyx_XDECREF(__pyx_r);
     2998-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_normal, __pyx_v_size, __pyx_v_floc, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1302; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     2999+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_normal, __pyx_v_size, __pyx_v_floc, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1302; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3000     __Pyx_GOTREF(__pyx_t_2);
     3001     __pyx_r = __pyx_t_2;
     3002     __pyx_t_2 = 0;
     3003     goto __pyx_L0;
     3004-    goto __pyx_L6;
     3005+    goto __pyx_L3;
     3006   }
     3007-  __pyx_L6:;
     3008+  __pyx_L3:;
     3009 
     3010   /* "mtrand.pyx":1304
     3011  *             return cont2_array_sc(self.internal_state, rk_normal, size, floc, fscale)
     3012@@ -7254,7 +7578,7 @@
     3013   __Pyx_GOTREF(__pyx_t_4);
     3014   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3015   __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1308; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3016-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     3017+  __Pyx_GOTREF(__pyx_t_2);
     3018   __Pyx_INCREF(((PyObject *)__pyx_v_oscale));
     3019   PyTuple_SET_ITEM(__pyx_t_2, 0, ((PyObject *)__pyx_v_oscale));
     3020   __Pyx_GIVEREF(((PyObject *)__pyx_v_oscale));
     3021@@ -7266,7 +7590,7 @@
     3022   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     3023   __Pyx_DECREF(((PyObject *)__pyx_t_2)); __pyx_t_2 = 0;
     3024   __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1308; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3025-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     3026+  __Pyx_GOTREF(__pyx_t_2);
     3027   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_5);
     3028   __Pyx_GIVEREF(__pyx_t_5);
     3029   __pyx_t_5 = 0;
     3030@@ -7290,9 +7614,9 @@
     3031     __Pyx_Raise(__pyx_t_5, 0, 0, 0);
     3032     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     3033     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1309; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3034-    goto __pyx_L8;
     3035+    goto __pyx_L5;
     3036   }
     3037-  __pyx_L8:;
     3038+  __pyx_L5:;
     3039 
     3040   /* "mtrand.pyx":1310
     3041  *         if np.any(np.less_equal(oscale, 0)):
     3042@@ -7302,7 +7626,7 @@
     3043  *     def beta(self, a, b, size=None):
     3044  */
     3045   __Pyx_XDECREF(__pyx_r);
     3046-  __pyx_t_5 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_normal, __pyx_v_size, __pyx_v_oloc, __pyx_v_oscale); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1310; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3047+  __pyx_t_5 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_normal, __pyx_v_size, __pyx_v_oloc, __pyx_v_oscale); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1310; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3048   __Pyx_GOTREF(__pyx_t_5);
     3049   __pyx_r = __pyx_t_5;
     3050   __pyx_t_5 = 0;
     3051@@ -7325,42 +7649,32 @@
     3052   return __pyx_r;
     3053 }
     3054 
     3055-/* "mtrand.pyx":1312
     3056- *         return cont2_array(self.internal_state, rk_normal, size, oloc, oscale)
     3057- *
     3058- *     def beta(self, a, b, size=None):             # <<<<<<<<<<<<<<
     3059- *         """
     3060- *         beta(a, b, size=None)
     3061- */
     3062-
     3063-static PyObject *__pyx_pf_6mtrand_11RandomState_18beta(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3064-static char __pyx_doc_6mtrand_11RandomState_18beta[] = "\n        beta(a, b, size=None)\n\n        The Beta distribution over ``[0, 1]``.\n\n        The Beta distribution is a special case of the Dirichlet distribution,\n        and is related to the Gamma distribution.  It has the probability\n        distribution function\n\n        .. math:: f(x; a,b) = \\frac{1}{B(\\alpha, \\beta)} x^{\\alpha - 1}\n                                                         (1 - x)^{\\beta - 1},\n\n        where the normalisation, B, is the beta function,\n\n        .. math:: B(\\alpha, \\beta) = \\int_0^1 t^{\\alpha - 1}\n                                     (1 - t)^{\\beta - 1} dt.\n\n        It is often seen in Bayesian inference and order statistics.\n\n        Parameters\n        ----------\n        a : float\n            Alpha, non-negative.\n        b : float\n            Beta, non-negative.\n        size : tuple of ints, optional\n            The number of samples to draw.  The ouput is packed according to\n            the size given.\n\n        Returns\n        -------\n        out : ndarray\n            Array of the given shape, containing values drawn from a\n            Beta distribution.\n\n        ";
     3065-static PyObject *__pyx_pf_6mtrand_11RandomState_18beta(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3066+/* Python wrapper */
     3067+static PyObject *__pyx_pw_6mtrand_11RandomState_37beta(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3068+static char __pyx_doc_6mtrand_11RandomState_36beta[] = "\n        beta(a, b, size=None)\n\n        The Beta distribution over ``[0, 1]``.\n\n        The Beta distribution is a special case of the Dirichlet distribution,\n        and is related to the Gamma distribution.  It has the probability\n        distribution function\n\n        .. math:: f(x; a,b) = \\frac{1}{B(\\alpha, \\beta)} x^{\\alpha - 1}\n                                                         (1 - x)^{\\beta - 1},\n\n        where the normalisation, B, is the beta function,\n\n        .. math:: B(\\alpha, \\beta) = \\int_0^1 t^{\\alpha - 1}\n                                     (1 - t)^{\\beta - 1} dt.\n\n        It is often seen in Bayesian inference and order statistics.\n\n        Parameters\n        ----------\n        a : float\n            Alpha, non-negative.\n        b : float\n            Beta, non-negative.\n        size : tuple of ints, optional\n            The number of samples to draw.  The ouput is packed according to\n            the size given.\n\n        Returns\n        -------\n        out : ndarray\n            Array of the given shape, containing values drawn from a\n            Beta distribution.\n\n        ";
     3069+static PyObject *__pyx_pw_6mtrand_11RandomState_37beta(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3070   PyObject *__pyx_v_a = 0;
     3071   PyObject *__pyx_v_b = 0;
     3072   PyObject *__pyx_v_size = 0;
     3073-  PyArrayObject *__pyx_v_oa = 0;
     3074-  PyArrayObject *__pyx_v_ob = 0;
     3075-  double __pyx_v_fa;
     3076-  double __pyx_v_fb;
     3077-  PyObject *__pyx_r = NULL;
     3078+  PyObject *__pyx_r = 0;
     3079   __Pyx_RefNannyDeclarations
     3080-  int __pyx_t_1;
     3081-  PyObject *__pyx_t_2 = NULL;
     3082-  PyObject *__pyx_t_3 = NULL;
     3083-  PyObject *__pyx_t_4 = NULL;
     3084-  PyObject *__pyx_t_5 = NULL;
     3085-  int __pyx_lineno = 0;
     3086-  const char *__pyx_filename = NULL;
     3087-  int __pyx_clineno = 0;
     3088-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__a,&__pyx_n_s__b,&__pyx_n_s__size,0};
     3089-  __Pyx_RefNannySetupContext("beta");
     3090+  __Pyx_RefNannySetupContext("beta (wrapper)", 0);
     3091   {
     3092+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__a,&__pyx_n_s__b,&__pyx_n_s__size,0};
     3093     PyObject* values[3] = {0,0,0};
     3094+
     3095+    /* "mtrand.pyx":1312
     3096+ *         return cont2_array(self.internal_state, rk_normal, size, oloc, oscale)
     3097+ *
     3098+ *     def beta(self, a, b, size=None):             # <<<<<<<<<<<<<<
     3099+ *         """
     3100+ *         beta(a, b, size=None)
     3101+ */
     3102     values[2] = ((PyObject *)Py_None);
     3103     if (unlikely(__pyx_kwds)) {
     3104       Py_ssize_t kw_args;
     3105-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     3106+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     3107+      switch (pos_args) {
     3108         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     3109         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     3110         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     3111@@ -7368,14 +7682,12 @@
     3112         default: goto __pyx_L5_argtuple_error;
     3113       }
     3114       kw_args = PyDict_Size(__pyx_kwds);
     3115-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     3116+      switch (pos_args) {
     3117         case  0:
     3118-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__a);
     3119-        if (likely(values[0])) kw_args--;
     3120+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__a)) != 0)) kw_args--;
     3121         else goto __pyx_L5_argtuple_error;
     3122         case  1:
     3123-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__b);
     3124-        if (likely(values[1])) kw_args--;
     3125+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__b)) != 0)) kw_args--;
     3126         else {
     3127           __Pyx_RaiseArgtupleInvalid("beta", 0, 2, 3, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1312; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     3128         }
     3129@@ -7386,7 +7698,7 @@
     3130         }
     3131       }
     3132       if (unlikely(kw_args > 0)) {
     3133-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "beta") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1312; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     3134+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "beta") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1312; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     3135       }
     3136     } else {
     3137       switch (PyTuple_GET_SIZE(__pyx_args)) {
     3138@@ -7409,6 +7721,27 @@
     3139   __Pyx_RefNannyFinishContext();
     3140   return NULL;
     3141   __pyx_L4_argument_unpacking_done:;
     3142+  __pyx_r = __pyx_pf_6mtrand_11RandomState_36beta(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_a, __pyx_v_b, __pyx_v_size);
     3143+  __Pyx_RefNannyFinishContext();
     3144+  return __pyx_r;
     3145+}
     3146+
     3147+static PyObject *__pyx_pf_6mtrand_11RandomState_36beta(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_a, PyObject *__pyx_v_b, PyObject *__pyx_v_size) {
     3148+  PyArrayObject *__pyx_v_oa = 0;
     3149+  PyArrayObject *__pyx_v_ob = 0;
     3150+  double __pyx_v_fa;
     3151+  double __pyx_v_fb;
     3152+  PyObject *__pyx_r = NULL;
     3153+  __Pyx_RefNannyDeclarations
     3154+  int __pyx_t_1;
     3155+  PyObject *__pyx_t_2 = NULL;
     3156+  PyObject *__pyx_t_3 = NULL;
     3157+  PyObject *__pyx_t_4 = NULL;
     3158+  PyObject *__pyx_t_5 = NULL;
     3159+  int __pyx_lineno = 0;
     3160+  const char *__pyx_filename = NULL;
     3161+  int __pyx_clineno = 0;
     3162+  __Pyx_RefNannySetupContext("beta", 0);
     3163 
     3164   /* "mtrand.pyx":1352
     3165  *         cdef double fa, fb
     3166@@ -7460,9 +7793,9 @@
     3167       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     3168       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3169       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1356; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3170-      goto __pyx_L7;
     3171+      goto __pyx_L4;
     3172     }
     3173-    __pyx_L7:;
     3174+    __pyx_L4:;
     3175 
     3176     /* "mtrand.pyx":1357
     3177  *             if fa <= 0:
     3178@@ -7486,9 +7819,9 @@
     3179       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     3180       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3181       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1358; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3182-      goto __pyx_L8;
     3183+      goto __pyx_L5;
     3184     }
     3185-    __pyx_L8:;
     3186+    __pyx_L5:;
     3187 
     3188     /* "mtrand.pyx":1359
     3189  *             if fb <= 0:
     3190@@ -7498,14 +7831,14 @@
     3191  *         PyErr_Clear()
     3192  */
     3193     __Pyx_XDECREF(__pyx_r);
     3194-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_beta, __pyx_v_size, __pyx_v_fa, __pyx_v_fb); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1359; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3195+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_beta, __pyx_v_size, __pyx_v_fa, __pyx_v_fb); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1359; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3196     __Pyx_GOTREF(__pyx_t_2);
     3197     __pyx_r = __pyx_t_2;
     3198     __pyx_t_2 = 0;
     3199     goto __pyx_L0;
     3200-    goto __pyx_L6;
     3201+    goto __pyx_L3;
     3202   }
     3203-  __pyx_L6:;
     3204+  __pyx_L3:;
     3205 
     3206   /* "mtrand.pyx":1361
     3207  *             return cont2_array_sc(self.internal_state, rk_beta, size, fa, fb)
     3208@@ -7560,7 +7893,7 @@
     3209   __Pyx_GOTREF(__pyx_t_4);
     3210   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3211   __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1365; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3212-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     3213+  __Pyx_GOTREF(__pyx_t_2);
     3214   __Pyx_INCREF(((PyObject *)__pyx_v_oa));
     3215   PyTuple_SET_ITEM(__pyx_t_2, 0, ((PyObject *)__pyx_v_oa));
     3216   __Pyx_GIVEREF(((PyObject *)__pyx_v_oa));
     3217@@ -7572,7 +7905,7 @@
     3218   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     3219   __Pyx_DECREF(((PyObject *)__pyx_t_2)); __pyx_t_2 = 0;
     3220   __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1365; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3221-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     3222+  __Pyx_GOTREF(__pyx_t_2);
     3223   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_5);
     3224   __Pyx_GIVEREF(__pyx_t_5);
     3225   __pyx_t_5 = 0;
     3226@@ -7596,9 +7929,9 @@
     3227     __Pyx_Raise(__pyx_t_5, 0, 0, 0);
     3228     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     3229     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1366; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3230-    goto __pyx_L9;
     3231+    goto __pyx_L6;
     3232   }
     3233-  __pyx_L9:;
     3234+  __pyx_L6:;
     3235 
     3236   /* "mtrand.pyx":1367
     3237  *         if np.any(np.less_equal(oa, 0)):
     3238@@ -7618,7 +7951,7 @@
     3239   __Pyx_GOTREF(__pyx_t_3);
     3240   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     3241   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1367; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3242-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     3243+  __Pyx_GOTREF(__pyx_t_5);
     3244   __Pyx_INCREF(((PyObject *)__pyx_v_ob));
     3245   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_ob));
     3246   __Pyx_GIVEREF(((PyObject *)__pyx_v_ob));
     3247@@ -7630,7 +7963,7 @@
     3248   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     3249   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     3250   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1367; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3251-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     3252+  __Pyx_GOTREF(__pyx_t_5);
     3253   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_4);
     3254   __Pyx_GIVEREF(__pyx_t_4);
     3255   __pyx_t_4 = 0;
     3256@@ -7654,9 +7987,9 @@
     3257     __Pyx_Raise(__pyx_t_4, 0, 0, 0);
     3258     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     3259     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1368; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3260-    goto __pyx_L10;
     3261+    goto __pyx_L7;
     3262   }
     3263-  __pyx_L10:;
     3264+  __pyx_L7:;
     3265 
     3266   /* "mtrand.pyx":1369
     3267  *         if np.any(np.less_equal(ob, 0)):
     3268@@ -7666,7 +7999,7 @@
     3269  *     def exponential(self, scale=1.0, size=None):
     3270  */
     3271   __Pyx_XDECREF(__pyx_r);
     3272-  __pyx_t_4 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_beta, __pyx_v_size, __pyx_v_oa, __pyx_v_ob); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1369; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3273+  __pyx_t_4 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_beta, __pyx_v_size, __pyx_v_oa, __pyx_v_ob); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1369; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3274   __Pyx_GOTREF(__pyx_t_4);
     3275   __pyx_r = __pyx_t_4;
     3276   __pyx_t_4 = 0;
     3277@@ -7689,47 +8022,39 @@
     3278   return __pyx_r;
     3279 }
     3280 
     3281-/* "mtrand.pyx":1371
     3282- *         return cont2_array(self.internal_state, rk_beta, size, oa, ob)
     3283- *
     3284- *     def exponential(self, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     3285- *         """
     3286- *         exponential(scale=1.0, size=None)
     3287- */
     3288-
     3289-static PyObject *__pyx_pf_6mtrand_11RandomState_19exponential(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3290-static char __pyx_doc_6mtrand_11RandomState_19exponential[] = "\n        exponential(scale=1.0, size=None)\n\n        Exponential distribution.\n\n        Its probability density function is\n\n        .. math:: f(x; \\frac{1}{\\beta}) = \\frac{1}{\\beta} \\exp(-\\frac{x}{\\beta}),\n\n        for ``x > 0`` and 0 elsewhere. :math:`\\beta` is the scale parameter,\n        which is the inverse of the rate parameter :math:`\\lambda = 1/\\beta`.\n        The rate parameter is an alternative, widely used parameterization\n        of the exponential distribution [3]_.\n\n        The exponential distribution is a continuous analogue of the\n        geometric distribution.  It describes many common situations, such as\n        the size of raindrops measured over many rainstorms [1]_, or the time\n        between page requests to Wikipedia [2]_.\n\n        Parameters\n        ----------\n        scale : float\n            The scale parameter, :math:`\\beta = 1/\\lambda`.\n        size : tuple of ints\n            Number of samples to draw.  The output is shaped\n            according to `size`.\n\n        References\n        ----------\n        .. [1] Peyton Z. Peebles Jr., \"Probability, Random Variables and\n               Random Signal Principles\", 4th ed, 2001, p. 57.\n        .. [2] \"Poisson Process\", Wikipedia,\n               http://en.wikipedia.org/wiki/Poisson_process\n        .. [3] \"Exponential Distribution, Wikipedia,\n               http://en.wikipedia.org/wiki/Exponential_distribution\n\n        ";
     3291-static PyObject *__pyx_pf_6mtrand_11RandomState_19exponential(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3292+/* Python wrapper */
     3293+static PyObject *__pyx_pw_6mtrand_11RandomState_39exponential(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3294+static char __pyx_doc_6mtrand_11RandomState_38exponential[] = "\n        exponential(scale=1.0, size=None)\n\n        Exponential distribution.\n\n        Its probability density function is\n\n        .. math:: f(x; \\frac{1}{\\beta}) = \\frac{1}{\\beta} \\exp(-\\frac{x}{\\beta}),\n\n        for ``x > 0`` and 0 elsewhere. :math:`\\beta` is the scale parameter,\n        which is the inverse of the rate parameter :math:`\\lambda = 1/\\beta`.\n        The rate parameter is an alternative, widely used parameterization\n        of the exponential distribution [3]_.\n\n        The exponential distribution is a continuous analogue of the\n        geometric distribution.  It describes many common situations, such as\n        the size of raindrops measured over many rainstorms [1]_, or the time\n        between page requests to Wikipedia [2]_.\n\n        Parameters\n        ----------\n        scale : float\n            The scale parameter, :math:`\\beta = 1/\\lambda`.\n        size : tuple of ints\n            Number of samples to draw.  The output is shaped\n            according to `size`.\n\n        References\n        ----------\n        .. [1] Peyton Z. Peebles Jr., \"Probability, Random Variables and\n               Random Signal Principles\", 4th ed, 2001, p. 57.\n        .. [2] \"Poisson Process\", Wikipedia,\n               http://en.wikipedia.org/wiki/Poisson_process\n        .. [3] \"Exponential Distribution, Wikipedia,\n               http://en.wikipedia.org/wiki/Exponential_distribution\n\n        ";
     3295+static PyObject *__pyx_pw_6mtrand_11RandomState_39exponential(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3296   PyObject *__pyx_v_scale = 0;
     3297   PyObject *__pyx_v_size = 0;
     3298-  PyArrayObject *__pyx_v_oscale = 0;
     3299-  double __pyx_v_fscale;
     3300-  PyObject *__pyx_r = NULL;
     3301+  PyObject *__pyx_r = 0;
     3302   __Pyx_RefNannyDeclarations
     3303-  int __pyx_t_1;
     3304-  PyObject *__pyx_t_2 = NULL;
     3305-  PyObject *__pyx_t_3 = NULL;
     3306-  PyObject *__pyx_t_4 = NULL;
     3307-  PyObject *__pyx_t_5 = NULL;
     3308-  int __pyx_lineno = 0;
     3309-  const char *__pyx_filename = NULL;
     3310-  int __pyx_clineno = 0;
     3311-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__scale,&__pyx_n_s__size,0};
     3312-  __Pyx_RefNannySetupContext("exponential");
     3313+  __Pyx_RefNannySetupContext("exponential (wrapper)", 0);
     3314   {
     3315+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__scale,&__pyx_n_s__size,0};
     3316     PyObject* values[2] = {0,0};
     3317     values[0] = __pyx_k_28;
     3318+
     3319+    /* "mtrand.pyx":1371
     3320+ *         return cont2_array(self.internal_state, rk_beta, size, oa, ob)
     3321+ *
     3322+ *     def exponential(self, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     3323+ *         """
     3324+ *         exponential(scale=1.0, size=None)
     3325+ */
     3326     values[1] = ((PyObject *)Py_None);
     3327     if (unlikely(__pyx_kwds)) {
     3328       Py_ssize_t kw_args;
     3329-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     3330+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     3331+      switch (pos_args) {
     3332         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     3333         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     3334         case  0: break;
     3335         default: goto __pyx_L5_argtuple_error;
     3336       }
     3337       kw_args = PyDict_Size(__pyx_kwds);
     3338-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     3339+      switch (pos_args) {
     3340         case  0:
     3341         if (kw_args > 0) {
     3342           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__scale);
     3343@@ -7742,7 +8067,7 @@
     3344         }
     3345       }
     3346       if (unlikely(kw_args > 0)) {
     3347-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "exponential") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1371; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     3348+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "exponential") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1371; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     3349       }
     3350     } else {
     3351       switch (PyTuple_GET_SIZE(__pyx_args)) {
     3352@@ -7763,7 +8088,26 @@
     3353   __Pyx_RefNannyFinishContext();
     3354   return NULL;
     3355   __pyx_L4_argument_unpacking_done:;
     3356-
     3357+  __pyx_r = __pyx_pf_6mtrand_11RandomState_38exponential(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_scale, __pyx_v_size);
     3358+  __Pyx_RefNannyFinishContext();
     3359+  return __pyx_r;
     3360+}
     3361+
     3362+static PyObject *__pyx_pf_6mtrand_11RandomState_38exponential(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_scale, PyObject *__pyx_v_size) {
     3363+  PyArrayObject *__pyx_v_oscale = 0;
     3364+  double __pyx_v_fscale;
     3365+  PyObject *__pyx_r = NULL;
     3366+  __Pyx_RefNannyDeclarations
     3367+  int __pyx_t_1;
     3368+  PyObject *__pyx_t_2 = NULL;
     3369+  PyObject *__pyx_t_3 = NULL;
     3370+  PyObject *__pyx_t_4 = NULL;
     3371+  PyObject *__pyx_t_5 = NULL;
     3372+  int __pyx_lineno = 0;
     3373+  const char *__pyx_filename = NULL;
     3374+  int __pyx_clineno = 0;
     3375+  __Pyx_RefNannySetupContext("exponential", 0);
     3376+
     3377   /* "mtrand.pyx":1412
     3378  *         cdef double fscale
     3379  *
     3380@@ -7805,9 +8149,9 @@
     3381       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     3382       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3383       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1415; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3384-      goto __pyx_L7;
     3385+      goto __pyx_L4;
     3386     }
     3387-    __pyx_L7:;
     3388+    __pyx_L4:;
     3389 
     3390     /* "mtrand.pyx":1416
     3391  *             if fscale <= 0:
     3392@@ -7817,14 +8161,14 @@
     3393  *         PyErr_Clear()
     3394  */
     3395     __Pyx_XDECREF(__pyx_r);
     3396-    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_exponential, __pyx_v_size, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1416; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3397+    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(__pyx_v_self->internal_state, rk_exponential, __pyx_v_size, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1416; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3398     __Pyx_GOTREF(__pyx_t_2);
     3399     __pyx_r = __pyx_t_2;
     3400     __pyx_t_2 = 0;
     3401     goto __pyx_L0;
     3402-    goto __pyx_L6;
     3403+    goto __pyx_L3;
     3404   }
     3405-  __pyx_L6:;
     3406+  __pyx_L3:;
     3407 
     3408   /* "mtrand.pyx":1418
     3409  *             return cont1_array_sc(self.internal_state, rk_exponential, size, fscale)
     3410@@ -7868,7 +8212,7 @@
     3411   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1421; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3412   __Pyx_GOTREF(__pyx_t_2);
     3413   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1421; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3414-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     3415+  __Pyx_GOTREF(__pyx_t_5);
     3416   __Pyx_INCREF(((PyObject *)__pyx_v_oscale));
     3417   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_oscale));
     3418   __Pyx_GIVEREF(((PyObject *)__pyx_v_oscale));
     3419@@ -7880,7 +8224,7 @@
     3420   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     3421   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     3422   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1421; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3423-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     3424+  __Pyx_GOTREF(__pyx_t_5);
     3425   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     3426   __Pyx_GIVEREF(__pyx_t_2);
     3427   __pyx_t_2 = 0;
     3428@@ -7904,9 +8248,9 @@
     3429     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     3430     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3431     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1422; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3432-    goto __pyx_L8;
     3433+    goto __pyx_L5;
     3434   }
     3435-  __pyx_L8:;
     3436+  __pyx_L5:;
     3437 
     3438   /* "mtrand.pyx":1423
     3439  *         if np.any(np.less_equal(oscale, 0.0)):
     3440@@ -7916,7 +8260,7 @@
     3441  *     def standard_exponential(self, size=None):
     3442  */
     3443   __Pyx_XDECREF(__pyx_r);
     3444-  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_exponential, __pyx_v_size, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1423; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3445+  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(__pyx_v_self->internal_state, rk_exponential, __pyx_v_size, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1423; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3446   __Pyx_GOTREF(__pyx_t_2);
     3447   __pyx_r = __pyx_t_2;
     3448   __pyx_t_2 = 0;
     3449@@ -7938,38 +8282,36 @@
     3450   return __pyx_r;
     3451 }
     3452 
     3453-/* "mtrand.pyx":1425
     3454+/* Python wrapper */
     3455+static PyObject *__pyx_pw_6mtrand_11RandomState_41standard_exponential(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3456+static char __pyx_doc_6mtrand_11RandomState_40standard_exponential[] = "\n        standard_exponential(size=None)\n\n        Draw samples from the standard exponential distribution.\n\n        `standard_exponential` is identical to the exponential distribution\n        with a scale parameter of 1.\n\n        Parameters\n        ----------\n        size : int or tuple of ints\n            Shape of the output.\n\n        Returns\n        -------\n        out : float or ndarray\n            Drawn samples.\n\n        Examples\n        --------\n        Output a 3x8000 array:\n\n        >>> n = np.random.standard_exponential((3, 8000))\n\n        ";
     3457+static PyObject *__pyx_pw_6mtrand_11RandomState_41standard_exponential(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3458+  PyObject *__pyx_v_size = 0;
     3459+  PyObject *__pyx_r = 0;
     3460+  __Pyx_RefNannyDeclarations
     3461+  __Pyx_RefNannySetupContext("standard_exponential (wrapper)", 0);
     3462+  {
     3463+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__size,0};
     3464+    PyObject* values[1] = {0};
     3465+
     3466+    /* "mtrand.pyx":1425
     3467  *         return cont1_array(self.internal_state, rk_exponential, size, oscale)
     3468  *
     3469  *     def standard_exponential(self, size=None):             # <<<<<<<<<<<<<<
     3470  *         """
     3471  *         standard_exponential(size=None)
     3472  */
     3473-
     3474-static PyObject *__pyx_pf_6mtrand_11RandomState_20standard_exponential(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3475-static char __pyx_doc_6mtrand_11RandomState_20standard_exponential[] = "\n        standard_exponential(size=None)\n\n        Draw samples from the standard exponential distribution.\n\n        `standard_exponential` is identical to the exponential distribution\n        with a scale parameter of 1.\n\n        Parameters\n        ----------\n        size : int or tuple of ints\n            Shape of the output.\n\n        Returns\n        -------\n        out : float or ndarray\n            Drawn samples.\n\n        Examples\n        --------\n        Output a 3x8000 array:\n\n        >>> n = np.random.standard_exponential((3, 8000))\n\n        ";
     3476-static PyObject *__pyx_pf_6mtrand_11RandomState_20standard_exponential(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3477-  PyObject *__pyx_v_size = 0;
     3478-  PyObject *__pyx_r = NULL;
     3479-  __Pyx_RefNannyDeclarations
     3480-  PyObject *__pyx_t_1 = NULL;
     3481-  int __pyx_lineno = 0;
     3482-  const char *__pyx_filename = NULL;
     3483-  int __pyx_clineno = 0;
     3484-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__size,0};
     3485-  __Pyx_RefNannySetupContext("standard_exponential");
     3486-  {
     3487-    PyObject* values[1] = {0};
     3488     values[0] = ((PyObject *)Py_None);
     3489     if (unlikely(__pyx_kwds)) {
     3490       Py_ssize_t kw_args;
     3491-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     3492+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     3493+      switch (pos_args) {
     3494         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     3495         case  0: break;
     3496         default: goto __pyx_L5_argtuple_error;
     3497       }
     3498       kw_args = PyDict_Size(__pyx_kwds);
     3499-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     3500+      switch (pos_args) {
     3501         case  0:
     3502         if (kw_args > 0) {
     3503           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__size);
     3504@@ -7977,7 +8319,7 @@
     3505         }
     3506       }
     3507       if (unlikely(kw_args > 0)) {
     3508-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "standard_exponential") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1425; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     3509+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "standard_exponential") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1425; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     3510       }
     3511     } else {
     3512       switch (PyTuple_GET_SIZE(__pyx_args)) {
     3513@@ -7996,6 +8338,19 @@
     3514   __Pyx_RefNannyFinishContext();
     3515   return NULL;
     3516   __pyx_L4_argument_unpacking_done:;
     3517+  __pyx_r = __pyx_pf_6mtrand_11RandomState_40standard_exponential(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_size);
     3518+  __Pyx_RefNannyFinishContext();
     3519+  return __pyx_r;
     3520+}
     3521+
     3522+static PyObject *__pyx_pf_6mtrand_11RandomState_40standard_exponential(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_size) {
     3523+  PyObject *__pyx_r = NULL;
     3524+  __Pyx_RefNannyDeclarations
     3525+  PyObject *__pyx_t_1 = NULL;
     3526+  int __pyx_lineno = 0;
     3527+  const char *__pyx_filename = NULL;
     3528+  int __pyx_clineno = 0;
     3529+  __Pyx_RefNannySetupContext("standard_exponential", 0);
     3530 
     3531   /* "mtrand.pyx":1451
     3532  *
     3533@@ -8005,7 +8360,7 @@
     3534  *     def standard_gamma(self, shape, size=None):
     3535  */
     3536   __Pyx_XDECREF(__pyx_r);
     3537-  __pyx_t_1 = __pyx_f_6mtrand_cont0_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_standard_exponential, __pyx_v_size); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1451; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3538+  __pyx_t_1 = __pyx_f_6mtrand_cont0_array(__pyx_v_self->internal_state, rk_standard_exponential, __pyx_v_size); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1451; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3539   __Pyx_GOTREF(__pyx_t_1);
     3540   __pyx_r = __pyx_t_1;
     3541   __pyx_t_1 = 0;
     3542@@ -8023,49 +8378,40 @@
     3543   return __pyx_r;
     3544 }
     3545 
     3546-/* "mtrand.pyx":1453
     3547+/* Python wrapper */
     3548+static PyObject *__pyx_pw_6mtrand_11RandomState_43standard_gamma(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3549+static char __pyx_doc_6mtrand_11RandomState_42standard_gamma[] = "\n        standard_gamma(shape, size=None)\n\n        Draw samples from a Standard Gamma distribution.\n\n        Samples are drawn from a Gamma distribution with specified parameters,\n        shape (sometimes designated \"k\") and scale=1.\n\n        Parameters\n        ----------\n        shape : float\n            Parameter, should be > 0.\n        size : int or tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : ndarray or scalar\n            The drawn samples.\n\n        See Also\n        --------\n        scipy.stats.distributions.gamma : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Gamma distribution is\n\n        .. math:: p(x) = x^{k-1}\\frac{e^{-x/\\theta}}{\\theta^k\\Gamma(k)},\n\n        where :math:`k` is the shape and :math:`\\theta` the scale,\n        and :math:`\\Gamma` is the Gamma function.\n\n        The Gamma distribution is often used to model the times to failure of\n        electronic components, and arises naturally in processes for which the\n        waiting times between Poisson distributed events are relevant.\n\n        References\n        ----------\n        .. [1] Weisstein, Eric W. \"Gamma Distribution.\" From MathWorld--A\n               Wolfram Web Resource.\n               http://mathworld.wolfram.com/GammaDistribution.html\n        .. [2] Wikipedia, \"Gamma-distribution\",\n               http://en.wikipedia.org/wiki/Gamma-distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> shape, scale = 2., 1. # mean and width\n        >>> s = np.random.standard_gamma(shape, 1000000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt""\n        >>> import scipy.special as sps\n        >>> count, bins, ignored = plt.hist(s, 50, normed=True)\n        >>> y = bins**(shape-1) * ((np.exp(-bins/scale))/ \\\n        ...                       (sps.gamma(shape) * scale**shape))\n        >>> plt.plot(bins, y, linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     3550+static PyObject *__pyx_pw_6mtrand_11RandomState_43standard_gamma(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3551+  PyObject *__pyx_v_shape = 0;
     3552+  PyObject *__pyx_v_size = 0;
     3553+  PyObject *__pyx_r = 0;
     3554+  __Pyx_RefNannyDeclarations
     3555+  __Pyx_RefNannySetupContext("standard_gamma (wrapper)", 0);
     3556+  {
     3557+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__shape,&__pyx_n_s__size,0};
     3558+    PyObject* values[2] = {0,0};
     3559+
     3560+    /* "mtrand.pyx":1453
     3561  *         return cont0_array(self.internal_state, rk_standard_exponential, size)
     3562  *
     3563  *     def standard_gamma(self, shape, size=None):             # <<<<<<<<<<<<<<
     3564  *         """
     3565  *         standard_gamma(shape, size=None)
     3566  */
     3567-
     3568-static PyObject *__pyx_pf_6mtrand_11RandomState_21standard_gamma(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3569-static char __pyx_doc_6mtrand_11RandomState_21standard_gamma[] = "\n        standard_gamma(shape, size=None)\n\n        Draw samples from a Standard Gamma distribution.\n\n        Samples are drawn from a Gamma distribution with specified parameters,\n        shape (sometimes designated \"k\") and scale=1.\n\n        Parameters\n        ----------\n        shape : float\n            Parameter, should be > 0.\n        size : int or tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : ndarray or scalar\n            The drawn samples.\n\n        See Also\n        --------\n        scipy.stats.distributions.gamma : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Gamma distribution is\n\n        .. math:: p(x) = x^{k-1}\\frac{e^{-x/\\theta}}{\\theta^k\\Gamma(k)},\n\n        where :math:`k` is the shape and :math:`\\theta` the scale,\n        and :math:`\\Gamma` is the Gamma function.\n\n        The Gamma distribution is often used to model the times to failure of\n        electronic components, and arises naturally in processes for which the\n        waiting times between Poisson distributed events are relevant.\n\n        References\n        ----------\n        .. [1] Weisstein, Eric W. \"Gamma Distribution.\" From MathWorld--A\n               Wolfram Web Resource.\n               http://mathworld.wolfram.com/GammaDistribution.html\n        .. [2] Wikipedia, \"Gamma-distribution\",\n               http://en.wikipedia.org/wiki/Gamma-distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> shape, scale = 2., 1. # mean and width\n        >>> s = np.random.standard_gamma(shape, 1000000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt""\n        >>> import scipy.special as sps\n        >>> count, bins, ignored = plt.hist(s, 50, normed=True)\n        >>> y = bins**(shape-1) * ((np.exp(-bins/scale))/ \\\n        ...                       (sps.gamma(shape) * scale**shape))\n        >>> plt.plot(bins, y, linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     3570-static PyObject *__pyx_pf_6mtrand_11RandomState_21standard_gamma(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3571-  PyObject *__pyx_v_shape = 0;
     3572-  PyObject *__pyx_v_size = 0;
     3573-  PyArrayObject *__pyx_v_oshape = 0;
     3574-  double __pyx_v_fshape;
     3575-  PyObject *__pyx_r = NULL;
     3576-  __Pyx_RefNannyDeclarations
     3577-  int __pyx_t_1;
     3578-  PyObject *__pyx_t_2 = NULL;
     3579-  PyObject *__pyx_t_3 = NULL;
     3580-  PyObject *__pyx_t_4 = NULL;
     3581-  PyObject *__pyx_t_5 = NULL;
     3582-  int __pyx_lineno = 0;
     3583-  const char *__pyx_filename = NULL;
     3584-  int __pyx_clineno = 0;
     3585-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__shape,&__pyx_n_s__size,0};
     3586-  __Pyx_RefNannySetupContext("standard_gamma");
     3587-  {
     3588-    PyObject* values[2] = {0,0};
     3589     values[1] = ((PyObject *)Py_None);
     3590     if (unlikely(__pyx_kwds)) {
     3591       Py_ssize_t kw_args;
     3592-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     3593+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     3594+      switch (pos_args) {
     3595         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     3596         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     3597         case  0: break;
     3598         default: goto __pyx_L5_argtuple_error;
     3599       }
     3600       kw_args = PyDict_Size(__pyx_kwds);
     3601-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     3602+      switch (pos_args) {
     3603         case  0:
     3604-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__shape);
     3605-        if (likely(values[0])) kw_args--;
     3606+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__shape)) != 0)) kw_args--;
     3607         else goto __pyx_L5_argtuple_error;
     3608         case  1:
     3609         if (kw_args > 0) {
     3610@@ -8074,7 +8420,7 @@
     3611         }
     3612       }
     3613       if (unlikely(kw_args > 0)) {
     3614-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "standard_gamma") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1453; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     3615+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "standard_gamma") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1453; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     3616       }
     3617     } else {
     3618       switch (PyTuple_GET_SIZE(__pyx_args)) {
     3619@@ -8095,6 +8441,25 @@
     3620   __Pyx_RefNannyFinishContext();
     3621   return NULL;
     3622   __pyx_L4_argument_unpacking_done:;
     3623+  __pyx_r = __pyx_pf_6mtrand_11RandomState_42standard_gamma(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_shape, __pyx_v_size);
     3624+  __Pyx_RefNannyFinishContext();
     3625+  return __pyx_r;
     3626+}
     3627+
     3628+static PyObject *__pyx_pf_6mtrand_11RandomState_42standard_gamma(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_shape, PyObject *__pyx_v_size) {
     3629+  PyArrayObject *__pyx_v_oshape = 0;
     3630+  double __pyx_v_fshape;
     3631+  PyObject *__pyx_r = NULL;
     3632+  __Pyx_RefNannyDeclarations
     3633+  int __pyx_t_1;
     3634+  PyObject *__pyx_t_2 = NULL;
     3635+  PyObject *__pyx_t_3 = NULL;
     3636+  PyObject *__pyx_t_4 = NULL;
     3637+  PyObject *__pyx_t_5 = NULL;
     3638+  int __pyx_lineno = 0;
     3639+  const char *__pyx_filename = NULL;
     3640+  int __pyx_clineno = 0;
     3641+  __Pyx_RefNannySetupContext("standard_gamma", 0);
     3642 
     3643   /* "mtrand.pyx":1523
     3644  *         cdef double fshape
     3645@@ -8137,9 +8502,9 @@
     3646       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     3647       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3648       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1526; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3649-      goto __pyx_L7;
     3650+      goto __pyx_L4;
     3651     }
     3652-    __pyx_L7:;
     3653+    __pyx_L4:;
     3654 
     3655     /* "mtrand.pyx":1527
     3656  *             if fshape <= 0:
     3657@@ -8149,14 +8514,14 @@
     3658  *         PyErr_Clear()
     3659  */
     3660     __Pyx_XDECREF(__pyx_r);
     3661-    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_standard_gamma, __pyx_v_size, __pyx_v_fshape); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1527; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3662+    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(__pyx_v_self->internal_state, rk_standard_gamma, __pyx_v_size, __pyx_v_fshape); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1527; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3663     __Pyx_GOTREF(__pyx_t_2);
     3664     __pyx_r = __pyx_t_2;
     3665     __pyx_t_2 = 0;
     3666     goto __pyx_L0;
     3667-    goto __pyx_L6;
     3668+    goto __pyx_L3;
     3669   }
     3670-  __pyx_L6:;
     3671+  __pyx_L3:;
     3672 
     3673   /* "mtrand.pyx":1529
     3674  *             return cont1_array_sc(self.internal_state, rk_standard_gamma, size, fshape)
     3675@@ -8200,7 +8565,7 @@
     3676   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1531; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3677   __Pyx_GOTREF(__pyx_t_2);
     3678   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1531; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3679-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     3680+  __Pyx_GOTREF(__pyx_t_5);
     3681   __Pyx_INCREF(((PyObject *)__pyx_v_oshape));
     3682   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_oshape));
     3683   __Pyx_GIVEREF(((PyObject *)__pyx_v_oshape));
     3684@@ -8212,7 +8577,7 @@
     3685   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     3686   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     3687   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1531; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3688-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     3689+  __Pyx_GOTREF(__pyx_t_5);
     3690   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     3691   __Pyx_GIVEREF(__pyx_t_2);
     3692   __pyx_t_2 = 0;
     3693@@ -8236,9 +8601,9 @@
     3694     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     3695     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3696     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1532; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3697-    goto __pyx_L8;
     3698+    goto __pyx_L5;
     3699   }
     3700-  __pyx_L8:;
     3701+  __pyx_L5:;
     3702 
     3703   /* "mtrand.pyx":1533
     3704  *         if np.any(np.less_equal(oshape, 0.0)):
     3705@@ -8248,7 +8613,7 @@
     3706  *     def gamma(self, shape, scale=1.0, size=None):
     3707  */
     3708   __Pyx_XDECREF(__pyx_r);
     3709-  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_standard_gamma, __pyx_v_size, __pyx_v_oshape); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1533; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3710+  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(__pyx_v_self->internal_state, rk_standard_gamma, __pyx_v_size, __pyx_v_oshape); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1533; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3711   __Pyx_GOTREF(__pyx_t_2);
     3712   __pyx_r = __pyx_t_2;
     3713   __pyx_t_2 = 0;
     3714@@ -8270,43 +8635,33 @@
     3715   return __pyx_r;
     3716 }
     3717 
     3718-/* "mtrand.pyx":1535
     3719- *         return cont1_array(self.internal_state, rk_standard_gamma, size, oshape)
     3720- *
     3721- *     def gamma(self, shape, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     3722- *         """
     3723- *         gamma(shape, scale=1.0, size=None)
     3724- */
     3725-
     3726-static PyObject *__pyx_pf_6mtrand_11RandomState_22gamma(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3727-static char __pyx_doc_6mtrand_11RandomState_22gamma[] = "\n        gamma(shape, scale=1.0, size=None)\n\n        Draw samples from a Gamma distribution.\n\n        Samples are drawn from a Gamma distribution with specified parameters,\n        `shape` (sometimes designated \"k\") and `scale` (sometimes designated\n        \"theta\"), where both parameters are > 0.\n\n        Parameters\n        ----------\n        shape : scalar > 0\n            The shape of the gamma distribution.\n        scale : scalar > 0, optional\n            The scale of the gamma distribution.  Default is equal to 1.\n        size : shape_tuple, optional\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        out : ndarray, float\n            Returns one sample unless `size` parameter is specified.\n\n        See Also\n        --------\n        scipy.stats.distributions.gamma : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Gamma distribution is\n\n        .. math:: p(x) = x^{k-1}\\frac{e^{-x/\\theta}}{\\theta^k\\Gamma(k)},\n\n        where :math:`k` is the shape and :math:`\\theta` the scale,\n        and :math:`\\Gamma` is the Gamma function.\n\n        The Gamma distribution is often used to model the times to failure of\n        electronic components, and arises naturally in processes for which the\n        waiting times between Poisson distributed events are relevant.\n\n        References\n        ----------\n        .. [1] Weisstein, Eric W. \"Gamma Distribution.\" From MathWorld--A\n               Wolfram Web Resource.\n               http://mathworld.wolfram.com/GammaDistribution.html\n        .. [2] Wikipedia, \"Gamma-distribution\",\n               http://en.wikipedia.org/wiki/Gamma-distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> shape, scale = 2.,"" 2. # mean and dispersion\n        >>> s = np.random.gamma(shape, scale, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> import scipy.special as sps\n        >>> count, bins, ignored = plt.hist(s, 50, normed=True)\n        >>> y = bins**(shape-1)*(np.exp(-bins/scale) /\n        ...                      (sps.gamma(shape)*scale**shape))\n        >>> plt.plot(bins, y, linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     3728-static PyObject *__pyx_pf_6mtrand_11RandomState_22gamma(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3729+/* Python wrapper */
     3730+static PyObject *__pyx_pw_6mtrand_11RandomState_45gamma(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3731+static char __pyx_doc_6mtrand_11RandomState_44gamma[] = "\n        gamma(shape, scale=1.0, size=None)\n\n        Draw samples from a Gamma distribution.\n\n        Samples are drawn from a Gamma distribution with specified parameters,\n        `shape` (sometimes designated \"k\") and `scale` (sometimes designated\n        \"theta\"), where both parameters are > 0.\n\n        Parameters\n        ----------\n        shape : scalar > 0\n            The shape of the gamma distribution.\n        scale : scalar > 0, optional\n            The scale of the gamma distribution.  Default is equal to 1.\n        size : shape_tuple, optional\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        out : ndarray, float\n            Returns one sample unless `size` parameter is specified.\n\n        See Also\n        --------\n        scipy.stats.distributions.gamma : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Gamma distribution is\n\n        .. math:: p(x) = x^{k-1}\\frac{e^{-x/\\theta}}{\\theta^k\\Gamma(k)},\n\n        where :math:`k` is the shape and :math:`\\theta` the scale,\n        and :math:`\\Gamma` is the Gamma function.\n\n        The Gamma distribution is often used to model the times to failure of\n        electronic components, and arises naturally in processes for which the\n        waiting times between Poisson distributed events are relevant.\n\n        References\n        ----------\n        .. [1] Weisstein, Eric W. \"Gamma Distribution.\" From MathWorld--A\n               Wolfram Web Resource.\n               http://mathworld.wolfram.com/GammaDistribution.html\n        .. [2] Wikipedia, \"Gamma-distribution\",\n               http://en.wikipedia.org/wiki/Gamma-distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> shape, scale = 2.,"" 2. # mean and dispersion\n        >>> s = np.random.gamma(shape, scale, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> import scipy.special as sps\n        >>> count, bins, ignored = plt.hist(s, 50, normed=True)\n        >>> y = bins**(shape-1)*(np.exp(-bins/scale) /\n        ...                      (sps.gamma(shape)*scale**shape))\n        >>> plt.plot(bins, y, linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     3732+static PyObject *__pyx_pw_6mtrand_11RandomState_45gamma(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3733   PyObject *__pyx_v_shape = 0;
     3734   PyObject *__pyx_v_scale = 0;
     3735   PyObject *__pyx_v_size = 0;
     3736-  PyArrayObject *__pyx_v_oshape = 0;
     3737-  PyArrayObject *__pyx_v_oscale = 0;
     3738-  double __pyx_v_fshape;
     3739-  double __pyx_v_fscale;
     3740-  PyObject *__pyx_r = NULL;
     3741+  PyObject *__pyx_r = 0;
     3742   __Pyx_RefNannyDeclarations
     3743-  int __pyx_t_1;
     3744-  PyObject *__pyx_t_2 = NULL;
     3745-  PyObject *__pyx_t_3 = NULL;
     3746-  PyObject *__pyx_t_4 = NULL;
     3747-  PyObject *__pyx_t_5 = NULL;
     3748-  int __pyx_lineno = 0;
     3749-  const char *__pyx_filename = NULL;
     3750-  int __pyx_clineno = 0;
     3751-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__shape,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     3752-  __Pyx_RefNannySetupContext("gamma");
     3753+  __Pyx_RefNannySetupContext("gamma (wrapper)", 0);
     3754   {
     3755+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__shape,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     3756     PyObject* values[3] = {0,0,0};
     3757     values[1] = __pyx_k_34;
     3758+
     3759+    /* "mtrand.pyx":1535
     3760+ *         return cont1_array(self.internal_state, rk_standard_gamma, size, oshape)
     3761+ *
     3762+ *     def gamma(self, shape, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     3763+ *         """
     3764+ *         gamma(shape, scale=1.0, size=None)
     3765+ */
     3766     values[2] = ((PyObject *)Py_None);
     3767     if (unlikely(__pyx_kwds)) {
     3768       Py_ssize_t kw_args;
     3769-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     3770+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     3771+      switch (pos_args) {
     3772         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     3773         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     3774         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     3775@@ -8314,10 +8669,9 @@
     3776         default: goto __pyx_L5_argtuple_error;
     3777       }
     3778       kw_args = PyDict_Size(__pyx_kwds);
     3779-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     3780+      switch (pos_args) {
     3781         case  0:
     3782-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__shape);
     3783-        if (likely(values[0])) kw_args--;
     3784+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__shape)) != 0)) kw_args--;
     3785         else goto __pyx_L5_argtuple_error;
     3786         case  1:
     3787         if (kw_args > 0) {
     3788@@ -8331,7 +8685,7 @@
     3789         }
     3790       }
     3791       if (unlikely(kw_args > 0)) {
     3792-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "gamma") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1535; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     3793+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "gamma") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1535; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     3794       }
     3795     } else {
     3796       switch (PyTuple_GET_SIZE(__pyx_args)) {
     3797@@ -8354,6 +8708,27 @@
     3798   __Pyx_RefNannyFinishContext();
     3799   return NULL;
     3800   __pyx_L4_argument_unpacking_done:;
     3801+  __pyx_r = __pyx_pf_6mtrand_11RandomState_44gamma(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_shape, __pyx_v_scale, __pyx_v_size);
     3802+  __Pyx_RefNannyFinishContext();
     3803+  return __pyx_r;
     3804+}
     3805+
     3806+static PyObject *__pyx_pf_6mtrand_11RandomState_44gamma(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_shape, PyObject *__pyx_v_scale, PyObject *__pyx_v_size) {
     3807+  PyArrayObject *__pyx_v_oshape = 0;
     3808+  PyArrayObject *__pyx_v_oscale = 0;
     3809+  double __pyx_v_fshape;
     3810+  double __pyx_v_fscale;
     3811+  PyObject *__pyx_r = NULL;
     3812+  __Pyx_RefNannyDeclarations
     3813+  int __pyx_t_1;
     3814+  PyObject *__pyx_t_2 = NULL;
     3815+  PyObject *__pyx_t_3 = NULL;
     3816+  PyObject *__pyx_t_4 = NULL;
     3817+  PyObject *__pyx_t_5 = NULL;
     3818+  int __pyx_lineno = 0;
     3819+  const char *__pyx_filename = NULL;
     3820+  int __pyx_clineno = 0;
     3821+  __Pyx_RefNannySetupContext("gamma", 0);
     3822 
     3823   /* "mtrand.pyx":1608
     3824  *         cdef double fshape, fscale
     3825@@ -8405,9 +8780,9 @@
     3826       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     3827       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3828       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1612; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3829-      goto __pyx_L7;
     3830+      goto __pyx_L4;
     3831     }
     3832-    __pyx_L7:;
     3833+    __pyx_L4:;
     3834 
     3835     /* "mtrand.pyx":1613
     3836  *             if fshape <= 0:
     3837@@ -8431,9 +8806,9 @@
     3838       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     3839       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3840       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1614; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3841-      goto __pyx_L8;
     3842+      goto __pyx_L5;
     3843     }
     3844-    __pyx_L8:;
     3845+    __pyx_L5:;
     3846 
     3847     /* "mtrand.pyx":1615
     3848  *             if fscale <= 0:
     3849@@ -8443,14 +8818,14 @@
     3850  *         PyErr_Clear()
     3851  */
     3852     __Pyx_XDECREF(__pyx_r);
     3853-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_gamma, __pyx_v_size, __pyx_v_fshape, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1615; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3854+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_gamma, __pyx_v_size, __pyx_v_fshape, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1615; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3855     __Pyx_GOTREF(__pyx_t_2);
     3856     __pyx_r = __pyx_t_2;
     3857     __pyx_t_2 = 0;
     3858     goto __pyx_L0;
     3859-    goto __pyx_L6;
     3860+    goto __pyx_L3;
     3861   }
     3862-  __pyx_L6:;
     3863+  __pyx_L3:;
     3864 
     3865   /* "mtrand.pyx":1617
     3866  *             return cont2_array_sc(self.internal_state, rk_gamma, size, fshape, fscale)
     3867@@ -8507,7 +8882,7 @@
     3868   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1620; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3869   __Pyx_GOTREF(__pyx_t_2);
     3870   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1620; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3871-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     3872+  __Pyx_GOTREF(__pyx_t_5);
     3873   __Pyx_INCREF(((PyObject *)__pyx_v_oshape));
     3874   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_oshape));
     3875   __Pyx_GIVEREF(((PyObject *)__pyx_v_oshape));
     3876@@ -8519,7 +8894,7 @@
     3877   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     3878   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     3879   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1620; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3880-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     3881+  __Pyx_GOTREF(__pyx_t_5);
     3882   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     3883   __Pyx_GIVEREF(__pyx_t_2);
     3884   __pyx_t_2 = 0;
     3885@@ -8543,9 +8918,9 @@
     3886     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     3887     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3888     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1621; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3889-    goto __pyx_L9;
     3890+    goto __pyx_L6;
     3891   }
     3892-  __pyx_L9:;
     3893+  __pyx_L6:;
     3894 
     3895   /* "mtrand.pyx":1622
     3896  *         if np.any(np.less_equal(oshape, 0.0)):
     3897@@ -8567,7 +8942,7 @@
     3898   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1622; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3899   __Pyx_GOTREF(__pyx_t_2);
     3900   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1622; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3901-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     3902+  __Pyx_GOTREF(__pyx_t_4);
     3903   __Pyx_INCREF(((PyObject *)__pyx_v_oscale));
     3904   PyTuple_SET_ITEM(__pyx_t_4, 0, ((PyObject *)__pyx_v_oscale));
     3905   __Pyx_GIVEREF(((PyObject *)__pyx_v_oscale));
     3906@@ -8579,7 +8954,7 @@
     3907   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     3908   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     3909   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1622; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3910-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     3911+  __Pyx_GOTREF(__pyx_t_4);
     3912   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     3913   __Pyx_GIVEREF(__pyx_t_2);
     3914   __pyx_t_2 = 0;
     3915@@ -8603,9 +8978,9 @@
     3916     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     3917     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     3918     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1623; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3919-    goto __pyx_L10;
     3920+    goto __pyx_L7;
     3921   }
     3922-  __pyx_L10:;
     3923+  __pyx_L7:;
     3924 
     3925   /* "mtrand.pyx":1624
     3926  *         if np.any(np.less_equal(oscale, 0.0)):
     3927@@ -8615,7 +8990,7 @@
     3928  *     def f(self, dfnum, dfden, size=None):
     3929  */
     3930   __Pyx_XDECREF(__pyx_r);
     3931-  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_gamma, __pyx_v_size, __pyx_v_oshape, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1624; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3932+  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_gamma, __pyx_v_size, __pyx_v_oshape, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1624; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     3933   __Pyx_GOTREF(__pyx_t_2);
     3934   __pyx_r = __pyx_t_2;
     3935   __pyx_t_2 = 0;
     3936@@ -8638,42 +9013,32 @@
     3937   return __pyx_r;
     3938 }
     3939 
     3940-/* "mtrand.pyx":1626
     3941- *         return cont2_array(self.internal_state, rk_gamma, size, oshape, oscale)
     3942- *
     3943- *     def f(self, dfnum, dfden, size=None):             # <<<<<<<<<<<<<<
     3944- *         """
     3945- *         f(dfnum, dfden, size=None)
     3946- */
     3947-
     3948-static PyObject *__pyx_pf_6mtrand_11RandomState_23f(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3949-static char __pyx_doc_6mtrand_11RandomState_23f[] = "\n        f(dfnum, dfden, size=None)\n\n        Draw samples from a F distribution.\n\n        Samples are drawn from an F distribution with specified parameters,\n        `dfnum` (degrees of freedom in numerator) and `dfden` (degrees of freedom\n        in denominator), where both parameters should be greater than zero.\n\n        The random variate of the F distribution (also known as the\n        Fisher distribution) is a continuous probability distribution\n        that arises in ANOVA tests, and is the ratio of two chi-square\n        variates.\n\n        Parameters\n        ----------\n        dfnum : float\n            Degrees of freedom in numerator. Should be greater than zero.\n        dfden : float\n            Degrees of freedom in denominator. Should be greater than zero.\n        size : {tuple, int}, optional\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``,\n            then ``m * n * k`` samples are drawn. By default only one sample\n            is returned.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n            Samples from the Fisher distribution.\n\n        See Also\n        --------\n        scipy.stats.distributions.f : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n\n        The F statistic is used to compare in-group variances to between-group\n        variances. Calculating the distribution depends on the sampling, and\n        so it is a function of the respective degrees of freedom in the\n        problem.  The variable `dfnum` is the number of samples minus one, the\n        between-groups degrees of freedom, while `dfden` is the within-groups\n        degrees of freedom, the sum of the number of samples in each group\n        minus the number of groups.\n\n        References\n        ----------\n        .. [1] Glantz, Stanton A. \"Primer of Biostatistics.\", McGraw-Hill,\n               Fifth Edition, 2002.""\n        .. [2] Wikipedia, \"F-distribution\",\n               http://en.wikipedia.org/wiki/F-distribution\n\n        Examples\n        --------\n        An example from Glantz[1], pp 47-40.\n        Two groups, children of diabetics (25 people) and children from people\n        without diabetes (25 controls). Fasting blood glucose was measured,\n        case group had a mean value of 86.1, controls had a mean value of\n        82.2. Standard deviations were 2.09 and 2.49 respectively. Are these\n        data consistent with the null hypothesis that the parents diabetic\n        status does not affect their children's blood glucose levels?\n        Calculating the F statistic from the data gives a value of 36.01.\n\n        Draw samples from the distribution:\n\n        >>> dfnum = 1. # between group degrees of freedom\n        >>> dfden = 48. # within groups degrees of freedom\n        >>> s = np.random.f(dfnum, dfden, 1000)\n\n        The lower bound for the top 1% of the samples is :\n\n        >>> sort(s)[-10]\n        7.61988120985\n\n        So there is about a 1% chance that the F statistic will exceed 7.62,\n        the measured value is 36, so the null hypothesis is rejected at the 1%\n        level.\n\n        ";
     3950-static PyObject *__pyx_pf_6mtrand_11RandomState_23f(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3951+/* Python wrapper */
     3952+static PyObject *__pyx_pw_6mtrand_11RandomState_47f(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     3953+static char __pyx_doc_6mtrand_11RandomState_46f[] = "\n        f(dfnum, dfden, size=None)\n\n        Draw samples from a F distribution.\n\n        Samples are drawn from an F distribution with specified parameters,\n        `dfnum` (degrees of freedom in numerator) and `dfden` (degrees of freedom\n        in denominator), where both parameters should be greater than zero.\n\n        The random variate of the F distribution (also known as the\n        Fisher distribution) is a continuous probability distribution\n        that arises in ANOVA tests, and is the ratio of two chi-square\n        variates.\n\n        Parameters\n        ----------\n        dfnum : float\n            Degrees of freedom in numerator. Should be greater than zero.\n        dfden : float\n            Degrees of freedom in denominator. Should be greater than zero.\n        size : {tuple, int}, optional\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``,\n            then ``m * n * k`` samples are drawn. By default only one sample\n            is returned.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n            Samples from the Fisher distribution.\n\n        See Also\n        --------\n        scipy.stats.distributions.f : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n\n        The F statistic is used to compare in-group variances to between-group\n        variances. Calculating the distribution depends on the sampling, and\n        so it is a function of the respective degrees of freedom in the\n        problem.  The variable `dfnum` is the number of samples minus one, the\n        between-groups degrees of freedom, while `dfden` is the within-groups\n        degrees of freedom, the sum of the number of samples in each group\n        minus the number of groups.\n\n        References\n        ----------\n        .. [1] Glantz, Stanton A. \"Primer of Biostatistics.\", McGraw-Hill,\n               Fifth Edition, 2002.""\n        .. [2] Wikipedia, \"F-distribution\",\n               http://en.wikipedia.org/wiki/F-distribution\n\n        Examples\n        --------\n        An example from Glantz[1], pp 47-40.\n        Two groups, children of diabetics (25 people) and children from people\n        without diabetes (25 controls). Fasting blood glucose was measured,\n        case group had a mean value of 86.1, controls had a mean value of\n        82.2. Standard deviations were 2.09 and 2.49 respectively. Are these\n        data consistent with the null hypothesis that the parents diabetic\n        status does not affect their children's blood glucose levels?\n        Calculating the F statistic from the data gives a value of 36.01.\n\n        Draw samples from the distribution:\n\n        >>> dfnum = 1. # between group degrees of freedom\n        >>> dfden = 48. # within groups degrees of freedom\n        >>> s = np.random.f(dfnum, dfden, 1000)\n\n        The lower bound for the top 1% of the samples is :\n\n        >>> sort(s)[-10]\n        7.61988120985\n\n        So there is about a 1% chance that the F statistic will exceed 7.62,\n        the measured value is 36, so the null hypothesis is rejected at the 1%\n        level.\n\n        ";
     3954+static PyObject *__pyx_pw_6mtrand_11RandomState_47f(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     3955   PyObject *__pyx_v_dfnum = 0;
     3956   PyObject *__pyx_v_dfden = 0;
     3957   PyObject *__pyx_v_size = 0;
     3958-  PyArrayObject *__pyx_v_odfnum = 0;
     3959-  PyArrayObject *__pyx_v_odfden = 0;
     3960-  double __pyx_v_fdfnum;
     3961-  double __pyx_v_fdfden;
     3962-  PyObject *__pyx_r = NULL;
     3963+  PyObject *__pyx_r = 0;
     3964   __Pyx_RefNannyDeclarations
     3965-  int __pyx_t_1;
     3966-  PyObject *__pyx_t_2 = NULL;
     3967-  PyObject *__pyx_t_3 = NULL;
     3968-  PyObject *__pyx_t_4 = NULL;
     3969-  PyObject *__pyx_t_5 = NULL;
     3970-  int __pyx_lineno = 0;
     3971-  const char *__pyx_filename = NULL;
     3972-  int __pyx_clineno = 0;
     3973-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__dfnum,&__pyx_n_s__dfden,&__pyx_n_s__size,0};
     3974-  __Pyx_RefNannySetupContext("f");
     3975+  __Pyx_RefNannySetupContext("f (wrapper)", 0);
     3976   {
     3977+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__dfnum,&__pyx_n_s__dfden,&__pyx_n_s__size,0};
     3978     PyObject* values[3] = {0,0,0};
     3979+
     3980+    /* "mtrand.pyx":1626
     3981+ *         return cont2_array(self.internal_state, rk_gamma, size, oshape, oscale)
     3982+ *
     3983+ *     def f(self, dfnum, dfden, size=None):             # <<<<<<<<<<<<<<
     3984+ *         """
     3985+ *         f(dfnum, dfden, size=None)
     3986+ */
     3987     values[2] = ((PyObject *)Py_None);
     3988     if (unlikely(__pyx_kwds)) {
     3989       Py_ssize_t kw_args;
     3990-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     3991+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     3992+      switch (pos_args) {
     3993         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     3994         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     3995         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     3996@@ -8681,14 +9046,12 @@
     3997         default: goto __pyx_L5_argtuple_error;
     3998       }
     3999       kw_args = PyDict_Size(__pyx_kwds);
     4000-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     4001+      switch (pos_args) {
     4002         case  0:
     4003-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__dfnum);
     4004-        if (likely(values[0])) kw_args--;
     4005+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__dfnum)) != 0)) kw_args--;
     4006         else goto __pyx_L5_argtuple_error;
     4007         case  1:
     4008-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__dfden);
     4009-        if (likely(values[1])) kw_args--;
     4010+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__dfden)) != 0)) kw_args--;
     4011         else {
     4012           __Pyx_RaiseArgtupleInvalid("f", 0, 2, 3, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1626; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4013         }
     4014@@ -8699,7 +9062,7 @@
     4015         }
     4016       }
     4017       if (unlikely(kw_args > 0)) {
     4018-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "f") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1626; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4019+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "f") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1626; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4020       }
     4021     } else {
     4022       switch (PyTuple_GET_SIZE(__pyx_args)) {
     4023@@ -8722,6 +9085,27 @@
     4024   __Pyx_RefNannyFinishContext();
     4025   return NULL;
     4026   __pyx_L4_argument_unpacking_done:;
     4027+  __pyx_r = __pyx_pf_6mtrand_11RandomState_46f(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_dfnum, __pyx_v_dfden, __pyx_v_size);
     4028+  __Pyx_RefNannyFinishContext();
     4029+  return __pyx_r;
     4030+}
     4031+
     4032+static PyObject *__pyx_pf_6mtrand_11RandomState_46f(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_dfnum, PyObject *__pyx_v_dfden, PyObject *__pyx_v_size) {
     4033+  PyArrayObject *__pyx_v_odfnum = 0;
     4034+  PyArrayObject *__pyx_v_odfden = 0;
     4035+  double __pyx_v_fdfnum;
     4036+  double __pyx_v_fdfden;
     4037+  PyObject *__pyx_r = NULL;
     4038+  __Pyx_RefNannyDeclarations
     4039+  int __pyx_t_1;
     4040+  PyObject *__pyx_t_2 = NULL;
     4041+  PyObject *__pyx_t_3 = NULL;
     4042+  PyObject *__pyx_t_4 = NULL;
     4043+  PyObject *__pyx_t_5 = NULL;
     4044+  int __pyx_lineno = 0;
     4045+  const char *__pyx_filename = NULL;
     4046+  int __pyx_clineno = 0;
     4047+  __Pyx_RefNannySetupContext("f", 0);
     4048 
     4049   /* "mtrand.pyx":1710
     4050  *         cdef double fdfnum, fdfden
     4051@@ -8773,9 +9157,9 @@
     4052       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4053       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4054       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1714; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4055-      goto __pyx_L7;
     4056+      goto __pyx_L4;
     4057     }
     4058-    __pyx_L7:;
     4059+    __pyx_L4:;
     4060 
     4061     /* "mtrand.pyx":1715
     4062  *             if fdfnum <= 0:
     4063@@ -8799,9 +9183,9 @@
     4064       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4065       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4066       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1716; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4067-      goto __pyx_L8;
     4068+      goto __pyx_L5;
     4069     }
     4070-    __pyx_L8:;
     4071+    __pyx_L5:;
     4072 
     4073     /* "mtrand.pyx":1717
     4074  *             if fdfden <= 0:
     4075@@ -8811,14 +9195,14 @@
     4076  *         PyErr_Clear()
     4077  */
     4078     __Pyx_XDECREF(__pyx_r);
     4079-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_f, __pyx_v_size, __pyx_v_fdfnum, __pyx_v_fdfden); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1717; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4080+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_f, __pyx_v_size, __pyx_v_fdfnum, __pyx_v_fdfden); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1717; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4081     __Pyx_GOTREF(__pyx_t_2);
     4082     __pyx_r = __pyx_t_2;
     4083     __pyx_t_2 = 0;
     4084     goto __pyx_L0;
     4085-    goto __pyx_L6;
     4086+    goto __pyx_L3;
     4087   }
     4088-  __pyx_L6:;
     4089+  __pyx_L3:;
     4090 
     4091   /* "mtrand.pyx":1719
     4092  *             return cont2_array_sc(self.internal_state, rk_f, size, fdfnum, fdfden)
     4093@@ -8875,7 +9259,7 @@
     4094   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1723; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4095   __Pyx_GOTREF(__pyx_t_2);
     4096   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1723; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4097-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     4098+  __Pyx_GOTREF(__pyx_t_5);
     4099   __Pyx_INCREF(((PyObject *)__pyx_v_odfnum));
     4100   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_odfnum));
     4101   __Pyx_GIVEREF(((PyObject *)__pyx_v_odfnum));
     4102@@ -8887,7 +9271,7 @@
     4103   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     4104   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     4105   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1723; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4106-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     4107+  __Pyx_GOTREF(__pyx_t_5);
     4108   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     4109   __Pyx_GIVEREF(__pyx_t_2);
     4110   __pyx_t_2 = 0;
     4111@@ -8911,9 +9295,9 @@
     4112     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4113     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4114     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1724; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4115-    goto __pyx_L9;
     4116+    goto __pyx_L6;
     4117   }
     4118-  __pyx_L9:;
     4119+  __pyx_L6:;
     4120 
     4121   /* "mtrand.pyx":1725
     4122  *         if np.any(np.less_equal(odfnum, 0.0)):
     4123@@ -8935,7 +9319,7 @@
     4124   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1725; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4125   __Pyx_GOTREF(__pyx_t_2);
     4126   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1725; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4127-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     4128+  __Pyx_GOTREF(__pyx_t_4);
     4129   __Pyx_INCREF(((PyObject *)__pyx_v_odfden));
     4130   PyTuple_SET_ITEM(__pyx_t_4, 0, ((PyObject *)__pyx_v_odfden));
     4131   __Pyx_GIVEREF(((PyObject *)__pyx_v_odfden));
     4132@@ -8947,7 +9331,7 @@
     4133   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     4134   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     4135   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1725; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4136-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     4137+  __Pyx_GOTREF(__pyx_t_4);
     4138   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     4139   __Pyx_GIVEREF(__pyx_t_2);
     4140   __pyx_t_2 = 0;
     4141@@ -8971,9 +9355,9 @@
     4142     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4143     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4144     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1726; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4145-    goto __pyx_L10;
     4146+    goto __pyx_L7;
     4147   }
     4148-  __pyx_L10:;
     4149+  __pyx_L7:;
     4150 
     4151   /* "mtrand.pyx":1727
     4152  *         if np.any(np.less_equal(odfden, 0.0)):
     4153@@ -8983,7 +9367,7 @@
     4154  *     def noncentral_f(self, dfnum, dfden, nonc, size=None):
     4155  */
     4156   __Pyx_XDECREF(__pyx_r);
     4157-  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_f, __pyx_v_size, __pyx_v_odfnum, __pyx_v_odfden); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1727; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4158+  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_f, __pyx_v_size, __pyx_v_odfnum, __pyx_v_odfden); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1727; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4159   __Pyx_GOTREF(__pyx_t_2);
     4160   __pyx_r = __pyx_t_2;
     4161   __pyx_t_2 = 0;
     4162@@ -9006,45 +9390,33 @@
     4163   return __pyx_r;
     4164 }
     4165 
     4166-/* "mtrand.pyx":1729
     4167+/* Python wrapper */
     4168+static PyObject *__pyx_pw_6mtrand_11RandomState_49noncentral_f(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     4169+static char __pyx_doc_6mtrand_11RandomState_48noncentral_f[] = "\n        noncentral_f(dfnum, dfden, nonc, size=None)\n\n        Draw samples from the noncentral F distribution.\n\n        Samples are drawn from an F distribution with specified parameters,\n        `dfnum` (degrees of freedom in numerator) and `dfden` (degrees of\n        freedom in denominator), where both parameters > 1.\n        `nonc` is the non-centrality parameter.\n\n        Parameters\n        ----------\n        dfnum : int\n            Parameter, should be > 1.\n        dfden : int\n            Parameter, should be > 1.\n        nonc : float\n            Parameter, should be >= 0.\n        size : int or tuple of ints\n            Output shape. If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : scalar or ndarray\n            Drawn samples.\n\n        Notes\n        -----\n        When calculating the power of an experiment (power = probability of\n        rejecting the null hypothesis when a specific alternative is true) the\n        non-central F statistic becomes important.  When the null hypothesis is\n        true, the F statistic follows a central F distribution. When the null\n        hypothesis is not true, then it follows a non-central F statistic.\n\n        References\n        ----------\n        Weisstein, Eric W. \"Noncentral F-Distribution.\" From MathWorld--A Wolfram\n        Web Resource.  http://mathworld.wolfram.com/NoncentralF-Distribution.html\n\n        Wikipedia, \"Noncentral F distribution\",\n        http://en.wikipedia.org/wiki/Noncentral_F-distribution\n\n        Examples\n        --------\n        In a study, testing for a specific alternative to the null hypothesis\n        requires use of the Noncentral F distribution. We need to calculate the\n        area in the tail of the distribution that exceeds the value of the F\n        distribution for the null hypothesis.  We'll plot the two probability\n        distributions for comp""arison.\n\n        >>> dfnum = 3 # between group deg of freedom\n        >>> dfden = 20 # within groups degrees of freedom\n        >>> nonc = 3.0\n        >>> nc_vals = np.random.noncentral_f(dfnum, dfden, nonc, 1000000)\n        >>> NF = np.histogram(nc_vals, bins=50, normed=True)\n        >>> c_vals = np.random.f(dfnum, dfden, 1000000)\n        >>> F = np.histogram(c_vals, bins=50, normed=True)\n        >>> plt.plot(F[1][1:], F[0])\n        >>> plt.plot(NF[1][1:], NF[0])\n        >>> plt.show()\n\n        ";
     4170+static PyObject *__pyx_pw_6mtrand_11RandomState_49noncentral_f(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     4171+  PyObject *__pyx_v_dfnum = 0;
     4172+  PyObject *__pyx_v_dfden = 0;
     4173+  PyObject *__pyx_v_nonc = 0;
     4174+  PyObject *__pyx_v_size = 0;
     4175+  PyObject *__pyx_r = 0;
     4176+  __Pyx_RefNannyDeclarations
     4177+  __Pyx_RefNannySetupContext("noncentral_f (wrapper)", 0);
     4178+  {
     4179+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__dfnum,&__pyx_n_s__dfden,&__pyx_n_s__nonc,&__pyx_n_s__size,0};
     4180+    PyObject* values[4] = {0,0,0,0};
     4181+
     4182+    /* "mtrand.pyx":1729
     4183  *         return cont2_array(self.internal_state, rk_f, size, odfnum, odfden)
     4184  *
     4185  *     def noncentral_f(self, dfnum, dfden, nonc, size=None):             # <<<<<<<<<<<<<<
     4186  *         """
     4187  *         noncentral_f(dfnum, dfden, nonc, size=None)
     4188  */
     4189-
     4190-static PyObject *__pyx_pf_6mtrand_11RandomState_24noncentral_f(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     4191-static char __pyx_doc_6mtrand_11RandomState_24noncentral_f[] = "\n        noncentral_f(dfnum, dfden, nonc, size=None)\n\n        Draw samples from the noncentral F distribution.\n\n        Samples are drawn from an F distribution with specified parameters,\n        `dfnum` (degrees of freedom in numerator) and `dfden` (degrees of\n        freedom in denominator), where both parameters > 1.\n        `nonc` is the non-centrality parameter.\n\n        Parameters\n        ----------\n        dfnum : int\n            Parameter, should be > 1.\n        dfden : int\n            Parameter, should be > 1.\n        nonc : float\n            Parameter, should be >= 0.\n        size : int or tuple of ints\n            Output shape. If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : scalar or ndarray\n            Drawn samples.\n\n        Notes\n        -----\n        When calculating the power of an experiment (power = probability of\n        rejecting the null hypothesis when a specific alternative is true) the\n        non-central F statistic becomes important.  When the null hypothesis is\n        true, the F statistic follows a central F distribution. When the null\n        hypothesis is not true, then it follows a non-central F statistic.\n\n        References\n        ----------\n        Weisstein, Eric W. \"Noncentral F-Distribution.\" From MathWorld--A Wolfram\n        Web Resource.  http://mathworld.wolfram.com/NoncentralF-Distribution.html\n\n        Wikipedia, \"Noncentral F distribution\",\n        http://en.wikipedia.org/wiki/Noncentral_F-distribution\n\n        Examples\n        --------\n        In a study, testing for a specific alternative to the null hypothesis\n        requires use of the Noncentral F distribution. We need to calculate the\n        area in the tail of the distribution that exceeds the value of the F\n        distribution for the null hypothesis.  We'll plot the two probability\n        distributions for comp""arison.\n\n        >>> dfnum = 3 # between group deg of freedom\n        >>> dfden = 20 # within groups degrees of freedom\n        >>> nonc = 3.0\n        >>> nc_vals = np.random.noncentral_f(dfnum, dfden, nonc, 1000000)\n        >>> NF = np.histogram(nc_vals, bins=50, normed=True)\n        >>> c_vals = np.random.f(dfnum, dfden, 1000000)\n        >>> F = np.histogram(c_vals, bins=50, normed=True)\n        >>> plt.plot(F[1][1:], F[0])\n        >>> plt.plot(NF[1][1:], NF[0])\n        >>> plt.show()\n\n        ";
     4192-static PyObject *__pyx_pf_6mtrand_11RandomState_24noncentral_f(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     4193-  PyObject *__pyx_v_dfnum = 0;
     4194-  PyObject *__pyx_v_dfden = 0;
     4195-  PyObject *__pyx_v_nonc = 0;
     4196-  PyObject *__pyx_v_size = 0;
     4197-  PyArrayObject *__pyx_v_odfnum = 0;
     4198-  PyArrayObject *__pyx_v_odfden = 0;
     4199-  PyArrayObject *__pyx_v_ononc = 0;
     4200-  double __pyx_v_fdfnum;
     4201-  double __pyx_v_fdfden;
     4202-  double __pyx_v_fnonc;
     4203-  PyObject *__pyx_r = NULL;
     4204-  __Pyx_RefNannyDeclarations
     4205-  int __pyx_t_1;
     4206-  PyObject *__pyx_t_2 = NULL;
     4207-  PyObject *__pyx_t_3 = NULL;
     4208-  PyObject *__pyx_t_4 = NULL;
     4209-  PyObject *__pyx_t_5 = NULL;
     4210-  int __pyx_lineno = 0;
     4211-  const char *__pyx_filename = NULL;
     4212-  int __pyx_clineno = 0;
     4213-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__dfnum,&__pyx_n_s__dfden,&__pyx_n_s__nonc,&__pyx_n_s__size,0};
     4214-  __Pyx_RefNannySetupContext("noncentral_f");
     4215-  {
     4216-    PyObject* values[4] = {0,0,0,0};
     4217     values[3] = ((PyObject *)Py_None);
     4218     if (unlikely(__pyx_kwds)) {
     4219       Py_ssize_t kw_args;
     4220-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     4221+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     4222+      switch (pos_args) {
     4223         case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
     4224         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     4225         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     4226@@ -9053,20 +9425,17 @@
     4227         default: goto __pyx_L5_argtuple_error;
     4228       }
     4229       kw_args = PyDict_Size(__pyx_kwds);
     4230-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     4231+      switch (pos_args) {
     4232         case  0:
     4233-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__dfnum);
     4234-        if (likely(values[0])) kw_args--;
     4235+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__dfnum)) != 0)) kw_args--;
     4236         else goto __pyx_L5_argtuple_error;
     4237         case  1:
     4238-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__dfden);
     4239-        if (likely(values[1])) kw_args--;
     4240+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__dfden)) != 0)) kw_args--;
     4241         else {
     4242           __Pyx_RaiseArgtupleInvalid("noncentral_f", 0, 3, 4, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1729; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4243         }
     4244         case  2:
     4245-        values[2] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__nonc);
     4246-        if (likely(values[2])) kw_args--;
     4247+        if (likely((values[2] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__nonc)) != 0)) kw_args--;
     4248         else {
     4249           __Pyx_RaiseArgtupleInvalid("noncentral_f", 0, 3, 4, 2); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1729; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4250         }
     4251@@ -9077,7 +9446,7 @@
     4252         }
     4253       }
     4254       if (unlikely(kw_args > 0)) {
     4255-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "noncentral_f") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1729; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4256+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "noncentral_f") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1729; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4257       }
     4258     } else {
     4259       switch (PyTuple_GET_SIZE(__pyx_args)) {
     4260@@ -9102,6 +9471,29 @@
     4261   __Pyx_RefNannyFinishContext();
     4262   return NULL;
     4263   __pyx_L4_argument_unpacking_done:;
     4264+  __pyx_r = __pyx_pf_6mtrand_11RandomState_48noncentral_f(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_dfnum, __pyx_v_dfden, __pyx_v_nonc, __pyx_v_size);
     4265+  __Pyx_RefNannyFinishContext();
     4266+  return __pyx_r;
     4267+}
     4268+
     4269+static PyObject *__pyx_pf_6mtrand_11RandomState_48noncentral_f(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_dfnum, PyObject *__pyx_v_dfden, PyObject *__pyx_v_nonc, PyObject *__pyx_v_size) {
     4270+  PyArrayObject *__pyx_v_odfnum = 0;
     4271+  PyArrayObject *__pyx_v_odfden = 0;
     4272+  PyArrayObject *__pyx_v_ononc = 0;
     4273+  double __pyx_v_fdfnum;
     4274+  double __pyx_v_fdfden;
     4275+  double __pyx_v_fnonc;
     4276+  PyObject *__pyx_r = NULL;
     4277+  __Pyx_RefNannyDeclarations
     4278+  int __pyx_t_1;
     4279+  PyObject *__pyx_t_2 = NULL;
     4280+  PyObject *__pyx_t_3 = NULL;
     4281+  PyObject *__pyx_t_4 = NULL;
     4282+  PyObject *__pyx_t_5 = NULL;
     4283+  int __pyx_lineno = 0;
     4284+  const char *__pyx_filename = NULL;
     4285+  int __pyx_clineno = 0;
     4286+  __Pyx_RefNannySetupContext("noncentral_f", 0);
     4287 
     4288   /* "mtrand.pyx":1796
     4289  *         cdef double fdfnum, fdfden, fnonc
     4290@@ -9162,9 +9554,9 @@
     4291       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4292       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4293       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1801; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4294-      goto __pyx_L7;
     4295+      goto __pyx_L4;
     4296     }
     4297-    __pyx_L7:;
     4298+    __pyx_L4:;
     4299 
     4300     /* "mtrand.pyx":1802
     4301  *             if fdfnum <= 1:
     4302@@ -9188,9 +9580,9 @@
     4303       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4304       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4305       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1803; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4306-      goto __pyx_L8;
     4307+      goto __pyx_L5;
     4308     }
     4309-    __pyx_L8:;
     4310+    __pyx_L5:;
     4311 
     4312     /* "mtrand.pyx":1804
     4313  *             if fdfden <= 0:
     4314@@ -9214,9 +9606,9 @@
     4315       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4316       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4317       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1805; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4318-      goto __pyx_L9;
     4319+      goto __pyx_L6;
     4320     }
     4321-    __pyx_L9:;
     4322+    __pyx_L6:;
     4323 
     4324     /* "mtrand.pyx":1806
     4325  *             if fnonc < 0:
     4326@@ -9234,14 +9626,14 @@
     4327  *
     4328  *         PyErr_Clear()
     4329  */
     4330-    __pyx_t_2 = __pyx_f_6mtrand_cont3_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_noncentral_f, __pyx_v_size, __pyx_v_fdfnum, __pyx_v_fdfden, __pyx_v_fnonc); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1806; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4331+    __pyx_t_2 = __pyx_f_6mtrand_cont3_array_sc(__pyx_v_self->internal_state, rk_noncentral_f, __pyx_v_size, __pyx_v_fdfnum, __pyx_v_fdfden, __pyx_v_fnonc); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1806; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4332     __Pyx_GOTREF(__pyx_t_2);
     4333     __pyx_r = __pyx_t_2;
     4334     __pyx_t_2 = 0;
     4335     goto __pyx_L0;
     4336-    goto __pyx_L6;
     4337+    goto __pyx_L3;
     4338   }
     4339-  __pyx_L6:;
     4340+  __pyx_L3:;
     4341 
     4342   /* "mtrand.pyx":1809
     4343  *                                   fdfnum, fdfden, fnonc)
     4344@@ -9311,7 +9703,7 @@
     4345   __pyx_t_2 = PyFloat_FromDouble(1.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1815; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4346   __Pyx_GOTREF(__pyx_t_2);
     4347   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1815; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4348-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     4349+  __Pyx_GOTREF(__pyx_t_5);
     4350   __Pyx_INCREF(((PyObject *)__pyx_v_odfnum));
     4351   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_odfnum));
     4352   __Pyx_GIVEREF(((PyObject *)__pyx_v_odfnum));
     4353@@ -9323,7 +9715,7 @@
     4354   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     4355   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     4356   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1815; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4357-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     4358+  __Pyx_GOTREF(__pyx_t_5);
     4359   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     4360   __Pyx_GIVEREF(__pyx_t_2);
     4361   __pyx_t_2 = 0;
     4362@@ -9347,9 +9739,9 @@
     4363     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4364     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4365     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1816; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4366-    goto __pyx_L10;
     4367+    goto __pyx_L7;
     4368   }
     4369-  __pyx_L10:;
     4370+  __pyx_L7:;
     4371 
     4372   /* "mtrand.pyx":1817
     4373  *         if np.any(np.less_equal(odfnum, 1.0)):
     4374@@ -9371,7 +9763,7 @@
     4375   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1817; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4376   __Pyx_GOTREF(__pyx_t_2);
     4377   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1817; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4378-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     4379+  __Pyx_GOTREF(__pyx_t_4);
     4380   __Pyx_INCREF(((PyObject *)__pyx_v_odfden));
     4381   PyTuple_SET_ITEM(__pyx_t_4, 0, ((PyObject *)__pyx_v_odfden));
     4382   __Pyx_GIVEREF(((PyObject *)__pyx_v_odfden));
     4383@@ -9383,7 +9775,7 @@
     4384   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     4385   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     4386   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1817; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4387-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     4388+  __Pyx_GOTREF(__pyx_t_4);
     4389   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     4390   __Pyx_GIVEREF(__pyx_t_2);
     4391   __pyx_t_2 = 0;
     4392@@ -9407,9 +9799,9 @@
     4393     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4394     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4395     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1818; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4396-    goto __pyx_L11;
     4397+    goto __pyx_L8;
     4398   }
     4399-  __pyx_L11:;
     4400+  __pyx_L8:;
     4401 
     4402   /* "mtrand.pyx":1819
     4403  *         if np.any(np.less_equal(odfden, 0.0)):
     4404@@ -9431,7 +9823,7 @@
     4405   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1819; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4406   __Pyx_GOTREF(__pyx_t_2);
     4407   __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1819; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4408-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     4409+  __Pyx_GOTREF(__pyx_t_3);
     4410   __Pyx_INCREF(((PyObject *)__pyx_v_ononc));
     4411   PyTuple_SET_ITEM(__pyx_t_3, 0, ((PyObject *)__pyx_v_ononc));
     4412   __Pyx_GIVEREF(((PyObject *)__pyx_v_ononc));
     4413@@ -9443,7 +9835,7 @@
     4414   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     4415   __Pyx_DECREF(((PyObject *)__pyx_t_3)); __pyx_t_3 = 0;
     4416   __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1819; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4417-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     4418+  __Pyx_GOTREF(__pyx_t_3);
     4419   PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_2);
     4420   __Pyx_GIVEREF(__pyx_t_2);
     4421   __pyx_t_2 = 0;
     4422@@ -9467,9 +9859,9 @@
     4423     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4424     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4425     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1820; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4426-    goto __pyx_L12;
     4427+    goto __pyx_L9;
     4428   }
     4429-  __pyx_L12:;
     4430+  __pyx_L9:;
     4431 
     4432   /* "mtrand.pyx":1821
     4433  *         if np.any(np.less(ononc, 0.0)):
     4434@@ -9487,7 +9879,7 @@
     4435  *
     4436  *     def chisquare(self, df, size=None):
     4437  */
     4438-  __pyx_t_2 = __pyx_f_6mtrand_cont3_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_noncentral_f, __pyx_v_size, __pyx_v_odfnum, __pyx_v_odfden, __pyx_v_ononc); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1821; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4439+  __pyx_t_2 = __pyx_f_6mtrand_cont3_array(__pyx_v_self->internal_state, rk_noncentral_f, __pyx_v_size, __pyx_v_odfnum, __pyx_v_odfden, __pyx_v_ononc); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1821; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4440   __Pyx_GOTREF(__pyx_t_2);
     4441   __pyx_r = __pyx_t_2;
     4442   __pyx_t_2 = 0;
     4443@@ -9511,49 +9903,40 @@
     4444   return __pyx_r;
     4445 }
     4446 
     4447-/* "mtrand.pyx":1824
     4448+/* Python wrapper */
     4449+static PyObject *__pyx_pw_6mtrand_11RandomState_51chisquare(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     4450+static char __pyx_doc_6mtrand_11RandomState_50chisquare[] = "\n        chisquare(df, size=None)\n\n        Draw samples from a chi-square distribution.\n\n        When `df` independent random variables, each with standard normal\n        distributions (mean 0, variance 1), are squared and summed, the\n        resulting distribution is chi-square (see Notes).  This distribution\n        is often used in hypothesis testing.\n\n        Parameters\n        ----------\n        df : int\n             Number of degrees of freedom.\n        size : tuple of ints, int, optional\n             Size of the returned array.  By default, a scalar is\n             returned.\n\n        Returns\n        -------\n        output : ndarray\n            Samples drawn from the distribution, packed in a `size`-shaped\n            array.\n\n        Raises\n        ------\n        ValueError\n            When `df` <= 0 or when an inappropriate `size` (e.g. ``size=-1``)\n            is given.\n\n        Notes\n        -----\n        The variable obtained by summing the squares of `df` independent,\n        standard normally distributed random variables:\n\n        .. math:: Q = \\sum_{i=0}^{\\mathtt{df}} X^2_i\n\n        is chi-square distributed, denoted\n\n        .. math:: Q \\sim \\chi^2_k.\n\n        The probability density function of the chi-squared distribution is\n\n        .. math:: p(x) = \\frac{(1/2)^{k/2}}{\\Gamma(k/2)}\n                         x^{k/2 - 1} e^{-x/2},\n\n        where :math:`\\Gamma` is the gamma function,\n\n        .. math:: \\Gamma(x) = \\int_0^{-\\infty} t^{x - 1} e^{-t} dt.\n\n        References\n        ----------\n        `NIST/SEMATECH e-Handbook of Statistical Methods\n        <http://www.itl.nist.gov/div898/handbook/eda/section3/eda3666.htm>`_\n\n        Examples\n        --------\n        >>> np.random.chisquare(2,4)\n        array([ 1.89920014,  9.00867716,  3.13710533,  5.62318272])\n\n        ";
     4451+static PyObject *__pyx_pw_6mtrand_11RandomState_51chisquare(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     4452+  PyObject *__pyx_v_df = 0;
     4453+  PyObject *__pyx_v_size = 0;
     4454+  PyObject *__pyx_r = 0;
     4455+  __Pyx_RefNannyDeclarations
     4456+  __Pyx_RefNannySetupContext("chisquare (wrapper)", 0);
     4457+  {
     4458+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__df,&__pyx_n_s__size,0};
     4459+    PyObject* values[2] = {0,0};
     4460+
     4461+    /* "mtrand.pyx":1824
     4462  *             odfden, ononc)
     4463  *
     4464  *     def chisquare(self, df, size=None):             # <<<<<<<<<<<<<<
     4465  *         """
     4466  *         chisquare(df, size=None)
     4467  */
     4468-
     4469-static PyObject *__pyx_pf_6mtrand_11RandomState_25chisquare(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     4470-static char __pyx_doc_6mtrand_11RandomState_25chisquare[] = "\n        chisquare(df, size=None)\n\n        Draw samples from a chi-square distribution.\n\n        When `df` independent random variables, each with standard normal\n        distributions (mean 0, variance 1), are squared and summed, the\n        resulting distribution is chi-square (see Notes).  This distribution\n        is often used in hypothesis testing.\n\n        Parameters\n        ----------\n        df : int\n             Number of degrees of freedom.\n        size : tuple of ints, int, optional\n             Size of the returned array.  By default, a scalar is\n             returned.\n\n        Returns\n        -------\n        output : ndarray\n            Samples drawn from the distribution, packed in a `size`-shaped\n            array.\n\n        Raises\n        ------\n        ValueError\n            When `df` <= 0 or when an inappropriate `size` (e.g. ``size=-1``)\n            is given.\n\n        Notes\n        -----\n        The variable obtained by summing the squares of `df` independent,\n        standard normally distributed random variables:\n\n        .. math:: Q = \\sum_{i=0}^{\\mathtt{df}} X^2_i\n\n        is chi-square distributed, denoted\n\n        .. math:: Q \\sim \\chi^2_k.\n\n        The probability density function of the chi-squared distribution is\n\n        .. math:: p(x) = \\frac{(1/2)^{k/2}}{\\Gamma(k/2)}\n                         x^{k/2 - 1} e^{-x/2},\n\n        where :math:`\\Gamma` is the gamma function,\n\n        .. math:: \\Gamma(x) = \\int_0^{-\\infty} t^{x - 1} e^{-t} dt.\n\n        References\n        ----------\n        `NIST/SEMATECH e-Handbook of Statistical Methods\n        <http://www.itl.nist.gov/div898/handbook/eda/section3/eda3666.htm>`_\n\n        Examples\n        --------\n        >>> np.random.chisquare(2,4)\n        array([ 1.89920014,  9.00867716,  3.13710533,  5.62318272])\n\n        ";
     4471-static PyObject *__pyx_pf_6mtrand_11RandomState_25chisquare(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     4472-  PyObject *__pyx_v_df = 0;
     4473-  PyObject *__pyx_v_size = 0;
     4474-  PyArrayObject *__pyx_v_odf = 0;
     4475-  double __pyx_v_fdf;
     4476-  PyObject *__pyx_r = NULL;
     4477-  __Pyx_RefNannyDeclarations
     4478-  int __pyx_t_1;
     4479-  PyObject *__pyx_t_2 = NULL;
     4480-  PyObject *__pyx_t_3 = NULL;
     4481-  PyObject *__pyx_t_4 = NULL;
     4482-  PyObject *__pyx_t_5 = NULL;
     4483-  int __pyx_lineno = 0;
     4484-  const char *__pyx_filename = NULL;
     4485-  int __pyx_clineno = 0;
     4486-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__df,&__pyx_n_s__size,0};
     4487-  __Pyx_RefNannySetupContext("chisquare");
     4488-  {
     4489-    PyObject* values[2] = {0,0};
     4490     values[1] = ((PyObject *)Py_None);
     4491     if (unlikely(__pyx_kwds)) {
     4492       Py_ssize_t kw_args;
     4493-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     4494+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     4495+      switch (pos_args) {
     4496         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     4497         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     4498         case  0: break;
     4499         default: goto __pyx_L5_argtuple_error;
     4500       }
     4501       kw_args = PyDict_Size(__pyx_kwds);
     4502-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     4503+      switch (pos_args) {
     4504         case  0:
     4505-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__df);
     4506-        if (likely(values[0])) kw_args--;
     4507+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__df)) != 0)) kw_args--;
     4508         else goto __pyx_L5_argtuple_error;
     4509         case  1:
     4510         if (kw_args > 0) {
     4511@@ -9562,7 +9945,7 @@
     4512         }
     4513       }
     4514       if (unlikely(kw_args > 0)) {
     4515-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "chisquare") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1824; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4516+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "chisquare") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1824; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4517       }
     4518     } else {
     4519       switch (PyTuple_GET_SIZE(__pyx_args)) {
     4520@@ -9583,6 +9966,25 @@
     4521   __Pyx_RefNannyFinishContext();
     4522   return NULL;
     4523   __pyx_L4_argument_unpacking_done:;
     4524+  __pyx_r = __pyx_pf_6mtrand_11RandomState_50chisquare(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_df, __pyx_v_size);
     4525+  __Pyx_RefNannyFinishContext();
     4526+  return __pyx_r;
     4527+}
     4528+
     4529+static PyObject *__pyx_pf_6mtrand_11RandomState_50chisquare(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_df, PyObject *__pyx_v_size) {
     4530+  PyArrayObject *__pyx_v_odf = 0;
     4531+  double __pyx_v_fdf;
     4532+  PyObject *__pyx_r = NULL;
     4533+  __Pyx_RefNannyDeclarations
     4534+  int __pyx_t_1;
     4535+  PyObject *__pyx_t_2 = NULL;
     4536+  PyObject *__pyx_t_3 = NULL;
     4537+  PyObject *__pyx_t_4 = NULL;
     4538+  PyObject *__pyx_t_5 = NULL;
     4539+  int __pyx_lineno = 0;
     4540+  const char *__pyx_filename = NULL;
     4541+  int __pyx_clineno = 0;
     4542+  __Pyx_RefNannySetupContext("chisquare", 0);
     4543 
     4544   /* "mtrand.pyx":1889
     4545  *         cdef double fdf
     4546@@ -9625,9 +10027,9 @@
     4547       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4548       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4549       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1892; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4550-      goto __pyx_L7;
     4551+      goto __pyx_L4;
     4552     }
     4553-    __pyx_L7:;
     4554+    __pyx_L4:;
     4555 
     4556     /* "mtrand.pyx":1893
     4557  *             if fdf <= 0:
     4558@@ -9637,14 +10039,14 @@
     4559  *         PyErr_Clear()
     4560  */
     4561     __Pyx_XDECREF(__pyx_r);
     4562-    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_chisquare, __pyx_v_size, __pyx_v_fdf); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1893; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4563+    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(__pyx_v_self->internal_state, rk_chisquare, __pyx_v_size, __pyx_v_fdf); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1893; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4564     __Pyx_GOTREF(__pyx_t_2);
     4565     __pyx_r = __pyx_t_2;
     4566     __pyx_t_2 = 0;
     4567     goto __pyx_L0;
     4568-    goto __pyx_L6;
     4569+    goto __pyx_L3;
     4570   }
     4571-  __pyx_L6:;
     4572+  __pyx_L3:;
     4573 
     4574   /* "mtrand.pyx":1895
     4575  *             return cont1_array_sc(self.internal_state, rk_chisquare, size, fdf)
     4576@@ -9688,7 +10090,7 @@
     4577   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1898; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4578   __Pyx_GOTREF(__pyx_t_2);
     4579   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1898; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4580-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     4581+  __Pyx_GOTREF(__pyx_t_5);
     4582   __Pyx_INCREF(((PyObject *)__pyx_v_odf));
     4583   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_odf));
     4584   __Pyx_GIVEREF(((PyObject *)__pyx_v_odf));
     4585@@ -9700,7 +10102,7 @@
     4586   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     4587   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     4588   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1898; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4589-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     4590+  __Pyx_GOTREF(__pyx_t_5);
     4591   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     4592   __Pyx_GIVEREF(__pyx_t_2);
     4593   __pyx_t_2 = 0;
     4594@@ -9724,9 +10126,9 @@
     4595     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4596     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4597     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1899; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4598-    goto __pyx_L8;
     4599+    goto __pyx_L5;
     4600   }
     4601-  __pyx_L8:;
     4602+  __pyx_L5:;
     4603 
     4604   /* "mtrand.pyx":1900
     4605  *         if np.any(np.less_equal(odf, 0.0)):
     4606@@ -9736,7 +10138,7 @@
     4607  *     def noncentral_chisquare(self, df, nonc, size=None):
     4608  */
     4609   __Pyx_XDECREF(__pyx_r);
     4610-  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_chisquare, __pyx_v_size, __pyx_v_odf); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1900; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4611+  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(__pyx_v_self->internal_state, rk_chisquare, __pyx_v_size, __pyx_v_odf); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1900; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4612   __Pyx_GOTREF(__pyx_t_2);
     4613   __pyx_r = __pyx_t_2;
     4614   __pyx_t_2 = 0;
     4615@@ -9758,42 +10160,32 @@
     4616   return __pyx_r;
     4617 }
     4618 
     4619-/* "mtrand.pyx":1902
     4620- *         return cont1_array(self.internal_state, rk_chisquare, size, odf)
     4621- *
     4622- *     def noncentral_chisquare(self, df, nonc, size=None):             # <<<<<<<<<<<<<<
     4623- *         """
     4624- *         noncentral_chisquare(df, nonc, size=None)
     4625- */
     4626-
     4627-static PyObject *__pyx_pf_6mtrand_11RandomState_26noncentral_chisquare(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     4628-static char __pyx_doc_6mtrand_11RandomState_26noncentral_chisquare[] = "\n        noncentral_chisquare(df, nonc, size=None)\n\n        Draw samples from a noncentral chi-square distribution.\n\n        The noncentral :math:`\\chi^2` distribution is a generalisation of\n        the :math:`\\chi^2` distribution.\n\n        Parameters\n        ----------\n        df : int\n            Degrees of freedom, should be >= 1.\n        nonc : float\n            Non-centrality, should be > 0.\n        size : int or tuple of ints\n            Shape of the output.\n\n        Notes\n        -----\n        The probability density function for the noncentral Chi-square distribution\n        is\n\n        .. math:: P(x;df,nonc) = \\sum^{\\infty}_{i=0}\n                               \\frac{e^{-nonc/2}(nonc/2)^{i}}{i!}P_{Y_{df+2i}}(x),\n\n        where :math:`Y_{q}` is the Chi-square with q degrees of freedom.\n\n        In Delhi (2007), it is noted that the noncentral chi-square is useful in\n        bombing and coverage problems, the probability of killing the point target\n        given by the noncentral chi-squared distribution.\n\n        References\n        ----------\n        .. [1] Delhi, M.S. Holla, \"On a noncentral chi-square distribution in the\n               analysis of weapon systems effectiveness\", Metrika, Volume 15,\n               Number 1 / December, 1970.\n        .. [2] Wikipedia, \"Noncentral chi-square distribution\"\n               http://en.wikipedia.org/wiki/Noncentral_chi-square_distribution\n\n        Examples\n        --------\n        Draw values from the distribution and plot the histogram\n\n        >>> import matplotlib.pyplot as plt\n        >>> values = plt.hist(np.random.noncentral_chisquare(3, 20, 100000),\n        ...                   bins=200, normed=True)\n        >>> plt.show()\n\n        Draw values from a noncentral chisquare with very small noncentrality,\n        and compare to a chisquare.\n\n        >>> plt.figure()\n        >>> values = plt.hist(np.random.noncentral_chisquare(3, .0000001, 100000),\n     ""   ...                   bins=np.arange(0., 25, .1), normed=True)\n        >>> values2 = plt.hist(np.random.chisquare(3, 100000),\n        ...                    bins=np.arange(0., 25, .1), normed=True)\n        >>> plt.plot(values[1][0:-1], values[0]-values2[0], 'ob')\n        >>> plt.show()\n\n        Demonstrate how large values of non-centrality lead to a more symmetric\n        distribution.\n\n        >>> plt.figure()\n        >>> values = plt.hist(np.random.noncentral_chisquare(3, 20, 100000),\n        ...                   bins=200, normed=True)\n        >>> plt.show()\n\n        ";
     4629-static PyObject *__pyx_pf_6mtrand_11RandomState_26noncentral_chisquare(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     4630+/* Python wrapper */
     4631+static PyObject *__pyx_pw_6mtrand_11RandomState_53noncentral_chisquare(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     4632+static char __pyx_doc_6mtrand_11RandomState_52noncentral_chisquare[] = "\n        noncentral_chisquare(df, nonc, size=None)\n\n        Draw samples from a noncentral chi-square distribution.\n\n        The noncentral :math:`\\chi^2` distribution is a generalisation of\n        the :math:`\\chi^2` distribution.\n\n        Parameters\n        ----------\n        df : int\n            Degrees of freedom, should be >= 1.\n        nonc : float\n            Non-centrality, should be > 0.\n        size : int or tuple of ints\n            Shape of the output.\n\n        Notes\n        -----\n        The probability density function for the noncentral Chi-square distribution\n        is\n\n        .. math:: P(x;df,nonc) = \\sum^{\\infty}_{i=0}\n                               \\frac{e^{-nonc/2}(nonc/2)^{i}}{i!}P_{Y_{df+2i}}(x),\n\n        where :math:`Y_{q}` is the Chi-square with q degrees of freedom.\n\n        In Delhi (2007), it is noted that the noncentral chi-square is useful in\n        bombing and coverage problems, the probability of killing the point target\n        given by the noncentral chi-squared distribution.\n\n        References\n        ----------\n        .. [1] Delhi, M.S. Holla, \"On a noncentral chi-square distribution in the\n               analysis of weapon systems effectiveness\", Metrika, Volume 15,\n               Number 1 / December, 1970.\n        .. [2] Wikipedia, \"Noncentral chi-square distribution\"\n               http://en.wikipedia.org/wiki/Noncentral_chi-square_distribution\n\n        Examples\n        --------\n        Draw values from the distribution and plot the histogram\n\n        >>> import matplotlib.pyplot as plt\n        >>> values = plt.hist(np.random.noncentral_chisquare(3, 20, 100000),\n        ...                   bins=200, normed=True)\n        >>> plt.show()\n\n        Draw values from a noncentral chisquare with very small noncentrality,\n        and compare to a chisquare.\n\n        >>> plt.figure()\n        >>> values = plt.hist(np.random.noncentral_chisquare(3, .0000001, 100000),\n     ""   ...                   bins=np.arange(0., 25, .1), normed=True)\n        >>> values2 = plt.hist(np.random.chisquare(3, 100000),\n        ...                    bins=np.arange(0., 25, .1), normed=True)\n        >>> plt.plot(values[1][0:-1], values[0]-values2[0], 'ob')\n        >>> plt.show()\n\n        Demonstrate how large values of non-centrality lead to a more symmetric\n        distribution.\n\n        >>> plt.figure()\n        >>> values = plt.hist(np.random.noncentral_chisquare(3, 20, 100000),\n        ...                   bins=200, normed=True)\n        >>> plt.show()\n\n        ";
     4633+static PyObject *__pyx_pw_6mtrand_11RandomState_53noncentral_chisquare(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     4634   PyObject *__pyx_v_df = 0;
     4635   PyObject *__pyx_v_nonc = 0;
     4636   PyObject *__pyx_v_size = 0;
     4637-  PyArrayObject *__pyx_v_odf = 0;
     4638-  PyArrayObject *__pyx_v_ononc = 0;
     4639-  double __pyx_v_fdf;
     4640-  double __pyx_v_fnonc;
     4641-  PyObject *__pyx_r = NULL;
     4642+  PyObject *__pyx_r = 0;
     4643   __Pyx_RefNannyDeclarations
     4644-  int __pyx_t_1;
     4645-  PyObject *__pyx_t_2 = NULL;
     4646-  PyObject *__pyx_t_3 = NULL;
     4647-  PyObject *__pyx_t_4 = NULL;
     4648-  PyObject *__pyx_t_5 = NULL;
     4649-  int __pyx_lineno = 0;
     4650-  const char *__pyx_filename = NULL;
     4651-  int __pyx_clineno = 0;
     4652-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__df,&__pyx_n_s__nonc,&__pyx_n_s__size,0};
     4653-  __Pyx_RefNannySetupContext("noncentral_chisquare");
     4654+  __Pyx_RefNannySetupContext("noncentral_chisquare (wrapper)", 0);
     4655   {
     4656+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__df,&__pyx_n_s__nonc,&__pyx_n_s__size,0};
     4657     PyObject* values[3] = {0,0,0};
     4658+
     4659+    /* "mtrand.pyx":1902
     4660+ *         return cont1_array(self.internal_state, rk_chisquare, size, odf)
     4661+ *
     4662+ *     def noncentral_chisquare(self, df, nonc, size=None):             # <<<<<<<<<<<<<<
     4663+ *         """
     4664+ *         noncentral_chisquare(df, nonc, size=None)
     4665+ */
     4666     values[2] = ((PyObject *)Py_None);
     4667     if (unlikely(__pyx_kwds)) {
     4668       Py_ssize_t kw_args;
     4669-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     4670+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     4671+      switch (pos_args) {
     4672         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     4673         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     4674         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     4675@@ -9801,14 +10193,12 @@
     4676         default: goto __pyx_L5_argtuple_error;
     4677       }
     4678       kw_args = PyDict_Size(__pyx_kwds);
     4679-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     4680+      switch (pos_args) {
     4681         case  0:
     4682-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__df);
     4683-        if (likely(values[0])) kw_args--;
     4684+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__df)) != 0)) kw_args--;
     4685         else goto __pyx_L5_argtuple_error;
     4686         case  1:
     4687-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__nonc);
     4688-        if (likely(values[1])) kw_args--;
     4689+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__nonc)) != 0)) kw_args--;
     4690         else {
     4691           __Pyx_RaiseArgtupleInvalid("noncentral_chisquare", 0, 2, 3, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1902; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4692         }
     4693@@ -9819,7 +10209,7 @@
     4694         }
     4695       }
     4696       if (unlikely(kw_args > 0)) {
     4697-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "noncentral_chisquare") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1902; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4698+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "noncentral_chisquare") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1902; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4699       }
     4700     } else {
     4701       switch (PyTuple_GET_SIZE(__pyx_args)) {
     4702@@ -9842,6 +10232,27 @@
     4703   __Pyx_RefNannyFinishContext();
     4704   return NULL;
     4705   __pyx_L4_argument_unpacking_done:;
     4706+  __pyx_r = __pyx_pf_6mtrand_11RandomState_52noncentral_chisquare(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_df, __pyx_v_nonc, __pyx_v_size);
     4707+  __Pyx_RefNannyFinishContext();
     4708+  return __pyx_r;
     4709+}
     4710+
     4711+static PyObject *__pyx_pf_6mtrand_11RandomState_52noncentral_chisquare(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_df, PyObject *__pyx_v_nonc, PyObject *__pyx_v_size) {
     4712+  PyArrayObject *__pyx_v_odf = 0;
     4713+  PyArrayObject *__pyx_v_ononc = 0;
     4714+  double __pyx_v_fdf;
     4715+  double __pyx_v_fnonc;
     4716+  PyObject *__pyx_r = NULL;
     4717+  __Pyx_RefNannyDeclarations
     4718+  int __pyx_t_1;
     4719+  PyObject *__pyx_t_2 = NULL;
     4720+  PyObject *__pyx_t_3 = NULL;
     4721+  PyObject *__pyx_t_4 = NULL;
     4722+  PyObject *__pyx_t_5 = NULL;
     4723+  int __pyx_lineno = 0;
     4724+  const char *__pyx_filename = NULL;
     4725+  int __pyx_clineno = 0;
     4726+  __Pyx_RefNannySetupContext("noncentral_chisquare", 0);
     4727 
     4728   /* "mtrand.pyx":1973
     4729  *         cdef ndarray odf, ononc
     4730@@ -9893,9 +10304,9 @@
     4731       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4732       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4733       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1977; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4734-      goto __pyx_L7;
     4735+      goto __pyx_L4;
     4736     }
     4737-    __pyx_L7:;
     4738+    __pyx_L4:;
     4739 
     4740     /* "mtrand.pyx":1978
     4741  *             if fdf <= 1:
     4742@@ -9919,9 +10330,9 @@
     4743       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4744       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4745       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1979; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4746-      goto __pyx_L8;
     4747+      goto __pyx_L5;
     4748     }
     4749-    __pyx_L8:;
     4750+    __pyx_L5:;
     4751 
     4752     /* "mtrand.pyx":1980
     4753  *             if fnonc <= 0:
     4754@@ -9939,14 +10350,14 @@
     4755  *
     4756  *         PyErr_Clear()
     4757  */
     4758-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_noncentral_chisquare, __pyx_v_size, __pyx_v_fdf, __pyx_v_fnonc); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1980; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4759+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_noncentral_chisquare, __pyx_v_size, __pyx_v_fdf, __pyx_v_fnonc); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1980; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4760     __Pyx_GOTREF(__pyx_t_2);
     4761     __pyx_r = __pyx_t_2;
     4762     __pyx_t_2 = 0;
     4763     goto __pyx_L0;
     4764-    goto __pyx_L6;
     4765+    goto __pyx_L3;
     4766   }
     4767-  __pyx_L6:;
     4768+  __pyx_L3:;
     4769 
     4770   /* "mtrand.pyx":1983
     4771  *                                   size, fdf, fnonc)
     4772@@ -10003,7 +10414,7 @@
     4773   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1987; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4774   __Pyx_GOTREF(__pyx_t_2);
     4775   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1987; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4776-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     4777+  __Pyx_GOTREF(__pyx_t_5);
     4778   __Pyx_INCREF(((PyObject *)__pyx_v_odf));
     4779   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_odf));
     4780   __Pyx_GIVEREF(((PyObject *)__pyx_v_odf));
     4781@@ -10015,7 +10426,7 @@
     4782   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     4783   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     4784   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1987; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4785-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     4786+  __Pyx_GOTREF(__pyx_t_5);
     4787   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     4788   __Pyx_GIVEREF(__pyx_t_2);
     4789   __pyx_t_2 = 0;
     4790@@ -10039,9 +10450,9 @@
     4791     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4792     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4793     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1988; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4794-    goto __pyx_L9;
     4795+    goto __pyx_L6;
     4796   }
     4797-  __pyx_L9:;
     4798+  __pyx_L6:;
     4799 
     4800   /* "mtrand.pyx":1989
     4801  *         if np.any(np.less_equal(odf, 0.0)):
     4802@@ -10063,7 +10474,7 @@
     4803   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1989; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4804   __Pyx_GOTREF(__pyx_t_2);
     4805   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1989; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4806-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     4807+  __Pyx_GOTREF(__pyx_t_4);
     4808   __Pyx_INCREF(((PyObject *)__pyx_v_ononc));
     4809   PyTuple_SET_ITEM(__pyx_t_4, 0, ((PyObject *)__pyx_v_ononc));
     4810   __Pyx_GIVEREF(((PyObject *)__pyx_v_ononc));
     4811@@ -10075,7 +10486,7 @@
     4812   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     4813   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     4814   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1989; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4815-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     4816+  __Pyx_GOTREF(__pyx_t_4);
     4817   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     4818   __Pyx_GIVEREF(__pyx_t_2);
     4819   __pyx_t_2 = 0;
     4820@@ -10099,9 +10510,9 @@
     4821     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     4822     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     4823     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1990; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4824-    goto __pyx_L10;
     4825+    goto __pyx_L7;
     4826   }
     4827-  __pyx_L10:;
     4828+  __pyx_L7:;
     4829 
     4830   /* "mtrand.pyx":1991
     4831  *         if np.any(np.less_equal(ononc, 0.0)):
     4832@@ -10119,7 +10530,7 @@
     4833  *
     4834  *     def standard_cauchy(self, size=None):
     4835  */
     4836-  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_noncentral_chisquare, __pyx_v_size, __pyx_v_odf, __pyx_v_ononc); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1991; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4837+  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_noncentral_chisquare, __pyx_v_size, __pyx_v_odf, __pyx_v_ononc); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1991; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4838   __Pyx_GOTREF(__pyx_t_2);
     4839   __pyx_r = __pyx_t_2;
     4840   __pyx_t_2 = 0;
     4841@@ -10142,38 +10553,36 @@
     4842   return __pyx_r;
     4843 }
     4844 
     4845-/* "mtrand.pyx":1994
     4846+/* Python wrapper */
     4847+static PyObject *__pyx_pw_6mtrand_11RandomState_55standard_cauchy(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     4848+static char __pyx_doc_6mtrand_11RandomState_54standard_cauchy[] = "\n        standard_cauchy(size=None)\n\n        Standard Cauchy distribution with mode = 0.\n\n        Also known as the Lorentz distribution.\n\n        Parameters\n        ----------\n        size : int or tuple of ints\n            Shape of the output.\n\n        Returns\n        -------\n        samples : ndarray or scalar\n            The drawn samples.\n\n        Notes\n        -----\n        The probability density function for the full Cauchy distribution is\n\n        .. math:: P(x; x_0, \\gamma) = \\frac{1}{\\pi \\gamma \\bigl[ 1+\n                  (\\frac{x-x_0}{\\gamma})^2 \\bigr] }\n\n        and the Standard Cauchy distribution just sets :math:`x_0=0` and\n        :math:`\\gamma=1`\n\n        The Cauchy distribution arises in the solution to the driven harmonic\n        oscillator problem, and also describes spectral line broadening. It\n        also describes the distribution of values at which a line tilted at\n        a random angle will cut the x axis.\n\n        When studying hypothesis tests that assume normality, seeing how the\n        tests perform on data from a Cauchy distribution is a good indicator of\n        their sensitivity to a heavy-tailed distribution, since the Cauchy looks\n        very much like a Gaussian distribution, but with heavier tails.\n\n        References\n        ----------\n        ..[1] NIST/SEMATECH e-Handbook of Statistical Methods, \"Cauchy\n              Distribution\",\n              http://www.itl.nist.gov/div898/handbook/eda/section3/eda3663.htm\n        ..[2] Weisstein, Eric W. \"Cauchy Distribution.\" From MathWorld--A\n              Wolfram Web Resource.\n              http://mathworld.wolfram.com/CauchyDistribution.html\n        ..[3] Wikipedia, \"Cauchy distribution\"\n              http://en.wikipedia.org/wiki/Cauchy_distribution\n\n        Examples\n        --------\n        Draw samples and plot the distribution:\n\n        >>> s = np.random.standard_cauchy(1000000)\n        >>> s = s[(s>-25) & (s<25)""]  # truncate distribution so it plots well\n        >>> plt.hist(s, bins=100)\n        >>> plt.show()\n\n        ";
     4849+static PyObject *__pyx_pw_6mtrand_11RandomState_55standard_cauchy(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     4850+  PyObject *__pyx_v_size = 0;
     4851+  PyObject *__pyx_r = 0;
     4852+  __Pyx_RefNannyDeclarations
     4853+  __Pyx_RefNannySetupContext("standard_cauchy (wrapper)", 0);
     4854+  {
     4855+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__size,0};
     4856+    PyObject* values[1] = {0};
     4857+
     4858+    /* "mtrand.pyx":1994
     4859  *             odf, ononc)
     4860  *
     4861  *     def standard_cauchy(self, size=None):             # <<<<<<<<<<<<<<
     4862  *         """
     4863  *         standard_cauchy(size=None)
     4864  */
     4865-
     4866-static PyObject *__pyx_pf_6mtrand_11RandomState_27standard_cauchy(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     4867-static char __pyx_doc_6mtrand_11RandomState_27standard_cauchy[] = "\n        standard_cauchy(size=None)\n\n        Standard Cauchy distribution with mode = 0.\n\n        Also known as the Lorentz distribution.\n\n        Parameters\n        ----------\n        size : int or tuple of ints\n            Shape of the output.\n\n        Returns\n        -------\n        samples : ndarray or scalar\n            The drawn samples.\n\n        Notes\n        -----\n        The probability density function for the full Cauchy distribution is\n\n        .. math:: P(x; x_0, \\gamma) = \\frac{1}{\\pi \\gamma \\bigl[ 1+\n                  (\\frac{x-x_0}{\\gamma})^2 \\bigr] }\n\n        and the Standard Cauchy distribution just sets :math:`x_0=0` and\n        :math:`\\gamma=1`\n\n        The Cauchy distribution arises in the solution to the driven harmonic\n        oscillator problem, and also describes spectral line broadening. It\n        also describes the distribution of values at which a line tilted at\n        a random angle will cut the x axis.\n\n        When studying hypothesis tests that assume normality, seeing how the\n        tests perform on data from a Cauchy distribution is a good indicator of\n        their sensitivity to a heavy-tailed distribution, since the Cauchy looks\n        very much like a Gaussian distribution, but with heavier tails.\n\n        References\n        ----------\n        ..[1] NIST/SEMATECH e-Handbook of Statistical Methods, \"Cauchy\n              Distribution\",\n              http://www.itl.nist.gov/div898/handbook/eda/section3/eda3663.htm\n        ..[2] Weisstein, Eric W. \"Cauchy Distribution.\" From MathWorld--A\n              Wolfram Web Resource.\n              http://mathworld.wolfram.com/CauchyDistribution.html\n        ..[3] Wikipedia, \"Cauchy distribution\"\n              http://en.wikipedia.org/wiki/Cauchy_distribution\n\n        Examples\n        --------\n        Draw samples and plot the distribution:\n\n        >>> s = np.random.standard_cauchy(1000000)\n        >>> s = s[(s>-25) & (s<25)""]  # truncate distribution so it plots well\n        >>> plt.hist(s, bins=100)\n        >>> plt.show()\n\n        ";
     4868-static PyObject *__pyx_pf_6mtrand_11RandomState_27standard_cauchy(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     4869-  PyObject *__pyx_v_size = 0;
     4870-  PyObject *__pyx_r = NULL;
     4871-  __Pyx_RefNannyDeclarations
     4872-  PyObject *__pyx_t_1 = NULL;
     4873-  int __pyx_lineno = 0;
     4874-  const char *__pyx_filename = NULL;
     4875-  int __pyx_clineno = 0;
     4876-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__size,0};
     4877-  __Pyx_RefNannySetupContext("standard_cauchy");
     4878-  {
     4879-    PyObject* values[1] = {0};
     4880     values[0] = ((PyObject *)Py_None);
     4881     if (unlikely(__pyx_kwds)) {
     4882       Py_ssize_t kw_args;
     4883-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     4884+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     4885+      switch (pos_args) {
     4886         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     4887         case  0: break;
     4888         default: goto __pyx_L5_argtuple_error;
     4889       }
     4890       kw_args = PyDict_Size(__pyx_kwds);
     4891-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     4892+      switch (pos_args) {
     4893         case  0:
     4894         if (kw_args > 0) {
     4895           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__size);
     4896@@ -10181,7 +10590,7 @@
     4897         }
     4898       }
     4899       if (unlikely(kw_args > 0)) {
     4900-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "standard_cauchy") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1994; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4901+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "standard_cauchy") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1994; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     4902       }
     4903     } else {
     4904       switch (PyTuple_GET_SIZE(__pyx_args)) {
     4905@@ -10200,6 +10609,19 @@
     4906   __Pyx_RefNannyFinishContext();
     4907   return NULL;
     4908   __pyx_L4_argument_unpacking_done:;
     4909+  __pyx_r = __pyx_pf_6mtrand_11RandomState_54standard_cauchy(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_size);
     4910+  __Pyx_RefNannyFinishContext();
     4911+  return __pyx_r;
     4912+}
     4913+
     4914+static PyObject *__pyx_pf_6mtrand_11RandomState_54standard_cauchy(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_size) {
     4915+  PyObject *__pyx_r = NULL;
     4916+  __Pyx_RefNannyDeclarations
     4917+  PyObject *__pyx_t_1 = NULL;
     4918+  int __pyx_lineno = 0;
     4919+  const char *__pyx_filename = NULL;
     4920+  int __pyx_clineno = 0;
     4921+  __Pyx_RefNannySetupContext("standard_cauchy", 0);
     4922 
     4923   /* "mtrand.pyx":2053
     4924  *
     4925@@ -10209,7 +10631,7 @@
     4926  *     def standard_t(self, df, size=None):
     4927  */
     4928   __Pyx_XDECREF(__pyx_r);
     4929-  __pyx_t_1 = __pyx_f_6mtrand_cont0_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_standard_cauchy, __pyx_v_size); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2053; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4930+  __pyx_t_1 = __pyx_f_6mtrand_cont0_array(__pyx_v_self->internal_state, rk_standard_cauchy, __pyx_v_size); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2053; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     4931   __Pyx_GOTREF(__pyx_t_1);
     4932   __pyx_r = __pyx_t_1;
     4933   __pyx_t_1 = 0;
     4934@@ -10227,49 +10649,40 @@
     4935   return __pyx_r;
     4936 }
     4937 
     4938-/* "mtrand.pyx":2055
     4939+/* Python wrapper */
     4940+static PyObject *__pyx_pw_6mtrand_11RandomState_57standard_t(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     4941+static char __pyx_doc_6mtrand_11RandomState_56standard_t[] = "\n        standard_t(df, size=None)\n\n        Standard Student's t distribution with df degrees of freedom.\n\n        A special case of the hyperbolic distribution.\n        As `df` gets large, the result resembles that of the standard normal\n        distribution (`standard_normal`).\n\n        Parameters\n        ----------\n        df : int\n            Degrees of freedom, should be > 0.\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single value is\n            returned.\n\n        Returns\n        -------\n        samples : ndarray or scalar\n            Drawn samples.\n\n        Notes\n        -----\n        The probability density function for the t distribution is\n\n        .. math:: P(x, df) = \\frac{\\Gamma(\\frac{df+1}{2})}{\\sqrt{\\pi df}\n                  \\Gamma(\\frac{df}{2})}\\Bigl( 1+\\frac{x^2}{df} \\Bigr)^{-(df+1)/2}\n\n        The t test is based on an assumption that the data come from a Normal\n        distribution. The t test provides a way to test whether the sample mean\n        (that is the mean calculated from the data) is a good estimate of the true\n        mean.\n\n        The derivation of the t-distribution was forst published in 1908 by William\n        Gisset while working for the Guinness Brewery in Dublin. Due to proprietary\n        issues, he had to publish under a pseudonym, and so he used the name\n        Student.\n\n        References\n        ----------\n        .. [1] Dalgaard, Peter, \"Introductory Statistics With R\",\n               Springer, 2002.\n        .. [2] Wikipedia, \"Student's t-distribution\"\n               http://en.wikipedia.org/wiki/Student's_t-distribution\n\n        Examples\n        --------\n        From Dalgaard page 83 [1]_, suppose the daily energy intake for 11\n        women in Kj is:\n\n        >>> intake = np.array([5260., 5470, 5640, 6180, 6390, 6515, 6805, 7515, \\\n        ...                    7515, 8230, 8770])\n\n        Doe""s their energy intake deviate systematically from the recommended\n        value of 7725 kJ?\n\n        We have 10 degrees of freedom, so is the sample mean within 95% of the\n        recommended value?\n\n        >>> s = np.random.standard_t(10, size=100000)\n        >>> np.mean(intake)\n        6753.636363636364\n        >>> intake.std(ddof=1)\n        1142.1232221373727\n\n        Calculate the t statistic, setting the ddof parameter to the unbiased\n        value so the divisor in the standard deviation will be degrees of\n        freedom, N-1.\n\n        >>> t = (np.mean(intake)-7725)/(intake.std(ddof=1)/np.sqrt(len(intake)))\n        >>> import matplotlib.pyplot as plt\n        >>> h = plt.hist(s, bins=100, normed=True)\n\n        For a one-sided t-test, how far out in the distribution does the t\n        statistic appear?\n\n        >>> >>> np.sum(s<t) / float(len(s))\n        0.0090699999999999999  #random\n\n        So the p-value is about 0.009, which says the null hypothesis has a\n        probability of about 99% of being true.\n\n        ";
     4942+static PyObject *__pyx_pw_6mtrand_11RandomState_57standard_t(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     4943+  PyObject *__pyx_v_df = 0;
     4944+  PyObject *__pyx_v_size = 0;
     4945+  PyObject *__pyx_r = 0;
     4946+  __Pyx_RefNannyDeclarations
     4947+  __Pyx_RefNannySetupContext("standard_t (wrapper)", 0);
     4948+  {
     4949+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__df,&__pyx_n_s__size,0};
     4950+    PyObject* values[2] = {0,0};
     4951+
     4952+    /* "mtrand.pyx":2055
     4953  *         return cont0_array(self.internal_state, rk_standard_cauchy, size)
     4954  *
     4955  *     def standard_t(self, df, size=None):             # <<<<<<<<<<<<<<
     4956  *         """
     4957  *         standard_t(df, size=None)
     4958  */
     4959-
     4960-static PyObject *__pyx_pf_6mtrand_11RandomState_28standard_t(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     4961-static char __pyx_doc_6mtrand_11RandomState_28standard_t[] = "\n        standard_t(df, size=None)\n\n        Standard Student's t distribution with df degrees of freedom.\n\n        A special case of the hyperbolic distribution.\n        As `df` gets large, the result resembles that of the standard normal\n        distribution (`standard_normal`).\n\n        Parameters\n        ----------\n        df : int\n            Degrees of freedom, should be > 0.\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single value is\n            returned.\n\n        Returns\n        -------\n        samples : ndarray or scalar\n            Drawn samples.\n\n        Notes\n        -----\n        The probability density function for the t distribution is\n\n        .. math:: P(x, df) = \\frac{\\Gamma(\\frac{df+1}{2})}{\\sqrt{\\pi df}\n                  \\Gamma(\\frac{df}{2})}\\Bigl( 1+\\frac{x^2}{df} \\Bigr)^{-(df+1)/2}\n\n        The t test is based on an assumption that the data come from a Normal\n        distribution. The t test provides a way to test whether the sample mean\n        (that is the mean calculated from the data) is a good estimate of the true\n        mean.\n\n        The derivation of the t-distribution was forst published in 1908 by William\n        Gisset while working for the Guinness Brewery in Dublin. Due to proprietary\n        issues, he had to publish under a pseudonym, and so he used the name\n        Student.\n\n        References\n        ----------\n        .. [1] Dalgaard, Peter, \"Introductory Statistics With R\",\n               Springer, 2002.\n        .. [2] Wikipedia, \"Student's t-distribution\"\n               http://en.wikipedia.org/wiki/Student's_t-distribution\n\n        Examples\n        --------\n        From Dalgaard page 83 [1]_, suppose the daily energy intake for 11\n        women in Kj is:\n\n        >>> intake = np.array([5260., 5470, 5640, 6180, 6390, 6515, 6805, 7515, \\\n        ...                    7515, 8230, 8770])\n\n        Doe""s their energy intake deviate systematically from the recommended\n        value of 7725 kJ?\n\n        We have 10 degrees of freedom, so is the sample mean within 95% of the\n        recommended value?\n\n        >>> s = np.random.standard_t(10, size=100000)\n        >>> np.mean(intake)\n        6753.636363636364\n        >>> intake.std(ddof=1)\n        1142.1232221373727\n\n        Calculate the t statistic, setting the ddof parameter to the unbiased\n        value so the divisor in the standard deviation will be degrees of\n        freedom, N-1.\n\n        >>> t = (np.mean(intake)-7725)/(intake.std(ddof=1)/np.sqrt(len(intake)))\n        >>> import matplotlib.pyplot as plt\n        >>> h = plt.hist(s, bins=100, normed=True)\n\n        For a one-sided t-test, how far out in the distribution does the t\n        statistic appear?\n\n        >>> >>> np.sum(s<t) / float(len(s))\n        0.0090699999999999999  #random\n\n        So the p-value is about 0.009, which says the null hypothesis has a\n        probability of about 99% of being true.\n\n        ";
     4962-static PyObject *__pyx_pf_6mtrand_11RandomState_28standard_t(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     4963-  PyObject *__pyx_v_df = 0;
     4964-  PyObject *__pyx_v_size = 0;
     4965-  PyArrayObject *__pyx_v_odf = 0;
     4966-  double __pyx_v_fdf;
     4967-  PyObject *__pyx_r = NULL;
     4968-  __Pyx_RefNannyDeclarations
     4969-  int __pyx_t_1;
     4970-  PyObject *__pyx_t_2 = NULL;
     4971-  PyObject *__pyx_t_3 = NULL;
     4972-  PyObject *__pyx_t_4 = NULL;
     4973-  PyObject *__pyx_t_5 = NULL;
     4974-  int __pyx_lineno = 0;
     4975-  const char *__pyx_filename = NULL;
     4976-  int __pyx_clineno = 0;
     4977-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__df,&__pyx_n_s__size,0};
     4978-  __Pyx_RefNannySetupContext("standard_t");
     4979-  {
     4980-    PyObject* values[2] = {0,0};
     4981     values[1] = ((PyObject *)Py_None);
     4982     if (unlikely(__pyx_kwds)) {
     4983       Py_ssize_t kw_args;
     4984-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     4985+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     4986+      switch (pos_args) {
     4987         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     4988         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     4989         case  0: break;
     4990         default: goto __pyx_L5_argtuple_error;
     4991       }
     4992       kw_args = PyDict_Size(__pyx_kwds);
     4993-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     4994+      switch (pos_args) {
     4995         case  0:
     4996-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__df);
     4997-        if (likely(values[0])) kw_args--;
     4998+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__df)) != 0)) kw_args--;
     4999         else goto __pyx_L5_argtuple_error;
     5000         case  1:
     5001         if (kw_args > 0) {
     5002@@ -10278,7 +10691,7 @@
     5003         }
     5004       }
     5005       if (unlikely(kw_args > 0)) {
     5006-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "standard_t") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2055; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5007+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "standard_t") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2055; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5008       }
     5009     } else {
     5010       switch (PyTuple_GET_SIZE(__pyx_args)) {
     5011@@ -10299,6 +10712,25 @@
     5012   __Pyx_RefNannyFinishContext();
     5013   return NULL;
     5014   __pyx_L4_argument_unpacking_done:;
     5015+  __pyx_r = __pyx_pf_6mtrand_11RandomState_56standard_t(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_df, __pyx_v_size);
     5016+  __Pyx_RefNannyFinishContext();
     5017+  return __pyx_r;
     5018+}
     5019+
     5020+static PyObject *__pyx_pf_6mtrand_11RandomState_56standard_t(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_df, PyObject *__pyx_v_size) {
     5021+  PyArrayObject *__pyx_v_odf = 0;
     5022+  double __pyx_v_fdf;
     5023+  PyObject *__pyx_r = NULL;
     5024+  __Pyx_RefNannyDeclarations
     5025+  int __pyx_t_1;
     5026+  PyObject *__pyx_t_2 = NULL;
     5027+  PyObject *__pyx_t_3 = NULL;
     5028+  PyObject *__pyx_t_4 = NULL;
     5029+  PyObject *__pyx_t_5 = NULL;
     5030+  int __pyx_lineno = 0;
     5031+  const char *__pyx_filename = NULL;
     5032+  int __pyx_clineno = 0;
     5033+  __Pyx_RefNannySetupContext("standard_t", 0);
     5034 
     5035   /* "mtrand.pyx":2143
     5036  *         cdef double fdf
     5037@@ -10341,9 +10773,9 @@
     5038       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5039       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5040       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2146; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5041-      goto __pyx_L7;
     5042+      goto __pyx_L4;
     5043     }
     5044-    __pyx_L7:;
     5045+    __pyx_L4:;
     5046 
     5047     /* "mtrand.pyx":2147
     5048  *             if fdf <= 0:
     5049@@ -10353,14 +10785,14 @@
     5050  *         PyErr_Clear()
     5051  */
     5052     __Pyx_XDECREF(__pyx_r);
     5053-    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_standard_t, __pyx_v_size, __pyx_v_fdf); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2147; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5054+    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(__pyx_v_self->internal_state, rk_standard_t, __pyx_v_size, __pyx_v_fdf); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2147; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5055     __Pyx_GOTREF(__pyx_t_2);
     5056     __pyx_r = __pyx_t_2;
     5057     __pyx_t_2 = 0;
     5058     goto __pyx_L0;
     5059-    goto __pyx_L6;
     5060+    goto __pyx_L3;
     5061   }
     5062-  __pyx_L6:;
     5063+  __pyx_L3:;
     5064 
     5065   /* "mtrand.pyx":2149
     5066  *             return cont1_array_sc(self.internal_state, rk_standard_t, size, fdf)
     5067@@ -10404,7 +10836,7 @@
     5068   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2152; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5069   __Pyx_GOTREF(__pyx_t_2);
     5070   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2152; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5071-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5072+  __Pyx_GOTREF(__pyx_t_5);
     5073   __Pyx_INCREF(((PyObject *)__pyx_v_odf));
     5074   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_odf));
     5075   __Pyx_GIVEREF(((PyObject *)__pyx_v_odf));
     5076@@ -10416,7 +10848,7 @@
     5077   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     5078   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     5079   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2152; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5080-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5081+  __Pyx_GOTREF(__pyx_t_5);
     5082   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     5083   __Pyx_GIVEREF(__pyx_t_2);
     5084   __pyx_t_2 = 0;
     5085@@ -10440,9 +10872,9 @@
     5086     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5087     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5088     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2153; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5089-    goto __pyx_L8;
     5090+    goto __pyx_L5;
     5091   }
     5092-  __pyx_L8:;
     5093+  __pyx_L5:;
     5094 
     5095   /* "mtrand.pyx":2154
     5096  *         if np.any(np.less_equal(odf, 0.0)):
     5097@@ -10452,7 +10884,7 @@
     5098  *     def vonmises(self, mu, kappa, size=None):
     5099  */
     5100   __Pyx_XDECREF(__pyx_r);
     5101-  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_standard_t, __pyx_v_size, __pyx_v_odf); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2154; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5102+  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(__pyx_v_self->internal_state, rk_standard_t, __pyx_v_size, __pyx_v_odf); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2154; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5103   __Pyx_GOTREF(__pyx_t_2);
     5104   __pyx_r = __pyx_t_2;
     5105   __pyx_t_2 = 0;
     5106@@ -10474,42 +10906,32 @@
     5107   return __pyx_r;
     5108 }
     5109 
     5110-/* "mtrand.pyx":2156
     5111- *         return cont1_array(self.internal_state, rk_standard_t, size, odf)
     5112- *
     5113- *     def vonmises(self, mu, kappa, size=None):             # <<<<<<<<<<<<<<
     5114- *         """
     5115- *         vonmises(mu, kappa, size=None)
     5116- */
     5117-
     5118-static PyObject *__pyx_pf_6mtrand_11RandomState_29vonmises(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     5119-static char __pyx_doc_6mtrand_11RandomState_29vonmises[] = "\n        vonmises(mu, kappa, size=None)\n\n        Draw samples from a von Mises distribution.\n\n        Samples are drawn from a von Mises distribution with specified mode\n        (mu) and dispersion (kappa), on the interval [-pi, pi].\n\n        The von Mises distribution (also known as the circular normal\n        distribution) is a continuous probability distribution on the unit\n        circle.  It may be thought of as the circular analogue of the normal\n        distribution.\n\n        Parameters\n        ----------\n        mu : float\n            Mode (\"center\") of the distribution.\n        kappa : float\n            Dispersion of the distribution, has to be >=0.\n        size : int or tuple of int\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : scalar or ndarray\n            The returned samples, which are in the interval [-pi, pi].\n\n        See Also\n        --------\n        scipy.stats.distributions.vonmises : probability density function,\n            distribution, or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the von Mises distribution is\n\n        .. math:: p(x) = \\frac{e^{\\kappa cos(x-\\mu)}}{2\\pi I_0(\\kappa)},\n\n        where :math:`\\mu` is the mode and :math:`\\kappa` the dispersion,\n        and :math:`I_0(\\kappa)` is the modified Bessel function of order 0.\n\n        The von Mises is named for Richard Edler von Mises, who was born in\n        Austria-Hungary, in what is now the Ukraine.  He fled to the United\n        States in 1939 and became a professor at Harvard.  He worked in\n        probability theory, aerodynamics, fluid mechanics, and philosophy of\n        science.\n\n        References\n        ----------\n        Abramowitz, M. and Stegun, I. A. (ed.), *Handbook of Mathematical\n        Functions*, New York: Dover, 1965.\n\n      ""  von Mises, R., *Mathematical Theory of Probability and Statistics*,\n        New York: Academic Press, 1964.\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> mu, kappa = 0.0, 4.0 # mean and dispersion\n        >>> s = np.random.vonmises(mu, kappa, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> import scipy.special as sps\n        >>> count, bins, ignored = plt.hist(s, 50, normed=True)\n        >>> x = np.arange(-np.pi, np.pi, 2*np.pi/50.)\n        >>> y = -np.exp(kappa*np.cos(x-mu))/(2*np.pi*sps.jn(0,kappa))\n        >>> plt.plot(x, y/max(y), linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     5120-static PyObject *__pyx_pf_6mtrand_11RandomState_29vonmises(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     5121+/* Python wrapper */
     5122+static PyObject *__pyx_pw_6mtrand_11RandomState_59vonmises(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     5123+static char __pyx_doc_6mtrand_11RandomState_58vonmises[] = "\n        vonmises(mu, kappa, size=None)\n\n        Draw samples from a von Mises distribution.\n\n        Samples are drawn from a von Mises distribution with specified mode\n        (mu) and dispersion (kappa), on the interval [-pi, pi].\n\n        The von Mises distribution (also known as the circular normal\n        distribution) is a continuous probability distribution on the unit\n        circle.  It may be thought of as the circular analogue of the normal\n        distribution.\n\n        Parameters\n        ----------\n        mu : float\n            Mode (\"center\") of the distribution.\n        kappa : float\n            Dispersion of the distribution, has to be >=0.\n        size : int or tuple of int\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : scalar or ndarray\n            The returned samples, which are in the interval [-pi, pi].\n\n        See Also\n        --------\n        scipy.stats.distributions.vonmises : probability density function,\n            distribution, or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the von Mises distribution is\n\n        .. math:: p(x) = \\frac{e^{\\kappa cos(x-\\mu)}}{2\\pi I_0(\\kappa)},\n\n        where :math:`\\mu` is the mode and :math:`\\kappa` the dispersion,\n        and :math:`I_0(\\kappa)` is the modified Bessel function of order 0.\n\n        The von Mises is named for Richard Edler von Mises, who was born in\n        Austria-Hungary, in what is now the Ukraine.  He fled to the United\n        States in 1939 and became a professor at Harvard.  He worked in\n        probability theory, aerodynamics, fluid mechanics, and philosophy of\n        science.\n\n        References\n        ----------\n        Abramowitz, M. and Stegun, I. A. (ed.), *Handbook of Mathematical\n        Functions*, New York: Dover, 1965.\n\n      ""  von Mises, R., *Mathematical Theory of Probability and Statistics*,\n        New York: Academic Press, 1964.\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> mu, kappa = 0.0, 4.0 # mean and dispersion\n        >>> s = np.random.vonmises(mu, kappa, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> import scipy.special as sps\n        >>> count, bins, ignored = plt.hist(s, 50, normed=True)\n        >>> x = np.arange(-np.pi, np.pi, 2*np.pi/50.)\n        >>> y = -np.exp(kappa*np.cos(x-mu))/(2*np.pi*sps.jn(0,kappa))\n        >>> plt.plot(x, y/max(y), linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     5124+static PyObject *__pyx_pw_6mtrand_11RandomState_59vonmises(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     5125   PyObject *__pyx_v_mu = 0;
     5126   PyObject *__pyx_v_kappa = 0;
     5127   PyObject *__pyx_v_size = 0;
     5128-  PyArrayObject *__pyx_v_omu = 0;
     5129-  PyArrayObject *__pyx_v_okappa = 0;
     5130-  double __pyx_v_fmu;
     5131-  double __pyx_v_fkappa;
     5132-  PyObject *__pyx_r = NULL;
     5133+  PyObject *__pyx_r = 0;
     5134   __Pyx_RefNannyDeclarations
     5135-  int __pyx_t_1;
     5136-  PyObject *__pyx_t_2 = NULL;
     5137-  PyObject *__pyx_t_3 = NULL;
     5138-  PyObject *__pyx_t_4 = NULL;
     5139-  PyObject *__pyx_t_5 = NULL;
     5140-  int __pyx_lineno = 0;
     5141-  const char *__pyx_filename = NULL;
     5142-  int __pyx_clineno = 0;
     5143-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__mu,&__pyx_n_s__kappa,&__pyx_n_s__size,0};
     5144-  __Pyx_RefNannySetupContext("vonmises");
     5145+  __Pyx_RefNannySetupContext("vonmises (wrapper)", 0);
     5146   {
     5147+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__mu,&__pyx_n_s__kappa,&__pyx_n_s__size,0};
     5148     PyObject* values[3] = {0,0,0};
     5149+
     5150+    /* "mtrand.pyx":2156
     5151+ *         return cont1_array(self.internal_state, rk_standard_t, size, odf)
     5152+ *
     5153+ *     def vonmises(self, mu, kappa, size=None):             # <<<<<<<<<<<<<<
     5154+ *         """
     5155+ *         vonmises(mu, kappa, size=None)
     5156+ */
     5157     values[2] = ((PyObject *)Py_None);
     5158     if (unlikely(__pyx_kwds)) {
     5159       Py_ssize_t kw_args;
     5160-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     5161+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     5162+      switch (pos_args) {
     5163         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     5164         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     5165         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     5166@@ -10517,14 +10939,12 @@
     5167         default: goto __pyx_L5_argtuple_error;
     5168       }
     5169       kw_args = PyDict_Size(__pyx_kwds);
     5170-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     5171+      switch (pos_args) {
     5172         case  0:
     5173-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__mu);
     5174-        if (likely(values[0])) kw_args--;
     5175+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__mu)) != 0)) kw_args--;
     5176         else goto __pyx_L5_argtuple_error;
     5177         case  1:
     5178-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__kappa);
     5179-        if (likely(values[1])) kw_args--;
     5180+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__kappa)) != 0)) kw_args--;
     5181         else {
     5182           __Pyx_RaiseArgtupleInvalid("vonmises", 0, 2, 3, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2156; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5183         }
     5184@@ -10535,7 +10955,7 @@
     5185         }
     5186       }
     5187       if (unlikely(kw_args > 0)) {
     5188-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "vonmises") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2156; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5189+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "vonmises") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2156; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5190       }
     5191     } else {
     5192       switch (PyTuple_GET_SIZE(__pyx_args)) {
     5193@@ -10558,6 +10978,27 @@
     5194   __Pyx_RefNannyFinishContext();
     5195   return NULL;
     5196   __pyx_L4_argument_unpacking_done:;
     5197+  __pyx_r = __pyx_pf_6mtrand_11RandomState_58vonmises(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_mu, __pyx_v_kappa, __pyx_v_size);
     5198+  __Pyx_RefNannyFinishContext();
     5199+  return __pyx_r;
     5200+}
     5201+
     5202+static PyObject *__pyx_pf_6mtrand_11RandomState_58vonmises(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_mu, PyObject *__pyx_v_kappa, PyObject *__pyx_v_size) {
     5203+  PyArrayObject *__pyx_v_omu = 0;
     5204+  PyArrayObject *__pyx_v_okappa = 0;
     5205+  double __pyx_v_fmu;
     5206+  double __pyx_v_fkappa;
     5207+  PyObject *__pyx_r = NULL;
     5208+  __Pyx_RefNannyDeclarations
     5209+  int __pyx_t_1;
     5210+  PyObject *__pyx_t_2 = NULL;
     5211+  PyObject *__pyx_t_3 = NULL;
     5212+  PyObject *__pyx_t_4 = NULL;
     5213+  PyObject *__pyx_t_5 = NULL;
     5214+  int __pyx_lineno = 0;
     5215+  const char *__pyx_filename = NULL;
     5216+  int __pyx_clineno = 0;
     5217+  __Pyx_RefNannySetupContext("vonmises", 0);
     5218 
     5219   /* "mtrand.pyx":2235
     5220  *         cdef double fmu, fkappa
     5221@@ -10609,9 +11050,9 @@
     5222       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5223       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5224       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2239; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5225-      goto __pyx_L7;
     5226+      goto __pyx_L4;
     5227     }
     5228-    __pyx_L7:;
     5229+    __pyx_L4:;
     5230 
     5231     /* "mtrand.pyx":2240
     5232  *             if fkappa < 0:
     5233@@ -10621,14 +11062,14 @@
     5234  *         PyErr_Clear()
     5235  */
     5236     __Pyx_XDECREF(__pyx_r);
     5237-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_vonmises, __pyx_v_size, __pyx_v_fmu, __pyx_v_fkappa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2240; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5238+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_vonmises, __pyx_v_size, __pyx_v_fmu, __pyx_v_fkappa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2240; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5239     __Pyx_GOTREF(__pyx_t_2);
     5240     __pyx_r = __pyx_t_2;
     5241     __pyx_t_2 = 0;
     5242     goto __pyx_L0;
     5243-    goto __pyx_L6;
     5244+    goto __pyx_L3;
     5245   }
     5246-  __pyx_L6:;
     5247+  __pyx_L3:;
     5248 
     5249   /* "mtrand.pyx":2242
     5250  *             return cont2_array_sc(self.internal_state, rk_vonmises, size, fmu, fkappa)
     5251@@ -10685,7 +11126,7 @@
     5252   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2246; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5253   __Pyx_GOTREF(__pyx_t_2);
     5254   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2246; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5255-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5256+  __Pyx_GOTREF(__pyx_t_5);
     5257   __Pyx_INCREF(((PyObject *)__pyx_v_okappa));
     5258   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_okappa));
     5259   __Pyx_GIVEREF(((PyObject *)__pyx_v_okappa));
     5260@@ -10697,7 +11138,7 @@
     5261   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     5262   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     5263   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2246; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5264-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5265+  __Pyx_GOTREF(__pyx_t_5);
     5266   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     5267   __Pyx_GIVEREF(__pyx_t_2);
     5268   __pyx_t_2 = 0;
     5269@@ -10721,9 +11162,9 @@
     5270     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5271     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5272     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2247; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5273-    goto __pyx_L8;
     5274+    goto __pyx_L5;
     5275   }
     5276-  __pyx_L8:;
     5277+  __pyx_L5:;
     5278 
     5279   /* "mtrand.pyx":2248
     5280  *         if np.any(np.less(okappa, 0.0)):
     5281@@ -10733,7 +11174,7 @@
     5282  *     def pareto(self, a, size=None):
     5283  */
     5284   __Pyx_XDECREF(__pyx_r);
     5285-  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_vonmises, __pyx_v_size, __pyx_v_omu, __pyx_v_okappa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2248; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5286+  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_vonmises, __pyx_v_size, __pyx_v_omu, __pyx_v_okappa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2248; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5287   __Pyx_GOTREF(__pyx_t_2);
     5288   __pyx_r = __pyx_t_2;
     5289   __pyx_t_2 = 0;
     5290@@ -10756,49 +11197,40 @@
     5291   return __pyx_r;
     5292 }
     5293 
     5294-/* "mtrand.pyx":2250
     5295+/* Python wrapper */
     5296+static PyObject *__pyx_pw_6mtrand_11RandomState_61pareto(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     5297+static char __pyx_doc_6mtrand_11RandomState_60pareto[] = "\n        pareto(a, size=None)\n\n        Draw samples from a Pareto II or Lomax distribution with specified shape.\n\n        The Lomax or Pareto II distribution is a shifted Pareto distribution. The\n        classical Pareto distribution can be obtained from the Lomax distribution\n        by adding the location parameter m, see below. The smallest value of the\n        Lomax distribution is zero while for the classical Pareto distribution it\n        is m, where the standard Pareto distribution has location m=1.\n        Lomax can also be considered as a simplified version of the Generalized\n        Pareto distribution (available in SciPy), with the scale set to one and\n        the location set to zero.\n\n        The Pareto distribution must be greater than zero, and is unbounded above.\n        It is also known as the \"80-20 rule\".  In this distribution, 80 percent of\n        the weights are in the lowest 20 percent of the range, while the other 20\n        percent fill the remaining 80 percent of the range.\n\n        Parameters\n        ----------\n        shape : float, > 0.\n            Shape of the distribution.\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        See Also\n        --------\n        scipy.stats.distributions.lomax.pdf : probability density function,\n            distribution or cumulative density function, etc.\n        scipy.stats.distributions.genpareto.pdf : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Pareto distribution is\n\n        .. math:: p(x) = \\frac{am^a}{x^{a+1}}\n\n        where :math:`a` is the shape and :math:`m` the location\n\n        The Pareto distribution, named after the Italian economist Vilfredo Pareto,\n        is a power law probability distribution useful in many real world probl""ems.\n        Outside the field of economics it is generally referred to as the Bradford\n        distribution. Pareto developed the distribution to describe the\n        distribution of wealth in an economy.  It has also found use in insurance,\n        web page access statistics, oil field sizes, and many other problems,\n        including the download frequency for projects in Sourceforge [1].  It is\n        one of the so-called \"fat-tailed\" distributions.\n\n\n        References\n        ----------\n        .. [1] Francis Hunt and Paul Johnson, On the Pareto Distribution of\n               Sourceforge projects.\n        .. [2] Pareto, V. (1896). Course of Political Economy. Lausanne.\n        .. [3] Reiss, R.D., Thomas, M.(2001), Statistical Analysis of Extreme\n               Values, Birkhauser Verlag, Basel, pp 23-30.\n        .. [4] Wikipedia, \"Pareto distribution\",\n               http://en.wikipedia.org/wiki/Pareto_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> a, m = 3., 1. # shape and mode\n        >>> s = np.random.pareto(a, 1000) + m\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 100, normed=True, align='center')\n        >>> fit = a*m**a/bins**(a+1)\n        >>> plt.plot(bins, max(count)*fit/max(fit),linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     5298+static PyObject *__pyx_pw_6mtrand_11RandomState_61pareto(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     5299+  PyObject *__pyx_v_a = 0;
     5300+  PyObject *__pyx_v_size = 0;
     5301+  PyObject *__pyx_r = 0;
     5302+  __Pyx_RefNannyDeclarations
     5303+  __Pyx_RefNannySetupContext("pareto (wrapper)", 0);
     5304+  {
     5305+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__a,&__pyx_n_s__size,0};
     5306+    PyObject* values[2] = {0,0};
     5307+
     5308+    /* "mtrand.pyx":2250
     5309  *         return cont2_array(self.internal_state, rk_vonmises, size, omu, okappa)
     5310  *
     5311  *     def pareto(self, a, size=None):             # <<<<<<<<<<<<<<
     5312  *         """
     5313  *         pareto(a, size=None)
     5314  */
     5315-
     5316-static PyObject *__pyx_pf_6mtrand_11RandomState_30pareto(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     5317-static char __pyx_doc_6mtrand_11RandomState_30pareto[] = "\n        pareto(a, size=None)\n\n        Draw samples from a Pareto II or Lomax distribution with specified shape.\n\n        The Lomax or Pareto II distribution is a shifted Pareto distribution. The\n        classical Pareto distribution can be obtained from the Lomax distribution\n        by adding the location parameter m, see below. The smallest value of the\n        Lomax distribution is zero while for the classical Pareto distribution it\n        is m, where the standard Pareto distribution has location m=1.\n        Lomax can also be considered as a simplified version of the Generalized\n        Pareto distribution (available in SciPy), with the scale set to one and\n        the location set to zero.\n\n        The Pareto distribution must be greater than zero, and is unbounded above.\n        It is also known as the \"80-20 rule\".  In this distribution, 80 percent of\n        the weights are in the lowest 20 percent of the range, while the other 20\n        percent fill the remaining 80 percent of the range.\n\n        Parameters\n        ----------\n        shape : float, > 0.\n            Shape of the distribution.\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        See Also\n        --------\n        scipy.stats.distributions.lomax.pdf : probability density function,\n            distribution or cumulative density function, etc.\n        scipy.stats.distributions.genpareto.pdf : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Pareto distribution is\n\n        .. math:: p(x) = \\frac{am^a}{x^{a+1}}\n\n        where :math:`a` is the shape and :math:`m` the location\n\n        The Pareto distribution, named after the Italian economist Vilfredo Pareto,\n        is a power law probability distribution useful in many real world probl""ems.\n        Outside the field of economics it is generally referred to as the Bradford\n        distribution. Pareto developed the distribution to describe the\n        distribution of wealth in an economy.  It has also found use in insurance,\n        web page access statistics, oil field sizes, and many other problems,\n        including the download frequency for projects in Sourceforge [1].  It is\n        one of the so-called \"fat-tailed\" distributions.\n\n\n        References\n        ----------\n        .. [1] Francis Hunt and Paul Johnson, On the Pareto Distribution of\n               Sourceforge projects.\n        .. [2] Pareto, V. (1896). Course of Political Economy. Lausanne.\n        .. [3] Reiss, R.D., Thomas, M.(2001), Statistical Analysis of Extreme\n               Values, Birkhauser Verlag, Basel, pp 23-30.\n        .. [4] Wikipedia, \"Pareto distribution\",\n               http://en.wikipedia.org/wiki/Pareto_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> a, m = 3., 1. # shape and mode\n        >>> s = np.random.pareto(a, 1000) + m\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 100, normed=True, align='center')\n        >>> fit = a*m**a/bins**(a+1)\n        >>> plt.plot(bins, max(count)*fit/max(fit),linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     5318-static PyObject *__pyx_pf_6mtrand_11RandomState_30pareto(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     5319-  PyObject *__pyx_v_a = 0;
     5320-  PyObject *__pyx_v_size = 0;
     5321-  PyArrayObject *__pyx_v_oa = 0;
     5322-  double __pyx_v_fa;
     5323-  PyObject *__pyx_r = NULL;
     5324-  __Pyx_RefNannyDeclarations
     5325-  int __pyx_t_1;
     5326-  PyObject *__pyx_t_2 = NULL;
     5327-  PyObject *__pyx_t_3 = NULL;
     5328-  PyObject *__pyx_t_4 = NULL;
     5329-  PyObject *__pyx_t_5 = NULL;
     5330-  int __pyx_lineno = 0;
     5331-  const char *__pyx_filename = NULL;
     5332-  int __pyx_clineno = 0;
     5333-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__a,&__pyx_n_s__size,0};
     5334-  __Pyx_RefNannySetupContext("pareto");
     5335-  {
     5336-    PyObject* values[2] = {0,0};
     5337     values[1] = ((PyObject *)Py_None);
     5338     if (unlikely(__pyx_kwds)) {
     5339       Py_ssize_t kw_args;
     5340-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     5341+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     5342+      switch (pos_args) {
     5343         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     5344         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     5345         case  0: break;
     5346         default: goto __pyx_L5_argtuple_error;
     5347       }
     5348       kw_args = PyDict_Size(__pyx_kwds);
     5349-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     5350+      switch (pos_args) {
     5351         case  0:
     5352-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__a);
     5353-        if (likely(values[0])) kw_args--;
     5354+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__a)) != 0)) kw_args--;
     5355         else goto __pyx_L5_argtuple_error;
     5356         case  1:
     5357         if (kw_args > 0) {
     5358@@ -10807,7 +11239,7 @@
     5359         }
     5360       }
     5361       if (unlikely(kw_args > 0)) {
     5362-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "pareto") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2250; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5363+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "pareto") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2250; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5364       }
     5365     } else {
     5366       switch (PyTuple_GET_SIZE(__pyx_args)) {
     5367@@ -10828,6 +11260,25 @@
     5368   __Pyx_RefNannyFinishContext();
     5369   return NULL;
     5370   __pyx_L4_argument_unpacking_done:;
     5371+  __pyx_r = __pyx_pf_6mtrand_11RandomState_60pareto(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_a, __pyx_v_size);
     5372+  __Pyx_RefNannyFinishContext();
     5373+  return __pyx_r;
     5374+}
     5375+
     5376+static PyObject *__pyx_pf_6mtrand_11RandomState_60pareto(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_a, PyObject *__pyx_v_size) {
     5377+  PyArrayObject *__pyx_v_oa = 0;
     5378+  double __pyx_v_fa;
     5379+  PyObject *__pyx_r = NULL;
     5380+  __Pyx_RefNannyDeclarations
     5381+  int __pyx_t_1;
     5382+  PyObject *__pyx_t_2 = NULL;
     5383+  PyObject *__pyx_t_3 = NULL;
     5384+  PyObject *__pyx_t_4 = NULL;
     5385+  PyObject *__pyx_t_5 = NULL;
     5386+  int __pyx_lineno = 0;
     5387+  const char *__pyx_filename = NULL;
     5388+  int __pyx_clineno = 0;
     5389+  __Pyx_RefNannySetupContext("pareto", 0);
     5390 
     5391   /* "mtrand.pyx":2333
     5392  *         cdef double fa
     5393@@ -10870,9 +11321,9 @@
     5394       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5395       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5396       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2336; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5397-      goto __pyx_L7;
     5398+      goto __pyx_L4;
     5399     }
     5400-    __pyx_L7:;
     5401+    __pyx_L4:;
     5402 
     5403     /* "mtrand.pyx":2337
     5404  *             if fa <= 0:
     5405@@ -10882,14 +11333,14 @@
     5406  *         PyErr_Clear()
     5407  */
     5408     __Pyx_XDECREF(__pyx_r);
     5409-    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_pareto, __pyx_v_size, __pyx_v_fa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2337; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5410+    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(__pyx_v_self->internal_state, rk_pareto, __pyx_v_size, __pyx_v_fa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2337; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5411     __Pyx_GOTREF(__pyx_t_2);
     5412     __pyx_r = __pyx_t_2;
     5413     __pyx_t_2 = 0;
     5414     goto __pyx_L0;
     5415-    goto __pyx_L6;
     5416+    goto __pyx_L3;
     5417   }
     5418-  __pyx_L6:;
     5419+  __pyx_L3:;
     5420 
     5421   /* "mtrand.pyx":2339
     5422  *             return cont1_array_sc(self.internal_state, rk_pareto, size, fa)
     5423@@ -10933,7 +11384,7 @@
     5424   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2342; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5425   __Pyx_GOTREF(__pyx_t_2);
     5426   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2342; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5427-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5428+  __Pyx_GOTREF(__pyx_t_5);
     5429   __Pyx_INCREF(((PyObject *)__pyx_v_oa));
     5430   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_oa));
     5431   __Pyx_GIVEREF(((PyObject *)__pyx_v_oa));
     5432@@ -10945,7 +11396,7 @@
     5433   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     5434   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     5435   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2342; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5436-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5437+  __Pyx_GOTREF(__pyx_t_5);
     5438   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     5439   __Pyx_GIVEREF(__pyx_t_2);
     5440   __pyx_t_2 = 0;
     5441@@ -10969,9 +11420,9 @@
     5442     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5443     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5444     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2343; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5445-    goto __pyx_L8;
     5446+    goto __pyx_L5;
     5447   }
     5448-  __pyx_L8:;
     5449+  __pyx_L5:;
     5450 
     5451   /* "mtrand.pyx":2344
     5452  *         if np.any(np.less_equal(oa, 0.0)):
     5453@@ -10981,7 +11432,7 @@
     5454  *     def weibull(self, a, size=None):
     5455  */
     5456   __Pyx_XDECREF(__pyx_r);
     5457-  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_pareto, __pyx_v_size, __pyx_v_oa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2344; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5458+  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(__pyx_v_self->internal_state, rk_pareto, __pyx_v_size, __pyx_v_oa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2344; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5459   __Pyx_GOTREF(__pyx_t_2);
     5460   __pyx_r = __pyx_t_2;
     5461   __pyx_t_2 = 0;
     5462@@ -11003,49 +11454,40 @@
     5463   return __pyx_r;
     5464 }
     5465 
     5466-/* "mtrand.pyx":2346
     5467+/* Python wrapper */
     5468+static PyObject *__pyx_pw_6mtrand_11RandomState_63weibull(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     5469+static char __pyx_doc_6mtrand_11RandomState_62weibull[] = "\n        weibull(a, size=None)\n\n        Weibull distribution.\n\n        Draw samples from a 1-parameter Weibull distribution with the given\n        shape parameter `a`.\n\n        .. math:: X = (-ln(U))^{1/a}\n\n        Here, U is drawn from the uniform distribution over (0,1].\n\n        The more common 2-parameter Weibull, including a scale parameter\n        :math:`\\lambda` is just :math:`X = \\lambda(-ln(U))^{1/a}`.\n\n        Parameters\n        ----------\n        a : float\n            Shape of the distribution.\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        See Also\n        --------\n        scipy.stats.distributions.weibull : probability density function,\n            distribution or cumulative density function, etc.\n\n        gumbel, scipy.stats.distributions.genextreme\n\n        Notes\n        -----\n        The Weibull (or Type III asymptotic extreme value distribution for smallest\n        values, SEV Type III, or Rosin-Rammler distribution) is one of a class of\n        Generalized Extreme Value (GEV) distributions used in modeling extreme\n        value problems.  This class includes the Gumbel and Frechet distributions.\n\n        The probability density for the Weibull distribution is\n\n        .. math:: p(x) = \\frac{a}\n                         {\\lambda}(\\frac{x}{\\lambda})^{a-1}e^{-(x/\\lambda)^a},\n\n        where :math:`a` is the shape and :math:`\\lambda` the scale.\n\n        The function has its peak (the mode) at\n        :math:`\\lambda(\\frac{a-1}{a})^{1/a}`.\n\n        When ``a = 1``, the Weibull distribution reduces to the exponential\n        distribution.\n\n        References\n        ----------\n        .. [1] Waloddi Weibull, Professor, Royal Technical University, Stockholm,\n               1939 \"A Statistical Theory Of The Strength Of Materials\",\n               Ingeniorsvetenskapsakademiens Handlingar"" Nr 151, 1939,\n               Generalstabens Litografiska Anstalts Forlag, Stockholm.\n        .. [2] Waloddi Weibull, 1951 \"A Statistical Distribution Function of Wide\n               Applicability\",  Journal Of Applied Mechanics ASME Paper.\n        .. [3] Wikipedia, \"Weibull distribution\",\n               http://en.wikipedia.org/wiki/Weibull_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> a = 5. # shape\n        >>> s = np.random.weibull(a, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> x = np.arange(1,100.)/50.\n        >>> def weib(x,n,a):\n        ...     return (a / n) * (x / n)**(a - 1) * np.exp(-(x / n)**a)\n\n        >>> count, bins, ignored = plt.hist(np.random.weibull(5.,1000))\n        >>> x = np.arange(1,100.)/50.\n        >>> scale = count.max()/weib(x, 1., 5.).max()\n        >>> plt.plot(x, weib(x, 1., 5.)*scale)\n        >>> plt.show()\n\n        ";
     5470+static PyObject *__pyx_pw_6mtrand_11RandomState_63weibull(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     5471+  PyObject *__pyx_v_a = 0;
     5472+  PyObject *__pyx_v_size = 0;
     5473+  PyObject *__pyx_r = 0;
     5474+  __Pyx_RefNannyDeclarations
     5475+  __Pyx_RefNannySetupContext("weibull (wrapper)", 0);
     5476+  {
     5477+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__a,&__pyx_n_s__size,0};
     5478+    PyObject* values[2] = {0,0};
     5479+
     5480+    /* "mtrand.pyx":2346
     5481  *         return cont1_array(self.internal_state, rk_pareto, size, oa)
     5482  *
     5483  *     def weibull(self, a, size=None):             # <<<<<<<<<<<<<<
     5484  *         """
     5485  *         weibull(a, size=None)
     5486  */
     5487-
     5488-static PyObject *__pyx_pf_6mtrand_11RandomState_31weibull(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     5489-static char __pyx_doc_6mtrand_11RandomState_31weibull[] = "\n        weibull(a, size=None)\n\n        Weibull distribution.\n\n        Draw samples from a 1-parameter Weibull distribution with the given\n        shape parameter `a`.\n\n        .. math:: X = (-ln(U))^{1/a}\n\n        Here, U is drawn from the uniform distribution over (0,1].\n\n        The more common 2-parameter Weibull, including a scale parameter\n        :math:`\\lambda` is just :math:`X = \\lambda(-ln(U))^{1/a}`.\n\n        Parameters\n        ----------\n        a : float\n            Shape of the distribution.\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        See Also\n        --------\n        scipy.stats.distributions.weibull : probability density function,\n            distribution or cumulative density function, etc.\n\n        gumbel, scipy.stats.distributions.genextreme\n\n        Notes\n        -----\n        The Weibull (or Type III asymptotic extreme value distribution for smallest\n        values, SEV Type III, or Rosin-Rammler distribution) is one of a class of\n        Generalized Extreme Value (GEV) distributions used in modeling extreme\n        value problems.  This class includes the Gumbel and Frechet distributions.\n\n        The probability density for the Weibull distribution is\n\n        .. math:: p(x) = \\frac{a}\n                         {\\lambda}(\\frac{x}{\\lambda})^{a-1}e^{-(x/\\lambda)^a},\n\n        where :math:`a` is the shape and :math:`\\lambda` the scale.\n\n        The function has its peak (the mode) at\n        :math:`\\lambda(\\frac{a-1}{a})^{1/a}`.\n\n        When ``a = 1``, the Weibull distribution reduces to the exponential\n        distribution.\n\n        References\n        ----------\n        .. [1] Waloddi Weibull, Professor, Royal Technical University, Stockholm,\n               1939 \"A Statistical Theory Of The Strength Of Materials\",\n               Ingeniorsvetenskapsakademiens Handlingar"" Nr 151, 1939,\n               Generalstabens Litografiska Anstalts Forlag, Stockholm.\n        .. [2] Waloddi Weibull, 1951 \"A Statistical Distribution Function of Wide\n               Applicability\",  Journal Of Applied Mechanics ASME Paper.\n        .. [3] Wikipedia, \"Weibull distribution\",\n               http://en.wikipedia.org/wiki/Weibull_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> a = 5. # shape\n        >>> s = np.random.weibull(a, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> x = np.arange(1,100.)/50.\n        >>> def weib(x,n,a):\n        ...     return (a / n) * (x / n)**(a - 1) * np.exp(-(x / n)**a)\n\n        >>> count, bins, ignored = plt.hist(np.random.weibull(5.,1000))\n        >>> x = np.arange(1,100.)/50.\n        >>> scale = count.max()/weib(x, 1., 5.).max()\n        >>> plt.plot(x, weib(x, 1., 5.)*scale)\n        >>> plt.show()\n\n        ";
     5490-static PyObject *__pyx_pf_6mtrand_11RandomState_31weibull(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     5491-  PyObject *__pyx_v_a = 0;
     5492-  PyObject *__pyx_v_size = 0;
     5493-  PyArrayObject *__pyx_v_oa = 0;
     5494-  double __pyx_v_fa;
     5495-  PyObject *__pyx_r = NULL;
     5496-  __Pyx_RefNannyDeclarations
     5497-  int __pyx_t_1;
     5498-  PyObject *__pyx_t_2 = NULL;
     5499-  PyObject *__pyx_t_3 = NULL;
     5500-  PyObject *__pyx_t_4 = NULL;
     5501-  PyObject *__pyx_t_5 = NULL;
     5502-  int __pyx_lineno = 0;
     5503-  const char *__pyx_filename = NULL;
     5504-  int __pyx_clineno = 0;
     5505-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__a,&__pyx_n_s__size,0};
     5506-  __Pyx_RefNannySetupContext("weibull");
     5507-  {
     5508-    PyObject* values[2] = {0,0};
     5509     values[1] = ((PyObject *)Py_None);
     5510     if (unlikely(__pyx_kwds)) {
     5511       Py_ssize_t kw_args;
     5512-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     5513+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     5514+      switch (pos_args) {
     5515         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     5516         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     5517         case  0: break;
     5518         default: goto __pyx_L5_argtuple_error;
     5519       }
     5520       kw_args = PyDict_Size(__pyx_kwds);
     5521-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     5522+      switch (pos_args) {
     5523         case  0:
     5524-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__a);
     5525-        if (likely(values[0])) kw_args--;
     5526+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__a)) != 0)) kw_args--;
     5527         else goto __pyx_L5_argtuple_error;
     5528         case  1:
     5529         if (kw_args > 0) {
     5530@@ -11054,7 +11496,7 @@
     5531         }
     5532       }
     5533       if (unlikely(kw_args > 0)) {
     5534-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "weibull") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2346; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5535+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "weibull") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2346; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5536       }
     5537     } else {
     5538       switch (PyTuple_GET_SIZE(__pyx_args)) {
     5539@@ -11075,6 +11517,25 @@
     5540   __Pyx_RefNannyFinishContext();
     5541   return NULL;
     5542   __pyx_L4_argument_unpacking_done:;
     5543+  __pyx_r = __pyx_pf_6mtrand_11RandomState_62weibull(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_a, __pyx_v_size);
     5544+  __Pyx_RefNannyFinishContext();
     5545+  return __pyx_r;
     5546+}
     5547+
     5548+static PyObject *__pyx_pf_6mtrand_11RandomState_62weibull(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_a, PyObject *__pyx_v_size) {
     5549+  PyArrayObject *__pyx_v_oa = 0;
     5550+  double __pyx_v_fa;
     5551+  PyObject *__pyx_r = NULL;
     5552+  __Pyx_RefNannyDeclarations
     5553+  int __pyx_t_1;
     5554+  PyObject *__pyx_t_2 = NULL;
     5555+  PyObject *__pyx_t_3 = NULL;
     5556+  PyObject *__pyx_t_4 = NULL;
     5557+  PyObject *__pyx_t_5 = NULL;
     5558+  int __pyx_lineno = 0;
     5559+  const char *__pyx_filename = NULL;
     5560+  int __pyx_clineno = 0;
     5561+  __Pyx_RefNannySetupContext("weibull", 0);
     5562 
     5563   /* "mtrand.pyx":2433
     5564  *         cdef double fa
     5565@@ -11117,9 +11578,9 @@
     5566       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5567       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5568       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2436; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5569-      goto __pyx_L7;
     5570+      goto __pyx_L4;
     5571     }
     5572-    __pyx_L7:;
     5573+    __pyx_L4:;
     5574 
     5575     /* "mtrand.pyx":2437
     5576  *             if fa <= 0:
     5577@@ -11129,14 +11590,14 @@
     5578  *         PyErr_Clear()
     5579  */
     5580     __Pyx_XDECREF(__pyx_r);
     5581-    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_weibull, __pyx_v_size, __pyx_v_fa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2437; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5582+    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(__pyx_v_self->internal_state, rk_weibull, __pyx_v_size, __pyx_v_fa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2437; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5583     __Pyx_GOTREF(__pyx_t_2);
     5584     __pyx_r = __pyx_t_2;
     5585     __pyx_t_2 = 0;
     5586     goto __pyx_L0;
     5587-    goto __pyx_L6;
     5588+    goto __pyx_L3;
     5589   }
     5590-  __pyx_L6:;
     5591+  __pyx_L3:;
     5592 
     5593   /* "mtrand.pyx":2439
     5594  *             return cont1_array_sc(self.internal_state, rk_weibull, size, fa)
     5595@@ -11180,7 +11641,7 @@
     5596   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2442; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5597   __Pyx_GOTREF(__pyx_t_2);
     5598   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2442; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5599-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5600+  __Pyx_GOTREF(__pyx_t_5);
     5601   __Pyx_INCREF(((PyObject *)__pyx_v_oa));
     5602   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_oa));
     5603   __Pyx_GIVEREF(((PyObject *)__pyx_v_oa));
     5604@@ -11192,7 +11653,7 @@
     5605   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     5606   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     5607   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2442; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5608-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5609+  __Pyx_GOTREF(__pyx_t_5);
     5610   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     5611   __Pyx_GIVEREF(__pyx_t_2);
     5612   __pyx_t_2 = 0;
     5613@@ -11216,9 +11677,9 @@
     5614     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5615     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5616     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2443; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5617-    goto __pyx_L8;
     5618+    goto __pyx_L5;
     5619   }
     5620-  __pyx_L8:;
     5621+  __pyx_L5:;
     5622 
     5623   /* "mtrand.pyx":2444
     5624  *         if np.any(np.less_equal(oa, 0.0)):
     5625@@ -11228,7 +11689,7 @@
     5626  *     def power(self, a, size=None):
     5627  */
     5628   __Pyx_XDECREF(__pyx_r);
     5629-  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_weibull, __pyx_v_size, __pyx_v_oa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2444; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5630+  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(__pyx_v_self->internal_state, rk_weibull, __pyx_v_size, __pyx_v_oa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2444; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5631   __Pyx_GOTREF(__pyx_t_2);
     5632   __pyx_r = __pyx_t_2;
     5633   __pyx_t_2 = 0;
     5634@@ -11250,49 +11711,40 @@
     5635   return __pyx_r;
     5636 }
     5637 
     5638-/* "mtrand.pyx":2446
     5639+/* Python wrapper */
     5640+static PyObject *__pyx_pw_6mtrand_11RandomState_65power(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     5641+static char __pyx_doc_6mtrand_11RandomState_64power[] = "\n        power(a, size=None)\n\n        Draws samples in [0, 1] from a power distribution with positive\n        exponent a - 1.\n\n        Also known as the power function distribution.\n\n        Parameters\n        ----------\n        a : float\n            parameter, > 0\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n                    ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n            The returned samples lie in [0, 1].\n\n        Raises\n        ------\n        ValueError\n            If a<1.\n\n        Notes\n        -----\n        The probability density function is\n\n        .. math:: P(x; a) = ax^{a-1}, 0 \\le x \\le 1, a>0.\n\n        The power function distribution is just the inverse of the Pareto\n        distribution. It may also be seen as a special case of the Beta\n        distribution.\n\n        It is used, for example, in modeling the over-reporting of insurance\n        claims.\n\n        References\n        ----------\n        .. [1] Christian Kleiber, Samuel Kotz, \"Statistical size distributions\n               in economics and actuarial sciences\", Wiley, 2003.\n        .. [2] Heckert, N. A. and Filliben, James J. (2003). NIST Handbook 148:\n               Dataplot Reference Manual, Volume 2: Let Subcommands and Library\n               Functions\", National Institute of Standards and Technology Handbook\n               Series, June 2003.\n               http://www.itl.nist.gov/div898/software/dataplot/refman2/auxillar/powpdf.pdf\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> a = 5. # shape\n        >>> samples = 1000\n        >>> s = np.random.power(a, samples)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, bins=""30)\n        >>> x = np.linspace(0, 1, 100)\n        >>> y = a*x**(a-1.)\n        >>> normed_y = samples*np.diff(bins)[0]*y\n        >>> plt.plot(x, normed_y)\n        >>> plt.show()\n\n        Compare the power function distribution to the inverse of the Pareto.\n\n        >>> from scipy import stats\n        >>> rvs = np.random.power(5, 1000000)\n        >>> rvsp = np.random.pareto(5, 1000000)\n        >>> xx = np.linspace(0,1,100)\n        >>> powpdf = stats.powerlaw.pdf(xx,5)\n\n        >>> plt.figure()\n        >>> plt.hist(rvs, bins=50, normed=True)\n        >>> plt.plot(xx,powpdf,'r-')\n        >>> plt.title('np.random.power(5)')\n\n        >>> plt.figure()\n        >>> plt.hist(1./(1.+rvsp), bins=50, normed=True)\n        >>> plt.plot(xx,powpdf,'r-')\n        >>> plt.title('inverse of 1 + np.random.pareto(5)')\n\n        >>> plt.figure()\n        >>> plt.hist(1./(1.+rvsp), bins=50, normed=True)\n        >>> plt.plot(xx,powpdf,'r-')\n        >>> plt.title('inverse of stats.pareto(5)')\n\n        ";
     5642+static PyObject *__pyx_pw_6mtrand_11RandomState_65power(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     5643+  PyObject *__pyx_v_a = 0;
     5644+  PyObject *__pyx_v_size = 0;
     5645+  PyObject *__pyx_r = 0;
     5646+  __Pyx_RefNannyDeclarations
     5647+  __Pyx_RefNannySetupContext("power (wrapper)", 0);
     5648+  {
     5649+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__a,&__pyx_n_s__size,0};
     5650+    PyObject* values[2] = {0,0};
     5651+
     5652+    /* "mtrand.pyx":2446
     5653  *         return cont1_array(self.internal_state, rk_weibull, size, oa)
     5654  *
     5655  *     def power(self, a, size=None):             # <<<<<<<<<<<<<<
     5656  *         """
     5657  *         power(a, size=None)
     5658  */
     5659-
     5660-static PyObject *__pyx_pf_6mtrand_11RandomState_32power(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     5661-static char __pyx_doc_6mtrand_11RandomState_32power[] = "\n        power(a, size=None)\n\n        Draws samples in [0, 1] from a power distribution with positive\n        exponent a - 1.\n\n        Also known as the power function distribution.\n\n        Parameters\n        ----------\n        a : float\n            parameter, > 0\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n                    ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n            The returned samples lie in [0, 1].\n\n        Raises\n        ------\n        ValueError\n            If a<1.\n\n        Notes\n        -----\n        The probability density function is\n\n        .. math:: P(x; a) = ax^{a-1}, 0 \\le x \\le 1, a>0.\n\n        The power function distribution is just the inverse of the Pareto\n        distribution. It may also be seen as a special case of the Beta\n        distribution.\n\n        It is used, for example, in modeling the over-reporting of insurance\n        claims.\n\n        References\n        ----------\n        .. [1] Christian Kleiber, Samuel Kotz, \"Statistical size distributions\n               in economics and actuarial sciences\", Wiley, 2003.\n        .. [2] Heckert, N. A. and Filliben, James J. (2003). NIST Handbook 148:\n               Dataplot Reference Manual, Volume 2: Let Subcommands and Library\n               Functions\", National Institute of Standards and Technology Handbook\n               Series, June 2003.\n               http://www.itl.nist.gov/div898/software/dataplot/refman2/auxillar/powpdf.pdf\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> a = 5. # shape\n        >>> samples = 1000\n        >>> s = np.random.power(a, samples)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, bins=""30)\n        >>> x = np.linspace(0, 1, 100)\n        >>> y = a*x**(a-1.)\n        >>> normed_y = samples*np.diff(bins)[0]*y\n        >>> plt.plot(x, normed_y)\n        >>> plt.show()\n\n        Compare the power function distribution to the inverse of the Pareto.\n\n        >>> from scipy import stats\n        >>> rvs = np.random.power(5, 1000000)\n        >>> rvsp = np.random.pareto(5, 1000000)\n        >>> xx = np.linspace(0,1,100)\n        >>> powpdf = stats.powerlaw.pdf(xx,5)\n\n        >>> plt.figure()\n        >>> plt.hist(rvs, bins=50, normed=True)\n        >>> plt.plot(xx,powpdf,'r-')\n        >>> plt.title('np.random.power(5)')\n\n        >>> plt.figure()\n        >>> plt.hist(1./(1.+rvsp), bins=50, normed=True)\n        >>> plt.plot(xx,powpdf,'r-')\n        >>> plt.title('inverse of 1 + np.random.pareto(5)')\n\n        >>> plt.figure()\n        >>> plt.hist(1./(1.+rvsp), bins=50, normed=True)\n        >>> plt.plot(xx,powpdf,'r-')\n        >>> plt.title('inverse of stats.pareto(5)')\n\n        ";
     5662-static PyObject *__pyx_pf_6mtrand_11RandomState_32power(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     5663-  PyObject *__pyx_v_a = 0;
     5664-  PyObject *__pyx_v_size = 0;
     5665-  PyArrayObject *__pyx_v_oa = 0;
     5666-  double __pyx_v_fa;
     5667-  PyObject *__pyx_r = NULL;
     5668-  __Pyx_RefNannyDeclarations
     5669-  int __pyx_t_1;
     5670-  PyObject *__pyx_t_2 = NULL;
     5671-  PyObject *__pyx_t_3 = NULL;
     5672-  PyObject *__pyx_t_4 = NULL;
     5673-  PyObject *__pyx_t_5 = NULL;
     5674-  int __pyx_lineno = 0;
     5675-  const char *__pyx_filename = NULL;
     5676-  int __pyx_clineno = 0;
     5677-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__a,&__pyx_n_s__size,0};
     5678-  __Pyx_RefNannySetupContext("power");
     5679-  {
     5680-    PyObject* values[2] = {0,0};
     5681     values[1] = ((PyObject *)Py_None);
     5682     if (unlikely(__pyx_kwds)) {
     5683       Py_ssize_t kw_args;
     5684-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     5685+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     5686+      switch (pos_args) {
     5687         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     5688         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     5689         case  0: break;
     5690         default: goto __pyx_L5_argtuple_error;
     5691       }
     5692       kw_args = PyDict_Size(__pyx_kwds);
     5693-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     5694+      switch (pos_args) {
     5695         case  0:
     5696-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__a);
     5697-        if (likely(values[0])) kw_args--;
     5698+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__a)) != 0)) kw_args--;
     5699         else goto __pyx_L5_argtuple_error;
     5700         case  1:
     5701         if (kw_args > 0) {
     5702@@ -11301,7 +11753,7 @@
     5703         }
     5704       }
     5705       if (unlikely(kw_args > 0)) {
     5706-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "power") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2446; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5707+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "power") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2446; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5708       }
     5709     } else {
     5710       switch (PyTuple_GET_SIZE(__pyx_args)) {
     5711@@ -11322,6 +11774,25 @@
     5712   __Pyx_RefNannyFinishContext();
     5713   return NULL;
     5714   __pyx_L4_argument_unpacking_done:;
     5715+  __pyx_r = __pyx_pf_6mtrand_11RandomState_64power(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_a, __pyx_v_size);
     5716+  __Pyx_RefNannyFinishContext();
     5717+  return __pyx_r;
     5718+}
     5719+
     5720+static PyObject *__pyx_pf_6mtrand_11RandomState_64power(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_a, PyObject *__pyx_v_size) {
     5721+  PyArrayObject *__pyx_v_oa = 0;
     5722+  double __pyx_v_fa;
     5723+  PyObject *__pyx_r = NULL;
     5724+  __Pyx_RefNannyDeclarations
     5725+  int __pyx_t_1;
     5726+  PyObject *__pyx_t_2 = NULL;
     5727+  PyObject *__pyx_t_3 = NULL;
     5728+  PyObject *__pyx_t_4 = NULL;
     5729+  PyObject *__pyx_t_5 = NULL;
     5730+  int __pyx_lineno = 0;
     5731+  const char *__pyx_filename = NULL;
     5732+  int __pyx_clineno = 0;
     5733+  __Pyx_RefNannySetupContext("power", 0);
     5734 
     5735   /* "mtrand.pyx":2542
     5736  *         cdef double fa
     5737@@ -11364,9 +11835,9 @@
     5738       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5739       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5740       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2545; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5741-      goto __pyx_L7;
     5742+      goto __pyx_L4;
     5743     }
     5744-    __pyx_L7:;
     5745+    __pyx_L4:;
     5746 
     5747     /* "mtrand.pyx":2546
     5748  *             if fa <= 0:
     5749@@ -11376,14 +11847,14 @@
     5750  *         PyErr_Clear()
     5751  */
     5752     __Pyx_XDECREF(__pyx_r);
     5753-    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_power, __pyx_v_size, __pyx_v_fa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2546; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5754+    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(__pyx_v_self->internal_state, rk_power, __pyx_v_size, __pyx_v_fa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2546; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5755     __Pyx_GOTREF(__pyx_t_2);
     5756     __pyx_r = __pyx_t_2;
     5757     __pyx_t_2 = 0;
     5758     goto __pyx_L0;
     5759-    goto __pyx_L6;
     5760+    goto __pyx_L3;
     5761   }
     5762-  __pyx_L6:;
     5763+  __pyx_L3:;
     5764 
     5765   /* "mtrand.pyx":2548
     5766  *             return cont1_array_sc(self.internal_state, rk_power, size, fa)
     5767@@ -11427,7 +11898,7 @@
     5768   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2551; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5769   __Pyx_GOTREF(__pyx_t_2);
     5770   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2551; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5771-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5772+  __Pyx_GOTREF(__pyx_t_5);
     5773   __Pyx_INCREF(((PyObject *)__pyx_v_oa));
     5774   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_oa));
     5775   __Pyx_GIVEREF(((PyObject *)__pyx_v_oa));
     5776@@ -11439,7 +11910,7 @@
     5777   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     5778   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     5779   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2551; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5780-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5781+  __Pyx_GOTREF(__pyx_t_5);
     5782   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     5783   __Pyx_GIVEREF(__pyx_t_2);
     5784   __pyx_t_2 = 0;
     5785@@ -11463,9 +11934,9 @@
     5786     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5787     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5788     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2552; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5789-    goto __pyx_L8;
     5790+    goto __pyx_L5;
     5791   }
     5792-  __pyx_L8:;
     5793+  __pyx_L5:;
     5794 
     5795   /* "mtrand.pyx":2553
     5796  *         if np.any(np.less_equal(oa, 0.0)):
     5797@@ -11475,7 +11946,7 @@
     5798  *     def laplace(self, loc=0.0, scale=1.0, size=None):
     5799  */
     5800   __Pyx_XDECREF(__pyx_r);
     5801-  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_power, __pyx_v_size, __pyx_v_oa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2553; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5802+  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(__pyx_v_self->internal_state, rk_power, __pyx_v_size, __pyx_v_oa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2553; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5803   __Pyx_GOTREF(__pyx_t_2);
     5804   __pyx_r = __pyx_t_2;
     5805   __pyx_t_2 = 0;
     5806@@ -11497,44 +11968,34 @@
     5807   return __pyx_r;
     5808 }
     5809 
     5810-/* "mtrand.pyx":2555
     5811- *         return cont1_array(self.internal_state, rk_power, size, oa)
     5812- *
     5813- *     def laplace(self, loc=0.0, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     5814- *         """
     5815- *         laplace(loc=0.0, scale=1.0, size=None)
     5816- */
     5817-
     5818-static PyObject *__pyx_pf_6mtrand_11RandomState_33laplace(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     5819-static char __pyx_doc_6mtrand_11RandomState_33laplace[] = "\n        laplace(loc=0.0, scale=1.0, size=None)\n\n        Draw samples from the Laplace or double exponential distribution with\n        specified location (or mean) and scale (decay).\n\n        The Laplace distribution is similar to the Gaussian/normal distribution,\n        but is sharper at the peak and has fatter tails. It represents the\n        difference between two independent, identically distributed exponential\n        random variables.\n\n        Parameters\n        ----------\n        loc : float\n            The position, :math:`\\mu`, of the distribution peak.\n        scale : float\n            :math:`\\lambda`, the exponential decay.\n\n        Notes\n        -----\n        It has the probability density function\n\n        .. math:: f(x; \\mu, \\lambda) = \\frac{1}{2\\lambda}\n                                       \\exp\\left(-\\frac{|x - \\mu|}{\\lambda}\\right).\n\n        The first law of Laplace, from 1774, states that the frequency of an error\n        can be expressed as an exponential function of the absolute magnitude of\n        the error, which leads to the Laplace distribution. For many problems in\n        Economics and Health sciences, this distribution seems to model the data\n        better than the standard Gaussian distribution\n\n\n        References\n        ----------\n        .. [1] Abramowitz, M. and Stegun, I. A. (Eds.). Handbook of Mathematical\n               Functions with Formulas, Graphs, and Mathematical Tables, 9th\n               printing.  New York: Dover, 1972.\n\n        .. [2] The Laplace distribution and generalizations\n               By Samuel Kotz, Tomasz J. Kozubowski, Krzysztof Podgorski,\n               Birkhauser, 2001.\n\n        .. [3] Weisstein, Eric W. \"Laplace Distribution.\"\n               From MathWorld--A Wolfram Web Resource.\n               http://mathworld.wolfram.com/LaplaceDistribution.html\n\n        .. [4] Wikipedia, \"Laplace distribution\",\n               http://en.wikipedia.org/wik""i/Laplace_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution\n\n        >>> loc, scale = 0., 1.\n        >>> s = np.random.laplace(loc, scale, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 30, normed=True)\n        >>> x = np.arange(-8., 8., .01)\n        >>> pdf = np.exp(-abs(x-loc/scale))/(2.*scale)\n        >>> plt.plot(x, pdf)\n\n        Plot Gaussian for comparison:\n\n        >>> g = (1/(scale * np.sqrt(2 * np.pi)) * \n        ...      np.exp( - (x - loc)**2 / (2 * scale**2) ))\n        >>> plt.plot(x,g)\n\n        ";
     5820-static PyObject *__pyx_pf_6mtrand_11RandomState_33laplace(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     5821+/* Python wrapper */
     5822+static PyObject *__pyx_pw_6mtrand_11RandomState_67laplace(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     5823+static char __pyx_doc_6mtrand_11RandomState_66laplace[] = "\n        laplace(loc=0.0, scale=1.0, size=None)\n\n        Draw samples from the Laplace or double exponential distribution with\n        specified location (or mean) and scale (decay).\n\n        The Laplace distribution is similar to the Gaussian/normal distribution,\n        but is sharper at the peak and has fatter tails. It represents the\n        difference between two independent, identically distributed exponential\n        random variables.\n\n        Parameters\n        ----------\n        loc : float\n            The position, :math:`\\mu`, of the distribution peak.\n        scale : float\n            :math:`\\lambda`, the exponential decay.\n\n        Notes\n        -----\n        It has the probability density function\n\n        .. math:: f(x; \\mu, \\lambda) = \\frac{1}{2\\lambda}\n                                       \\exp\\left(-\\frac{|x - \\mu|}{\\lambda}\\right).\n\n        The first law of Laplace, from 1774, states that the frequency of an error\n        can be expressed as an exponential function of the absolute magnitude of\n        the error, which leads to the Laplace distribution. For many problems in\n        Economics and Health sciences, this distribution seems to model the data\n        better than the standard Gaussian distribution\n\n\n        References\n        ----------\n        .. [1] Abramowitz, M. and Stegun, I. A. (Eds.). Handbook of Mathematical\n               Functions with Formulas, Graphs, and Mathematical Tables, 9th\n               printing.  New York: Dover, 1972.\n\n        .. [2] The Laplace distribution and generalizations\n               By Samuel Kotz, Tomasz J. Kozubowski, Krzysztof Podgorski,\n               Birkhauser, 2001.\n\n        .. [3] Weisstein, Eric W. \"Laplace Distribution.\"\n               From MathWorld--A Wolfram Web Resource.\n               http://mathworld.wolfram.com/LaplaceDistribution.html\n\n        .. [4] Wikipedia, \"Laplace distribution\",\n               http://en.wikipedia.org/wik""i/Laplace_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution\n\n        >>> loc, scale = 0., 1.\n        >>> s = np.random.laplace(loc, scale, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 30, normed=True)\n        >>> x = np.arange(-8., 8., .01)\n        >>> pdf = np.exp(-abs(x-loc/scale))/(2.*scale)\n        >>> plt.plot(x, pdf)\n\n        Plot Gaussian for comparison:\n\n        >>> g = (1/(scale * np.sqrt(2 * np.pi)) * \n        ...      np.exp( - (x - loc)**2 / (2 * scale**2) ))\n        >>> plt.plot(x,g)\n\n        ";
     5824+static PyObject *__pyx_pw_6mtrand_11RandomState_67laplace(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     5825   PyObject *__pyx_v_loc = 0;
     5826   PyObject *__pyx_v_scale = 0;
     5827   PyObject *__pyx_v_size = 0;
     5828-  PyArrayObject *__pyx_v_oloc = 0;
     5829-  PyArrayObject *__pyx_v_oscale = 0;
     5830-  double __pyx_v_floc;
     5831-  double __pyx_v_fscale;
     5832-  PyObject *__pyx_r = NULL;
     5833+  PyObject *__pyx_r = 0;
     5834   __Pyx_RefNannyDeclarations
     5835-  int __pyx_t_1;
     5836-  PyObject *__pyx_t_2 = NULL;
     5837-  PyObject *__pyx_t_3 = NULL;
     5838-  PyObject *__pyx_t_4 = NULL;
     5839-  PyObject *__pyx_t_5 = NULL;
     5840-  int __pyx_lineno = 0;
     5841-  const char *__pyx_filename = NULL;
     5842-  int __pyx_clineno = 0;
     5843-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__loc,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     5844-  __Pyx_RefNannySetupContext("laplace");
     5845+  __Pyx_RefNannySetupContext("laplace (wrapper)", 0);
     5846   {
     5847+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__loc,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     5848     PyObject* values[3] = {0,0,0};
     5849     values[0] = __pyx_k_73;
     5850     values[1] = __pyx_k_74;
     5851+
     5852+    /* "mtrand.pyx":2555
     5853+ *         return cont1_array(self.internal_state, rk_power, size, oa)
     5854+ *
     5855+ *     def laplace(self, loc=0.0, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     5856+ *         """
     5857+ *         laplace(loc=0.0, scale=1.0, size=None)
     5858+ */
     5859     values[2] = ((PyObject *)Py_None);
     5860     if (unlikely(__pyx_kwds)) {
     5861       Py_ssize_t kw_args;
     5862-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     5863+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     5864+      switch (pos_args) {
     5865         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     5866         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     5867         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     5868@@ -11542,7 +12003,7 @@
     5869         default: goto __pyx_L5_argtuple_error;
     5870       }
     5871       kw_args = PyDict_Size(__pyx_kwds);
     5872-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     5873+      switch (pos_args) {
     5874         case  0:
     5875         if (kw_args > 0) {
     5876           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__loc);
     5877@@ -11560,7 +12021,7 @@
     5878         }
     5879       }
     5880       if (unlikely(kw_args > 0)) {
     5881-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "laplace") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2555; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5882+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "laplace") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2555; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     5883       }
     5884     } else {
     5885       switch (PyTuple_GET_SIZE(__pyx_args)) {
     5886@@ -11583,6 +12044,27 @@
     5887   __Pyx_RefNannyFinishContext();
     5888   return NULL;
     5889   __pyx_L4_argument_unpacking_done:;
     5890+  __pyx_r = __pyx_pf_6mtrand_11RandomState_66laplace(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_loc, __pyx_v_scale, __pyx_v_size);
     5891+  __Pyx_RefNannyFinishContext();
     5892+  return __pyx_r;
     5893+}
     5894+
     5895+static PyObject *__pyx_pf_6mtrand_11RandomState_66laplace(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_loc, PyObject *__pyx_v_scale, PyObject *__pyx_v_size) {
     5896+  PyArrayObject *__pyx_v_oloc = 0;
     5897+  PyArrayObject *__pyx_v_oscale = 0;
     5898+  double __pyx_v_floc;
     5899+  double __pyx_v_fscale;
     5900+  PyObject *__pyx_r = NULL;
     5901+  __Pyx_RefNannyDeclarations
     5902+  int __pyx_t_1;
     5903+  PyObject *__pyx_t_2 = NULL;
     5904+  PyObject *__pyx_t_3 = NULL;
     5905+  PyObject *__pyx_t_4 = NULL;
     5906+  PyObject *__pyx_t_5 = NULL;
     5907+  int __pyx_lineno = 0;
     5908+  const char *__pyx_filename = NULL;
     5909+  int __pyx_clineno = 0;
     5910+  __Pyx_RefNannySetupContext("laplace", 0);
     5911 
     5912   /* "mtrand.pyx":2631
     5913  *         cdef double floc, fscale
     5914@@ -11634,9 +12116,9 @@
     5915       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5916       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5917       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2635; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5918-      goto __pyx_L7;
     5919+      goto __pyx_L4;
     5920     }
     5921-    __pyx_L7:;
     5922+    __pyx_L4:;
     5923 
     5924     /* "mtrand.pyx":2636
     5925  *             if fscale <= 0:
     5926@@ -11646,14 +12128,14 @@
     5927  *         PyErr_Clear()
     5928  */
     5929     __Pyx_XDECREF(__pyx_r);
     5930-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_laplace, __pyx_v_size, __pyx_v_floc, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2636; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5931+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_laplace, __pyx_v_size, __pyx_v_floc, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2636; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5932     __Pyx_GOTREF(__pyx_t_2);
     5933     __pyx_r = __pyx_t_2;
     5934     __pyx_t_2 = 0;
     5935     goto __pyx_L0;
     5936-    goto __pyx_L6;
     5937+    goto __pyx_L3;
     5938   }
     5939-  __pyx_L6:;
     5940+  __pyx_L3:;
     5941 
     5942   /* "mtrand.pyx":2638
     5943  *             return cont2_array_sc(self.internal_state, rk_laplace, size, floc, fscale)
     5944@@ -11710,7 +12192,7 @@
     5945   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2641; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5946   __Pyx_GOTREF(__pyx_t_2);
     5947   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2641; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5948-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5949+  __Pyx_GOTREF(__pyx_t_5);
     5950   __Pyx_INCREF(((PyObject *)__pyx_v_oscale));
     5951   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_oscale));
     5952   __Pyx_GIVEREF(((PyObject *)__pyx_v_oscale));
     5953@@ -11722,7 +12204,7 @@
     5954   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     5955   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     5956   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2641; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5957-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     5958+  __Pyx_GOTREF(__pyx_t_5);
     5959   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     5960   __Pyx_GIVEREF(__pyx_t_2);
     5961   __pyx_t_2 = 0;
     5962@@ -11746,9 +12228,9 @@
     5963     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     5964     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     5965     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2642; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5966-    goto __pyx_L8;
     5967+    goto __pyx_L5;
     5968   }
     5969-  __pyx_L8:;
     5970+  __pyx_L5:;
     5971 
     5972   /* "mtrand.pyx":2643
     5973  *         if np.any(np.less_equal(oscale, 0.0)):
     5974@@ -11758,7 +12240,7 @@
     5975  *     def gumbel(self, loc=0.0, scale=1.0, size=None):
     5976  */
     5977   __Pyx_XDECREF(__pyx_r);
     5978-  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_laplace, __pyx_v_size, __pyx_v_oloc, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2643; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5979+  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_laplace, __pyx_v_size, __pyx_v_oloc, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2643; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     5980   __Pyx_GOTREF(__pyx_t_2);
     5981   __pyx_r = __pyx_t_2;
     5982   __pyx_t_2 = 0;
     5983@@ -11781,44 +12263,34 @@
     5984   return __pyx_r;
     5985 }
     5986 
     5987-/* "mtrand.pyx":2645
     5988- *         return cont2_array(self.internal_state, rk_laplace, size, oloc, oscale)
     5989- *
     5990- *     def gumbel(self, loc=0.0, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     5991- *         """
     5992- *         gumbel(loc=0.0, scale=1.0, size=None)
     5993- */
     5994-
     5995-static PyObject *__pyx_pf_6mtrand_11RandomState_34gumbel(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     5996-static char __pyx_doc_6mtrand_11RandomState_34gumbel[] = "\n        gumbel(loc=0.0, scale=1.0, size=None)\n\n        Gumbel distribution.\n\n        Draw samples from a Gumbel distribution with specified location and scale.\n        For more information on the Gumbel distribution, see Notes and References\n        below.\n\n        Parameters\n        ----------\n        loc : float\n            The location of the mode of the distribution.\n        scale : float\n            The scale parameter of the distribution.\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        out : ndarray\n            The samples\n\n        See Also\n        --------\n        scipy.stats.gumbel_l\n        scipy.stats.gumbel_r\n        scipy.stats.genextreme\n            probability density function, distribution, or cumulative density\n            function, etc. for each of the above\n        weibull\n\n        Notes\n        -----\n        The Gumbel (or Smallest Extreme Value (SEV) or the Smallest Extreme Value\n        Type I) distribution is one of a class of Generalized Extreme Value (GEV)\n        distributions used in modeling extreme value problems.  The Gumbel is a\n        special case of the Extreme Value Type I distribution for maximums from\n        distributions with \"exponential-like\" tails.\n\n        The probability density for the Gumbel distribution is\n\n        .. math:: p(x) = \\frac{e^{-(x - \\mu)/ \\beta}}{\\beta} e^{ -e^{-(x - \\mu)/\n                  \\beta}},\n\n        where :math:`\\mu` is the mode, a location parameter, and :math:`\\beta` is\n        the scale parameter.\n\n        The Gumbel (named for German mathematician Emil Julius Gumbel) was used\n        very early in the hydrology literature, for modeling the occurrence of\n        flood events. It is also used for modeling maximum wind speed and rainfall\n        rates.  It is a \"fat-tailed\" distribution - the ""probability of an event in\n        the tail of the distribution is larger than if one used a Gaussian, hence\n        the surprisingly frequent occurrence of 100-year floods. Floods were\n        initially modeled as a Gaussian process, which underestimated the frequency\n        of extreme events.\n\n\n        It is one of a class of extreme value distributions, the Generalized\n        Extreme Value (GEV) distributions, which also includes the Weibull and\n        Frechet.\n\n        The function has a mean of :math:`\\mu + 0.57721\\beta` and a variance of\n        :math:`\\frac{\\pi^2}{6}\\beta^2`.\n\n        References\n        ----------\n        Gumbel, E. J., *Statistics of Extremes*, New York: Columbia University\n        Press, 1958.\n\n        Reiss, R.-D. and Thomas, M., *Statistical Analysis of Extreme Values from\n        Insurance, Finance, Hydrology and Other Fields*, Basel: Birkhauser Verlag,\n        2001.\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> mu, beta = 0, 0.1 # location and scale\n        >>> s = np.random.gumbel(mu, beta, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 30, normed=True)\n        >>> plt.plot(bins, (1/beta)*np.exp(-(bins - mu)/beta)\n        ...          * np.exp( -np.exp( -(bins - mu) /beta) ),\n        ...          linewidth=2, color='r')\n        >>> plt.show()\n\n        Show how an extreme value distribution can arise from a Gaussian process\n        and compare to a Gaussian:\n\n        >>> means = []\n        >>> maxima = []\n        >>> for i in range(0,1000) :\n        ...    a = np.random.normal(mu, beta, 1000)\n        ...    means.append(a.mean())\n        ...    maxima.append(a.max())\n        >>> count, bins, ignored = plt.hist(maxima, 30, normed=True)\n        >>> beta = np.std(maxima)*np.pi/np.sqrt(6)""\n        >>> mu = np.mean(maxima) - 0.57721*beta\n        >>> plt.plot(bins, (1/beta)*np.exp(-(bins - mu)/beta)\n        ...          * np.exp(-np.exp(-(bins - mu)/beta)),\n        ...          linewidth=2, color='r')\n        >>> plt.plot(bins, 1/(beta * np.sqrt(2 * np.pi))\n        ...          * np.exp(-(bins - mu)**2 / (2 * beta**2)),\n        ...          linewidth=2, color='g')\n        >>> plt.show()\n\n        ";
     5997-static PyObject *__pyx_pf_6mtrand_11RandomState_34gumbel(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     5998+/* Python wrapper */
     5999+static PyObject *__pyx_pw_6mtrand_11RandomState_69gumbel(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     6000+static char __pyx_doc_6mtrand_11RandomState_68gumbel[] = "\n        gumbel(loc=0.0, scale=1.0, size=None)\n\n        Gumbel distribution.\n\n        Draw samples from a Gumbel distribution with specified location and scale.\n        For more information on the Gumbel distribution, see Notes and References\n        below.\n\n        Parameters\n        ----------\n        loc : float\n            The location of the mode of the distribution.\n        scale : float\n            The scale parameter of the distribution.\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        out : ndarray\n            The samples\n\n        See Also\n        --------\n        scipy.stats.gumbel_l\n        scipy.stats.gumbel_r\n        scipy.stats.genextreme\n            probability density function, distribution, or cumulative density\n            function, etc. for each of the above\n        weibull\n\n        Notes\n        -----\n        The Gumbel (or Smallest Extreme Value (SEV) or the Smallest Extreme Value\n        Type I) distribution is one of a class of Generalized Extreme Value (GEV)\n        distributions used in modeling extreme value problems.  The Gumbel is a\n        special case of the Extreme Value Type I distribution for maximums from\n        distributions with \"exponential-like\" tails.\n\n        The probability density for the Gumbel distribution is\n\n        .. math:: p(x) = \\frac{e^{-(x - \\mu)/ \\beta}}{\\beta} e^{ -e^{-(x - \\mu)/\n                  \\beta}},\n\n        where :math:`\\mu` is the mode, a location parameter, and :math:`\\beta` is\n        the scale parameter.\n\n        The Gumbel (named for German mathematician Emil Julius Gumbel) was used\n        very early in the hydrology literature, for modeling the occurrence of\n        flood events. It is also used for modeling maximum wind speed and rainfall\n        rates.  It is a \"fat-tailed\" distribution - the ""probability of an event in\n        the tail of the distribution is larger than if one used a Gaussian, hence\n        the surprisingly frequent occurrence of 100-year floods. Floods were\n        initially modeled as a Gaussian process, which underestimated the frequency\n        of extreme events.\n\n\n        It is one of a class of extreme value distributions, the Generalized\n        Extreme Value (GEV) distributions, which also includes the Weibull and\n        Frechet.\n\n        The function has a mean of :math:`\\mu + 0.57721\\beta` and a variance of\n        :math:`\\frac{\\pi^2}{6}\\beta^2`.\n\n        References\n        ----------\n        Gumbel, E. J., *Statistics of Extremes*, New York: Columbia University\n        Press, 1958.\n\n        Reiss, R.-D. and Thomas, M., *Statistical Analysis of Extreme Values from\n        Insurance, Finance, Hydrology and Other Fields*, Basel: Birkhauser Verlag,\n        2001.\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> mu, beta = 0, 0.1 # location and scale\n        >>> s = np.random.gumbel(mu, beta, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 30, normed=True)\n        >>> plt.plot(bins, (1/beta)*np.exp(-(bins - mu)/beta)\n        ...          * np.exp( -np.exp( -(bins - mu) /beta) ),\n        ...          linewidth=2, color='r')\n        >>> plt.show()\n\n        Show how an extreme value distribution can arise from a Gaussian process\n        and compare to a Gaussian:\n\n        >>> means = []\n        >>> maxima = []\n        >>> for i in range(0,1000) :\n        ...    a = np.random.normal(mu, beta, 1000)\n        ...    means.append(a.mean())\n        ...    maxima.append(a.max())\n        >>> count, bins, ignored = plt.hist(maxima, 30, normed=True)\n        >>> beta = np.std(maxima)*np.pi/np.sqrt(6)""\n        >>> mu = np.mean(maxima) - 0.57721*beta\n        >>> plt.plot(bins, (1/beta)*np.exp(-(bins - mu)/beta)\n        ...          * np.exp(-np.exp(-(bins - mu)/beta)),\n        ...          linewidth=2, color='r')\n        >>> plt.plot(bins, 1/(beta * np.sqrt(2 * np.pi))\n        ...          * np.exp(-(bins - mu)**2 / (2 * beta**2)),\n        ...          linewidth=2, color='g')\n        >>> plt.show()\n\n        ";
     6001+static PyObject *__pyx_pw_6mtrand_11RandomState_69gumbel(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     6002   PyObject *__pyx_v_loc = 0;
     6003   PyObject *__pyx_v_scale = 0;
     6004   PyObject *__pyx_v_size = 0;
     6005-  PyArrayObject *__pyx_v_oloc = 0;
     6006-  PyArrayObject *__pyx_v_oscale = 0;
     6007-  double __pyx_v_floc;
     6008-  double __pyx_v_fscale;
     6009-  PyObject *__pyx_r = NULL;
     6010+  PyObject *__pyx_r = 0;
     6011   __Pyx_RefNannyDeclarations
     6012-  int __pyx_t_1;
     6013-  PyObject *__pyx_t_2 = NULL;
     6014-  PyObject *__pyx_t_3 = NULL;
     6015-  PyObject *__pyx_t_4 = NULL;
     6016-  PyObject *__pyx_t_5 = NULL;
     6017-  int __pyx_lineno = 0;
     6018-  const char *__pyx_filename = NULL;
     6019-  int __pyx_clineno = 0;
     6020-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__loc,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     6021-  __Pyx_RefNannySetupContext("gumbel");
     6022+  __Pyx_RefNannySetupContext("gumbel (wrapper)", 0);
     6023   {
     6024+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__loc,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     6025     PyObject* values[3] = {0,0,0};
     6026     values[0] = __pyx_k_77;
     6027     values[1] = __pyx_k_78;
     6028+
     6029+    /* "mtrand.pyx":2645
     6030+ *         return cont2_array(self.internal_state, rk_laplace, size, oloc, oscale)
     6031+ *
     6032+ *     def gumbel(self, loc=0.0, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     6033+ *         """
     6034+ *         gumbel(loc=0.0, scale=1.0, size=None)
     6035+ */
     6036     values[2] = ((PyObject *)Py_None);
     6037     if (unlikely(__pyx_kwds)) {
     6038       Py_ssize_t kw_args;
     6039-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6040+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     6041+      switch (pos_args) {
     6042         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     6043         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     6044         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     6045@@ -11826,7 +12298,7 @@
     6046         default: goto __pyx_L5_argtuple_error;
     6047       }
     6048       kw_args = PyDict_Size(__pyx_kwds);
     6049-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6050+      switch (pos_args) {
     6051         case  0:
     6052         if (kw_args > 0) {
     6053           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__loc);
     6054@@ -11844,7 +12316,7 @@
     6055         }
     6056       }
     6057       if (unlikely(kw_args > 0)) {
     6058-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "gumbel") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2645; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6059+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "gumbel") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2645; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6060       }
     6061     } else {
     6062       switch (PyTuple_GET_SIZE(__pyx_args)) {
     6063@@ -11867,6 +12339,27 @@
     6064   __Pyx_RefNannyFinishContext();
     6065   return NULL;
     6066   __pyx_L4_argument_unpacking_done:;
     6067+  __pyx_r = __pyx_pf_6mtrand_11RandomState_68gumbel(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_loc, __pyx_v_scale, __pyx_v_size);
     6068+  __Pyx_RefNannyFinishContext();
     6069+  return __pyx_r;
     6070+}
     6071+
     6072+static PyObject *__pyx_pf_6mtrand_11RandomState_68gumbel(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_loc, PyObject *__pyx_v_scale, PyObject *__pyx_v_size) {
     6073+  PyArrayObject *__pyx_v_oloc = 0;
     6074+  PyArrayObject *__pyx_v_oscale = 0;
     6075+  double __pyx_v_floc;
     6076+  double __pyx_v_fscale;
     6077+  PyObject *__pyx_r = NULL;
     6078+  __Pyx_RefNannyDeclarations
     6079+  int __pyx_t_1;
     6080+  PyObject *__pyx_t_2 = NULL;
     6081+  PyObject *__pyx_t_3 = NULL;
     6082+  PyObject *__pyx_t_4 = NULL;
     6083+  PyObject *__pyx_t_5 = NULL;
     6084+  int __pyx_lineno = 0;
     6085+  const char *__pyx_filename = NULL;
     6086+  int __pyx_clineno = 0;
     6087+  __Pyx_RefNannySetupContext("gumbel", 0);
     6088 
     6089   /* "mtrand.pyx":2762
     6090  *         cdef double floc, fscale
     6091@@ -11918,9 +12411,9 @@
     6092       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6093       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6094       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2766; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6095-      goto __pyx_L7;
     6096+      goto __pyx_L4;
     6097     }
     6098-    __pyx_L7:;
     6099+    __pyx_L4:;
     6100 
     6101     /* "mtrand.pyx":2767
     6102  *             if fscale <= 0:
     6103@@ -11930,14 +12423,14 @@
     6104  *         PyErr_Clear()
     6105  */
     6106     __Pyx_XDECREF(__pyx_r);
     6107-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_gumbel, __pyx_v_size, __pyx_v_floc, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2767; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6108+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_gumbel, __pyx_v_size, __pyx_v_floc, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2767; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6109     __Pyx_GOTREF(__pyx_t_2);
     6110     __pyx_r = __pyx_t_2;
     6111     __pyx_t_2 = 0;
     6112     goto __pyx_L0;
     6113-    goto __pyx_L6;
     6114+    goto __pyx_L3;
     6115   }
     6116-  __pyx_L6:;
     6117+  __pyx_L3:;
     6118 
     6119   /* "mtrand.pyx":2769
     6120  *             return cont2_array_sc(self.internal_state, rk_gumbel, size, floc, fscale)
     6121@@ -11994,7 +12487,7 @@
     6122   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2772; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6123   __Pyx_GOTREF(__pyx_t_2);
     6124   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2772; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6125-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     6126+  __Pyx_GOTREF(__pyx_t_5);
     6127   __Pyx_INCREF(((PyObject *)__pyx_v_oscale));
     6128   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_oscale));
     6129   __Pyx_GIVEREF(((PyObject *)__pyx_v_oscale));
     6130@@ -12006,7 +12499,7 @@
     6131   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     6132   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     6133   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2772; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6134-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     6135+  __Pyx_GOTREF(__pyx_t_5);
     6136   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     6137   __Pyx_GIVEREF(__pyx_t_2);
     6138   __pyx_t_2 = 0;
     6139@@ -12030,9 +12523,9 @@
     6140     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6141     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6142     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2773; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6143-    goto __pyx_L8;
     6144+    goto __pyx_L5;
     6145   }
     6146-  __pyx_L8:;
     6147+  __pyx_L5:;
     6148 
     6149   /* "mtrand.pyx":2774
     6150  *         if np.any(np.less_equal(oscale, 0.0)):
     6151@@ -12042,7 +12535,7 @@
     6152  *     def logistic(self, loc=0.0, scale=1.0, size=None):
     6153  */
     6154   __Pyx_XDECREF(__pyx_r);
     6155-  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_gumbel, __pyx_v_size, __pyx_v_oloc, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2774; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6156+  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_gumbel, __pyx_v_size, __pyx_v_oloc, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2774; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6157   __Pyx_GOTREF(__pyx_t_2);
     6158   __pyx_r = __pyx_t_2;
     6159   __pyx_t_2 = 0;
     6160@@ -12065,44 +12558,34 @@
     6161   return __pyx_r;
     6162 }
     6163 
     6164-/* "mtrand.pyx":2776
     6165- *         return cont2_array(self.internal_state, rk_gumbel, size, oloc, oscale)
     6166- *
     6167- *     def logistic(self, loc=0.0, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     6168- *         """
     6169- *         logistic(loc=0.0, scale=1.0, size=None)
     6170- */
     6171-
     6172-static PyObject *__pyx_pf_6mtrand_11RandomState_35logistic(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     6173-static char __pyx_doc_6mtrand_11RandomState_35logistic[] = "\n        logistic(loc=0.0, scale=1.0, size=None)\n\n        Draw samples from a Logistic distribution.\n\n        Samples are drawn from a Logistic distribution with specified\n        parameters, loc (location or mean, also median), and scale (>0).\n\n        Parameters\n        ----------\n        loc : float\n\n        scale : float > 0.\n\n        size : {tuple, int}\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n                  where the values are all integers in  [0, n].\n\n        See Also\n        --------\n        scipy.stats.distributions.logistic : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Logistic distribution is\n\n        .. math:: P(x) = P(x) = \\frac{e^{-(x-\\mu)/s}}{s(1+e^{-(x-\\mu)/s})^2},\n\n        where :math:`\\mu` = location and :math:`s` = scale.\n\n        The Logistic distribution is used in Extreme Value problems where it\n        can act as a mixture of Gumbel distributions, in Epidemiology, and by\n        the World Chess Federation (FIDE) where it is used in the Elo ranking\n        system, assuming the performance of each player is a logistically\n        distributed random variable.\n\n        References\n        ----------\n        .. [1] Reiss, R.-D. and Thomas M. (2001), Statistical Analysis of Extreme\n               Values, from Insurance, Finance, Hydrology and Other Fields,\n               Birkhauser Verlag, Basel, pp 132-133.\n        .. [2] Weisstein, Eric W. \"Logistic Distribution.\" From\n               MathWorld--A Wolfram Web Resource.\n               http://mathworld.wolfram.com/LogisticDistribution.html\n        .. [3] Wikipedia, \"Logistic-distribution\",\n               http://en.wikipedia.org/wiki/Logistic-distribution\n\n        Examples\n   ""     --------\n        Draw samples from the distribution:\n\n        >>> loc, scale = 10, 1\n        >>> s = np.random.logistic(loc, scale, 10000)\n        >>> count, bins, ignored = plt.hist(s, bins=50)\n\n        #   plot against distribution\n\n        >>> def logist(x, loc, scale):\n        ...     return exp((loc-x)/scale)/(scale*(1+exp((loc-x)/scale))**2)\n        >>> plt.plot(bins, logist(bins, loc, scale)*count.max()/\\\n        ... logist(bins, loc, scale).max())\n        >>> plt.show()\n\n        ";
     6174-static PyObject *__pyx_pf_6mtrand_11RandomState_35logistic(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     6175+/* Python wrapper */
     6176+static PyObject *__pyx_pw_6mtrand_11RandomState_71logistic(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     6177+static char __pyx_doc_6mtrand_11RandomState_70logistic[] = "\n        logistic(loc=0.0, scale=1.0, size=None)\n\n        Draw samples from a Logistic distribution.\n\n        Samples are drawn from a Logistic distribution with specified\n        parameters, loc (location or mean, also median), and scale (>0).\n\n        Parameters\n        ----------\n        loc : float\n\n        scale : float > 0.\n\n        size : {tuple, int}\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n                  where the values are all integers in  [0, n].\n\n        See Also\n        --------\n        scipy.stats.distributions.logistic : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Logistic distribution is\n\n        .. math:: P(x) = P(x) = \\frac{e^{-(x-\\mu)/s}}{s(1+e^{-(x-\\mu)/s})^2},\n\n        where :math:`\\mu` = location and :math:`s` = scale.\n\n        The Logistic distribution is used in Extreme Value problems where it\n        can act as a mixture of Gumbel distributions, in Epidemiology, and by\n        the World Chess Federation (FIDE) where it is used in the Elo ranking\n        system, assuming the performance of each player is a logistically\n        distributed random variable.\n\n        References\n        ----------\n        .. [1] Reiss, R.-D. and Thomas M. (2001), Statistical Analysis of Extreme\n               Values, from Insurance, Finance, Hydrology and Other Fields,\n               Birkhauser Verlag, Basel, pp 132-133.\n        .. [2] Weisstein, Eric W. \"Logistic Distribution.\" From\n               MathWorld--A Wolfram Web Resource.\n               http://mathworld.wolfram.com/LogisticDistribution.html\n        .. [3] Wikipedia, \"Logistic-distribution\",\n               http://en.wikipedia.org/wiki/Logistic-distribution\n\n        Examples\n   ""     --------\n        Draw samples from the distribution:\n\n        >>> loc, scale = 10, 1\n        >>> s = np.random.logistic(loc, scale, 10000)\n        >>> count, bins, ignored = plt.hist(s, bins=50)\n\n        #   plot against distribution\n\n        >>> def logist(x, loc, scale):\n        ...     return exp((loc-x)/scale)/(scale*(1+exp((loc-x)/scale))**2)\n        >>> plt.plot(bins, logist(bins, loc, scale)*count.max()/\\\n        ... logist(bins, loc, scale).max())\n        >>> plt.show()\n\n        ";
     6178+static PyObject *__pyx_pw_6mtrand_11RandomState_71logistic(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     6179   PyObject *__pyx_v_loc = 0;
     6180   PyObject *__pyx_v_scale = 0;
     6181   PyObject *__pyx_v_size = 0;
     6182-  PyArrayObject *__pyx_v_oloc = 0;
     6183-  PyArrayObject *__pyx_v_oscale = 0;
     6184-  double __pyx_v_floc;
     6185-  double __pyx_v_fscale;
     6186-  PyObject *__pyx_r = NULL;
     6187+  PyObject *__pyx_r = 0;
     6188   __Pyx_RefNannyDeclarations
     6189-  int __pyx_t_1;
     6190-  PyObject *__pyx_t_2 = NULL;
     6191-  PyObject *__pyx_t_3 = NULL;
     6192-  PyObject *__pyx_t_4 = NULL;
     6193-  PyObject *__pyx_t_5 = NULL;
     6194-  int __pyx_lineno = 0;
     6195-  const char *__pyx_filename = NULL;
     6196-  int __pyx_clineno = 0;
     6197-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__loc,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     6198-  __Pyx_RefNannySetupContext("logistic");
     6199+  __Pyx_RefNannySetupContext("logistic (wrapper)", 0);
     6200   {
     6201+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__loc,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     6202     PyObject* values[3] = {0,0,0};
     6203     values[0] = __pyx_k_81;
     6204     values[1] = __pyx_k_82;
     6205+
     6206+    /* "mtrand.pyx":2776
     6207+ *         return cont2_array(self.internal_state, rk_gumbel, size, oloc, oscale)
     6208+ *
     6209+ *     def logistic(self, loc=0.0, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     6210+ *         """
     6211+ *         logistic(loc=0.0, scale=1.0, size=None)
     6212+ */
     6213     values[2] = ((PyObject *)Py_None);
     6214     if (unlikely(__pyx_kwds)) {
     6215       Py_ssize_t kw_args;
     6216-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6217+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     6218+      switch (pos_args) {
     6219         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     6220         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     6221         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     6222@@ -12110,7 +12593,7 @@
     6223         default: goto __pyx_L5_argtuple_error;
     6224       }
     6225       kw_args = PyDict_Size(__pyx_kwds);
     6226-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6227+      switch (pos_args) {
     6228         case  0:
     6229         if (kw_args > 0) {
     6230           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__loc);
     6231@@ -12128,7 +12611,7 @@
     6232         }
     6233       }
     6234       if (unlikely(kw_args > 0)) {
     6235-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "logistic") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2776; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6236+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "logistic") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2776; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6237       }
     6238     } else {
     6239       switch (PyTuple_GET_SIZE(__pyx_args)) {
     6240@@ -12151,6 +12634,27 @@
     6241   __Pyx_RefNannyFinishContext();
     6242   return NULL;
     6243   __pyx_L4_argument_unpacking_done:;
     6244+  __pyx_r = __pyx_pf_6mtrand_11RandomState_70logistic(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_loc, __pyx_v_scale, __pyx_v_size);
     6245+  __Pyx_RefNannyFinishContext();
     6246+  return __pyx_r;
     6247+}
     6248+
     6249+static PyObject *__pyx_pf_6mtrand_11RandomState_70logistic(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_loc, PyObject *__pyx_v_scale, PyObject *__pyx_v_size) {
     6250+  PyArrayObject *__pyx_v_oloc = 0;
     6251+  PyArrayObject *__pyx_v_oscale = 0;
     6252+  double __pyx_v_floc;
     6253+  double __pyx_v_fscale;
     6254+  PyObject *__pyx_r = NULL;
     6255+  __Pyx_RefNannyDeclarations
     6256+  int __pyx_t_1;
     6257+  PyObject *__pyx_t_2 = NULL;
     6258+  PyObject *__pyx_t_3 = NULL;
     6259+  PyObject *__pyx_t_4 = NULL;
     6260+  PyObject *__pyx_t_5 = NULL;
     6261+  int __pyx_lineno = 0;
     6262+  const char *__pyx_filename = NULL;
     6263+  int __pyx_clineno = 0;
     6264+  __Pyx_RefNannySetupContext("logistic", 0);
     6265 
     6266   /* "mtrand.pyx":2850
     6267  *         cdef double floc, fscale
     6268@@ -12202,9 +12706,9 @@
     6269       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6270       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6271       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2854; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6272-      goto __pyx_L7;
     6273+      goto __pyx_L4;
     6274     }
     6275-    __pyx_L7:;
     6276+    __pyx_L4:;
     6277 
     6278     /* "mtrand.pyx":2855
     6279  *             if fscale <= 0:
     6280@@ -12214,14 +12718,14 @@
     6281  *         PyErr_Clear()
     6282  */
     6283     __Pyx_XDECREF(__pyx_r);
     6284-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_logistic, __pyx_v_size, __pyx_v_floc, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2855; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6285+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_logistic, __pyx_v_size, __pyx_v_floc, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2855; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6286     __Pyx_GOTREF(__pyx_t_2);
     6287     __pyx_r = __pyx_t_2;
     6288     __pyx_t_2 = 0;
     6289     goto __pyx_L0;
     6290-    goto __pyx_L6;
     6291+    goto __pyx_L3;
     6292   }
     6293-  __pyx_L6:;
     6294+  __pyx_L3:;
     6295 
     6296   /* "mtrand.pyx":2857
     6297  *             return cont2_array_sc(self.internal_state, rk_logistic, size, floc, fscale)
     6298@@ -12278,7 +12782,7 @@
     6299   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2860; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6300   __Pyx_GOTREF(__pyx_t_2);
     6301   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2860; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6302-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     6303+  __Pyx_GOTREF(__pyx_t_5);
     6304   __Pyx_INCREF(((PyObject *)__pyx_v_oscale));
     6305   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_oscale));
     6306   __Pyx_GIVEREF(((PyObject *)__pyx_v_oscale));
     6307@@ -12290,7 +12794,7 @@
     6308   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     6309   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     6310   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2860; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6311-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     6312+  __Pyx_GOTREF(__pyx_t_5);
     6313   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     6314   __Pyx_GIVEREF(__pyx_t_2);
     6315   __pyx_t_2 = 0;
     6316@@ -12314,9 +12818,9 @@
     6317     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6318     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6319     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2861; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6320-    goto __pyx_L8;
     6321+    goto __pyx_L5;
     6322   }
     6323-  __pyx_L8:;
     6324+  __pyx_L5:;
     6325 
     6326   /* "mtrand.pyx":2862
     6327  *         if np.any(np.less_equal(oscale, 0.0)):
     6328@@ -12326,7 +12830,7 @@
     6329  *     def lognormal(self, mean=0.0, sigma=1.0, size=None):
     6330  */
     6331   __Pyx_XDECREF(__pyx_r);
     6332-  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_logistic, __pyx_v_size, __pyx_v_oloc, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2862; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6333+  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_logistic, __pyx_v_size, __pyx_v_oloc, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2862; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6334   __Pyx_GOTREF(__pyx_t_2);
     6335   __pyx_r = __pyx_t_2;
     6336   __pyx_t_2 = 0;
     6337@@ -12349,44 +12853,34 @@
     6338   return __pyx_r;
     6339 }
     6340 
     6341-/* "mtrand.pyx":2864
     6342- *         return cont2_array(self.internal_state, rk_logistic, size, oloc, oscale)
     6343- *
     6344- *     def lognormal(self, mean=0.0, sigma=1.0, size=None):             # <<<<<<<<<<<<<<
     6345- *         """
     6346- *         lognormal(mean=0.0, sigma=1.0, size=None)
     6347- */
     6348-
     6349-static PyObject *__pyx_pf_6mtrand_11RandomState_36lognormal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     6350-static char __pyx_doc_6mtrand_11RandomState_36lognormal[] = "\n        lognormal(mean=0.0, sigma=1.0, size=None)\n\n        Return samples drawn from a log-normal distribution.\n\n        Draw samples from a log-normal distribution with specified mean, standard\n        deviation, and shape. Note that the mean and standard deviation are not the\n        values for the distribution itself, but of the underlying normal\n        distribution it is derived from.\n\n\n        Parameters\n        ----------\n        mean : float\n            Mean value of the underlying normal distribution\n        sigma : float, >0.\n            Standard deviation of the underlying normal distribution\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        See Also\n        --------\n        scipy.stats.lognorm : probability density function, distribution,\n            cumulative density function, etc.\n\n        Notes\n        -----\n        A variable `x` has a log-normal distribution if `log(x)` is normally\n        distributed.\n\n        The probability density function for the log-normal distribution is\n\n        .. math:: p(x) = \\frac{1}{\\sigma x \\sqrt{2\\pi}}\n                         e^{(-\\frac{(ln(x)-\\mu)^2}{2\\sigma^2})}\n\n        where :math:`\\mu` is the mean and :math:`\\sigma` is the standard deviation\n        of the normally distributed logarithm of the variable.\n\n        A log-normal distribution results if a random variable is the *product* of\n        a large number of independent, identically-distributed variables in the\n        same way that a normal distribution results if the variable is the *sum*\n        of a large number of independent, identically-distributed variables\n        (see the last example). It is one of the so-called \"fat-tailed\"\n        distributions.\n\n        The log-normal distribution is commonly used to model the lifespan of units\n        with fatigue-stress failure modes. Since thi""s includes\n        most mechanical systems, the log-normal distribution has widespread\n        application.\n\n        It is also commonly used to model oil field sizes, species abundance, and\n        latent periods of infectious diseases.\n\n        References\n        ----------\n        .. [1] Eckhard Limpert, Werner A. Stahel, and Markus Abbt, \"Log-normal\n               Distributions across the Sciences: Keys and Clues\", May 2001\n               Vol. 51 No. 5 BioScience\n               http://stat.ethz.ch/~stahel/lognormal/bioscience.pdf\n        .. [2] Reiss, R.D., Thomas, M.(2001), Statistical Analysis of Extreme\n               Values, Birkhauser Verlag, Basel, pp 31-32.\n        .. [3] Wikipedia, \"Lognormal distribution\",\n               http://en.wikipedia.org/wiki/Lognormal_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> mu, sigma = 3., 1. # mean and standard deviation\n        >>> s = np.random.lognormal(mu, sigma, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 100, normed=True, align='mid')\n\n        >>> x = np.linspace(min(bins), max(bins), 10000)\n        >>> pdf = (np.exp(-(np.log(x) - mu)**2 / (2 * sigma**2))\n        ...        / (x * sigma * np.sqrt(2 * np.pi)))\n\n        >>> plt.plot(x, pdf, linewidth=2, color='r')\n        >>> plt.axis('tight')\n        >>> plt.show()\n\n        Demonstrate that taking the products of random samples from a uniform\n        distribution can be fit well by a log-normal probability density function.\n\n        >>> # Generate a thousand samples: each is the product of 100 random\n        >>> # values, drawn from a normal distribution.\n        >>> b = []\n        >>> for i in range(1000):\n        ...    a = 10. + np.random.random(100)\n        ...    b.append(np.product(a))\n\n        >>> b"" = np.array(b) / np.min(b) # scale values to be positive\n\n        >>> count, bins, ignored = plt.hist(b, 100, normed=True, align='center')\n\n        >>> sigma = np.std(np.log(b))\n        >>> mu = np.mean(np.log(b))\n\n        >>> x = np.linspace(min(bins), max(bins), 10000)\n        >>> pdf = (np.exp(-(np.log(x) - mu)**2 / (2 * sigma**2))\n        ...        / (x * sigma * np.sqrt(2 * np.pi)))\n\n        >>> plt.plot(x, pdf, color='r', linewidth=2)\n        >>> plt.show()\n\n        ";
     6351-static PyObject *__pyx_pf_6mtrand_11RandomState_36lognormal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     6352+/* Python wrapper */
     6353+static PyObject *__pyx_pw_6mtrand_11RandomState_73lognormal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     6354+static char __pyx_doc_6mtrand_11RandomState_72lognormal[] = "\n        lognormal(mean=0.0, sigma=1.0, size=None)\n\n        Return samples drawn from a log-normal distribution.\n\n        Draw samples from a log-normal distribution with specified mean, standard\n        deviation, and shape. Note that the mean and standard deviation are not the\n        values for the distribution itself, but of the underlying normal\n        distribution it is derived from.\n\n\n        Parameters\n        ----------\n        mean : float\n            Mean value of the underlying normal distribution\n        sigma : float, >0.\n            Standard deviation of the underlying normal distribution\n        size : tuple of ints\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        See Also\n        --------\n        scipy.stats.lognorm : probability density function, distribution,\n            cumulative density function, etc.\n\n        Notes\n        -----\n        A variable `x` has a log-normal distribution if `log(x)` is normally\n        distributed.\n\n        The probability density function for the log-normal distribution is\n\n        .. math:: p(x) = \\frac{1}{\\sigma x \\sqrt{2\\pi}}\n                         e^{(-\\frac{(ln(x)-\\mu)^2}{2\\sigma^2})}\n\n        where :math:`\\mu` is the mean and :math:`\\sigma` is the standard deviation\n        of the normally distributed logarithm of the variable.\n\n        A log-normal distribution results if a random variable is the *product* of\n        a large number of independent, identically-distributed variables in the\n        same way that a normal distribution results if the variable is the *sum*\n        of a large number of independent, identically-distributed variables\n        (see the last example). It is one of the so-called \"fat-tailed\"\n        distributions.\n\n        The log-normal distribution is commonly used to model the lifespan of units\n        with fatigue-stress failure modes. Since thi""s includes\n        most mechanical systems, the log-normal distribution has widespread\n        application.\n\n        It is also commonly used to model oil field sizes, species abundance, and\n        latent periods of infectious diseases.\n\n        References\n        ----------\n        .. [1] Eckhard Limpert, Werner A. Stahel, and Markus Abbt, \"Log-normal\n               Distributions across the Sciences: Keys and Clues\", May 2001\n               Vol. 51 No. 5 BioScience\n               http://stat.ethz.ch/~stahel/lognormal/bioscience.pdf\n        .. [2] Reiss, R.D., Thomas, M.(2001), Statistical Analysis of Extreme\n               Values, Birkhauser Verlag, Basel, pp 31-32.\n        .. [3] Wikipedia, \"Lognormal distribution\",\n               http://en.wikipedia.org/wiki/Lognormal_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> mu, sigma = 3., 1. # mean and standard deviation\n        >>> s = np.random.lognormal(mu, sigma, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 100, normed=True, align='mid')\n\n        >>> x = np.linspace(min(bins), max(bins), 10000)\n        >>> pdf = (np.exp(-(np.log(x) - mu)**2 / (2 * sigma**2))\n        ...        / (x * sigma * np.sqrt(2 * np.pi)))\n\n        >>> plt.plot(x, pdf, linewidth=2, color='r')\n        >>> plt.axis('tight')\n        >>> plt.show()\n\n        Demonstrate that taking the products of random samples from a uniform\n        distribution can be fit well by a log-normal probability density function.\n\n        >>> # Generate a thousand samples: each is the product of 100 random\n        >>> # values, drawn from a normal distribution.\n        >>> b = []\n        >>> for i in range(1000):\n        ...    a = 10. + np.random.random(100)\n        ...    b.append(np.product(a))\n\n        >>> b"" = np.array(b) / np.min(b) # scale values to be positive\n\n        >>> count, bins, ignored = plt.hist(b, 100, normed=True, align='center')\n\n        >>> sigma = np.std(np.log(b))\n        >>> mu = np.mean(np.log(b))\n\n        >>> x = np.linspace(min(bins), max(bins), 10000)\n        >>> pdf = (np.exp(-(np.log(x) - mu)**2 / (2 * sigma**2))\n        ...        / (x * sigma * np.sqrt(2 * np.pi)))\n\n        >>> plt.plot(x, pdf, color='r', linewidth=2)\n        >>> plt.show()\n\n        ";
     6355+static PyObject *__pyx_pw_6mtrand_11RandomState_73lognormal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     6356   PyObject *__pyx_v_mean = 0;
     6357   PyObject *__pyx_v_sigma = 0;
     6358   PyObject *__pyx_v_size = 0;
     6359-  PyArrayObject *__pyx_v_omean = 0;
     6360-  PyArrayObject *__pyx_v_osigma = 0;
     6361-  double __pyx_v_fmean;
     6362-  double __pyx_v_fsigma;
     6363-  PyObject *__pyx_r = NULL;
     6364+  PyObject *__pyx_r = 0;
     6365   __Pyx_RefNannyDeclarations
     6366-  int __pyx_t_1;
     6367-  PyObject *__pyx_t_2 = NULL;
     6368-  PyObject *__pyx_t_3 = NULL;
     6369-  PyObject *__pyx_t_4 = NULL;
     6370-  PyObject *__pyx_t_5 = NULL;
     6371-  int __pyx_lineno = 0;
     6372-  const char *__pyx_filename = NULL;
     6373-  int __pyx_clineno = 0;
     6374-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__mean,&__pyx_n_s__sigma,&__pyx_n_s__size,0};
     6375-  __Pyx_RefNannySetupContext("lognormal");
     6376+  __Pyx_RefNannySetupContext("lognormal (wrapper)", 0);
     6377   {
     6378+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__mean,&__pyx_n_s__sigma,&__pyx_n_s__size,0};
     6379     PyObject* values[3] = {0,0,0};
     6380     values[0] = __pyx_k_85;
     6381     values[1] = __pyx_k_86;
     6382+
     6383+    /* "mtrand.pyx":2864
     6384+ *         return cont2_array(self.internal_state, rk_logistic, size, oloc, oscale)
     6385+ *
     6386+ *     def lognormal(self, mean=0.0, sigma=1.0, size=None):             # <<<<<<<<<<<<<<
     6387+ *         """
     6388+ *         lognormal(mean=0.0, sigma=1.0, size=None)
     6389+ */
     6390     values[2] = ((PyObject *)Py_None);
     6391     if (unlikely(__pyx_kwds)) {
     6392       Py_ssize_t kw_args;
     6393-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6394+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     6395+      switch (pos_args) {
     6396         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     6397         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     6398         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     6399@@ -12394,7 +12888,7 @@
     6400         default: goto __pyx_L5_argtuple_error;
     6401       }
     6402       kw_args = PyDict_Size(__pyx_kwds);
     6403-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6404+      switch (pos_args) {
     6405         case  0:
     6406         if (kw_args > 0) {
     6407           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__mean);
     6408@@ -12412,7 +12906,7 @@
     6409         }
     6410       }
     6411       if (unlikely(kw_args > 0)) {
     6412-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "lognormal") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2864; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6413+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "lognormal") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2864; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6414       }
     6415     } else {
     6416       switch (PyTuple_GET_SIZE(__pyx_args)) {
     6417@@ -12435,6 +12929,27 @@
     6418   __Pyx_RefNannyFinishContext();
     6419   return NULL;
     6420   __pyx_L4_argument_unpacking_done:;
     6421+  __pyx_r = __pyx_pf_6mtrand_11RandomState_72lognormal(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_mean, __pyx_v_sigma, __pyx_v_size);
     6422+  __Pyx_RefNannyFinishContext();
     6423+  return __pyx_r;
     6424+}
     6425+
     6426+static PyObject *__pyx_pf_6mtrand_11RandomState_72lognormal(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_mean, PyObject *__pyx_v_sigma, PyObject *__pyx_v_size) {
     6427+  PyArrayObject *__pyx_v_omean = 0;
     6428+  PyArrayObject *__pyx_v_osigma = 0;
     6429+  double __pyx_v_fmean;
     6430+  double __pyx_v_fsigma;
     6431+  PyObject *__pyx_r = NULL;
     6432+  __Pyx_RefNannyDeclarations
     6433+  int __pyx_t_1;
     6434+  PyObject *__pyx_t_2 = NULL;
     6435+  PyObject *__pyx_t_3 = NULL;
     6436+  PyObject *__pyx_t_4 = NULL;
     6437+  PyObject *__pyx_t_5 = NULL;
     6438+  int __pyx_lineno = 0;
     6439+  const char *__pyx_filename = NULL;
     6440+  int __pyx_clineno = 0;
     6441+  __Pyx_RefNannySetupContext("lognormal", 0);
     6442 
     6443   /* "mtrand.pyx":2979
     6444  *         cdef double fmean, fsigma
     6445@@ -12486,9 +13001,9 @@
     6446       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6447       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6448       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2984; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6449-      goto __pyx_L7;
     6450+      goto __pyx_L4;
     6451     }
     6452-    __pyx_L7:;
     6453+    __pyx_L4:;
     6454 
     6455     /* "mtrand.pyx":2985
     6456  *             if fsigma <= 0:
     6457@@ -12498,14 +13013,14 @@
     6458  *         PyErr_Clear()
     6459  */
     6460     __Pyx_XDECREF(__pyx_r);
     6461-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_lognormal, __pyx_v_size, __pyx_v_fmean, __pyx_v_fsigma); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2985; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6462+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_lognormal, __pyx_v_size, __pyx_v_fmean, __pyx_v_fsigma); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2985; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6463     __Pyx_GOTREF(__pyx_t_2);
     6464     __pyx_r = __pyx_t_2;
     6465     __pyx_t_2 = 0;
     6466     goto __pyx_L0;
     6467-    goto __pyx_L6;
     6468+    goto __pyx_L3;
     6469   }
     6470-  __pyx_L6:;
     6471+  __pyx_L3:;
     6472 
     6473   /* "mtrand.pyx":2987
     6474  *             return cont2_array_sc(self.internal_state, rk_lognormal, size, fmean, fsigma)
     6475@@ -12562,7 +13077,7 @@
     6476   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2991; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6477   __Pyx_GOTREF(__pyx_t_2);
     6478   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2991; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6479-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     6480+  __Pyx_GOTREF(__pyx_t_5);
     6481   __Pyx_INCREF(((PyObject *)__pyx_v_osigma));
     6482   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_osigma));
     6483   __Pyx_GIVEREF(((PyObject *)__pyx_v_osigma));
     6484@@ -12574,7 +13089,7 @@
     6485   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     6486   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     6487   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2991; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6488-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     6489+  __Pyx_GOTREF(__pyx_t_5);
     6490   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     6491   __Pyx_GIVEREF(__pyx_t_2);
     6492   __pyx_t_2 = 0;
     6493@@ -12598,9 +13113,9 @@
     6494     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6495     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6496     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2992; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6497-    goto __pyx_L8;
     6498+    goto __pyx_L5;
     6499   }
     6500-  __pyx_L8:;
     6501+  __pyx_L5:;
     6502 
     6503   /* "mtrand.pyx":2993
     6504  *         if np.any(np.less_equal(osigma, 0.0)):
     6505@@ -12610,7 +13125,7 @@
     6506  *     def rayleigh(self, scale=1.0, size=None):
     6507  */
     6508   __Pyx_XDECREF(__pyx_r);
     6509-  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_lognormal, __pyx_v_size, __pyx_v_omean, __pyx_v_osigma); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2993; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6510+  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_lognormal, __pyx_v_size, __pyx_v_omean, __pyx_v_osigma); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2993; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6511   __Pyx_GOTREF(__pyx_t_2);
     6512   __pyx_r = __pyx_t_2;
     6513   __pyx_t_2 = 0;
     6514@@ -12633,47 +13148,39 @@
     6515   return __pyx_r;
     6516 }
     6517 
     6518-/* "mtrand.pyx":2995
     6519- *         return cont2_array(self.internal_state, rk_lognormal, size, omean, osigma)
     6520- *
     6521- *     def rayleigh(self, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     6522- *         """
     6523- *         rayleigh(scale=1.0, size=None)
     6524- */
     6525-
     6526-static PyObject *__pyx_pf_6mtrand_11RandomState_37rayleigh(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     6527-static char __pyx_doc_6mtrand_11RandomState_37rayleigh[] = "\n        rayleigh(scale=1.0, size=None)\n\n        Draw samples from a Rayleigh distribution.\n\n        The :math:`\\chi` and Weibull distributions are generalizations of the\n        Rayleigh.\n\n        Parameters\n        ----------\n        scale : scalar\n            Scale, also equals the mode. Should be >= 0.\n        size : int or tuple of ints, optional\n            Shape of the output. Default is None, in which case a single\n            value is returned.\n\n        Notes\n        -----\n        The probability density function for the Rayleigh distribution is\n\n        .. math:: P(x;scale) = \\frac{x}{scale^2}e^{\\frac{-x^2}{2 \\cdotp scale^2}}\n\n        The Rayleigh distribution arises if the wind speed and wind direction are\n        both gaussian variables, then the vector wind velocity forms a Rayleigh\n        distribution. The Rayleigh distribution is used to model the expected\n        output from wind turbines.\n\n        References\n        ----------\n        ..[1] Brighton Webs Ltd., Rayleigh Distribution,\n              http://www.brighton-webs.co.uk/distributions/rayleigh.asp\n        ..[2] Wikipedia, \"Rayleigh distribution\"\n              http://en.wikipedia.org/wiki/Rayleigh_distribution\n\n        Examples\n        --------\n        Draw values from the distribution and plot the histogram\n\n        >>> values = hist(np.random.rayleigh(3, 100000), bins=200, normed=True)\n\n        Wave heights tend to follow a Rayleigh distribution. If the mean wave\n        height is 1 meter, what fraction of waves are likely to be larger than 3\n        meters?\n\n        >>> meanvalue = 1\n        >>> modevalue = np.sqrt(2 / np.pi) * meanvalue\n        >>> s = np.random.rayleigh(modevalue, 1000000)\n\n        The percentage of waves larger than 3 meters is:\n\n        >>> 100.*sum(s>3)/1000000.\n        0.087300000000000003\n\n        ";
     6528-static PyObject *__pyx_pf_6mtrand_11RandomState_37rayleigh(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     6529+/* Python wrapper */
     6530+static PyObject *__pyx_pw_6mtrand_11RandomState_75rayleigh(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     6531+static char __pyx_doc_6mtrand_11RandomState_74rayleigh[] = "\n        rayleigh(scale=1.0, size=None)\n\n        Draw samples from a Rayleigh distribution.\n\n        The :math:`\\chi` and Weibull distributions are generalizations of the\n        Rayleigh.\n\n        Parameters\n        ----------\n        scale : scalar\n            Scale, also equals the mode. Should be >= 0.\n        size : int or tuple of ints, optional\n            Shape of the output. Default is None, in which case a single\n            value is returned.\n\n        Notes\n        -----\n        The probability density function for the Rayleigh distribution is\n\n        .. math:: P(x;scale) = \\frac{x}{scale^2}e^{\\frac{-x^2}{2 \\cdotp scale^2}}\n\n        The Rayleigh distribution arises if the wind speed and wind direction are\n        both gaussian variables, then the vector wind velocity forms a Rayleigh\n        distribution. The Rayleigh distribution is used to model the expected\n        output from wind turbines.\n\n        References\n        ----------\n        ..[1] Brighton Webs Ltd., Rayleigh Distribution,\n              http://www.brighton-webs.co.uk/distributions/rayleigh.asp\n        ..[2] Wikipedia, \"Rayleigh distribution\"\n              http://en.wikipedia.org/wiki/Rayleigh_distribution\n\n        Examples\n        --------\n        Draw values from the distribution and plot the histogram\n\n        >>> values = hist(np.random.rayleigh(3, 100000), bins=200, normed=True)\n\n        Wave heights tend to follow a Rayleigh distribution. If the mean wave\n        height is 1 meter, what fraction of waves are likely to be larger than 3\n        meters?\n\n        >>> meanvalue = 1\n        >>> modevalue = np.sqrt(2 / np.pi) * meanvalue\n        >>> s = np.random.rayleigh(modevalue, 1000000)\n\n        The percentage of waves larger than 3 meters is:\n\n        >>> 100.*sum(s>3)/1000000.\n        0.087300000000000003\n\n        ";
     6532+static PyObject *__pyx_pw_6mtrand_11RandomState_75rayleigh(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     6533   PyObject *__pyx_v_scale = 0;
     6534   PyObject *__pyx_v_size = 0;
     6535-  PyArrayObject *__pyx_v_oscale = 0;
     6536-  double __pyx_v_fscale;
     6537-  PyObject *__pyx_r = NULL;
     6538+  PyObject *__pyx_r = 0;
     6539   __Pyx_RefNannyDeclarations
     6540-  int __pyx_t_1;
     6541-  PyObject *__pyx_t_2 = NULL;
     6542-  PyObject *__pyx_t_3 = NULL;
     6543-  PyObject *__pyx_t_4 = NULL;
     6544-  PyObject *__pyx_t_5 = NULL;
     6545-  int __pyx_lineno = 0;
     6546-  const char *__pyx_filename = NULL;
     6547-  int __pyx_clineno = 0;
     6548-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__scale,&__pyx_n_s__size,0};
     6549-  __Pyx_RefNannySetupContext("rayleigh");
     6550+  __Pyx_RefNannySetupContext("rayleigh (wrapper)", 0);
     6551   {
     6552+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__scale,&__pyx_n_s__size,0};
     6553     PyObject* values[2] = {0,0};
     6554     values[0] = __pyx_k_91;
     6555+
     6556+    /* "mtrand.pyx":2995
     6557+ *         return cont2_array(self.internal_state, rk_lognormal, size, omean, osigma)
     6558+ *
     6559+ *     def rayleigh(self, scale=1.0, size=None):             # <<<<<<<<<<<<<<
     6560+ *         """
     6561+ *         rayleigh(scale=1.0, size=None)
     6562+ */
     6563     values[1] = ((PyObject *)Py_None);
     6564     if (unlikely(__pyx_kwds)) {
     6565       Py_ssize_t kw_args;
     6566-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6567+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     6568+      switch (pos_args) {
     6569         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     6570         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     6571         case  0: break;
     6572         default: goto __pyx_L5_argtuple_error;
     6573       }
     6574       kw_args = PyDict_Size(__pyx_kwds);
     6575-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6576+      switch (pos_args) {
     6577         case  0:
     6578         if (kw_args > 0) {
     6579           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__scale);
     6580@@ -12686,7 +13193,7 @@
     6581         }
     6582       }
     6583       if (unlikely(kw_args > 0)) {
     6584-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "rayleigh") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2995; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6585+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "rayleigh") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2995; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6586       }
     6587     } else {
     6588       switch (PyTuple_GET_SIZE(__pyx_args)) {
     6589@@ -12707,6 +13214,25 @@
     6590   __Pyx_RefNannyFinishContext();
     6591   return NULL;
     6592   __pyx_L4_argument_unpacking_done:;
     6593+  __pyx_r = __pyx_pf_6mtrand_11RandomState_74rayleigh(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_scale, __pyx_v_size);
     6594+  __Pyx_RefNannyFinishContext();
     6595+  return __pyx_r;
     6596+}
     6597+
     6598+static PyObject *__pyx_pf_6mtrand_11RandomState_74rayleigh(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_scale, PyObject *__pyx_v_size) {
     6599+  PyArrayObject *__pyx_v_oscale = 0;
     6600+  double __pyx_v_fscale;
     6601+  PyObject *__pyx_r = NULL;
     6602+  __Pyx_RefNannyDeclarations
     6603+  int __pyx_t_1;
     6604+  PyObject *__pyx_t_2 = NULL;
     6605+  PyObject *__pyx_t_3 = NULL;
     6606+  PyObject *__pyx_t_4 = NULL;
     6607+  PyObject *__pyx_t_5 = NULL;
     6608+  int __pyx_lineno = 0;
     6609+  const char *__pyx_filename = NULL;
     6610+  int __pyx_clineno = 0;
     6611+  __Pyx_RefNannySetupContext("rayleigh", 0);
     6612 
     6613   /* "mtrand.pyx":3053
     6614  *         cdef double fscale
     6615@@ -12749,9 +13275,9 @@
     6616       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6617       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6618       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3057; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6619-      goto __pyx_L7;
     6620+      goto __pyx_L4;
     6621     }
     6622-    __pyx_L7:;
     6623+    __pyx_L4:;
     6624 
     6625     /* "mtrand.pyx":3058
     6626  *             if fscale <= 0:
     6627@@ -12761,14 +13287,14 @@
     6628  *         PyErr_Clear()
     6629  */
     6630     __Pyx_XDECREF(__pyx_r);
     6631-    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_rayleigh, __pyx_v_size, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3058; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6632+    __pyx_t_2 = __pyx_f_6mtrand_cont1_array_sc(__pyx_v_self->internal_state, rk_rayleigh, __pyx_v_size, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3058; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6633     __Pyx_GOTREF(__pyx_t_2);
     6634     __pyx_r = __pyx_t_2;
     6635     __pyx_t_2 = 0;
     6636     goto __pyx_L0;
     6637-    goto __pyx_L6;
     6638+    goto __pyx_L3;
     6639   }
     6640-  __pyx_L6:;
     6641+  __pyx_L3:;
     6642 
     6643   /* "mtrand.pyx":3060
     6644  *             return cont1_array_sc(self.internal_state, rk_rayleigh, size, fscale)
     6645@@ -12812,7 +13338,7 @@
     6646   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3063; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6647   __Pyx_GOTREF(__pyx_t_2);
     6648   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3063; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6649-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     6650+  __Pyx_GOTREF(__pyx_t_5);
     6651   __Pyx_INCREF(((PyObject *)__pyx_v_oscale));
     6652   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_oscale));
     6653   __Pyx_GIVEREF(((PyObject *)__pyx_v_oscale));
     6654@@ -12824,7 +13350,7 @@
     6655   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     6656   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     6657   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3063; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6658-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     6659+  __Pyx_GOTREF(__pyx_t_5);
     6660   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     6661   __Pyx_GIVEREF(__pyx_t_2);
     6662   __pyx_t_2 = 0;
     6663@@ -12848,9 +13374,9 @@
     6664     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6665     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6666     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3064; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6667-    goto __pyx_L8;
     6668+    goto __pyx_L5;
     6669   }
     6670-  __pyx_L8:;
     6671+  __pyx_L5:;
     6672 
     6673   /* "mtrand.pyx":3065
     6674  *         if np.any(np.less_equal(oscale, 0.0)):
     6675@@ -12860,7 +13386,7 @@
     6676  *     def wald(self, mean, scale, size=None):
     6677  */
     6678   __Pyx_XDECREF(__pyx_r);
     6679-  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_rayleigh, __pyx_v_size, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3065; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6680+  __pyx_t_2 = __pyx_f_6mtrand_cont1_array(__pyx_v_self->internal_state, rk_rayleigh, __pyx_v_size, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3065; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6681   __Pyx_GOTREF(__pyx_t_2);
     6682   __pyx_r = __pyx_t_2;
     6683   __pyx_t_2 = 0;
     6684@@ -12882,42 +13408,32 @@
     6685   return __pyx_r;
     6686 }
     6687 
     6688-/* "mtrand.pyx":3067
     6689- *         return cont1_array(self.internal_state, rk_rayleigh, size, oscale)
     6690- *
     6691- *     def wald(self, mean, scale, size=None):             # <<<<<<<<<<<<<<
     6692- *         """
     6693- *         wald(mean, scale, size=None)
     6694- */
     6695-
     6696-static PyObject *__pyx_pf_6mtrand_11RandomState_38wald(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     6697-static char __pyx_doc_6mtrand_11RandomState_38wald[] = "\n        wald(mean, scale, size=None)\n\n        Draw samples from a Wald, or Inverse Gaussian, distribution.\n\n        As the scale approaches infinity, the distribution becomes more like a\n        Gaussian.\n\n        Some references claim that the Wald is an Inverse Gaussian with mean=1, but\n        this is by no means universal.\n\n        The Inverse Gaussian distribution was first studied in relationship to\n        Brownian motion. In 1956 M.C.K. Tweedie used the name Inverse Gaussian\n        because there is an inverse relationship between the time to cover a unit\n        distance and distance covered in unit time.\n\n        Parameters\n        ----------\n        mean : scalar\n            Distribution mean, should be > 0.\n        scale : scalar\n            Scale parameter, should be >= 0.\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single value is\n            returned.\n\n        Returns\n        -------\n        samples : ndarray or scalar\n            Drawn sample, all greater than zero.\n\n        Notes\n        -----\n        The probability density function for the Wald distribution is\n\n        .. math:: P(x;mean,scale) = \\sqrt{\\frac{scale}{2\\pi x^3}}e^\n                                    \\frac{-scale(x-mean)^2}{2\\cdotp mean^2x}\n\n        As noted above the Inverse Gaussian distribution first arise from attempts\n        to model Brownian Motion. It is also a competitor to the Weibull for use in\n        reliability modeling and modeling stock returns and interest rate\n        processes.\n\n        References\n        ----------\n        ..[1] Brighton Webs Ltd., Wald Distribution,\n              http://www.brighton-webs.co.uk/distributions/wald.asp\n        ..[2] Chhikara, Raj S., and Folks, J. Leroy, \"The Inverse Gaussian\n              Distribution: Theory : Methodology, and Applications\", CRC Press,\n              1988.\n        ..[3] Wikipedia, \"Wald distributio""n\"\n              http://en.wikipedia.org/wiki/Wald_distribution\n\n        Examples\n        --------\n        Draw values from the distribution and plot the histogram:\n\n        >>> import matplotlib.pyplot as plt\n        >>> h = plt.hist(np.random.wald(3, 2, 100000), bins=200, normed=True)\n        >>> plt.show()\n\n        ";
     6698-static PyObject *__pyx_pf_6mtrand_11RandomState_38wald(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     6699+/* Python wrapper */
     6700+static PyObject *__pyx_pw_6mtrand_11RandomState_77wald(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     6701+static char __pyx_doc_6mtrand_11RandomState_76wald[] = "\n        wald(mean, scale, size=None)\n\n        Draw samples from a Wald, or Inverse Gaussian, distribution.\n\n        As the scale approaches infinity, the distribution becomes more like a\n        Gaussian.\n\n        Some references claim that the Wald is an Inverse Gaussian with mean=1, but\n        this is by no means universal.\n\n        The Inverse Gaussian distribution was first studied in relationship to\n        Brownian motion. In 1956 M.C.K. Tweedie used the name Inverse Gaussian\n        because there is an inverse relationship between the time to cover a unit\n        distance and distance covered in unit time.\n\n        Parameters\n        ----------\n        mean : scalar\n            Distribution mean, should be > 0.\n        scale : scalar\n            Scale parameter, should be >= 0.\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single value is\n            returned.\n\n        Returns\n        -------\n        samples : ndarray or scalar\n            Drawn sample, all greater than zero.\n\n        Notes\n        -----\n        The probability density function for the Wald distribution is\n\n        .. math:: P(x;mean,scale) = \\sqrt{\\frac{scale}{2\\pi x^3}}e^\n                                    \\frac{-scale(x-mean)^2}{2\\cdotp mean^2x}\n\n        As noted above the Inverse Gaussian distribution first arise from attempts\n        to model Brownian Motion. It is also a competitor to the Weibull for use in\n        reliability modeling and modeling stock returns and interest rate\n        processes.\n\n        References\n        ----------\n        ..[1] Brighton Webs Ltd., Wald Distribution,\n              http://www.brighton-webs.co.uk/distributions/wald.asp\n        ..[2] Chhikara, Raj S., and Folks, J. Leroy, \"The Inverse Gaussian\n              Distribution: Theory : Methodology, and Applications\", CRC Press,\n              1988.\n        ..[3] Wikipedia, \"Wald distributio""n\"\n              http://en.wikipedia.org/wiki/Wald_distribution\n\n        Examples\n        --------\n        Draw values from the distribution and plot the histogram:\n\n        >>> import matplotlib.pyplot as plt\n        >>> h = plt.hist(np.random.wald(3, 2, 100000), bins=200, normed=True)\n        >>> plt.show()\n\n        ";
     6702+static PyObject *__pyx_pw_6mtrand_11RandomState_77wald(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     6703   PyObject *__pyx_v_mean = 0;
     6704   PyObject *__pyx_v_scale = 0;
     6705   PyObject *__pyx_v_size = 0;
     6706-  PyArrayObject *__pyx_v_omean = 0;
     6707-  PyArrayObject *__pyx_v_oscale = 0;
     6708-  double __pyx_v_fmean;
     6709-  double __pyx_v_fscale;
     6710-  PyObject *__pyx_r = NULL;
     6711+  PyObject *__pyx_r = 0;
     6712   __Pyx_RefNannyDeclarations
     6713-  int __pyx_t_1;
     6714-  PyObject *__pyx_t_2 = NULL;
     6715-  PyObject *__pyx_t_3 = NULL;
     6716-  PyObject *__pyx_t_4 = NULL;
     6717-  PyObject *__pyx_t_5 = NULL;
     6718-  int __pyx_lineno = 0;
     6719-  const char *__pyx_filename = NULL;
     6720-  int __pyx_clineno = 0;
     6721-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__mean,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     6722-  __Pyx_RefNannySetupContext("wald");
     6723+  __Pyx_RefNannySetupContext("wald (wrapper)", 0);
     6724   {
     6725+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__mean,&__pyx_n_s__scale,&__pyx_n_s__size,0};
     6726     PyObject* values[3] = {0,0,0};
     6727+
     6728+    /* "mtrand.pyx":3067
     6729+ *         return cont1_array(self.internal_state, rk_rayleigh, size, oscale)
     6730+ *
     6731+ *     def wald(self, mean, scale, size=None):             # <<<<<<<<<<<<<<
     6732+ *         """
     6733+ *         wald(mean, scale, size=None)
     6734+ */
     6735     values[2] = ((PyObject *)Py_None);
     6736     if (unlikely(__pyx_kwds)) {
     6737       Py_ssize_t kw_args;
     6738-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6739+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     6740+      switch (pos_args) {
     6741         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     6742         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     6743         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     6744@@ -12925,14 +13441,12 @@
     6745         default: goto __pyx_L5_argtuple_error;
     6746       }
     6747       kw_args = PyDict_Size(__pyx_kwds);
     6748-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6749+      switch (pos_args) {
     6750         case  0:
     6751-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__mean);
     6752-        if (likely(values[0])) kw_args--;
     6753+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__mean)) != 0)) kw_args--;
     6754         else goto __pyx_L5_argtuple_error;
     6755         case  1:
     6756-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__scale);
     6757-        if (likely(values[1])) kw_args--;
     6758+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__scale)) != 0)) kw_args--;
     6759         else {
     6760           __Pyx_RaiseArgtupleInvalid("wald", 0, 2, 3, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3067; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6761         }
     6762@@ -12943,7 +13457,7 @@
     6763         }
     6764       }
     6765       if (unlikely(kw_args > 0)) {
     6766-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "wald") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3067; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6767+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "wald") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3067; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6768       }
     6769     } else {
     6770       switch (PyTuple_GET_SIZE(__pyx_args)) {
     6771@@ -12966,6 +13480,27 @@
     6772   __Pyx_RefNannyFinishContext();
     6773   return NULL;
     6774   __pyx_L4_argument_unpacking_done:;
     6775+  __pyx_r = __pyx_pf_6mtrand_11RandomState_76wald(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_mean, __pyx_v_scale, __pyx_v_size);
     6776+  __Pyx_RefNannyFinishContext();
     6777+  return __pyx_r;
     6778+}
     6779+
     6780+static PyObject *__pyx_pf_6mtrand_11RandomState_76wald(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_mean, PyObject *__pyx_v_scale, PyObject *__pyx_v_size) {
     6781+  PyArrayObject *__pyx_v_omean = 0;
     6782+  PyArrayObject *__pyx_v_oscale = 0;
     6783+  double __pyx_v_fmean;
     6784+  double __pyx_v_fscale;
     6785+  PyObject *__pyx_r = NULL;
     6786+  __Pyx_RefNannyDeclarations
     6787+  int __pyx_t_1;
     6788+  PyObject *__pyx_t_2 = NULL;
     6789+  PyObject *__pyx_t_3 = NULL;
     6790+  PyObject *__pyx_t_4 = NULL;
     6791+  PyObject *__pyx_t_5 = NULL;
     6792+  int __pyx_lineno = 0;
     6793+  const char *__pyx_filename = NULL;
     6794+  int __pyx_clineno = 0;
     6795+  __Pyx_RefNannySetupContext("wald", 0);
     6796 
     6797   /* "mtrand.pyx":3133
     6798  *         cdef double fmean, fscale
     6799@@ -13017,9 +13552,9 @@
     6800       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6801       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6802       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3137; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6803-      goto __pyx_L7;
     6804+      goto __pyx_L4;
     6805     }
     6806-    __pyx_L7:;
     6807+    __pyx_L4:;
     6808 
     6809     /* "mtrand.pyx":3138
     6810  *             if fmean <= 0:
     6811@@ -13043,9 +13578,9 @@
     6812       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6813       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6814       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3139; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6815-      goto __pyx_L8;
     6816+      goto __pyx_L5;
     6817     }
     6818-    __pyx_L8:;
     6819+    __pyx_L5:;
     6820 
     6821     /* "mtrand.pyx":3140
     6822  *             if fscale <= 0:
     6823@@ -13055,14 +13590,14 @@
     6824  *         PyErr_Clear()
     6825  */
     6826     __Pyx_XDECREF(__pyx_r);
     6827-    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_wald, __pyx_v_size, __pyx_v_fmean, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3140; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6828+    __pyx_t_2 = __pyx_f_6mtrand_cont2_array_sc(__pyx_v_self->internal_state, rk_wald, __pyx_v_size, __pyx_v_fmean, __pyx_v_fscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3140; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6829     __Pyx_GOTREF(__pyx_t_2);
     6830     __pyx_r = __pyx_t_2;
     6831     __pyx_t_2 = 0;
     6832     goto __pyx_L0;
     6833-    goto __pyx_L6;
     6834+    goto __pyx_L3;
     6835   }
     6836-  __pyx_L6:;
     6837+  __pyx_L3:;
     6838 
     6839   /* "mtrand.pyx":3142
     6840  *             return cont2_array_sc(self.internal_state, rk_wald, size, fmean, fscale)
     6841@@ -13119,7 +13654,7 @@
     6842   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3145; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6843   __Pyx_GOTREF(__pyx_t_2);
     6844   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3145; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6845-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     6846+  __Pyx_GOTREF(__pyx_t_5);
     6847   __Pyx_INCREF(((PyObject *)__pyx_v_omean));
     6848   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_omean));
     6849   __Pyx_GIVEREF(((PyObject *)__pyx_v_omean));
     6850@@ -13131,7 +13666,7 @@
     6851   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     6852   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     6853   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3145; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6854-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     6855+  __Pyx_GOTREF(__pyx_t_5);
     6856   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     6857   __Pyx_GIVEREF(__pyx_t_2);
     6858   __pyx_t_2 = 0;
     6859@@ -13155,7 +13690,7 @@
     6860     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6861     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6862     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3146; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6863-    goto __pyx_L9;
     6864+    goto __pyx_L6;
     6865   }
     6866 
     6867   /* "mtrand.pyx":3147
     6868@@ -13178,7 +13713,7 @@
     6869   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3147; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6870   __Pyx_GOTREF(__pyx_t_2);
     6871   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3147; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6872-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     6873+  __Pyx_GOTREF(__pyx_t_4);
     6874   __Pyx_INCREF(((PyObject *)__pyx_v_oscale));
     6875   PyTuple_SET_ITEM(__pyx_t_4, 0, ((PyObject *)__pyx_v_oscale));
     6876   __Pyx_GIVEREF(((PyObject *)__pyx_v_oscale));
     6877@@ -13190,7 +13725,7 @@
     6878   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     6879   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     6880   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3147; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6881-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     6882+  __Pyx_GOTREF(__pyx_t_4);
     6883   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     6884   __Pyx_GIVEREF(__pyx_t_2);
     6885   __pyx_t_2 = 0;
     6886@@ -13214,9 +13749,9 @@
     6887     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     6888     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     6889     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3148; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6890-    goto __pyx_L9;
     6891+    goto __pyx_L6;
     6892   }
     6893-  __pyx_L9:;
     6894+  __pyx_L6:;
     6895 
     6896   /* "mtrand.pyx":3149
     6897  *         elif np.any(np.less_equal(oscale,0.0)):
     6898@@ -13226,7 +13761,7 @@
     6899  *
     6900  */
     6901   __Pyx_XDECREF(__pyx_r);
     6902-  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_wald, __pyx_v_size, __pyx_v_omean, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3149; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6903+  __pyx_t_2 = __pyx_f_6mtrand_cont2_array(__pyx_v_self->internal_state, rk_wald, __pyx_v_size, __pyx_v_omean, __pyx_v_oscale); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3149; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     6904   __Pyx_GOTREF(__pyx_t_2);
     6905   __pyx_r = __pyx_t_2;
     6906   __pyx_t_2 = 0;
     6907@@ -13249,45 +13784,33 @@
     6908   return __pyx_r;
     6909 }
     6910 
     6911-/* "mtrand.pyx":3153
     6912- *
     6913- *
     6914- *     def triangular(self, left, mode, right, size=None):             # <<<<<<<<<<<<<<
     6915- *         """
     6916- *         triangular(left, mode, right, size=None)
     6917- */
     6918-
     6919-static PyObject *__pyx_pf_6mtrand_11RandomState_39triangular(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     6920-static char __pyx_doc_6mtrand_11RandomState_39triangular[] = "\n        triangular(left, mode, right, size=None)\n\n        Draw samples from the triangular distribution.\n\n        The triangular distribution is a continuous probability distribution with\n        lower limit left, peak at mode, and upper limit right. Unlike the other\n        distributions, these parameters directly define the shape of the pdf.\n\n        Parameters\n        ----------\n        left : scalar\n            Lower limit.\n        mode : scalar\n            The value where the peak of the distribution occurs.\n            The value should fulfill the condition ``left <= mode <= right``.\n        right : scalar\n            Upper limit, should be larger than `left`.\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single value is\n            returned.\n\n        Returns\n        -------\n        samples : ndarray or scalar\n            The returned samples all lie in the interval [left, right].\n\n        Notes\n        -----\n        The probability density function for the Triangular distribution is\n\n        .. math:: P(x;l, m, r) = \\begin{cases}\n                  \\frac{2(x-l)}{(r-l)(m-l)}& \\text{for $l \\leq x \\leq m$},\\\\\n                  \\frac{2(m-x)}{(r-l)(r-m)}& \\text{for $m \\leq x \\leq r$},\\\\\n                  0& \\text{otherwise}.\n                  \\end{cases}\n\n        The triangular distribution is often used in ill-defined problems where the\n        underlying distribution is not known, but some knowledge of the limits and\n        mode exists. Often it is used in simulations.\n\n        References\n        ----------\n        ..[1] Wikipedia, \"Triangular distribution\"\n              http://en.wikipedia.org/wiki/Triangular_distribution\n\n        Examples\n        --------\n        Draw values from the distribution and plot the histogram:\n\n        >>> import matplotlib.pyplot as plt\n        >>> h = plt.hist(np.random.triangular(-3, 0, 8, 100000), bins=2""00,\n        ...              normed=True)\n        >>> plt.show()\n\n        ";
     6921-static PyObject *__pyx_pf_6mtrand_11RandomState_39triangular(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     6922+/* Python wrapper */
     6923+static PyObject *__pyx_pw_6mtrand_11RandomState_79triangular(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     6924+static char __pyx_doc_6mtrand_11RandomState_78triangular[] = "\n        triangular(left, mode, right, size=None)\n\n        Draw samples from the triangular distribution.\n\n        The triangular distribution is a continuous probability distribution with\n        lower limit left, peak at mode, and upper limit right. Unlike the other\n        distributions, these parameters directly define the shape of the pdf.\n\n        Parameters\n        ----------\n        left : scalar\n            Lower limit.\n        mode : scalar\n            The value where the peak of the distribution occurs.\n            The value should fulfill the condition ``left <= mode <= right``.\n        right : scalar\n            Upper limit, should be larger than `left`.\n        size : int or tuple of ints, optional\n            Output shape. Default is None, in which case a single value is\n            returned.\n\n        Returns\n        -------\n        samples : ndarray or scalar\n            The returned samples all lie in the interval [left, right].\n\n        Notes\n        -----\n        The probability density function for the Triangular distribution is\n\n        .. math:: P(x;l, m, r) = \\begin{cases}\n                  \\frac{2(x-l)}{(r-l)(m-l)}& \\text{for $l \\leq x \\leq m$},\\\\\n                  \\frac{2(m-x)}{(r-l)(r-m)}& \\text{for $m \\leq x \\leq r$},\\\\\n                  0& \\text{otherwise}.\n                  \\end{cases}\n\n        The triangular distribution is often used in ill-defined problems where the\n        underlying distribution is not known, but some knowledge of the limits and\n        mode exists. Often it is used in simulations.\n\n        References\n        ----------\n        ..[1] Wikipedia, \"Triangular distribution\"\n              http://en.wikipedia.org/wiki/Triangular_distribution\n\n        Examples\n        --------\n        Draw values from the distribution and plot the histogram:\n\n        >>> import matplotlib.pyplot as plt\n        >>> h = plt.hist(np.random.triangular(-3, 0, 8, 100000), bins=2""00,\n        ...              normed=True)\n        >>> plt.show()\n\n        ";
     6925+static PyObject *__pyx_pw_6mtrand_11RandomState_79triangular(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     6926   PyObject *__pyx_v_left = 0;
     6927   PyObject *__pyx_v_mode = 0;
     6928   PyObject *__pyx_v_right = 0;
     6929   PyObject *__pyx_v_size = 0;
     6930-  PyArrayObject *__pyx_v_oleft = 0;
     6931-  PyArrayObject *__pyx_v_omode = 0;
     6932-  PyArrayObject *__pyx_v_oright = 0;
     6933-  double __pyx_v_fleft;
     6934-  double __pyx_v_fmode;
     6935-  double __pyx_v_fright;
     6936-  PyObject *__pyx_r = NULL;
     6937+  PyObject *__pyx_r = 0;
     6938   __Pyx_RefNannyDeclarations
     6939-  int __pyx_t_1;
     6940-  PyObject *__pyx_t_2 = NULL;
     6941-  PyObject *__pyx_t_3 = NULL;
     6942-  PyObject *__pyx_t_4 = NULL;
     6943-  PyObject *__pyx_t_5 = NULL;
     6944-  int __pyx_lineno = 0;
     6945-  const char *__pyx_filename = NULL;
     6946-  int __pyx_clineno = 0;
     6947-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__left,&__pyx_n_s__mode,&__pyx_n_s__right,&__pyx_n_s__size,0};
     6948-  __Pyx_RefNannySetupContext("triangular");
     6949+  __Pyx_RefNannySetupContext("triangular (wrapper)", 0);
     6950   {
     6951+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__left,&__pyx_n_s__mode,&__pyx_n_s__right,&__pyx_n_s__size,0};
     6952     PyObject* values[4] = {0,0,0,0};
     6953+
     6954+    /* "mtrand.pyx":3153
     6955+ *
     6956+ *
     6957+ *     def triangular(self, left, mode, right, size=None):             # <<<<<<<<<<<<<<
     6958+ *         """
     6959+ *         triangular(left, mode, right, size=None)
     6960+ */
     6961     values[3] = ((PyObject *)Py_None);
     6962     if (unlikely(__pyx_kwds)) {
     6963       Py_ssize_t kw_args;
     6964-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6965+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     6966+      switch (pos_args) {
     6967         case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
     6968         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     6969         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     6970@@ -13296,20 +13819,17 @@
     6971         default: goto __pyx_L5_argtuple_error;
     6972       }
     6973       kw_args = PyDict_Size(__pyx_kwds);
     6974-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     6975+      switch (pos_args) {
     6976         case  0:
     6977-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__left);
     6978-        if (likely(values[0])) kw_args--;
     6979+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__left)) != 0)) kw_args--;
     6980         else goto __pyx_L5_argtuple_error;
     6981         case  1:
     6982-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__mode);
     6983-        if (likely(values[1])) kw_args--;
     6984+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__mode)) != 0)) kw_args--;
     6985         else {
     6986           __Pyx_RaiseArgtupleInvalid("triangular", 0, 3, 4, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3153; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6987         }
     6988         case  2:
     6989-        values[2] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__right);
     6990-        if (likely(values[2])) kw_args--;
     6991+        if (likely((values[2] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__right)) != 0)) kw_args--;
     6992         else {
     6993           __Pyx_RaiseArgtupleInvalid("triangular", 0, 3, 4, 2); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3153; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     6994         }
     6995@@ -13320,7 +13840,7 @@
     6996         }
     6997       }
     6998       if (unlikely(kw_args > 0)) {
     6999-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "triangular") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3153; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     7000+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "triangular") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3153; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     7001       }
     7002     } else {
     7003       switch (PyTuple_GET_SIZE(__pyx_args)) {
     7004@@ -13345,6 +13865,29 @@
     7005   __Pyx_RefNannyFinishContext();
     7006   return NULL;
     7007   __pyx_L4_argument_unpacking_done:;
     7008+  __pyx_r = __pyx_pf_6mtrand_11RandomState_78triangular(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_left, __pyx_v_mode, __pyx_v_right, __pyx_v_size);
     7009+  __Pyx_RefNannyFinishContext();
     7010+  return __pyx_r;
     7011+}
     7012+
     7013+static PyObject *__pyx_pf_6mtrand_11RandomState_78triangular(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_left, PyObject *__pyx_v_mode, PyObject *__pyx_v_right, PyObject *__pyx_v_size) {
     7014+  PyArrayObject *__pyx_v_oleft = 0;
     7015+  PyArrayObject *__pyx_v_omode = 0;
     7016+  PyArrayObject *__pyx_v_oright = 0;
     7017+  double __pyx_v_fleft;
     7018+  double __pyx_v_fmode;
     7019+  double __pyx_v_fright;
     7020+  PyObject *__pyx_r = NULL;
     7021+  __Pyx_RefNannyDeclarations
     7022+  int __pyx_t_1;
     7023+  PyObject *__pyx_t_2 = NULL;
     7024+  PyObject *__pyx_t_3 = NULL;
     7025+  PyObject *__pyx_t_4 = NULL;
     7026+  PyObject *__pyx_t_5 = NULL;
     7027+  int __pyx_lineno = 0;
     7028+  const char *__pyx_filename = NULL;
     7029+  int __pyx_clineno = 0;
     7030+  __Pyx_RefNannySetupContext("triangular", 0);
     7031 
     7032   /* "mtrand.pyx":3213
     7033  *         cdef double fleft, fmode, fright
     7034@@ -13405,9 +13948,9 @@
     7035       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     7036       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7037       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3218; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7038-      goto __pyx_L7;
     7039+      goto __pyx_L4;
     7040     }
     7041-    __pyx_L7:;
     7042+    __pyx_L4:;
     7043 
     7044     /* "mtrand.pyx":3219
     7045  *             if fleft > fmode:
     7046@@ -13431,9 +13974,9 @@
     7047       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     7048       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7049       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3220; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7050-      goto __pyx_L8;
     7051+      goto __pyx_L5;
     7052     }
     7053-    __pyx_L8:;
     7054+    __pyx_L5:;
     7055 
     7056     /* "mtrand.pyx":3221
     7057  *             if fmode > fright:
     7058@@ -13457,9 +14000,9 @@
     7059       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     7060       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7061       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3222; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7062-      goto __pyx_L9;
     7063+      goto __pyx_L6;
     7064     }
     7065-    __pyx_L9:;
     7066+    __pyx_L6:;
     7067 
     7068     /* "mtrand.pyx":3223
     7069  *             if fleft == fright:
     7070@@ -13477,14 +14020,14 @@
     7071  *
     7072  *         PyErr_Clear()
     7073  */
     7074-    __pyx_t_2 = __pyx_f_6mtrand_cont3_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_triangular, __pyx_v_size, __pyx_v_fleft, __pyx_v_fmode, __pyx_v_fright); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3223; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7075+    __pyx_t_2 = __pyx_f_6mtrand_cont3_array_sc(__pyx_v_self->internal_state, rk_triangular, __pyx_v_size, __pyx_v_fleft, __pyx_v_fmode, __pyx_v_fright); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3223; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7076     __Pyx_GOTREF(__pyx_t_2);
     7077     __pyx_r = __pyx_t_2;
     7078     __pyx_t_2 = 0;
     7079     goto __pyx_L0;
     7080-    goto __pyx_L6;
     7081+    goto __pyx_L3;
     7082   }
     7083-  __pyx_L6:;
     7084+  __pyx_L3:;
     7085 
     7086   /* "mtrand.pyx":3226
     7087  *                                   fmode, fright)
     7088@@ -13552,7 +14095,7 @@
     7089   __Pyx_GOTREF(__pyx_t_4);
     7090   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7091   __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3231; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7092-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     7093+  __Pyx_GOTREF(__pyx_t_2);
     7094   __Pyx_INCREF(((PyObject *)__pyx_v_oleft));
     7095   PyTuple_SET_ITEM(__pyx_t_2, 0, ((PyObject *)__pyx_v_oleft));
     7096   __Pyx_GIVEREF(((PyObject *)__pyx_v_oleft));
     7097@@ -13564,7 +14107,7 @@
     7098   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     7099   __Pyx_DECREF(((PyObject *)__pyx_t_2)); __pyx_t_2 = 0;
     7100   __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3231; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7101-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     7102+  __Pyx_GOTREF(__pyx_t_2);
     7103   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_5);
     7104   __Pyx_GIVEREF(__pyx_t_5);
     7105   __pyx_t_5 = 0;
     7106@@ -13588,9 +14131,9 @@
     7107     __Pyx_Raise(__pyx_t_5, 0, 0, 0);
     7108     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     7109     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3232; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7110-    goto __pyx_L10;
     7111+    goto __pyx_L7;
     7112   }
     7113-  __pyx_L10:;
     7114+  __pyx_L7:;
     7115 
     7116   /* "mtrand.pyx":3233
     7117  *         if np.any(np.greater(oleft, omode)):
     7118@@ -13610,7 +14153,7 @@
     7119   __Pyx_GOTREF(__pyx_t_3);
     7120   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     7121   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3233; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7122-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     7123+  __Pyx_GOTREF(__pyx_t_5);
     7124   __Pyx_INCREF(((PyObject *)__pyx_v_omode));
     7125   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_omode));
     7126   __Pyx_GIVEREF(((PyObject *)__pyx_v_omode));
     7127@@ -13622,7 +14165,7 @@
     7128   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     7129   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     7130   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3233; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7131-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     7132+  __Pyx_GOTREF(__pyx_t_5);
     7133   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_4);
     7134   __Pyx_GIVEREF(__pyx_t_4);
     7135   __pyx_t_4 = 0;
     7136@@ -13646,9 +14189,9 @@
     7137     __Pyx_Raise(__pyx_t_4, 0, 0, 0);
     7138     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     7139     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3234; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7140-    goto __pyx_L11;
     7141+    goto __pyx_L8;
     7142   }
     7143-  __pyx_L11:;
     7144+  __pyx_L8:;
     7145 
     7146   /* "mtrand.pyx":3235
     7147  *         if np.any(np.greater(omode, oright)):
     7148@@ -13668,7 +14211,7 @@
     7149   __Pyx_GOTREF(__pyx_t_2);
     7150   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     7151   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3235; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7152-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     7153+  __Pyx_GOTREF(__pyx_t_4);
     7154   __Pyx_INCREF(((PyObject *)__pyx_v_oleft));
     7155   PyTuple_SET_ITEM(__pyx_t_4, 0, ((PyObject *)__pyx_v_oleft));
     7156   __Pyx_GIVEREF(((PyObject *)__pyx_v_oleft));
     7157@@ -13680,7 +14223,7 @@
     7158   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7159   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     7160   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3235; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7161-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     7162+  __Pyx_GOTREF(__pyx_t_4);
     7163   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_3);
     7164   __Pyx_GIVEREF(__pyx_t_3);
     7165   __pyx_t_3 = 0;
     7166@@ -13704,9 +14247,9 @@
     7167     __Pyx_Raise(__pyx_t_3, 0, 0, 0);
     7168     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     7169     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3236; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7170-    goto __pyx_L12;
     7171+    goto __pyx_L9;
     7172   }
     7173-  __pyx_L12:;
     7174+  __pyx_L9:;
     7175 
     7176   /* "mtrand.pyx":3237
     7177  *         if np.any(np.equal(oleft, oright)):
     7178@@ -13724,7 +14267,7 @@
     7179  *
     7180  *     # Complicated, discrete distributions:
     7181  */
     7182-  __pyx_t_3 = __pyx_f_6mtrand_cont3_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_triangular, __pyx_v_size, __pyx_v_oleft, __pyx_v_omode, __pyx_v_oright); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3237; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7183+  __pyx_t_3 = __pyx_f_6mtrand_cont3_array(__pyx_v_self->internal_state, rk_triangular, __pyx_v_size, __pyx_v_oleft, __pyx_v_omode, __pyx_v_oright); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3237; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7184   __Pyx_GOTREF(__pyx_t_3);
     7185   __pyx_r = __pyx_t_3;
     7186   __pyx_t_3 = 0;
     7187@@ -13748,42 +14291,32 @@
     7188   return __pyx_r;
     7189 }
     7190 
     7191-/* "mtrand.pyx":3241
     7192- *
     7193- *     # Complicated, discrete distributions:
     7194- *     def binomial(self, n, p, size=None):             # <<<<<<<<<<<<<<
     7195- *         """
     7196- *         binomial(n, p, size=None)
     7197- */
     7198-
     7199-static PyObject *__pyx_pf_6mtrand_11RandomState_40binomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     7200-static char __pyx_doc_6mtrand_11RandomState_40binomial[] = "\n        binomial(n, p, size=None)\n\n        Draw samples from a binomial distribution.\n\n        Samples are drawn from a Binomial distribution with specified\n        parameters, n trials and p probability of success where\n        n an integer > 0 and p is in the interval [0,1]. (n may be\n        input as a float, but it is truncated to an integer in use)\n\n        Parameters\n        ----------\n        n : float (but truncated to an integer)\n                parameter, > 0.\n        p : float\n                parameter, >= 0 and <=1.\n        size : {tuple, int}\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n                  where the values are all integers in  [0, n].\n\n        See Also\n        --------\n        scipy.stats.distributions.binom : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Binomial distribution is\n\n        .. math:: P(N) = \\binom{n}{N}p^N(1-p)^{n-N},\n\n        where :math:`n` is the number of trials, :math:`p` is the probability\n        of success, and :math:`N` is the number of successes.\n\n        When estimating the standard error of a proportion in a population by\n        using a random sample, the normal distribution works well unless the\n        product p*n <=5, where p = population proportion estimate, and n =\n        number of samples, in which case the binomial distribution is used\n        instead. For example, a sample of 15 people shows 4 who are left\n        handed, and 11 who are right handed. Then p = 4/15 = 27%. 0.27*15 = 4,\n        so the binomial distribution should be used in this case.\n\n        References\n        ----------\n        .. [1] Dalgaard, Peter, \"Introductory Statistics with R\",\n               Springer-Verlag, 2002.\n    ""    .. [2] Glantz, Stanton A. \"Primer of Biostatistics.\", McGraw-Hill,\n               Fifth Edition, 2002.\n        .. [3] Lentner, Marvin, \"Elementary Applied Statistics\", Bogden\n               and Quigley, 1972.\n        .. [4] Weisstein, Eric W. \"Binomial Distribution.\" From MathWorld--A\n               Wolfram Web Resource.\n               http://mathworld.wolfram.com/BinomialDistribution.html\n        .. [5] Wikipedia, \"Binomial-distribution\",\n               http://en.wikipedia.org/wiki/Binomial_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> n, p = 10, .5 # number of trials, probability of each trial\n        >>> s = np.random.binomial(n, p, 1000)\n        # result of flipping a coin 10 times, tested 1000 times.\n\n        A real world example. A company drills 9 wild-cat oil exploration\n        wells, each with an estimated probability of success of 0.1. All nine\n        wells fail. What is the probability of that happening?\n\n        Let's do 20,000 trials of the model, and count the number that\n        generate zero positive results.\n\n        >>> sum(np.random.binomial(9,0.1,20000)==0)/20000.\n        answer = 0.38885, or 38%.\n\n        ";
     7201-static PyObject *__pyx_pf_6mtrand_11RandomState_40binomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     7202+/* Python wrapper */
     7203+static PyObject *__pyx_pw_6mtrand_11RandomState_81binomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     7204+static char __pyx_doc_6mtrand_11RandomState_80binomial[] = "\n        binomial(n, p, size=None)\n\n        Draw samples from a binomial distribution.\n\n        Samples are drawn from a Binomial distribution with specified\n        parameters, n trials and p probability of success where\n        n an integer > 0 and p is in the interval [0,1]. (n may be\n        input as a float, but it is truncated to an integer in use)\n\n        Parameters\n        ----------\n        n : float (but truncated to an integer)\n                parameter, > 0.\n        p : float\n                parameter, >= 0 and <=1.\n        size : {tuple, int}\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n                  where the values are all integers in  [0, n].\n\n        See Also\n        --------\n        scipy.stats.distributions.binom : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Binomial distribution is\n\n        .. math:: P(N) = \\binom{n}{N}p^N(1-p)^{n-N},\n\n        where :math:`n` is the number of trials, :math:`p` is the probability\n        of success, and :math:`N` is the number of successes.\n\n        When estimating the standard error of a proportion in a population by\n        using a random sample, the normal distribution works well unless the\n        product p*n <=5, where p = population proportion estimate, and n =\n        number of samples, in which case the binomial distribution is used\n        instead. For example, a sample of 15 people shows 4 who are left\n        handed, and 11 who are right handed. Then p = 4/15 = 27%. 0.27*15 = 4,\n        so the binomial distribution should be used in this case.\n\n        References\n        ----------\n        .. [1] Dalgaard, Peter, \"Introductory Statistics with R\",\n               Springer-Verlag, 2002.\n    ""    .. [2] Glantz, Stanton A. \"Primer of Biostatistics.\", McGraw-Hill,\n               Fifth Edition, 2002.\n        .. [3] Lentner, Marvin, \"Elementary Applied Statistics\", Bogden\n               and Quigley, 1972.\n        .. [4] Weisstein, Eric W. \"Binomial Distribution.\" From MathWorld--A\n               Wolfram Web Resource.\n               http://mathworld.wolfram.com/BinomialDistribution.html\n        .. [5] Wikipedia, \"Binomial-distribution\",\n               http://en.wikipedia.org/wiki/Binomial_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> n, p = 10, .5 # number of trials, probability of each trial\n        >>> s = np.random.binomial(n, p, 1000)\n        # result of flipping a coin 10 times, tested 1000 times.\n\n        A real world example. A company drills 9 wild-cat oil exploration\n        wells, each with an estimated probability of success of 0.1. All nine\n        wells fail. What is the probability of that happening?\n\n        Let's do 20,000 trials of the model, and count the number that\n        generate zero positive results.\n\n        >>> sum(np.random.binomial(9,0.1,20000)==0)/20000.\n        answer = 0.38885, or 38%.\n\n        ";
     7205+static PyObject *__pyx_pw_6mtrand_11RandomState_81binomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     7206   PyObject *__pyx_v_n = 0;
     7207   PyObject *__pyx_v_p = 0;
     7208   PyObject *__pyx_v_size = 0;
     7209-  PyArrayObject *__pyx_v_on = 0;
     7210-  PyArrayObject *__pyx_v_op = 0;
     7211-  long __pyx_v_ln;
     7212-  double __pyx_v_fp;
     7213-  PyObject *__pyx_r = NULL;
     7214+  PyObject *__pyx_r = 0;
     7215   __Pyx_RefNannyDeclarations
     7216-  int __pyx_t_1;
     7217-  PyObject *__pyx_t_2 = NULL;
     7218-  PyObject *__pyx_t_3 = NULL;
     7219-  PyObject *__pyx_t_4 = NULL;
     7220-  PyObject *__pyx_t_5 = NULL;
     7221-  int __pyx_lineno = 0;
     7222-  const char *__pyx_filename = NULL;
     7223-  int __pyx_clineno = 0;
     7224-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__n,&__pyx_n_s__p,&__pyx_n_s__size,0};
     7225-  __Pyx_RefNannySetupContext("binomial");
     7226+  __Pyx_RefNannySetupContext("binomial (wrapper)", 0);
     7227   {
     7228+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__n,&__pyx_n_s__p,&__pyx_n_s__size,0};
     7229     PyObject* values[3] = {0,0,0};
     7230+
     7231+    /* "mtrand.pyx":3241
     7232+ *
     7233+ *     # Complicated, discrete distributions:
     7234+ *     def binomial(self, n, p, size=None):             # <<<<<<<<<<<<<<
     7235+ *         """
     7236+ *         binomial(n, p, size=None)
     7237+ */
     7238     values[2] = ((PyObject *)Py_None);
     7239     if (unlikely(__pyx_kwds)) {
     7240       Py_ssize_t kw_args;
     7241-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     7242+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     7243+      switch (pos_args) {
     7244         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     7245         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     7246         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     7247@@ -13791,14 +14324,12 @@
     7248         default: goto __pyx_L5_argtuple_error;
     7249       }
     7250       kw_args = PyDict_Size(__pyx_kwds);
     7251-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     7252+      switch (pos_args) {
     7253         case  0:
     7254-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__n);
     7255-        if (likely(values[0])) kw_args--;
     7256+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__n)) != 0)) kw_args--;
     7257         else goto __pyx_L5_argtuple_error;
     7258         case  1:
     7259-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__p);
     7260-        if (likely(values[1])) kw_args--;
     7261+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__p)) != 0)) kw_args--;
     7262         else {
     7263           __Pyx_RaiseArgtupleInvalid("binomial", 0, 2, 3, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3241; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     7264         }
     7265@@ -13809,7 +14340,7 @@
     7266         }
     7267       }
     7268       if (unlikely(kw_args > 0)) {
     7269-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "binomial") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3241; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     7270+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "binomial") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3241; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     7271       }
     7272     } else {
     7273       switch (PyTuple_GET_SIZE(__pyx_args)) {
     7274@@ -13832,6 +14363,27 @@
     7275   __Pyx_RefNannyFinishContext();
     7276   return NULL;
     7277   __pyx_L4_argument_unpacking_done:;
     7278+  __pyx_r = __pyx_pf_6mtrand_11RandomState_80binomial(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_n, __pyx_v_p, __pyx_v_size);
     7279+  __Pyx_RefNannyFinishContext();
     7280+  return __pyx_r;
     7281+}
     7282+
     7283+static PyObject *__pyx_pf_6mtrand_11RandomState_80binomial(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_n, PyObject *__pyx_v_p, PyObject *__pyx_v_size) {
     7284+  PyArrayObject *__pyx_v_on = 0;
     7285+  PyArrayObject *__pyx_v_op = 0;
     7286+  long __pyx_v_ln;
     7287+  double __pyx_v_fp;
     7288+  PyObject *__pyx_r = NULL;
     7289+  __Pyx_RefNannyDeclarations
     7290+  int __pyx_t_1;
     7291+  PyObject *__pyx_t_2 = NULL;
     7292+  PyObject *__pyx_t_3 = NULL;
     7293+  PyObject *__pyx_t_4 = NULL;
     7294+  PyObject *__pyx_t_5 = NULL;
     7295+  int __pyx_lineno = 0;
     7296+  const char *__pyx_filename = NULL;
     7297+  int __pyx_clineno = 0;
     7298+  __Pyx_RefNannySetupContext("binomial", 0);
     7299 
     7300   /* "mtrand.pyx":3326
     7301  *         cdef double fp
     7302@@ -13883,9 +14435,9 @@
     7303       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     7304       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7305       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3330; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7306-      goto __pyx_L7;
     7307+      goto __pyx_L4;
     7308     }
     7309-    __pyx_L7:;
     7310+    __pyx_L4:;
     7311 
     7312     /* "mtrand.pyx":3331
     7313  *             if ln <= 0:
     7314@@ -13909,7 +14461,7 @@
     7315       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     7316       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7317       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3332; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7318-      goto __pyx_L8;
     7319+      goto __pyx_L5;
     7320     }
     7321 
     7322     /* "mtrand.pyx":3333
     7323@@ -13934,9 +14486,9 @@
     7324       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     7325       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7326       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3334; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7327-      goto __pyx_L8;
     7328+      goto __pyx_L5;
     7329     }
     7330-    __pyx_L8:;
     7331+    __pyx_L5:;
     7332 
     7333     /* "mtrand.pyx":3335
     7334  *             elif fp > 1:
     7335@@ -13946,14 +14498,14 @@
     7336  *         PyErr_Clear()
     7337  */
     7338     __Pyx_XDECREF(__pyx_r);
     7339-    __pyx_t_2 = __pyx_f_6mtrand_discnp_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_binomial, __pyx_v_size, __pyx_v_ln, __pyx_v_fp); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3335; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7340+    __pyx_t_2 = __pyx_f_6mtrand_discnp_array_sc(__pyx_v_self->internal_state, rk_binomial, __pyx_v_size, __pyx_v_ln, __pyx_v_fp); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3335; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7341     __Pyx_GOTREF(__pyx_t_2);
     7342     __pyx_r = __pyx_t_2;
     7343     __pyx_t_2 = 0;
     7344     goto __pyx_L0;
     7345-    goto __pyx_L6;
     7346+    goto __pyx_L3;
     7347   }
     7348-  __pyx_L6:;
     7349+  __pyx_L3:;
     7350 
     7351   /* "mtrand.pyx":3337
     7352  *             return discnp_array_sc(self.internal_state, rk_binomial, size, ln, fp)
     7353@@ -14008,7 +14560,7 @@
     7354   __Pyx_GOTREF(__pyx_t_4);
     7355   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7356   __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3341; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7357-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     7358+  __Pyx_GOTREF(__pyx_t_2);
     7359   __Pyx_INCREF(__pyx_v_n);
     7360   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_n);
     7361   __Pyx_GIVEREF(__pyx_v_n);
     7362@@ -14020,7 +14572,7 @@
     7363   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     7364   __Pyx_DECREF(((PyObject *)__pyx_t_2)); __pyx_t_2 = 0;
     7365   __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3341; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7366-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     7367+  __Pyx_GOTREF(__pyx_t_2);
     7368   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_5);
     7369   __Pyx_GIVEREF(__pyx_t_5);
     7370   __pyx_t_5 = 0;
     7371@@ -14044,9 +14596,9 @@
     7372     __Pyx_Raise(__pyx_t_5, 0, 0, 0);
     7373     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     7374     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3342; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7375-    goto __pyx_L9;
     7376+    goto __pyx_L6;
     7377   }
     7378-  __pyx_L9:;
     7379+  __pyx_L6:;
     7380 
     7381   /* "mtrand.pyx":3343
     7382  *         if np.any(np.less_equal(n, 0)):
     7383@@ -14066,7 +14618,7 @@
     7384   __Pyx_GOTREF(__pyx_t_3);
     7385   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     7386   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3343; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7387-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     7388+  __Pyx_GOTREF(__pyx_t_5);
     7389   __Pyx_INCREF(__pyx_v_p);
     7390   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_v_p);
     7391   __Pyx_GIVEREF(__pyx_v_p);
     7392@@ -14078,7 +14630,7 @@
     7393   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     7394   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     7395   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3343; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7396-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     7397+  __Pyx_GOTREF(__pyx_t_5);
     7398   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_4);
     7399   __Pyx_GIVEREF(__pyx_t_4);
     7400   __pyx_t_4 = 0;
     7401@@ -14102,9 +14654,9 @@
     7402     __Pyx_Raise(__pyx_t_4, 0, 0, 0);
     7403     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     7404     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3344; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7405-    goto __pyx_L10;
     7406+    goto __pyx_L7;
     7407   }
     7408-  __pyx_L10:;
     7409+  __pyx_L7:;
     7410 
     7411   /* "mtrand.pyx":3345
     7412  *         if np.any(np.less(p, 0)):
     7413@@ -14124,7 +14676,7 @@
     7414   __Pyx_GOTREF(__pyx_t_2);
     7415   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     7416   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3345; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7417-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     7418+  __Pyx_GOTREF(__pyx_t_4);
     7419   __Pyx_INCREF(__pyx_v_p);
     7420   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_p);
     7421   __Pyx_GIVEREF(__pyx_v_p);
     7422@@ -14136,7 +14688,7 @@
     7423   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7424   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     7425   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3345; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7426-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     7427+  __Pyx_GOTREF(__pyx_t_4);
     7428   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_3);
     7429   __Pyx_GIVEREF(__pyx_t_3);
     7430   __pyx_t_3 = 0;
     7431@@ -14160,9 +14712,9 @@
     7432     __Pyx_Raise(__pyx_t_3, 0, 0, 0);
     7433     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     7434     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3346; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7435-    goto __pyx_L11;
     7436+    goto __pyx_L8;
     7437   }
     7438-  __pyx_L11:;
     7439+  __pyx_L8:;
     7440 
     7441   /* "mtrand.pyx":3347
     7442  *         if np.any(np.greater(p, 1)):
     7443@@ -14172,7 +14724,7 @@
     7444  *     def negative_binomial(self, n, p, size=None):
     7445  */
     7446   __Pyx_XDECREF(__pyx_r);
     7447-  __pyx_t_3 = __pyx_f_6mtrand_discnp_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_binomial, __pyx_v_size, __pyx_v_on, __pyx_v_op); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3347; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7448+  __pyx_t_3 = __pyx_f_6mtrand_discnp_array(__pyx_v_self->internal_state, rk_binomial, __pyx_v_size, __pyx_v_on, __pyx_v_op); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3347; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7449   __Pyx_GOTREF(__pyx_t_3);
     7450   __pyx_r = __pyx_t_3;
     7451   __pyx_t_3 = 0;
     7452@@ -14195,42 +14747,32 @@
     7453   return __pyx_r;
     7454 }
     7455 
     7456-/* "mtrand.pyx":3349
     7457- *         return discnp_array(self.internal_state, rk_binomial, size, on, op)
     7458- *
     7459- *     def negative_binomial(self, n, p, size=None):             # <<<<<<<<<<<<<<
     7460- *         """
     7461- *         negative_binomial(n, p, size=None)
     7462- */
     7463-
     7464-static PyObject *__pyx_pf_6mtrand_11RandomState_41negative_binomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     7465-static char __pyx_doc_6mtrand_11RandomState_41negative_binomial[] = "\n        negative_binomial(n, p, size=None)\n\n        Draw samples from a negative_binomial distribution.\n\n        Samples are drawn from a negative_Binomial distribution with specified\n        parameters, `n` trials and `p` probability of success where `n` is an\n        integer > 0 and `p` is in the interval [0, 1].\n\n        Parameters\n        ----------\n        n : int\n            Parameter, > 0.\n        p : float\n            Parameter, >= 0 and <=1.\n        size : int or tuple of ints\n            Output shape. If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : int or ndarray of ints\n            Drawn samples.\n\n        Notes\n        -----\n        The probability density for the Negative Binomial distribution is\n\n        .. math:: P(N;n,p) = \\binom{N+n-1}{n-1}p^{n}(1-p)^{N},\n\n        where :math:`n-1` is the number of successes, :math:`p` is the probability\n        of success, and :math:`N+n-1` is the number of trials.\n\n        The negative binomial distribution gives the probability of n-1 successes\n        and N failures in N+n-1 trials, and success on the (N+n)th trial.\n\n        If one throws a die repeatedly until the third time a \"1\" appears, then the\n        probability distribution of the number of non-\"1\"s that appear before the\n        third \"1\" is a negative binomial distribution.\n\n        References\n        ----------\n        .. [1] Weisstein, Eric W. \"Negative Binomial Distribution.\" From\n               MathWorld--A Wolfram Web Resource.\n               http://mathworld.wolfram.com/NegativeBinomialDistribution.html\n        .. [2] Wikipedia, \"Negative binomial distribution\",\n               http://en.wikipedia.org/wiki/Negative_binomial_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        A real world example. A company drills wild-cat oil exploration well""s, each\n        with an estimated probability of success of 0.1.  What is the probability\n        of having one success for each successive well, that is what is the\n        probability of a single success after drilling 5 wells, after 6 wells,\n        etc.?\n\n        >>> s = np.random.negative_binomial(1, 0.1, 100000)\n        >>> for i in range(1, 11):\n        ...    probability = sum(s<i) / 100000.\n        ...    print i, \"wells drilled, probability of one success =\", probability\n\n        ";
     7466-static PyObject *__pyx_pf_6mtrand_11RandomState_41negative_binomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     7467+/* Python wrapper */
     7468+static PyObject *__pyx_pw_6mtrand_11RandomState_83negative_binomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     7469+static char __pyx_doc_6mtrand_11RandomState_82negative_binomial[] = "\n        negative_binomial(n, p, size=None)\n\n        Draw samples from a negative_binomial distribution.\n\n        Samples are drawn from a negative_Binomial distribution with specified\n        parameters, `n` trials and `p` probability of success where `n` is an\n        integer > 0 and `p` is in the interval [0, 1].\n\n        Parameters\n        ----------\n        n : int\n            Parameter, > 0.\n        p : float\n            Parameter, >= 0 and <=1.\n        size : int or tuple of ints\n            Output shape. If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : int or ndarray of ints\n            Drawn samples.\n\n        Notes\n        -----\n        The probability density for the Negative Binomial distribution is\n\n        .. math:: P(N;n,p) = \\binom{N+n-1}{n-1}p^{n}(1-p)^{N},\n\n        where :math:`n-1` is the number of successes, :math:`p` is the probability\n        of success, and :math:`N+n-1` is the number of trials.\n\n        The negative binomial distribution gives the probability of n-1 successes\n        and N failures in N+n-1 trials, and success on the (N+n)th trial.\n\n        If one throws a die repeatedly until the third time a \"1\" appears, then the\n        probability distribution of the number of non-\"1\"s that appear before the\n        third \"1\" is a negative binomial distribution.\n\n        References\n        ----------\n        .. [1] Weisstein, Eric W. \"Negative Binomial Distribution.\" From\n               MathWorld--A Wolfram Web Resource.\n               http://mathworld.wolfram.com/NegativeBinomialDistribution.html\n        .. [2] Wikipedia, \"Negative binomial distribution\",\n               http://en.wikipedia.org/wiki/Negative_binomial_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        A real world example. A company drills wild-cat oil exploration well""s, each\n        with an estimated probability of success of 0.1.  What is the probability\n        of having one success for each successive well, that is what is the\n        probability of a single success after drilling 5 wells, after 6 wells,\n        etc.?\n\n        >>> s = np.random.negative_binomial(1, 0.1, 100000)\n        >>> for i in range(1, 11):\n        ...    probability = sum(s<i) / 100000.\n        ...    print i, \"wells drilled, probability of one success =\", probability\n\n        ";
     7470+static PyObject *__pyx_pw_6mtrand_11RandomState_83negative_binomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     7471   PyObject *__pyx_v_n = 0;
     7472   PyObject *__pyx_v_p = 0;
     7473   PyObject *__pyx_v_size = 0;
     7474-  PyArrayObject *__pyx_v_on = 0;
     7475-  PyArrayObject *__pyx_v_op = 0;
     7476-  double __pyx_v_fn;
     7477-  double __pyx_v_fp;
     7478-  PyObject *__pyx_r = NULL;
     7479+  PyObject *__pyx_r = 0;
     7480   __Pyx_RefNannyDeclarations
     7481-  int __pyx_t_1;
     7482-  PyObject *__pyx_t_2 = NULL;
     7483-  PyObject *__pyx_t_3 = NULL;
     7484-  PyObject *__pyx_t_4 = NULL;
     7485-  PyObject *__pyx_t_5 = NULL;
     7486-  int __pyx_lineno = 0;
     7487-  const char *__pyx_filename = NULL;
     7488-  int __pyx_clineno = 0;
     7489-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__n,&__pyx_n_s__p,&__pyx_n_s__size,0};
     7490-  __Pyx_RefNannySetupContext("negative_binomial");
     7491+  __Pyx_RefNannySetupContext("negative_binomial (wrapper)", 0);
     7492   {
     7493+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__n,&__pyx_n_s__p,&__pyx_n_s__size,0};
     7494     PyObject* values[3] = {0,0,0};
     7495+
     7496+    /* "mtrand.pyx":3349
     7497+ *         return discnp_array(self.internal_state, rk_binomial, size, on, op)
     7498+ *
     7499+ *     def negative_binomial(self, n, p, size=None):             # <<<<<<<<<<<<<<
     7500+ *         """
     7501+ *         negative_binomial(n, p, size=None)
     7502+ */
     7503     values[2] = ((PyObject *)Py_None);
     7504     if (unlikely(__pyx_kwds)) {
     7505       Py_ssize_t kw_args;
     7506-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     7507+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     7508+      switch (pos_args) {
     7509         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     7510         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     7511         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     7512@@ -14238,14 +14780,12 @@
     7513         default: goto __pyx_L5_argtuple_error;
     7514       }
     7515       kw_args = PyDict_Size(__pyx_kwds);
     7516-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     7517+      switch (pos_args) {
     7518         case  0:
     7519-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__n);
     7520-        if (likely(values[0])) kw_args--;
     7521+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__n)) != 0)) kw_args--;
     7522         else goto __pyx_L5_argtuple_error;
     7523         case  1:
     7524-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__p);
     7525-        if (likely(values[1])) kw_args--;
     7526+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__p)) != 0)) kw_args--;
     7527         else {
     7528           __Pyx_RaiseArgtupleInvalid("negative_binomial", 0, 2, 3, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3349; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     7529         }
     7530@@ -14256,7 +14796,7 @@
     7531         }
     7532       }
     7533       if (unlikely(kw_args > 0)) {
     7534-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "negative_binomial") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3349; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     7535+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "negative_binomial") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3349; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     7536       }
     7537     } else {
     7538       switch (PyTuple_GET_SIZE(__pyx_args)) {
     7539@@ -14279,7 +14819,28 @@
     7540   __Pyx_RefNannyFinishContext();
     7541   return NULL;
     7542   __pyx_L4_argument_unpacking_done:;
     7543-
     7544+  __pyx_r = __pyx_pf_6mtrand_11RandomState_82negative_binomial(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_n, __pyx_v_p, __pyx_v_size);
     7545+  __Pyx_RefNannyFinishContext();
     7546+  return __pyx_r;
     7547+}
     7548+
     7549+static PyObject *__pyx_pf_6mtrand_11RandomState_82negative_binomial(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_n, PyObject *__pyx_v_p, PyObject *__pyx_v_size) {
     7550+  PyArrayObject *__pyx_v_on = 0;
     7551+  PyArrayObject *__pyx_v_op = 0;
     7552+  double __pyx_v_fn;
     7553+  double __pyx_v_fp;
     7554+  PyObject *__pyx_r = NULL;
     7555+  __Pyx_RefNannyDeclarations
     7556+  int __pyx_t_1;
     7557+  PyObject *__pyx_t_2 = NULL;
     7558+  PyObject *__pyx_t_3 = NULL;
     7559+  PyObject *__pyx_t_4 = NULL;
     7560+  PyObject *__pyx_t_5 = NULL;
     7561+  int __pyx_lineno = 0;
     7562+  const char *__pyx_filename = NULL;
     7563+  int __pyx_clineno = 0;
     7564+  __Pyx_RefNannySetupContext("negative_binomial", 0);
     7565+
     7566   /* "mtrand.pyx":3419
     7567  *         cdef double fp
     7568  *
     7569@@ -14330,9 +14891,9 @@
     7570       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     7571       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7572       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3423; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7573-      goto __pyx_L7;
     7574+      goto __pyx_L4;
     7575     }
     7576-    __pyx_L7:;
     7577+    __pyx_L4:;
     7578 
     7579     /* "mtrand.pyx":3424
     7580  *             if fn <= 0:
     7581@@ -14356,7 +14917,7 @@
     7582       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     7583       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7584       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3425; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7585-      goto __pyx_L8;
     7586+      goto __pyx_L5;
     7587     }
     7588 
     7589     /* "mtrand.pyx":3426
     7590@@ -14381,9 +14942,9 @@
     7591       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     7592       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7593       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3427; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7594-      goto __pyx_L8;
     7595+      goto __pyx_L5;
     7596     }
     7597-    __pyx_L8:;
     7598+    __pyx_L5:;
     7599 
     7600     /* "mtrand.pyx":3428
     7601  *             elif fp > 1:
     7602@@ -14401,14 +14962,14 @@
     7603  *
     7604  *         PyErr_Clear()
     7605  */
     7606-    __pyx_t_2 = __pyx_f_6mtrand_discdd_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_negative_binomial, __pyx_v_size, __pyx_v_fn, __pyx_v_fp); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3428; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7607+    __pyx_t_2 = __pyx_f_6mtrand_discdd_array_sc(__pyx_v_self->internal_state, rk_negative_binomial, __pyx_v_size, __pyx_v_fn, __pyx_v_fp); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3428; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7608     __Pyx_GOTREF(__pyx_t_2);
     7609     __pyx_r = __pyx_t_2;
     7610     __pyx_t_2 = 0;
     7611     goto __pyx_L0;
     7612-    goto __pyx_L6;
     7613+    goto __pyx_L3;
     7614   }
     7615-  __pyx_L6:;
     7616+  __pyx_L3:;
     7617 
     7618   /* "mtrand.pyx":3431
     7619  *                                    size, fn, fp)
     7620@@ -14463,7 +15024,7 @@
     7621   __Pyx_GOTREF(__pyx_t_4);
     7622   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7623   __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3435; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7624-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     7625+  __Pyx_GOTREF(__pyx_t_2);
     7626   __Pyx_INCREF(__pyx_v_n);
     7627   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_n);
     7628   __Pyx_GIVEREF(__pyx_v_n);
     7629@@ -14475,7 +15036,7 @@
     7630   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     7631   __Pyx_DECREF(((PyObject *)__pyx_t_2)); __pyx_t_2 = 0;
     7632   __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3435; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7633-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     7634+  __Pyx_GOTREF(__pyx_t_2);
     7635   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_5);
     7636   __Pyx_GIVEREF(__pyx_t_5);
     7637   __pyx_t_5 = 0;
     7638@@ -14499,9 +15060,9 @@
     7639     __Pyx_Raise(__pyx_t_5, 0, 0, 0);
     7640     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     7641     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3436; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7642-    goto __pyx_L9;
     7643+    goto __pyx_L6;
     7644   }
     7645-  __pyx_L9:;
     7646+  __pyx_L6:;
     7647 
     7648   /* "mtrand.pyx":3437
     7649  *         if np.any(np.less_equal(n, 0)):
     7650@@ -14521,7 +15082,7 @@
     7651   __Pyx_GOTREF(__pyx_t_3);
     7652   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     7653   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3437; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7654-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     7655+  __Pyx_GOTREF(__pyx_t_5);
     7656   __Pyx_INCREF(__pyx_v_p);
     7657   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_v_p);
     7658   __Pyx_GIVEREF(__pyx_v_p);
     7659@@ -14533,7 +15094,7 @@
     7660   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     7661   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     7662   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3437; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7663-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     7664+  __Pyx_GOTREF(__pyx_t_5);
     7665   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_4);
     7666   __Pyx_GIVEREF(__pyx_t_4);
     7667   __pyx_t_4 = 0;
     7668@@ -14557,9 +15118,9 @@
     7669     __Pyx_Raise(__pyx_t_4, 0, 0, 0);
     7670     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     7671     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3438; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7672-    goto __pyx_L10;
     7673+    goto __pyx_L7;
     7674   }
     7675-  __pyx_L10:;
     7676+  __pyx_L7:;
     7677 
     7678   /* "mtrand.pyx":3439
     7679  *         if np.any(np.less(p, 0)):
     7680@@ -14579,7 +15140,7 @@
     7681   __Pyx_GOTREF(__pyx_t_2);
     7682   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     7683   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3439; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7684-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     7685+  __Pyx_GOTREF(__pyx_t_4);
     7686   __Pyx_INCREF(__pyx_v_p);
     7687   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_p);
     7688   __Pyx_GIVEREF(__pyx_v_p);
     7689@@ -14591,7 +15152,7 @@
     7690   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7691   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     7692   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3439; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7693-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     7694+  __Pyx_GOTREF(__pyx_t_4);
     7695   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_3);
     7696   __Pyx_GIVEREF(__pyx_t_3);
     7697   __pyx_t_3 = 0;
     7698@@ -14615,9 +15176,9 @@
     7699     __Pyx_Raise(__pyx_t_3, 0, 0, 0);
     7700     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     7701     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3440; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7702-    goto __pyx_L11;
     7703+    goto __pyx_L8;
     7704   }
     7705-  __pyx_L11:;
     7706+  __pyx_L8:;
     7707 
     7708   /* "mtrand.pyx":3441
     7709  *         if np.any(np.greater(p, 1)):
     7710@@ -14635,7 +15196,7 @@
     7711  *
     7712  *     def poisson(self, lam=1.0, size=None):
     7713  */
     7714-  __pyx_t_3 = __pyx_f_6mtrand_discdd_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_negative_binomial, __pyx_v_size, __pyx_v_on, __pyx_v_op); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3441; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7715+  __pyx_t_3 = __pyx_f_6mtrand_discdd_array(__pyx_v_self->internal_state, rk_negative_binomial, __pyx_v_size, __pyx_v_on, __pyx_v_op); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3441; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7716   __Pyx_GOTREF(__pyx_t_3);
     7717   __pyx_r = __pyx_t_3;
     7718   __pyx_t_3 = 0;
     7719@@ -14658,47 +15219,39 @@
     7720   return __pyx_r;
     7721 }
     7722 
     7723-/* "mtrand.pyx":3444
     7724- *                             on, op)
     7725- *
     7726- *     def poisson(self, lam=1.0, size=None):             # <<<<<<<<<<<<<<
     7727- *         """
     7728- *         poisson(lam=1.0, size=None)
     7729- */
     7730-
     7731-static PyObject *__pyx_pf_6mtrand_11RandomState_42poisson(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     7732-static char __pyx_doc_6mtrand_11RandomState_42poisson[] = "\n        poisson(lam=1.0, size=None)\n\n        Draw samples from a Poisson distribution.\n\n        The Poisson distribution is the limit of the Binomial\n        distribution for large N.\n\n        Parameters\n        ----------\n        lam : float\n            Expectation of interval, should be >= 0.\n        size : int or tuple of ints, optional\n            Output shape. If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Notes\n        -----\n        The Poisson distribution\n\n        .. math:: f(k; \\lambda)=\\frac{\\lambda^k e^{-\\lambda}}{k!}\n\n        For events with an expected separation :math:`\\lambda` the Poisson\n        distribution :math:`f(k; \\lambda)` describes the probability of\n        :math:`k` events occurring within the observed interval :math:`\\lambda`.\n\n        Because the output is limited to the range of the C long type, a\n        ValueError is raised when `lam` is within 10 sigma of the maximum\n        representable value.\n\n        References\n        ----------\n        .. [1] Weisstein, Eric W. \"Poisson Distribution.\" From MathWorld--A Wolfram\n               Web Resource. http://mathworld.wolfram.com/PoissonDistribution.html\n        .. [2] Wikipedia, \"Poisson distribution\",\n           http://en.wikipedia.org/wiki/Poisson_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> import numpy as np\n        >>> s = np.random.poisson(5, 10000)\n\n        Display histogram of the sample:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 14, normed=True)\n        >>> plt.show()\n\n        ";
     7733-static PyObject *__pyx_pf_6mtrand_11RandomState_42poisson(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     7734+/* Python wrapper */
     7735+static PyObject *__pyx_pw_6mtrand_11RandomState_85poisson(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     7736+static char __pyx_doc_6mtrand_11RandomState_84poisson[] = "\n        poisson(lam=1.0, size=None)\n\n        Draw samples from a Poisson distribution.\n\n        The Poisson distribution is the limit of the Binomial\n        distribution for large N.\n\n        Parameters\n        ----------\n        lam : float\n            Expectation of interval, should be >= 0.\n        size : int or tuple of ints, optional\n            Output shape. If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Notes\n        -----\n        The Poisson distribution\n\n        .. math:: f(k; \\lambda)=\\frac{\\lambda^k e^{-\\lambda}}{k!}\n\n        For events with an expected separation :math:`\\lambda` the Poisson\n        distribution :math:`f(k; \\lambda)` describes the probability of\n        :math:`k` events occurring within the observed interval :math:`\\lambda`.\n\n        Because the output is limited to the range of the C long type, a\n        ValueError is raised when `lam` is within 10 sigma of the maximum\n        representable value.\n\n        References\n        ----------\n        .. [1] Weisstein, Eric W. \"Poisson Distribution.\" From MathWorld--A Wolfram\n               Web Resource. http://mathworld.wolfram.com/PoissonDistribution.html\n        .. [2] Wikipedia, \"Poisson distribution\",\n           http://en.wikipedia.org/wiki/Poisson_distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> import numpy as np\n        >>> s = np.random.poisson(5, 10000)\n\n        Display histogram of the sample:\n\n        >>> import matplotlib.pyplot as plt\n        >>> count, bins, ignored = plt.hist(s, 14, normed=True)\n        >>> plt.show()\n\n        ";
     7737+static PyObject *__pyx_pw_6mtrand_11RandomState_85poisson(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     7738   PyObject *__pyx_v_lam = 0;
     7739   PyObject *__pyx_v_size = 0;
     7740-  PyArrayObject *__pyx_v_olam = 0;
     7741-  double __pyx_v_flam;
     7742-  PyObject *__pyx_r = NULL;
     7743+  PyObject *__pyx_r = 0;
     7744   __Pyx_RefNannyDeclarations
     7745-  int __pyx_t_1;
     7746-  PyObject *__pyx_t_2 = NULL;
     7747-  PyObject *__pyx_t_3 = NULL;
     7748-  PyObject *__pyx_t_4 = NULL;
     7749-  PyObject *__pyx_t_5 = NULL;
     7750-  int __pyx_lineno = 0;
     7751-  const char *__pyx_filename = NULL;
     7752-  int __pyx_clineno = 0;
     7753-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__lam,&__pyx_n_s__size,0};
     7754-  __Pyx_RefNannySetupContext("poisson");
     7755+  __Pyx_RefNannySetupContext("poisson (wrapper)", 0);
     7756   {
     7757+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__lam,&__pyx_n_s__size,0};
     7758     PyObject* values[2] = {0,0};
     7759     values[0] = __pyx_k_125;
     7760+
     7761+    /* "mtrand.pyx":3444
     7762+ *                             on, op)
     7763+ *
     7764+ *     def poisson(self, lam=1.0, size=None):             # <<<<<<<<<<<<<<
     7765+ *         """
     7766+ *         poisson(lam=1.0, size=None)
     7767+ */
     7768     values[1] = ((PyObject *)Py_None);
     7769     if (unlikely(__pyx_kwds)) {
     7770       Py_ssize_t kw_args;
     7771-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     7772+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     7773+      switch (pos_args) {
     7774         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     7775         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     7776         case  0: break;
     7777         default: goto __pyx_L5_argtuple_error;
     7778       }
     7779       kw_args = PyDict_Size(__pyx_kwds);
     7780-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     7781+      switch (pos_args) {
     7782         case  0:
     7783         if (kw_args > 0) {
     7784           PyObject* value = PyDict_GetItem(__pyx_kwds, __pyx_n_s__lam);
     7785@@ -14711,7 +15264,7 @@
     7786         }
     7787       }
     7788       if (unlikely(kw_args > 0)) {
     7789-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "poisson") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3444; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     7790+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "poisson") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3444; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     7791       }
     7792     } else {
     7793       switch (PyTuple_GET_SIZE(__pyx_args)) {
     7794@@ -14732,6 +15285,25 @@
     7795   __Pyx_RefNannyFinishContext();
     7796   return NULL;
     7797   __pyx_L4_argument_unpacking_done:;
     7798+  __pyx_r = __pyx_pf_6mtrand_11RandomState_84poisson(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_lam, __pyx_v_size);
     7799+  __Pyx_RefNannyFinishContext();
     7800+  return __pyx_r;
     7801+}
     7802+
     7803+static PyObject *__pyx_pf_6mtrand_11RandomState_84poisson(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_lam, PyObject *__pyx_v_size) {
     7804+  PyArrayObject *__pyx_v_olam = 0;
     7805+  double __pyx_v_flam;
     7806+  PyObject *__pyx_r = NULL;
     7807+  __Pyx_RefNannyDeclarations
     7808+  int __pyx_t_1;
     7809+  PyObject *__pyx_t_2 = NULL;
     7810+  PyObject *__pyx_t_3 = NULL;
     7811+  PyObject *__pyx_t_4 = NULL;
     7812+  PyObject *__pyx_t_5 = NULL;
     7813+  int __pyx_lineno = 0;
     7814+  const char *__pyx_filename = NULL;
     7815+  int __pyx_clineno = 0;
     7816+  __Pyx_RefNannySetupContext("poisson", 0);
     7817 
     7818   /* "mtrand.pyx":3498
     7819  *         cdef ndarray olam
     7820@@ -14759,8 +15331,7 @@
     7821  *                 raise ValueError("lam < 0")
     7822  *             if lam > self.poisson_lam_max:
     7823  */
     7824-    __pyx_t_2 = PyObject_RichCompare(__pyx_v_lam, __pyx_int_0, Py_LT); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3500; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7825-    __Pyx_GOTREF(__pyx_t_2);
     7826+    __pyx_t_2 = PyObject_RichCompare(__pyx_v_lam, __pyx_int_0, Py_LT); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3500; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7827     __pyx_t_1 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_1 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3500; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7828     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7829     if (__pyx_t_1) {
     7830@@ -14777,9 +15348,9 @@
     7831       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     7832       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7833       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3501; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7834-      goto __pyx_L7;
     7835+      goto __pyx_L4;
     7836     }
     7837-    __pyx_L7:;
     7838+    __pyx_L4:;
     7839 
     7840     /* "mtrand.pyx":3502
     7841  *             if lam < 0:
     7842@@ -14788,10 +15359,9 @@
     7843  *                 raise ValueError("lam value too large")
     7844  *             return discd_array_sc(self.internal_state, rk_poisson, size, flam)
     7845  */
     7846-    __pyx_t_2 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__poisson_lam_max); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3502; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7847+    __pyx_t_2 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__poisson_lam_max); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3502; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7848     __Pyx_GOTREF(__pyx_t_2);
     7849-    __pyx_t_3 = PyObject_RichCompare(__pyx_v_lam, __pyx_t_2, Py_GT); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3502; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7850-    __Pyx_GOTREF(__pyx_t_3);
     7851+    __pyx_t_3 = PyObject_RichCompare(__pyx_v_lam, __pyx_t_2, Py_GT); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3502; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7852     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7853     __pyx_t_1 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_1 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3502; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7854     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     7855@@ -14809,9 +15379,9 @@
     7856       __Pyx_Raise(__pyx_t_3, 0, 0, 0);
     7857       __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     7858       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3503; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7859-      goto __pyx_L8;
     7860+      goto __pyx_L5;
     7861     }
     7862-    __pyx_L8:;
     7863+    __pyx_L5:;
     7864 
     7865     /* "mtrand.pyx":3504
     7866  *             if lam > self.poisson_lam_max:
     7867@@ -14821,14 +15391,14 @@
     7868  *         PyErr_Clear()
     7869  */
     7870     __Pyx_XDECREF(__pyx_r);
     7871-    __pyx_t_3 = __pyx_f_6mtrand_discd_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_poisson, __pyx_v_size, __pyx_v_flam); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3504; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7872+    __pyx_t_3 = __pyx_f_6mtrand_discd_array_sc(__pyx_v_self->internal_state, rk_poisson, __pyx_v_size, __pyx_v_flam); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3504; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7873     __Pyx_GOTREF(__pyx_t_3);
     7874     __pyx_r = __pyx_t_3;
     7875     __pyx_t_3 = 0;
     7876     goto __pyx_L0;
     7877-    goto __pyx_L6;
     7878+    goto __pyx_L3;
     7879   }
     7880-  __pyx_L6:;
     7881+  __pyx_L3:;
     7882 
     7883   /* "mtrand.pyx":3506
     7884  *             return discd_array_sc(self.internal_state, rk_poisson, size, flam)
     7885@@ -14870,7 +15440,7 @@
     7886   __Pyx_GOTREF(__pyx_t_4);
     7887   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     7888   __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3509; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7889-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     7890+  __Pyx_GOTREF(__pyx_t_3);
     7891   __Pyx_INCREF(((PyObject *)__pyx_v_olam));
     7892   PyTuple_SET_ITEM(__pyx_t_3, 0, ((PyObject *)__pyx_v_olam));
     7893   __Pyx_GIVEREF(((PyObject *)__pyx_v_olam));
     7894@@ -14882,7 +15452,7 @@
     7895   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     7896   __Pyx_DECREF(((PyObject *)__pyx_t_3)); __pyx_t_3 = 0;
     7897   __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3509; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7898-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     7899+  __Pyx_GOTREF(__pyx_t_3);
     7900   PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_5);
     7901   __Pyx_GIVEREF(__pyx_t_5);
     7902   __pyx_t_5 = 0;
     7903@@ -14906,9 +15476,9 @@
     7904     __Pyx_Raise(__pyx_t_5, 0, 0, 0);
     7905     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     7906     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3510; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7907-    goto __pyx_L9;
     7908+    goto __pyx_L6;
     7909   }
     7910-  __pyx_L9:;
     7911+  __pyx_L6:;
     7912 
     7913   /* "mtrand.pyx":3511
     7914  *         if np.any(np.less(olam, 0)):
     7915@@ -14927,10 +15497,10 @@
     7916   __pyx_t_2 = PyObject_GetAttr(__pyx_t_5, __pyx_n_s__greater); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3511; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7917   __Pyx_GOTREF(__pyx_t_2);
     7918   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     7919-  __pyx_t_5 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__poisson_lam_max); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3511; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7920+  __pyx_t_5 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__poisson_lam_max); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3511; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7921   __Pyx_GOTREF(__pyx_t_5);
     7922   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3511; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7923-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     7924+  __Pyx_GOTREF(__pyx_t_4);
     7925   __Pyx_INCREF(((PyObject *)__pyx_v_olam));
     7926   PyTuple_SET_ITEM(__pyx_t_4, 0, ((PyObject *)__pyx_v_olam));
     7927   __Pyx_GIVEREF(((PyObject *)__pyx_v_olam));
     7928@@ -14942,7 +15512,7 @@
     7929   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     7930   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     7931   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3511; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7932-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     7933+  __Pyx_GOTREF(__pyx_t_4);
     7934   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_5);
     7935   __Pyx_GIVEREF(__pyx_t_5);
     7936   __pyx_t_5 = 0;
     7937@@ -14966,9 +15536,9 @@
     7938     __Pyx_Raise(__pyx_t_5, 0, 0, 0);
     7939     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     7940     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3512; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7941-    goto __pyx_L10;
     7942+    goto __pyx_L7;
     7943   }
     7944-  __pyx_L10:;
     7945+  __pyx_L7:;
     7946 
     7947   /* "mtrand.pyx":3513
     7948  *         if np.any(np.greater(olam, self.poisson_lam_max)):
     7949@@ -14978,7 +15548,7 @@
     7950  *     def zipf(self, a, size=None):
     7951  */
     7952   __Pyx_XDECREF(__pyx_r);
     7953-  __pyx_t_5 = __pyx_f_6mtrand_discd_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_poisson, __pyx_v_size, __pyx_v_olam); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3513; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7954+  __pyx_t_5 = __pyx_f_6mtrand_discd_array(__pyx_v_self->internal_state, rk_poisson, __pyx_v_size, __pyx_v_olam); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3513; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     7955   __Pyx_GOTREF(__pyx_t_5);
     7956   __pyx_r = __pyx_t_5;
     7957   __pyx_t_5 = 0;
     7958@@ -15000,49 +15570,40 @@
     7959   return __pyx_r;
     7960 }
     7961 
     7962-/* "mtrand.pyx":3515
     7963+/* Python wrapper */
     7964+static PyObject *__pyx_pw_6mtrand_11RandomState_87zipf(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     7965+static char __pyx_doc_6mtrand_11RandomState_86zipf[] = "\n        zipf(a, size=None)\n\n        Draw samples from a Zipf distribution.\n\n        Samples are drawn from a Zipf distribution with specified parameter\n        `a` > 1.\n\n        The Zipf distribution (also known as the zeta distribution) is a\n        continuous probability distribution that satisfies Zipf's law: the\n        frequency of an item is inversely proportional to its rank in a\n        frequency table.\n\n        Parameters\n        ----------\n        a : float > 1\n            Distribution parameter.\n        size : int or tuple of int, optional\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn; a single integer is equivalent in\n            its result to providing a mono-tuple, i.e., a 1-D array of length\n            *size* is returned.  The default is None, in which case a single\n            scalar is returned.\n\n        Returns\n        -------\n        samples : scalar or ndarray\n            The returned samples are greater than or equal to one.\n\n        See Also\n        --------\n        scipy.stats.distributions.zipf : probability density function,\n            distribution, or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Zipf distribution is\n\n        .. math:: p(x) = \\frac{x^{-a}}{\\zeta(a)},\n\n        where :math:`\\zeta` is the Riemann Zeta function.\n\n        It is named for the American linguist George Kingsley Zipf, who noted\n        that the frequency of any word in a sample of a language is inversely\n        proportional to its rank in the frequency table.\n\n        References\n        ----------\n        Zipf, G. K., *Selected Studies of the Principle of Relative Frequency\n        in Language*, Cambridge, MA: Harvard Univ. Press, 1932.\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> a = 2. # parameter\n        >>> s = np.random.zipf""(a, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> import scipy.special as sps\n        Truncate s values at 50 so plot is interesting\n        >>> count, bins, ignored = plt.hist(s[s<50], 50, normed=True)\n        >>> x = np.arange(1., 50.)\n        >>> y = x**(-a)/sps.zetac(a)\n        >>> plt.plot(x, y/max(y), linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     7966+static PyObject *__pyx_pw_6mtrand_11RandomState_87zipf(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     7967+  PyObject *__pyx_v_a = 0;
     7968+  PyObject *__pyx_v_size = 0;
     7969+  PyObject *__pyx_r = 0;
     7970+  __Pyx_RefNannyDeclarations
     7971+  __Pyx_RefNannySetupContext("zipf (wrapper)", 0);
     7972+  {
     7973+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__a,&__pyx_n_s__size,0};
     7974+    PyObject* values[2] = {0,0};
     7975+
     7976+    /* "mtrand.pyx":3515
     7977  *         return discd_array(self.internal_state, rk_poisson, size, olam)
     7978  *
     7979  *     def zipf(self, a, size=None):             # <<<<<<<<<<<<<<
     7980  *         """
     7981  *         zipf(a, size=None)
     7982  */
     7983-
     7984-static PyObject *__pyx_pf_6mtrand_11RandomState_43zipf(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     7985-static char __pyx_doc_6mtrand_11RandomState_43zipf[] = "\n        zipf(a, size=None)\n\n        Draw samples from a Zipf distribution.\n\n        Samples are drawn from a Zipf distribution with specified parameter\n        `a` > 1.\n\n        The Zipf distribution (also known as the zeta distribution) is a\n        continuous probability distribution that satisfies Zipf's law: the\n        frequency of an item is inversely proportional to its rank in a\n        frequency table.\n\n        Parameters\n        ----------\n        a : float > 1\n            Distribution parameter.\n        size : int or tuple of int, optional\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn; a single integer is equivalent in\n            its result to providing a mono-tuple, i.e., a 1-D array of length\n            *size* is returned.  The default is None, in which case a single\n            scalar is returned.\n\n        Returns\n        -------\n        samples : scalar or ndarray\n            The returned samples are greater than or equal to one.\n\n        See Also\n        --------\n        scipy.stats.distributions.zipf : probability density function,\n            distribution, or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Zipf distribution is\n\n        .. math:: p(x) = \\frac{x^{-a}}{\\zeta(a)},\n\n        where :math:`\\zeta` is the Riemann Zeta function.\n\n        It is named for the American linguist George Kingsley Zipf, who noted\n        that the frequency of any word in a sample of a language is inversely\n        proportional to its rank in the frequency table.\n\n        References\n        ----------\n        Zipf, G. K., *Selected Studies of the Principle of Relative Frequency\n        in Language*, Cambridge, MA: Harvard Univ. Press, 1932.\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> a = 2. # parameter\n        >>> s = np.random.zipf""(a, 1000)\n\n        Display the histogram of the samples, along with\n        the probability density function:\n\n        >>> import matplotlib.pyplot as plt\n        >>> import scipy.special as sps\n        Truncate s values at 50 so plot is interesting\n        >>> count, bins, ignored = plt.hist(s[s<50], 50, normed=True)\n        >>> x = np.arange(1., 50.)\n        >>> y = x**(-a)/sps.zetac(a)\n        >>> plt.plot(x, y/max(y), linewidth=2, color='r')\n        >>> plt.show()\n\n        ";
     7986-static PyObject *__pyx_pf_6mtrand_11RandomState_43zipf(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     7987-  PyObject *__pyx_v_a = 0;
     7988-  PyObject *__pyx_v_size = 0;
     7989-  PyArrayObject *__pyx_v_oa = 0;
     7990-  double __pyx_v_fa;
     7991-  PyObject *__pyx_r = NULL;
     7992-  __Pyx_RefNannyDeclarations
     7993-  int __pyx_t_1;
     7994-  PyObject *__pyx_t_2 = NULL;
     7995-  PyObject *__pyx_t_3 = NULL;
     7996-  PyObject *__pyx_t_4 = NULL;
     7997-  PyObject *__pyx_t_5 = NULL;
     7998-  int __pyx_lineno = 0;
     7999-  const char *__pyx_filename = NULL;
     8000-  int __pyx_clineno = 0;
     8001-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__a,&__pyx_n_s__size,0};
     8002-  __Pyx_RefNannySetupContext("zipf");
     8003-  {
     8004-    PyObject* values[2] = {0,0};
     8005     values[1] = ((PyObject *)Py_None);
     8006     if (unlikely(__pyx_kwds)) {
     8007       Py_ssize_t kw_args;
     8008-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     8009+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     8010+      switch (pos_args) {
     8011         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     8012         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     8013         case  0: break;
     8014         default: goto __pyx_L5_argtuple_error;
     8015       }
     8016       kw_args = PyDict_Size(__pyx_kwds);
     8017-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     8018+      switch (pos_args) {
     8019         case  0:
     8020-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__a);
     8021-        if (likely(values[0])) kw_args--;
     8022+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__a)) != 0)) kw_args--;
     8023         else goto __pyx_L5_argtuple_error;
     8024         case  1:
     8025         if (kw_args > 0) {
     8026@@ -15051,7 +15612,7 @@
     8027         }
     8028       }
     8029       if (unlikely(kw_args > 0)) {
     8030-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "zipf") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3515; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     8031+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "zipf") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3515; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     8032       }
     8033     } else {
     8034       switch (PyTuple_GET_SIZE(__pyx_args)) {
     8035@@ -15072,6 +15633,25 @@
     8036   __Pyx_RefNannyFinishContext();
     8037   return NULL;
     8038   __pyx_L4_argument_unpacking_done:;
     8039+  __pyx_r = __pyx_pf_6mtrand_11RandomState_86zipf(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_a, __pyx_v_size);
     8040+  __Pyx_RefNannyFinishContext();
     8041+  return __pyx_r;
     8042+}
     8043+
     8044+static PyObject *__pyx_pf_6mtrand_11RandomState_86zipf(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_a, PyObject *__pyx_v_size) {
     8045+  PyArrayObject *__pyx_v_oa = 0;
     8046+  double __pyx_v_fa;
     8047+  PyObject *__pyx_r = NULL;
     8048+  __Pyx_RefNannyDeclarations
     8049+  int __pyx_t_1;
     8050+  PyObject *__pyx_t_2 = NULL;
     8051+  PyObject *__pyx_t_3 = NULL;
     8052+  PyObject *__pyx_t_4 = NULL;
     8053+  PyObject *__pyx_t_5 = NULL;
     8054+  int __pyx_lineno = 0;
     8055+  const char *__pyx_filename = NULL;
     8056+  int __pyx_clineno = 0;
     8057+  __Pyx_RefNannySetupContext("zipf", 0);
     8058 
     8059   /* "mtrand.pyx":3590
     8060  *         cdef double fa
     8061@@ -15114,9 +15694,9 @@
     8062       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8063       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8064       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3593; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8065-      goto __pyx_L7;
     8066+      goto __pyx_L4;
     8067     }
     8068-    __pyx_L7:;
     8069+    __pyx_L4:;
     8070 
     8071     /* "mtrand.pyx":3594
     8072  *             if fa <= 1.0:
     8073@@ -15126,14 +15706,14 @@
     8074  *         PyErr_Clear()
     8075  */
     8076     __Pyx_XDECREF(__pyx_r);
     8077-    __pyx_t_2 = __pyx_f_6mtrand_discd_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_zipf, __pyx_v_size, __pyx_v_fa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3594; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8078+    __pyx_t_2 = __pyx_f_6mtrand_discd_array_sc(__pyx_v_self->internal_state, rk_zipf, __pyx_v_size, __pyx_v_fa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3594; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8079     __Pyx_GOTREF(__pyx_t_2);
     8080     __pyx_r = __pyx_t_2;
     8081     __pyx_t_2 = 0;
     8082     goto __pyx_L0;
     8083-    goto __pyx_L6;
     8084+    goto __pyx_L3;
     8085   }
     8086-  __pyx_L6:;
     8087+  __pyx_L3:;
     8088 
     8089   /* "mtrand.pyx":3596
     8090  *             return discd_array_sc(self.internal_state, rk_zipf, size, fa)
     8091@@ -15177,7 +15757,7 @@
     8092   __pyx_t_2 = PyFloat_FromDouble(1.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3599; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8093   __Pyx_GOTREF(__pyx_t_2);
     8094   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3599; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8095-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     8096+  __Pyx_GOTREF(__pyx_t_5);
     8097   __Pyx_INCREF(((PyObject *)__pyx_v_oa));
     8098   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_oa));
     8099   __Pyx_GIVEREF(((PyObject *)__pyx_v_oa));
     8100@@ -15189,7 +15769,7 @@
     8101   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     8102   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     8103   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3599; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8104-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     8105+  __Pyx_GOTREF(__pyx_t_5);
     8106   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     8107   __Pyx_GIVEREF(__pyx_t_2);
     8108   __pyx_t_2 = 0;
     8109@@ -15213,9 +15793,9 @@
     8110     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8111     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8112     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3600; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8113-    goto __pyx_L8;
     8114+    goto __pyx_L5;
     8115   }
     8116-  __pyx_L8:;
     8117+  __pyx_L5:;
     8118 
     8119   /* "mtrand.pyx":3601
     8120  *         if np.any(np.less_equal(oa, 1.0)):
     8121@@ -15225,7 +15805,7 @@
     8122  *     def geometric(self, p, size=None):
     8123  */
     8124   __Pyx_XDECREF(__pyx_r);
     8125-  __pyx_t_2 = __pyx_f_6mtrand_discd_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_zipf, __pyx_v_size, __pyx_v_oa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3601; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8126+  __pyx_t_2 = __pyx_f_6mtrand_discd_array(__pyx_v_self->internal_state, rk_zipf, __pyx_v_size, __pyx_v_oa); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3601; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8127   __Pyx_GOTREF(__pyx_t_2);
     8128   __pyx_r = __pyx_t_2;
     8129   __pyx_t_2 = 0;
     8130@@ -15247,49 +15827,40 @@
     8131   return __pyx_r;
     8132 }
     8133 
     8134-/* "mtrand.pyx":3603
     8135+/* Python wrapper */
     8136+static PyObject *__pyx_pw_6mtrand_11RandomState_89geometric(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     8137+static char __pyx_doc_6mtrand_11RandomState_88geometric[] = "\n        geometric(p, size=None)\n\n        Draw samples from the geometric distribution.\n\n        Bernoulli trials are experiments with one of two outcomes:\n        success or failure (an example of such an experiment is flipping\n        a coin).  The geometric distribution models the number of trials\n        that must be run in order to achieve success.  It is therefore\n        supported on the positive integers, ``k = 1, 2, ...``.\n\n        The probability mass function of the geometric distribution is\n\n        .. math:: f(k) = (1 - p)^{k - 1} p\n\n        where `p` is the probability of success of an individual trial.\n\n        Parameters\n        ----------\n        p : float\n            The probability of success of an individual trial.\n        size : tuple of ints\n            Number of values to draw from the distribution.  The output\n            is shaped according to `size`.\n\n        Returns\n        -------\n        out : ndarray\n            Samples from the geometric distribution, shaped according to\n            `size`.\n\n        Examples\n        --------\n        Draw ten thousand values from the geometric distribution,\n        with the probability of an individual success equal to 0.35:\n\n        >>> z = np.random.geometric(p=0.35, size=10000)\n\n        How many trials succeeded after a single run?\n\n        >>> (z == 1).sum() / 10000.\n        0.34889999999999999 #random\n\n        ";
     8138+static PyObject *__pyx_pw_6mtrand_11RandomState_89geometric(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     8139+  PyObject *__pyx_v_p = 0;
     8140+  PyObject *__pyx_v_size = 0;
     8141+  PyObject *__pyx_r = 0;
     8142+  __Pyx_RefNannyDeclarations
     8143+  __Pyx_RefNannySetupContext("geometric (wrapper)", 0);
     8144+  {
     8145+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__p,&__pyx_n_s__size,0};
     8146+    PyObject* values[2] = {0,0};
     8147+
     8148+    /* "mtrand.pyx":3603
     8149  *         return discd_array(self.internal_state, rk_zipf, size, oa)
     8150  *
     8151  *     def geometric(self, p, size=None):             # <<<<<<<<<<<<<<
     8152  *         """
     8153  *         geometric(p, size=None)
     8154  */
     8155-
     8156-static PyObject *__pyx_pf_6mtrand_11RandomState_44geometric(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     8157-static char __pyx_doc_6mtrand_11RandomState_44geometric[] = "\n        geometric(p, size=None)\n\n        Draw samples from the geometric distribution.\n\n        Bernoulli trials are experiments with one of two outcomes:\n        success or failure (an example of such an experiment is flipping\n        a coin).  The geometric distribution models the number of trials\n        that must be run in order to achieve success.  It is therefore\n        supported on the positive integers, ``k = 1, 2, ...``.\n\n        The probability mass function of the geometric distribution is\n\n        .. math:: f(k) = (1 - p)^{k - 1} p\n\n        where `p` is the probability of success of an individual trial.\n\n        Parameters\n        ----------\n        p : float\n            The probability of success of an individual trial.\n        size : tuple of ints\n            Number of values to draw from the distribution.  The output\n            is shaped according to `size`.\n\n        Returns\n        -------\n        out : ndarray\n            Samples from the geometric distribution, shaped according to\n            `size`.\n\n        Examples\n        --------\n        Draw ten thousand values from the geometric distribution,\n        with the probability of an individual success equal to 0.35:\n\n        >>> z = np.random.geometric(p=0.35, size=10000)\n\n        How many trials succeeded after a single run?\n\n        >>> (z == 1).sum() / 10000.\n        0.34889999999999999 #random\n\n        ";
     8158-static PyObject *__pyx_pf_6mtrand_11RandomState_44geometric(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     8159-  PyObject *__pyx_v_p = 0;
     8160-  PyObject *__pyx_v_size = 0;
     8161-  PyArrayObject *__pyx_v_op = 0;
     8162-  double __pyx_v_fp;
     8163-  PyObject *__pyx_r = NULL;
     8164-  __Pyx_RefNannyDeclarations
     8165-  int __pyx_t_1;
     8166-  PyObject *__pyx_t_2 = NULL;
     8167-  PyObject *__pyx_t_3 = NULL;
     8168-  PyObject *__pyx_t_4 = NULL;
     8169-  PyObject *__pyx_t_5 = NULL;
     8170-  int __pyx_lineno = 0;
     8171-  const char *__pyx_filename = NULL;
     8172-  int __pyx_clineno = 0;
     8173-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__p,&__pyx_n_s__size,0};
     8174-  __Pyx_RefNannySetupContext("geometric");
     8175-  {
     8176-    PyObject* values[2] = {0,0};
     8177     values[1] = ((PyObject *)Py_None);
     8178     if (unlikely(__pyx_kwds)) {
     8179       Py_ssize_t kw_args;
     8180-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     8181+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     8182+      switch (pos_args) {
     8183         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     8184         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     8185         case  0: break;
     8186         default: goto __pyx_L5_argtuple_error;
     8187       }
     8188       kw_args = PyDict_Size(__pyx_kwds);
     8189-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     8190+      switch (pos_args) {
     8191         case  0:
     8192-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__p);
     8193-        if (likely(values[0])) kw_args--;
     8194+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__p)) != 0)) kw_args--;
     8195         else goto __pyx_L5_argtuple_error;
     8196         case  1:
     8197         if (kw_args > 0) {
     8198@@ -15298,7 +15869,7 @@
     8199         }
     8200       }
     8201       if (unlikely(kw_args > 0)) {
     8202-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "geometric") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3603; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     8203+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "geometric") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3603; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     8204       }
     8205     } else {
     8206       switch (PyTuple_GET_SIZE(__pyx_args)) {
     8207@@ -15319,6 +15890,25 @@
     8208   __Pyx_RefNannyFinishContext();
     8209   return NULL;
     8210   __pyx_L4_argument_unpacking_done:;
     8211+  __pyx_r = __pyx_pf_6mtrand_11RandomState_88geometric(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_p, __pyx_v_size);
     8212+  __Pyx_RefNannyFinishContext();
     8213+  return __pyx_r;
     8214+}
     8215+
     8216+static PyObject *__pyx_pf_6mtrand_11RandomState_88geometric(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_p, PyObject *__pyx_v_size) {
     8217+  PyArrayObject *__pyx_v_op = 0;
     8218+  double __pyx_v_fp;
     8219+  PyObject *__pyx_r = NULL;
     8220+  __Pyx_RefNannyDeclarations
     8221+  int __pyx_t_1;
     8222+  PyObject *__pyx_t_2 = NULL;
     8223+  PyObject *__pyx_t_3 = NULL;
     8224+  PyObject *__pyx_t_4 = NULL;
     8225+  PyObject *__pyx_t_5 = NULL;
     8226+  int __pyx_lineno = 0;
     8227+  const char *__pyx_filename = NULL;
     8228+  int __pyx_clineno = 0;
     8229+  __Pyx_RefNannySetupContext("geometric", 0);
     8230 
     8231   /* "mtrand.pyx":3651
     8232  *         cdef double fp
     8233@@ -15361,9 +15951,9 @@
     8234       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8235       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8236       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3654; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8237-      goto __pyx_L7;
     8238+      goto __pyx_L4;
     8239     }
     8240-    __pyx_L7:;
     8241+    __pyx_L4:;
     8242 
     8243     /* "mtrand.pyx":3655
     8244  *             if fp < 0.0:
     8245@@ -15387,9 +15977,9 @@
     8246       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8247       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8248       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3656; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8249-      goto __pyx_L8;
     8250+      goto __pyx_L5;
     8251     }
     8252-    __pyx_L8:;
     8253+    __pyx_L5:;
     8254 
     8255     /* "mtrand.pyx":3657
     8256  *             if fp > 1.0:
     8257@@ -15399,14 +15989,14 @@
     8258  *         PyErr_Clear()
     8259  */
     8260     __Pyx_XDECREF(__pyx_r);
     8261-    __pyx_t_2 = __pyx_f_6mtrand_discd_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_geometric, __pyx_v_size, __pyx_v_fp); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3657; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8262+    __pyx_t_2 = __pyx_f_6mtrand_discd_array_sc(__pyx_v_self->internal_state, rk_geometric, __pyx_v_size, __pyx_v_fp); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3657; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8263     __Pyx_GOTREF(__pyx_t_2);
     8264     __pyx_r = __pyx_t_2;
     8265     __pyx_t_2 = 0;
     8266     goto __pyx_L0;
     8267-    goto __pyx_L6;
     8268+    goto __pyx_L3;
     8269   }
     8270-  __pyx_L6:;
     8271+  __pyx_L3:;
     8272 
     8273   /* "mtrand.pyx":3659
     8274  *             return discd_array_sc(self.internal_state, rk_geometric, size, fp)
     8275@@ -15450,7 +16040,7 @@
     8276   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3663; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8277   __Pyx_GOTREF(__pyx_t_2);
     8278   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3663; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8279-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     8280+  __Pyx_GOTREF(__pyx_t_5);
     8281   __Pyx_INCREF(((PyObject *)__pyx_v_op));
     8282   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_op));
     8283   __Pyx_GIVEREF(((PyObject *)__pyx_v_op));
     8284@@ -15462,7 +16052,7 @@
     8285   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     8286   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     8287   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3663; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8288-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     8289+  __Pyx_GOTREF(__pyx_t_5);
     8290   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     8291   __Pyx_GIVEREF(__pyx_t_2);
     8292   __pyx_t_2 = 0;
     8293@@ -15486,9 +16076,9 @@
     8294     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8295     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8296     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3664; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8297-    goto __pyx_L9;
     8298+    goto __pyx_L6;
     8299   }
     8300-  __pyx_L9:;
     8301+  __pyx_L6:;
     8302 
     8303   /* "mtrand.pyx":3665
     8304  *         if np.any(np.less(op, 0.0)):
     8305@@ -15510,7 +16100,7 @@
     8306   __pyx_t_2 = PyFloat_FromDouble(1.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3665; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8307   __Pyx_GOTREF(__pyx_t_2);
     8308   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3665; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8309-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     8310+  __Pyx_GOTREF(__pyx_t_4);
     8311   __Pyx_INCREF(((PyObject *)__pyx_v_op));
     8312   PyTuple_SET_ITEM(__pyx_t_4, 0, ((PyObject *)__pyx_v_op));
     8313   __Pyx_GIVEREF(((PyObject *)__pyx_v_op));
     8314@@ -15522,7 +16112,7 @@
     8315   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     8316   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     8317   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3665; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8318-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     8319+  __Pyx_GOTREF(__pyx_t_4);
     8320   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     8321   __Pyx_GIVEREF(__pyx_t_2);
     8322   __pyx_t_2 = 0;
     8323@@ -15546,9 +16136,9 @@
     8324     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8325     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8326     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3666; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8327-    goto __pyx_L10;
     8328+    goto __pyx_L7;
     8329   }
     8330-  __pyx_L10:;
     8331+  __pyx_L7:;
     8332 
     8333   /* "mtrand.pyx":3667
     8334  *         if np.any(np.greater(op, 1.0)):
     8335@@ -15558,7 +16148,7 @@
     8336  *     def hypergeometric(self, ngood, nbad, nsample, size=None):
     8337  */
     8338   __Pyx_XDECREF(__pyx_r);
     8339-  __pyx_t_2 = __pyx_f_6mtrand_discd_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_geometric, __pyx_v_size, __pyx_v_op); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3667; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8340+  __pyx_t_2 = __pyx_f_6mtrand_discd_array(__pyx_v_self->internal_state, rk_geometric, __pyx_v_size, __pyx_v_op); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3667; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8341   __Pyx_GOTREF(__pyx_t_2);
     8342   __pyx_r = __pyx_t_2;
     8343   __pyx_t_2 = 0;
     8344@@ -15580,46 +16170,33 @@
     8345   return __pyx_r;
     8346 }
     8347 
     8348-/* "mtrand.pyx":3669
     8349- *         return discd_array(self.internal_state, rk_geometric, size, op)
     8350- *
     8351- *     def hypergeometric(self, ngood, nbad, nsample, size=None):             # <<<<<<<<<<<<<<
     8352- *         """
     8353- *         hypergeometric(ngood, nbad, nsample, size=None)
     8354- */
     8355-
     8356-static PyObject *__pyx_pf_6mtrand_11RandomState_45hypergeometric(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     8357-static char __pyx_doc_6mtrand_11RandomState_45hypergeometric[] = "\n        hypergeometric(ngood, nbad, nsample, size=None)\n\n        Draw samples from a Hypergeometric distribution.\n\n        Samples are drawn from a Hypergeometric distribution with specified\n        parameters, ngood (ways to make a good selection), nbad (ways to make\n        a bad selection), and nsample = number of items sampled, which is less\n        than or equal to the sum ngood + nbad.\n\n        Parameters\n        ----------\n        ngood : float (but truncated to an integer)\n                parameter, > 0.\n        nbad  : float\n                parameter, >= 0.\n        nsample  : float\n                   parameter, > 0 and <= ngood+nbad\n        size : {tuple, int}\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n                  where the values are all integers in  [0, n].\n\n        See Also\n        --------\n        scipy.stats.distributions.hypergeom : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Hypergeometric distribution is\n\n        .. math:: P(x) = \\frac{\\binom{m}{n}\\binom{N-m}{n-x}}{\\binom{N}{n}},\n\n        where :math:`0 \\le x \\le m` and :math:`n+m-N \\le x \\le n`\n\n        for P(x) the probability of x successes, n = ngood, m = nbad, and\n        N = number of samples.\n\n        Consider an urn with black and white marbles in it, ngood of them\n        black and nbad are white. If you draw nsample balls without\n        replacement, then the Hypergeometric distribution describes the\n        distribution of black balls in the drawn sample.\n\n        Note that this distribution is very similar to the Binomial\n        distribution, except that in this case, samples are drawn without\n        replacement, whereas in the Binomial case samples are drawn wit""h\n        replacement (or the sample space is infinite). As the sample space\n        becomes large, this distribution approaches the Binomial.\n\n        References\n        ----------\n        .. [1] Lentner, Marvin, \"Elementary Applied Statistics\", Bogden\n               and Quigley, 1972.\n        .. [2] Weisstein, Eric W. \"Hypergeometric Distribution.\" From\n               MathWorld--A Wolfram Web Resource.\n               http://mathworld.wolfram.com/HypergeometricDistribution.html\n        .. [3] Wikipedia, \"Hypergeometric-distribution\",\n               http://en.wikipedia.org/wiki/Hypergeometric-distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> ngood, nbad, nsamp = 100, 2, 10\n        # number of good, number of bad, and number of samples\n        >>> s = np.random.hypergeometric(ngood, nbad, nsamp, 1000)\n        >>> hist(s)\n        #   note that it is very unlikely to grab both bad items\n\n        Suppose you have an urn with 15 white and 15 black marbles.\n        If you pull 15 marbles at random, how likely is it that\n        12 or more of them are one color?\n\n        >>> s = np.random.hypergeometric(15, 15, 15, 100000)\n        >>> sum(s>=12)/100000. + sum(s<=3)/100000.\n        #   answer = 0.003 ... pretty unlikely!\n\n        ";
     8358-static PyObject *__pyx_pf_6mtrand_11RandomState_45hypergeometric(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     8359+/* Python wrapper */
     8360+static PyObject *__pyx_pw_6mtrand_11RandomState_91hypergeometric(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     8361+static char __pyx_doc_6mtrand_11RandomState_90hypergeometric[] = "\n        hypergeometric(ngood, nbad, nsample, size=None)\n\n        Draw samples from a Hypergeometric distribution.\n\n        Samples are drawn from a Hypergeometric distribution with specified\n        parameters, ngood (ways to make a good selection), nbad (ways to make\n        a bad selection), and nsample = number of items sampled, which is less\n        than or equal to the sum ngood + nbad.\n\n        Parameters\n        ----------\n        ngood : float (but truncated to an integer)\n                parameter, > 0.\n        nbad  : float\n                parameter, >= 0.\n        nsample  : float\n                   parameter, > 0 and <= ngood+nbad\n        size : {tuple, int}\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n                  where the values are all integers in  [0, n].\n\n        See Also\n        --------\n        scipy.stats.distributions.hypergeom : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Hypergeometric distribution is\n\n        .. math:: P(x) = \\frac{\\binom{m}{n}\\binom{N-m}{n-x}}{\\binom{N}{n}},\n\n        where :math:`0 \\le x \\le m` and :math:`n+m-N \\le x \\le n`\n\n        for P(x) the probability of x successes, n = ngood, m = nbad, and\n        N = number of samples.\n\n        Consider an urn with black and white marbles in it, ngood of them\n        black and nbad are white. If you draw nsample balls without\n        replacement, then the Hypergeometric distribution describes the\n        distribution of black balls in the drawn sample.\n\n        Note that this distribution is very similar to the Binomial\n        distribution, except that in this case, samples are drawn without\n        replacement, whereas in the Binomial case samples are drawn wit""h\n        replacement (or the sample space is infinite). As the sample space\n        becomes large, this distribution approaches the Binomial.\n\n        References\n        ----------\n        .. [1] Lentner, Marvin, \"Elementary Applied Statistics\", Bogden\n               and Quigley, 1972.\n        .. [2] Weisstein, Eric W. \"Hypergeometric Distribution.\" From\n               MathWorld--A Wolfram Web Resource.\n               http://mathworld.wolfram.com/HypergeometricDistribution.html\n        .. [3] Wikipedia, \"Hypergeometric-distribution\",\n               http://en.wikipedia.org/wiki/Hypergeometric-distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> ngood, nbad, nsamp = 100, 2, 10\n        # number of good, number of bad, and number of samples\n        >>> s = np.random.hypergeometric(ngood, nbad, nsamp, 1000)\n        >>> hist(s)\n        #   note that it is very unlikely to grab both bad items\n\n        Suppose you have an urn with 15 white and 15 black marbles.\n        If you pull 15 marbles at random, how likely is it that\n        12 or more of them are one color?\n\n        >>> s = np.random.hypergeometric(15, 15, 15, 100000)\n        >>> sum(s>=12)/100000. + sum(s<=3)/100000.\n        #   answer = 0.003 ... pretty unlikely!\n\n        ";
     8362+static PyObject *__pyx_pw_6mtrand_11RandomState_91hypergeometric(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     8363   PyObject *__pyx_v_ngood = 0;
     8364   PyObject *__pyx_v_nbad = 0;
     8365   PyObject *__pyx_v_nsample = 0;
     8366   PyObject *__pyx_v_size = 0;
     8367-  PyArrayObject *__pyx_v_ongood = 0;
     8368-  PyArrayObject *__pyx_v_onbad = 0;
     8369-  PyArrayObject *__pyx_v_onsample = 0;
     8370-  long __pyx_v_lngood;
     8371-  long __pyx_v_lnbad;
     8372-  long __pyx_v_lnsample;
     8373-  PyObject *__pyx_r = NULL;
     8374+  PyObject *__pyx_r = 0;
     8375   __Pyx_RefNannyDeclarations
     8376-  int __pyx_t_1;
     8377-  PyObject *__pyx_t_2 = NULL;
     8378-  PyObject *__pyx_t_3 = NULL;
     8379-  PyObject *__pyx_t_4 = NULL;
     8380-  PyObject *__pyx_t_5 = NULL;
     8381-  PyObject *__pyx_t_6 = NULL;
     8382-  int __pyx_lineno = 0;
     8383-  const char *__pyx_filename = NULL;
     8384-  int __pyx_clineno = 0;
     8385-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__ngood,&__pyx_n_s__nbad,&__pyx_n_s__nsample,&__pyx_n_s__size,0};
     8386-  __Pyx_RefNannySetupContext("hypergeometric");
     8387+  __Pyx_RefNannySetupContext("hypergeometric (wrapper)", 0);
     8388   {
     8389+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__ngood,&__pyx_n_s__nbad,&__pyx_n_s__nsample,&__pyx_n_s__size,0};
     8390     PyObject* values[4] = {0,0,0,0};
     8391+
     8392+    /* "mtrand.pyx":3669
     8393+ *         return discd_array(self.internal_state, rk_geometric, size, op)
     8394+ *
     8395+ *     def hypergeometric(self, ngood, nbad, nsample, size=None):             # <<<<<<<<<<<<<<
     8396+ *         """
     8397+ *         hypergeometric(ngood, nbad, nsample, size=None)
     8398+ */
     8399     values[3] = ((PyObject *)Py_None);
     8400     if (unlikely(__pyx_kwds)) {
     8401       Py_ssize_t kw_args;
     8402-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     8403+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     8404+      switch (pos_args) {
     8405         case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
     8406         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     8407         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     8408@@ -15628,20 +16205,17 @@
     8409         default: goto __pyx_L5_argtuple_error;
     8410       }
     8411       kw_args = PyDict_Size(__pyx_kwds);
     8412-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     8413+      switch (pos_args) {
     8414         case  0:
     8415-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__ngood);
     8416-        if (likely(values[0])) kw_args--;
     8417+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__ngood)) != 0)) kw_args--;
     8418         else goto __pyx_L5_argtuple_error;
     8419         case  1:
     8420-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__nbad);
     8421-        if (likely(values[1])) kw_args--;
     8422+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__nbad)) != 0)) kw_args--;
     8423         else {
     8424           __Pyx_RaiseArgtupleInvalid("hypergeometric", 0, 3, 4, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3669; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     8425         }
     8426         case  2:
     8427-        values[2] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__nsample);
     8428-        if (likely(values[2])) kw_args--;
     8429+        if (likely((values[2] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__nsample)) != 0)) kw_args--;
     8430         else {
     8431           __Pyx_RaiseArgtupleInvalid("hypergeometric", 0, 3, 4, 2); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3669; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     8432         }
     8433@@ -15652,7 +16226,7 @@
     8434         }
     8435       }
     8436       if (unlikely(kw_args > 0)) {
     8437-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "hypergeometric") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3669; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     8438+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "hypergeometric") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3669; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     8439       }
     8440     } else {
     8441       switch (PyTuple_GET_SIZE(__pyx_args)) {
     8442@@ -15677,6 +16251,30 @@
     8443   __Pyx_RefNannyFinishContext();
     8444   return NULL;
     8445   __pyx_L4_argument_unpacking_done:;
     8446+  __pyx_r = __pyx_pf_6mtrand_11RandomState_90hypergeometric(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_ngood, __pyx_v_nbad, __pyx_v_nsample, __pyx_v_size);
     8447+  __Pyx_RefNannyFinishContext();
     8448+  return __pyx_r;
     8449+}
     8450+
     8451+static PyObject *__pyx_pf_6mtrand_11RandomState_90hypergeometric(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_ngood, PyObject *__pyx_v_nbad, PyObject *__pyx_v_nsample, PyObject *__pyx_v_size) {
     8452+  PyArrayObject *__pyx_v_ongood = 0;
     8453+  PyArrayObject *__pyx_v_onbad = 0;
     8454+  PyArrayObject *__pyx_v_onsample = 0;
     8455+  long __pyx_v_lngood;
     8456+  long __pyx_v_lnbad;
     8457+  long __pyx_v_lnsample;
     8458+  PyObject *__pyx_r = NULL;
     8459+  __Pyx_RefNannyDeclarations
     8460+  int __pyx_t_1;
     8461+  PyObject *__pyx_t_2 = NULL;
     8462+  PyObject *__pyx_t_3 = NULL;
     8463+  PyObject *__pyx_t_4 = NULL;
     8464+  PyObject *__pyx_t_5 = NULL;
     8465+  PyObject *__pyx_t_6 = NULL;
     8466+  int __pyx_lineno = 0;
     8467+  const char *__pyx_filename = NULL;
     8468+  int __pyx_clineno = 0;
     8469+  __Pyx_RefNannySetupContext("hypergeometric", 0);
     8470 
     8471   /* "mtrand.pyx":3756
     8472  *         cdef long lngood, lnbad, lnsample
     8473@@ -15722,8 +16320,7 @@
     8474  *                 raise ValueError("ngood < 1")
     8475  *             if nbad < 1:
     8476  */
     8477-    __pyx_t_2 = PyObject_RichCompare(__pyx_v_ngood, __pyx_int_1, Py_LT); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3760; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8478-    __Pyx_GOTREF(__pyx_t_2);
     8479+    __pyx_t_2 = PyObject_RichCompare(__pyx_v_ngood, __pyx_int_1, Py_LT); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3760; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8480     __pyx_t_1 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_1 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3760; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8481     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8482     if (__pyx_t_1) {
     8483@@ -15740,9 +16337,9 @@
     8484       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8485       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8486       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3761; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8487-      goto __pyx_L7;
     8488+      goto __pyx_L4;
     8489     }
     8490-    __pyx_L7:;
     8491+    __pyx_L4:;
     8492 
     8493     /* "mtrand.pyx":3762
     8494  *             if ngood < 1:
     8495@@ -15751,8 +16348,7 @@
     8496  *                 raise ValueError("nbad < 1")
     8497  *             if nsample < 1:
     8498  */
     8499-    __pyx_t_2 = PyObject_RichCompare(__pyx_v_nbad, __pyx_int_1, Py_LT); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3762; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8500-    __Pyx_GOTREF(__pyx_t_2);
     8501+    __pyx_t_2 = PyObject_RichCompare(__pyx_v_nbad, __pyx_int_1, Py_LT); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3762; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8502     __pyx_t_1 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_1 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3762; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8503     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8504     if (__pyx_t_1) {
     8505@@ -15769,9 +16365,9 @@
     8506       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8507       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8508       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3763; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8509-      goto __pyx_L8;
     8510+      goto __pyx_L5;
     8511     }
     8512-    __pyx_L8:;
     8513+    __pyx_L5:;
     8514 
     8515     /* "mtrand.pyx":3764
     8516  *             if nbad < 1:
     8517@@ -15780,8 +16376,7 @@
     8518  *                 raise ValueError("nsample < 1")
     8519  *             if ngood + nbad < nsample:
     8520  */
     8521-    __pyx_t_2 = PyObject_RichCompare(__pyx_v_nsample, __pyx_int_1, Py_LT); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3764; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8522-    __Pyx_GOTREF(__pyx_t_2);
     8523+    __pyx_t_2 = PyObject_RichCompare(__pyx_v_nsample, __pyx_int_1, Py_LT); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3764; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8524     __pyx_t_1 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_1 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3764; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8525     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8526     if (__pyx_t_1) {
     8527@@ -15798,9 +16393,9 @@
     8528       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8529       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8530       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3765; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8531-      goto __pyx_L9;
     8532+      goto __pyx_L6;
     8533     }
     8534-    __pyx_L9:;
     8535+    __pyx_L6:;
     8536 
     8537     /* "mtrand.pyx":3766
     8538  *             if nsample < 1:
     8539@@ -15811,8 +16406,7 @@
     8540  */
     8541     __pyx_t_2 = PyNumber_Add(__pyx_v_ngood, __pyx_v_nbad); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3766; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8542     __Pyx_GOTREF(__pyx_t_2);
     8543-    __pyx_t_3 = PyObject_RichCompare(__pyx_t_2, __pyx_v_nsample, Py_LT); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3766; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8544-    __Pyx_GOTREF(__pyx_t_3);
     8545+    __pyx_t_3 = PyObject_RichCompare(__pyx_t_2, __pyx_v_nsample, Py_LT); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3766; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8546     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8547     __pyx_t_1 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_1 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3766; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8548     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     8549@@ -15830,9 +16424,9 @@
     8550       __Pyx_Raise(__pyx_t_3, 0, 0, 0);
     8551       __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     8552       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3767; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8553-      goto __pyx_L10;
     8554+      goto __pyx_L7;
     8555     }
     8556-    __pyx_L10:;
     8557+    __pyx_L7:;
     8558 
     8559     /* "mtrand.pyx":3768
     8560  *             if ngood + nbad < nsample:
     8561@@ -15850,14 +16444,14 @@
     8562  *
     8563  *
     8564  */
     8565-    __pyx_t_3 = __pyx_f_6mtrand_discnmN_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_hypergeometric, __pyx_v_size, __pyx_v_lngood, __pyx_v_lnbad, __pyx_v_lnsample); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3768; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8566+    __pyx_t_3 = __pyx_f_6mtrand_discnmN_array_sc(__pyx_v_self->internal_state, rk_hypergeometric, __pyx_v_size, __pyx_v_lngood, __pyx_v_lnbad, __pyx_v_lnsample); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3768; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8567     __Pyx_GOTREF(__pyx_t_3);
     8568     __pyx_r = __pyx_t_3;
     8569     __pyx_t_3 = 0;
     8570     goto __pyx_L0;
     8571-    goto __pyx_L6;
     8572+    goto __pyx_L3;
     8573   }
     8574-  __pyx_L6:;
     8575+  __pyx_L3:;
     8576 
     8577   /* "mtrand.pyx":3772
     8578  *
     8579@@ -15925,7 +16519,7 @@
     8580   __Pyx_GOTREF(__pyx_t_4);
     8581   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     8582   __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3777; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8583-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     8584+  __Pyx_GOTREF(__pyx_t_3);
     8585   __Pyx_INCREF(((PyObject *)__pyx_v_ongood));
     8586   PyTuple_SET_ITEM(__pyx_t_3, 0, ((PyObject *)__pyx_v_ongood));
     8587   __Pyx_GIVEREF(((PyObject *)__pyx_v_ongood));
     8588@@ -15937,7 +16531,7 @@
     8589   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     8590   __Pyx_DECREF(((PyObject *)__pyx_t_3)); __pyx_t_3 = 0;
     8591   __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3777; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8592-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     8593+  __Pyx_GOTREF(__pyx_t_3);
     8594   PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_5);
     8595   __Pyx_GIVEREF(__pyx_t_5);
     8596   __pyx_t_5 = 0;
     8597@@ -15961,9 +16555,9 @@
     8598     __Pyx_Raise(__pyx_t_5, 0, 0, 0);
     8599     __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     8600     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3778; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8601-    goto __pyx_L11;
     8602+    goto __pyx_L8;
     8603   }
     8604-  __pyx_L11:;
     8605+  __pyx_L8:;
     8606 
     8607   /* "mtrand.pyx":3779
     8608  *         if np.any(np.less(ongood, 1)):
     8609@@ -15983,7 +16577,7 @@
     8610   __Pyx_GOTREF(__pyx_t_2);
     8611   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     8612   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3779; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8613-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     8614+  __Pyx_GOTREF(__pyx_t_5);
     8615   __Pyx_INCREF(((PyObject *)__pyx_v_onbad));
     8616   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_onbad));
     8617   __Pyx_GIVEREF(((PyObject *)__pyx_v_onbad));
     8618@@ -15995,7 +16589,7 @@
     8619   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8620   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     8621   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3779; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8622-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     8623+  __Pyx_GOTREF(__pyx_t_5);
     8624   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_4);
     8625   __Pyx_GIVEREF(__pyx_t_4);
     8626   __pyx_t_4 = 0;
     8627@@ -16019,9 +16613,9 @@
     8628     __Pyx_Raise(__pyx_t_4, 0, 0, 0);
     8629     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     8630     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3780; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8631-    goto __pyx_L12;
     8632+    goto __pyx_L9;
     8633   }
     8634-  __pyx_L12:;
     8635+  __pyx_L9:;
     8636 
     8637   /* "mtrand.pyx":3781
     8638  *         if np.any(np.less(onbad, 1)):
     8639@@ -16041,7 +16635,7 @@
     8640   __Pyx_GOTREF(__pyx_t_3);
     8641   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     8642   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3781; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8643-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     8644+  __Pyx_GOTREF(__pyx_t_4);
     8645   __Pyx_INCREF(((PyObject *)__pyx_v_onsample));
     8646   PyTuple_SET_ITEM(__pyx_t_4, 0, ((PyObject *)__pyx_v_onsample));
     8647   __Pyx_GIVEREF(((PyObject *)__pyx_v_onsample));
     8648@@ -16053,7 +16647,7 @@
     8649   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     8650   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     8651   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3781; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8652-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     8653+  __Pyx_GOTREF(__pyx_t_4);
     8654   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     8655   __Pyx_GIVEREF(__pyx_t_2);
     8656   __pyx_t_2 = 0;
     8657@@ -16077,9 +16671,9 @@
     8658     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8659     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8660     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3782; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8661-    goto __pyx_L13;
     8662+    goto __pyx_L10;
     8663   }
     8664-  __pyx_L13:;
     8665+  __pyx_L10:;
     8666 
     8667   /* "mtrand.pyx":3783
     8668  *         if np.any(np.less(onsample, 1)):
     8669@@ -16104,7 +16698,7 @@
     8670   __Pyx_GOTREF(__pyx_t_3);
     8671   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8672   __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3783; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8673-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     8674+  __Pyx_GOTREF(__pyx_t_2);
     8675   __Pyx_INCREF(((PyObject *)__pyx_v_ongood));
     8676   PyTuple_SET_ITEM(__pyx_t_2, 0, ((PyObject *)__pyx_v_ongood));
     8677   __Pyx_GIVEREF(((PyObject *)__pyx_v_ongood));
     8678@@ -16116,7 +16710,7 @@
     8679   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     8680   __Pyx_DECREF(((PyObject *)__pyx_t_2)); __pyx_t_2 = 0;
     8681   __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3783; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8682-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     8683+  __Pyx_GOTREF(__pyx_t_2);
     8684   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_6);
     8685   __Pyx_GIVEREF(__pyx_t_6);
     8686   __Pyx_INCREF(((PyObject *)__pyx_v_onsample));
     8687@@ -16128,7 +16722,7 @@
     8688   __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     8689   __Pyx_DECREF(((PyObject *)__pyx_t_2)); __pyx_t_2 = 0;
     8690   __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3783; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8691-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     8692+  __Pyx_GOTREF(__pyx_t_2);
     8693   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_6);
     8694   __Pyx_GIVEREF(__pyx_t_6);
     8695   __pyx_t_6 = 0;
     8696@@ -16152,9 +16746,9 @@
     8697     __Pyx_Raise(__pyx_t_6, 0, 0, 0);
     8698     __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     8699     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3784; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8700-    goto __pyx_L14;
     8701+    goto __pyx_L11;
     8702   }
     8703-  __pyx_L14:;
     8704+  __pyx_L11:;
     8705 
     8706   /* "mtrand.pyx":3785
     8707  *         if np.any(np.less(np.add(ongood, onbad),onsample)):
     8708@@ -16172,7 +16766,7 @@
     8709  *
     8710  *     def logseries(self, p, size=None):
     8711  */
     8712-  __pyx_t_6 = __pyx_f_6mtrand_discnmN_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_hypergeometric, __pyx_v_size, __pyx_v_ongood, __pyx_v_onbad, __pyx_v_onsample); if (unlikely(!__pyx_t_6)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3785; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8713+  __pyx_t_6 = __pyx_f_6mtrand_discnmN_array(__pyx_v_self->internal_state, rk_hypergeometric, __pyx_v_size, __pyx_v_ongood, __pyx_v_onbad, __pyx_v_onsample); if (unlikely(!__pyx_t_6)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3785; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8714   __Pyx_GOTREF(__pyx_t_6);
     8715   __pyx_r = __pyx_t_6;
     8716   __pyx_t_6 = 0;
     8717@@ -16197,49 +16791,40 @@
     8718   return __pyx_r;
     8719 }
     8720 
     8721-/* "mtrand.pyx":3788
     8722+/* Python wrapper */
     8723+static PyObject *__pyx_pw_6mtrand_11RandomState_93logseries(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     8724+static char __pyx_doc_6mtrand_11RandomState_92logseries[] = "\n        logseries(p, size=None)\n\n        Draw samples from a Logarithmic Series distribution.\n\n        Samples are drawn from a Log Series distribution with specified\n        parameter, p (probability, 0 < p < 1).\n\n        Parameters\n        ----------\n        loc : float\n\n        scale : float > 0.\n\n        size : {tuple, int}\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n                  where the values are all integers in  [0, n].\n\n        See Also\n        --------\n        scipy.stats.distributions.logser : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Log Series distribution is\n\n        .. math:: P(k) = \\frac{-p^k}{k \\ln(1-p)},\n\n        where p = probability.\n\n        The Log Series distribution is frequently used to represent species\n        richness and occurrence, first proposed by Fisher, Corbet, and\n        Williams in 1943 [2].  It may also be used to model the numbers of\n        occupants seen in cars [3].\n\n        References\n        ----------\n        .. [1] Buzas, Martin A.; Culver, Stephen J.,  Understanding regional\n               species diversity through the log series distribution of\n               occurrences: BIODIVERSITY RESEARCH Diversity & Distributions,\n               Volume 5, Number 5, September 1999 , pp. 187-195(9).\n        .. [2] Fisher, R.A,, A.S. Corbet, and C.B. Williams. 1943. The\n               relation between the number of species and the number of\n               individuals in a random sample of an animal population.\n               Journal of Animal Ecology, 12:42-58.\n        .. [3] D. J. Hand, F. Daly, D. Lunn, E. Ostrowski, A Handbook of Small\n               Data Sets, CRC Press, 1994.\n        .. [4] Wikipedia, \"Log""arithmic-distribution\",\n               http://en.wikipedia.org/wiki/Logarithmic-distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> a = .6\n        >>> s = np.random.logseries(a, 10000)\n        >>> count, bins, ignored = plt.hist(s)\n\n        #   plot against distribution\n\n        >>> def logseries(k, p):\n        ...     return -p**k/(k*log(1-p))\n        >>> plt.plot(bins, logseries(bins, a)*count.max()/\n                     logseries(bins, a).max(), 'r')\n        >>> plt.show()\n\n        ";
     8725+static PyObject *__pyx_pw_6mtrand_11RandomState_93logseries(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     8726+  PyObject *__pyx_v_p = 0;
     8727+  PyObject *__pyx_v_size = 0;
     8728+  PyObject *__pyx_r = 0;
     8729+  __Pyx_RefNannyDeclarations
     8730+  __Pyx_RefNannySetupContext("logseries (wrapper)", 0);
     8731+  {
     8732+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__p,&__pyx_n_s__size,0};
     8733+    PyObject* values[2] = {0,0};
     8734+
     8735+    /* "mtrand.pyx":3788
     8736  *             ongood, onbad, onsample)
     8737  *
     8738  *     def logseries(self, p, size=None):             # <<<<<<<<<<<<<<
     8739  *         """
     8740  *         logseries(p, size=None)
     8741  */
     8742-
     8743-static PyObject *__pyx_pf_6mtrand_11RandomState_46logseries(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     8744-static char __pyx_doc_6mtrand_11RandomState_46logseries[] = "\n        logseries(p, size=None)\n\n        Draw samples from a Logarithmic Series distribution.\n\n        Samples are drawn from a Log Series distribution with specified\n        parameter, p (probability, 0 < p < 1).\n\n        Parameters\n        ----------\n        loc : float\n\n        scale : float > 0.\n\n        size : {tuple, int}\n            Output shape.  If the given shape is, e.g., ``(m, n, k)``, then\n            ``m * n * k`` samples are drawn.\n\n        Returns\n        -------\n        samples : {ndarray, scalar}\n                  where the values are all integers in  [0, n].\n\n        See Also\n        --------\n        scipy.stats.distributions.logser : probability density function,\n            distribution or cumulative density function, etc.\n\n        Notes\n        -----\n        The probability density for the Log Series distribution is\n\n        .. math:: P(k) = \\frac{-p^k}{k \\ln(1-p)},\n\n        where p = probability.\n\n        The Log Series distribution is frequently used to represent species\n        richness and occurrence, first proposed by Fisher, Corbet, and\n        Williams in 1943 [2].  It may also be used to model the numbers of\n        occupants seen in cars [3].\n\n        References\n        ----------\n        .. [1] Buzas, Martin A.; Culver, Stephen J.,  Understanding regional\n               species diversity through the log series distribution of\n               occurrences: BIODIVERSITY RESEARCH Diversity & Distributions,\n               Volume 5, Number 5, September 1999 , pp. 187-195(9).\n        .. [2] Fisher, R.A,, A.S. Corbet, and C.B. Williams. 1943. The\n               relation between the number of species and the number of\n               individuals in a random sample of an animal population.\n               Journal of Animal Ecology, 12:42-58.\n        .. [3] D. J. Hand, F. Daly, D. Lunn, E. Ostrowski, A Handbook of Small\n               Data Sets, CRC Press, 1994.\n        .. [4] Wikipedia, \"Log""arithmic-distribution\",\n               http://en.wikipedia.org/wiki/Logarithmic-distribution\n\n        Examples\n        --------\n        Draw samples from the distribution:\n\n        >>> a = .6\n        >>> s = np.random.logseries(a, 10000)\n        >>> count, bins, ignored = plt.hist(s)\n\n        #   plot against distribution\n\n        >>> def logseries(k, p):\n        ...     return -p**k/(k*log(1-p))\n        >>> plt.plot(bins, logseries(bins, a)*count.max()/\n                     logseries(bins, a).max(), 'r')\n        >>> plt.show()\n\n        ";
     8745-static PyObject *__pyx_pf_6mtrand_11RandomState_46logseries(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     8746-  PyObject *__pyx_v_p = 0;
     8747-  PyObject *__pyx_v_size = 0;
     8748-  PyArrayObject *__pyx_v_op = 0;
     8749-  double __pyx_v_fp;
     8750-  PyObject *__pyx_r = NULL;
     8751-  __Pyx_RefNannyDeclarations
     8752-  int __pyx_t_1;
     8753-  PyObject *__pyx_t_2 = NULL;
     8754-  PyObject *__pyx_t_3 = NULL;
     8755-  PyObject *__pyx_t_4 = NULL;
     8756-  PyObject *__pyx_t_5 = NULL;
     8757-  int __pyx_lineno = 0;
     8758-  const char *__pyx_filename = NULL;
     8759-  int __pyx_clineno = 0;
     8760-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__p,&__pyx_n_s__size,0};
     8761-  __Pyx_RefNannySetupContext("logseries");
     8762-  {
     8763-    PyObject* values[2] = {0,0};
     8764     values[1] = ((PyObject *)Py_None);
     8765     if (unlikely(__pyx_kwds)) {
     8766       Py_ssize_t kw_args;
     8767-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     8768+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     8769+      switch (pos_args) {
     8770         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     8771         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     8772         case  0: break;
     8773         default: goto __pyx_L5_argtuple_error;
     8774       }
     8775       kw_args = PyDict_Size(__pyx_kwds);
     8776-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     8777+      switch (pos_args) {
     8778         case  0:
     8779-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__p);
     8780-        if (likely(values[0])) kw_args--;
     8781+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__p)) != 0)) kw_args--;
     8782         else goto __pyx_L5_argtuple_error;
     8783         case  1:
     8784         if (kw_args > 0) {
     8785@@ -16248,7 +16833,7 @@
     8786         }
     8787       }
     8788       if (unlikely(kw_args > 0)) {
     8789-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "logseries") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3788; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     8790+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "logseries") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3788; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     8791       }
     8792     } else {
     8793       switch (PyTuple_GET_SIZE(__pyx_args)) {
     8794@@ -16269,6 +16854,25 @@
     8795   __Pyx_RefNannyFinishContext();
     8796   return NULL;
     8797   __pyx_L4_argument_unpacking_done:;
     8798+  __pyx_r = __pyx_pf_6mtrand_11RandomState_92logseries(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_p, __pyx_v_size);
     8799+  __Pyx_RefNannyFinishContext();
     8800+  return __pyx_r;
     8801+}
     8802+
     8803+static PyObject *__pyx_pf_6mtrand_11RandomState_92logseries(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_p, PyObject *__pyx_v_size) {
     8804+  PyArrayObject *__pyx_v_op = 0;
     8805+  double __pyx_v_fp;
     8806+  PyObject *__pyx_r = NULL;
     8807+  __Pyx_RefNannyDeclarations
     8808+  int __pyx_t_1;
     8809+  PyObject *__pyx_t_2 = NULL;
     8810+  PyObject *__pyx_t_3 = NULL;
     8811+  PyObject *__pyx_t_4 = NULL;
     8812+  PyObject *__pyx_t_5 = NULL;
     8813+  int __pyx_lineno = 0;
     8814+  const char *__pyx_filename = NULL;
     8815+  int __pyx_clineno = 0;
     8816+  __Pyx_RefNannySetupContext("logseries", 0);
     8817 
     8818   /* "mtrand.pyx":3865
     8819  *         cdef double fp
     8820@@ -16311,9 +16915,9 @@
     8821       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8822       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8823       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3868; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8824-      goto __pyx_L7;
     8825+      goto __pyx_L4;
     8826     }
     8827-    __pyx_L7:;
     8828+    __pyx_L4:;
     8829 
     8830     /* "mtrand.pyx":3869
     8831  *             if fp <= 0.0:
     8832@@ -16337,9 +16941,9 @@
     8833       __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8834       __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8835       {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3870; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8836-      goto __pyx_L8;
     8837+      goto __pyx_L5;
     8838     }
     8839-    __pyx_L8:;
     8840+    __pyx_L5:;
     8841 
     8842     /* "mtrand.pyx":3871
     8843  *             if fp >= 1.0:
     8844@@ -16349,14 +16953,14 @@
     8845  *         PyErr_Clear()
     8846  */
     8847     __Pyx_XDECREF(__pyx_r);
     8848-    __pyx_t_2 = __pyx_f_6mtrand_discd_array_sc(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_logseries, __pyx_v_size, __pyx_v_fp); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3871; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8849+    __pyx_t_2 = __pyx_f_6mtrand_discd_array_sc(__pyx_v_self->internal_state, rk_logseries, __pyx_v_size, __pyx_v_fp); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3871; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8850     __Pyx_GOTREF(__pyx_t_2);
     8851     __pyx_r = __pyx_t_2;
     8852     __pyx_t_2 = 0;
     8853     goto __pyx_L0;
     8854-    goto __pyx_L6;
     8855+    goto __pyx_L3;
     8856   }
     8857-  __pyx_L6:;
     8858+  __pyx_L3:;
     8859 
     8860   /* "mtrand.pyx":3873
     8861  *             return discd_array_sc(self.internal_state, rk_logseries, size, fp)
     8862@@ -16400,7 +17004,7 @@
     8863   __pyx_t_2 = PyFloat_FromDouble(0.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3876; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8864   __Pyx_GOTREF(__pyx_t_2);
     8865   __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3876; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8866-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     8867+  __Pyx_GOTREF(__pyx_t_5);
     8868   __Pyx_INCREF(((PyObject *)__pyx_v_op));
     8869   PyTuple_SET_ITEM(__pyx_t_5, 0, ((PyObject *)__pyx_v_op));
     8870   __Pyx_GIVEREF(((PyObject *)__pyx_v_op));
     8871@@ -16412,7 +17016,7 @@
     8872   __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     8873   __Pyx_DECREF(((PyObject *)__pyx_t_5)); __pyx_t_5 = 0;
     8874   __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3876; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8875-  __Pyx_GOTREF(((PyObject *)__pyx_t_5));
     8876+  __Pyx_GOTREF(__pyx_t_5);
     8877   PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_2);
     8878   __Pyx_GIVEREF(__pyx_t_2);
     8879   __pyx_t_2 = 0;
     8880@@ -16436,9 +17040,9 @@
     8881     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8882     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8883     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3877; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8884-    goto __pyx_L9;
     8885+    goto __pyx_L6;
     8886   }
     8887-  __pyx_L9:;
     8888+  __pyx_L6:;
     8889 
     8890   /* "mtrand.pyx":3878
     8891  *         if np.any(np.less_equal(op, 0.0)):
     8892@@ -16460,7 +17064,7 @@
     8893   __pyx_t_2 = PyFloat_FromDouble(1.0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3878; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8894   __Pyx_GOTREF(__pyx_t_2);
     8895   __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3878; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8896-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     8897+  __Pyx_GOTREF(__pyx_t_4);
     8898   __Pyx_INCREF(((PyObject *)__pyx_v_op));
     8899   PyTuple_SET_ITEM(__pyx_t_4, 0, ((PyObject *)__pyx_v_op));
     8900   __Pyx_GIVEREF(((PyObject *)__pyx_v_op));
     8901@@ -16472,7 +17076,7 @@
     8902   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     8903   __Pyx_DECREF(((PyObject *)__pyx_t_4)); __pyx_t_4 = 0;
     8904   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3878; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8905-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     8906+  __Pyx_GOTREF(__pyx_t_4);
     8907   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     8908   __Pyx_GIVEREF(__pyx_t_2);
     8909   __pyx_t_2 = 0;
     8910@@ -16496,9 +17100,9 @@
     8911     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     8912     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     8913     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3879; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8914-    goto __pyx_L10;
     8915+    goto __pyx_L7;
     8916   }
     8917-  __pyx_L10:;
     8918+  __pyx_L7:;
     8919 
     8920   /* "mtrand.pyx":3880
     8921  *         if np.any(np.greater_equal(op, 1.0)):
     8922@@ -16508,7 +17112,7 @@
     8923  *     # Multivariate distributions:
     8924  */
     8925   __Pyx_XDECREF(__pyx_r);
     8926-  __pyx_t_2 = __pyx_f_6mtrand_discd_array(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, rk_logseries, __pyx_v_size, __pyx_v_op); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3880; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8927+  __pyx_t_2 = __pyx_f_6mtrand_discd_array(__pyx_v_self->internal_state, rk_logseries, __pyx_v_size, __pyx_v_op); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3880; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     8928   __Pyx_GOTREF(__pyx_t_2);
     8929   __pyx_r = __pyx_t_2;
     8930   __pyx_t_2 = 0;
     8931@@ -16530,51 +17134,32 @@
     8932   return __pyx_r;
     8933 }
     8934 
     8935-/* "mtrand.pyx":3883
     8936- *
     8937- *     # Multivariate distributions:
     8938- *     def multivariate_normal(self, mean, cov, size=None):             # <<<<<<<<<<<<<<
     8939- *         """
     8940- *         multivariate_normal(mean, cov[, size])
     8941- */
     8942-
     8943-static PyObject *__pyx_pf_6mtrand_11RandomState_47multivariate_normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     8944-static char __pyx_doc_6mtrand_11RandomState_47multivariate_normal[] = "\n        multivariate_normal(mean, cov[, size])\n\n        Draw random samples from a multivariate normal distribution.\n\n        The multivariate normal, multinormal or Gaussian distribution is a\n        generalization of the one-dimensional normal distribution to higher\n        dimensions.  Such a distribution is specified by its mean and\n        covariance matrix.  These parameters are analogous to the mean\n        (average or \"center\") and variance (standard deviation, or \"width,\"\n        squared) of the one-dimensional normal distribution.\n\n        Parameters\n        ----------\n        mean : 1-D array_like, of length N\n            Mean of the N-dimensional distribution.\n        cov : 2-D array_like, of shape (N, N)\n            Covariance matrix of the distribution.  Must be symmetric and\n            positive semi-definite for \"physically meaningful\" results.\n        size : tuple of ints, optional\n            Given a shape of, for example, ``(m,n,k)``, ``m*n*k`` samples are\n            generated, and packed in an `m`-by-`n`-by-`k` arrangement.  Because\n            each sample is `N`-dimensional, the output shape is ``(m,n,k,N)``.\n            If no shape is specified, a single (`N`-D) sample is returned.\n\n        Returns\n        -------\n        out : ndarray\n            The drawn samples, of shape *size*, if that was provided.  If not,\n            the shape is ``(N,)``.\n\n            In other words, each entry ``out[i,j,...,:]`` is an N-dimensional\n            value drawn from the distribution.\n\n        Notes\n        -----\n        The mean is a coordinate in N-dimensional space, which represents the\n        location where samples are most likely to be generated.  This is\n        analogous to the peak of the bell curve for the one-dimensional or\n        univariate normal distribution.\n\n        Covariance indicates the level to which two variables vary together.\n        From the multivariate normal distribution, we draw ""N-dimensional\n        samples, :math:`X = [x_1, x_2, ... x_N]`.  The covariance matrix\n        element :math:`C_{ij}` is the covariance of :math:`x_i` and :math:`x_j`.\n        The element :math:`C_{ii}` is the variance of :math:`x_i` (i.e. its\n        \"spread\").\n\n        Instead of specifying the full covariance matrix, popular\n        approximations include:\n\n          - Spherical covariance (*cov* is a multiple of the identity matrix)\n          - Diagonal covariance (*cov* has non-negative elements, and only on\n            the diagonal)\n\n        This geometrical property can be seen in two dimensions by plotting\n        generated data-points:\n\n        >>> mean = [0,0]\n        >>> cov = [[1,0],[0,100]] # diagonal covariance, points lie on x or y-axis\n\n        >>> import matplotlib.pyplot as plt\n        >>> x,y = np.random.multivariate_normal(mean,cov,5000).T\n        >>> plt.plot(x,y,'x'); plt.axis('equal'); plt.show()\n\n        Note that the covariance matrix must be non-negative definite.\n\n        References\n        ----------\n        Papoulis, A., *Probability, Random Variables, and Stochastic Processes*,\n        3rd ed., New York: McGraw-Hill, 1991.\n\n        Duda, R. O., Hart, P. E., and Stork, D. G., *Pattern Classification*,\n        2nd ed., New York: Wiley, 2001.\n\n        Examples\n        --------\n        >>> mean = (1,2)\n        >>> cov = [[1,0],[1,0]]\n        >>> x = np.random.multivariate_normal(mean,cov,(3,3))\n        >>> x.shape\n        (3, 3, 2)\n\n        The following is probably true, given that 0.6 is roughly twice the\n        standard deviation:\n\n        >>> print list( (x[0,0,:] - mean) < 0.6 )\n        [True, True]\n\n        ";
     8945-static PyObject *__pyx_pf_6mtrand_11RandomState_47multivariate_normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     8946+/* Python wrapper */
     8947+static PyObject *__pyx_pw_6mtrand_11RandomState_95multivariate_normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     8948+static char __pyx_doc_6mtrand_11RandomState_94multivariate_normal[] = "\n        multivariate_normal(mean, cov[, size])\n\n        Draw random samples from a multivariate normal distribution.\n\n        The multivariate normal, multinormal or Gaussian distribution is a\n        generalization of the one-dimensional normal distribution to higher\n        dimensions.  Such a distribution is specified by its mean and\n        covariance matrix.  These parameters are analogous to the mean\n        (average or \"center\") and variance (standard deviation, or \"width,\"\n        squared) of the one-dimensional normal distribution.\n\n        Parameters\n        ----------\n        mean : 1-D array_like, of length N\n            Mean of the N-dimensional distribution.\n        cov : 2-D array_like, of shape (N, N)\n            Covariance matrix of the distribution.  Must be symmetric and\n            positive semi-definite for \"physically meaningful\" results.\n        size : tuple of ints, optional\n            Given a shape of, for example, ``(m,n,k)``, ``m*n*k`` samples are\n            generated, and packed in an `m`-by-`n`-by-`k` arrangement.  Because\n            each sample is `N`-dimensional, the output shape is ``(m,n,k,N)``.\n            If no shape is specified, a single (`N`-D) sample is returned.\n\n        Returns\n        -------\n        out : ndarray\n            The drawn samples, of shape *size*, if that was provided.  If not,\n            the shape is ``(N,)``.\n\n            In other words, each entry ``out[i,j,...,:]`` is an N-dimensional\n            value drawn from the distribution.\n\n        Notes\n        -----\n        The mean is a coordinate in N-dimensional space, which represents the\n        location where samples are most likely to be generated.  This is\n        analogous to the peak of the bell curve for the one-dimensional or\n        univariate normal distribution.\n\n        Covariance indicates the level to which two variables vary together.\n        From the multivariate normal distribution, we draw ""N-dimensional\n        samples, :math:`X = [x_1, x_2, ... x_N]`.  The covariance matrix\n        element :math:`C_{ij}` is the covariance of :math:`x_i` and :math:`x_j`.\n        The element :math:`C_{ii}` is the variance of :math:`x_i` (i.e. its\n        \"spread\").\n\n        Instead of specifying the full covariance matrix, popular\n        approximations include:\n\n          - Spherical covariance (*cov* is a multiple of the identity matrix)\n          - Diagonal covariance (*cov* has non-negative elements, and only on\n            the diagonal)\n\n        This geometrical property can be seen in two dimensions by plotting\n        generated data-points:\n\n        >>> mean = [0,0]\n        >>> cov = [[1,0],[0,100]] # diagonal covariance, points lie on x or y-axis\n\n        >>> import matplotlib.pyplot as plt\n        >>> x,y = np.random.multivariate_normal(mean,cov,5000).T\n        >>> plt.plot(x,y,'x'); plt.axis('equal'); plt.show()\n\n        Note that the covariance matrix must be non-negative definite.\n\n        References\n        ----------\n        Papoulis, A., *Probability, Random Variables, and Stochastic Processes*,\n        3rd ed., New York: McGraw-Hill, 1991.\n\n        Duda, R. O., Hart, P. E., and Stork, D. G., *Pattern Classification*,\n        2nd ed., New York: Wiley, 2001.\n\n        Examples\n        --------\n        >>> mean = (1,2)\n        >>> cov = [[1,0],[1,0]]\n        >>> x = np.random.multivariate_normal(mean,cov,(3,3))\n        >>> x.shape\n        (3, 3, 2)\n\n        The following is probably true, given that 0.6 is roughly twice the\n        standard deviation:\n\n        >>> print list( (x[0,0,:] - mean) < 0.6 )\n        [True, True]\n\n        ";
     8949+static PyObject *__pyx_pw_6mtrand_11RandomState_95multivariate_normal(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     8950   PyObject *__pyx_v_mean = 0;
     8951   PyObject *__pyx_v_cov = 0;
     8952   PyObject *__pyx_v_size = 0;
     8953-  PyObject *__pyx_v_shape = NULL;
     8954-  PyObject *__pyx_v_final_shape = NULL;
     8955-  PyObject *__pyx_v_x = NULL;
     8956-  PyObject *__pyx_v_svd = NULL;
     8957-  PyObject *__pyx_v_u = NULL;
     8958-  PyObject *__pyx_v_s = NULL;
     8959-  PyObject *__pyx_v_v = NULL;
     8960-  PyObject *__pyx_r = NULL;
     8961+  PyObject *__pyx_r = 0;
     8962   __Pyx_RefNannyDeclarations
     8963-  PyObject *__pyx_t_1 = NULL;
     8964-  PyObject *__pyx_t_2 = NULL;
     8965-  PyObject *__pyx_t_3 = NULL;
     8966-  int __pyx_t_4;
     8967-  Py_ssize_t __pyx_t_5;
     8968-  int __pyx_t_6;
     8969-  int __pyx_t_7;
     8970-  int __pyx_t_8;
     8971-  PyObject *__pyx_t_9 = NULL;
     8972-  PyObject *__pyx_t_10 = NULL;
     8973-  PyObject *(*__pyx_t_11)(PyObject *);
     8974-  int __pyx_lineno = 0;
     8975-  const char *__pyx_filename = NULL;
     8976-  int __pyx_clineno = 0;
     8977-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__mean,&__pyx_n_s__cov,&__pyx_n_s__size,0};
     8978-  __Pyx_RefNannySetupContext("multivariate_normal");
     8979+  __Pyx_RefNannySetupContext("multivariate_normal (wrapper)", 0);
     8980   {
     8981+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__mean,&__pyx_n_s__cov,&__pyx_n_s__size,0};
     8982     PyObject* values[3] = {0,0,0};
     8983+
     8984+    /* "mtrand.pyx":3883
     8985+ *
     8986+ *     # Multivariate distributions:
     8987+ *     def multivariate_normal(self, mean, cov, size=None):             # <<<<<<<<<<<<<<
     8988+ *         """
     8989+ *         multivariate_normal(mean, cov[, size])
     8990+ */
     8991     values[2] = ((PyObject *)Py_None);
     8992     if (unlikely(__pyx_kwds)) {
     8993       Py_ssize_t kw_args;
     8994-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     8995+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     8996+      switch (pos_args) {
     8997         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     8998         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     8999         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     9000@@ -16582,14 +17167,12 @@
     9001         default: goto __pyx_L5_argtuple_error;
     9002       }
     9003       kw_args = PyDict_Size(__pyx_kwds);
     9004-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     9005+      switch (pos_args) {
     9006         case  0:
     9007-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__mean);
     9008-        if (likely(values[0])) kw_args--;
     9009+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__mean)) != 0)) kw_args--;
     9010         else goto __pyx_L5_argtuple_error;
     9011         case  1:
     9012-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__cov);
     9013-        if (likely(values[1])) kw_args--;
     9014+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__cov)) != 0)) kw_args--;
     9015         else {
     9016           __Pyx_RaiseArgtupleInvalid("multivariate_normal", 0, 2, 3, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3883; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     9017         }
     9018@@ -16600,7 +17183,7 @@
     9019         }
     9020       }
     9021       if (unlikely(kw_args > 0)) {
     9022-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "multivariate_normal") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3883; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     9023+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "multivariate_normal") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3883; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     9024       }
     9025     } else {
     9026       switch (PyTuple_GET_SIZE(__pyx_args)) {
     9027@@ -16623,6 +17206,36 @@
     9028   __Pyx_RefNannyFinishContext();
     9029   return NULL;
     9030   __pyx_L4_argument_unpacking_done:;
     9031+  __pyx_r = __pyx_pf_6mtrand_11RandomState_94multivariate_normal(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_mean, __pyx_v_cov, __pyx_v_size);
     9032+  __Pyx_RefNannyFinishContext();
     9033+  return __pyx_r;
     9034+}
     9035+
     9036+static PyObject *__pyx_pf_6mtrand_11RandomState_94multivariate_normal(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_mean, PyObject *__pyx_v_cov, PyObject *__pyx_v_size) {
     9037+  PyObject *__pyx_v_shape = NULL;
     9038+  PyObject *__pyx_v_final_shape = NULL;
     9039+  PyObject *__pyx_v_x = NULL;
     9040+  PyObject *__pyx_v_svd = NULL;
     9041+  CYTHON_UNUSED PyObject *__pyx_v_u = NULL;
     9042+  PyObject *__pyx_v_s = NULL;
     9043+  PyObject *__pyx_v_v = NULL;
     9044+  PyObject *__pyx_r = NULL;
     9045+  __Pyx_RefNannyDeclarations
     9046+  PyObject *__pyx_t_1 = NULL;
     9047+  PyObject *__pyx_t_2 = NULL;
     9048+  PyObject *__pyx_t_3 = NULL;
     9049+  int __pyx_t_4;
     9050+  Py_ssize_t __pyx_t_5;
     9051+  int __pyx_t_6;
     9052+  int __pyx_t_7;
     9053+  int __pyx_t_8;
     9054+  PyObject *__pyx_t_9 = NULL;
     9055+  PyObject *__pyx_t_10 = NULL;
     9056+  PyObject *(*__pyx_t_11)(PyObject *);
     9057+  int __pyx_lineno = 0;
     9058+  const char *__pyx_filename = NULL;
     9059+  int __pyx_clineno = 0;
     9060+  __Pyx_RefNannySetupContext("multivariate_normal", 0);
     9061   __Pyx_INCREF(__pyx_v_mean);
     9062   __Pyx_INCREF(__pyx_v_cov);
     9063 
     9064@@ -16639,7 +17252,7 @@
     9065   __Pyx_GOTREF(__pyx_t_2);
     9066   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     9067   __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3975; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9068-  __Pyx_GOTREF(((PyObject *)__pyx_t_1));
     9069+  __Pyx_GOTREF(__pyx_t_1);
     9070   __Pyx_INCREF(__pyx_v_mean);
     9071   PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_v_mean);
     9072   __Pyx_GIVEREF(__pyx_v_mean);
     9073@@ -16664,7 +17277,7 @@
     9074   __Pyx_GOTREF(__pyx_t_1);
     9075   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     9076   __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3976; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9077-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     9078+  __Pyx_GOTREF(__pyx_t_3);
     9079   __Pyx_INCREF(__pyx_v_cov);
     9080   PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_v_cov);
     9081   __Pyx_GIVEREF(__pyx_v_cov);
     9082@@ -16694,10 +17307,10 @@
     9083  *             shape = size
     9084  */
     9085     __pyx_t_2 = PyList_New(0); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3978; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9086-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     9087+    __Pyx_GOTREF(__pyx_t_2);
     9088     __pyx_v_shape = ((PyObject *)__pyx_t_2);
     9089     __pyx_t_2 = 0;
     9090-    goto __pyx_L6;
     9091+    goto __pyx_L3;
     9092   }
     9093   /*else*/ {
     9094 
     9095@@ -16711,7 +17324,7 @@
     9096     __Pyx_INCREF(__pyx_v_size);
     9097     __pyx_v_shape = __pyx_v_size;
     9098   }
     9099-  __pyx_L6:;
     9100+  __pyx_L3:;
     9101 
     9102   /* "mtrand.pyx":3981
     9103  *         else:
     9104@@ -16739,9 +17352,9 @@
     9105     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     9106     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     9107     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3982; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9108-    goto __pyx_L7;
     9109+    goto __pyx_L4;
     9110   }
     9111-  __pyx_L7:;
     9112+  __pyx_L4:;
     9113 
     9114   /* "mtrand.pyx":3983
     9115  *         if len(mean.shape) != 1:
     9116@@ -16766,8 +17379,7 @@
     9117     __pyx_t_1 = __Pyx_GetItemInt(__pyx_t_2, 1, sizeof(long), PyInt_FromLong); if (!__pyx_t_1) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3983; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9118     __Pyx_GOTREF(__pyx_t_1);
     9119     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     9120-    __pyx_t_2 = PyObject_RichCompare(__pyx_t_3, __pyx_t_1, Py_NE); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3983; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9121-    __Pyx_GOTREF(__pyx_t_2);
     9122+    __pyx_t_2 = PyObject_RichCompare(__pyx_t_3, __pyx_t_1, Py_NE); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3983; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9123     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     9124     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     9125     __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_6 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3983; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9126@@ -16790,9 +17402,9 @@
     9127     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     9128     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     9129     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3984; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9130-    goto __pyx_L8;
     9131+    goto __pyx_L5;
     9132   }
     9133-  __pyx_L8:;
     9134+  __pyx_L5:;
     9135 
     9136   /* "mtrand.pyx":3985
     9137  *         if (len(cov.shape) != 2) or (cov.shape[0] != cov.shape[1]):
     9138@@ -16811,8 +17423,7 @@
     9139   __pyx_t_3 = __Pyx_GetItemInt(__pyx_t_2, 0, sizeof(long), PyInt_FromLong); if (!__pyx_t_3) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3985; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9140   __Pyx_GOTREF(__pyx_t_3);
     9141   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     9142-  __pyx_t_2 = PyObject_RichCompare(__pyx_t_1, __pyx_t_3, Py_NE); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3985; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9143-  __Pyx_GOTREF(__pyx_t_2);
     9144+  __pyx_t_2 = PyObject_RichCompare(__pyx_t_1, __pyx_t_3, Py_NE); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3985; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9145   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     9146   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     9147   __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_7 < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3985; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9148@@ -16831,9 +17442,9 @@
     9149     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     9150     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     9151     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3986; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9152-    goto __pyx_L9;
     9153+    goto __pyx_L6;
     9154   }
     9155-  __pyx_L9:;
     9156+  __pyx_L6:;
     9157 
     9158   /* "mtrand.pyx":3988
     9159  *                raise ValueError("mean and cov must have same length")
     9160@@ -16856,16 +17467,16 @@
     9161  *         final_shape.append(mean.shape[0])
     9162  */
     9163     __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3989; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9164-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     9165+    __Pyx_GOTREF(__pyx_t_2);
     9166     __Pyx_INCREF(__pyx_v_shape);
     9167     PyList_SET_ITEM(__pyx_t_2, 0, __pyx_v_shape);
     9168     __Pyx_GIVEREF(__pyx_v_shape);
     9169     __Pyx_DECREF(__pyx_v_shape);
     9170     __pyx_v_shape = ((PyObject *)__pyx_t_2);
     9171     __pyx_t_2 = 0;
     9172-    goto __pyx_L10;
     9173+    goto __pyx_L7;
     9174   }
     9175-  __pyx_L10:;
     9176+  __pyx_L7:;
     9177 
     9178   /* "mtrand.pyx":3990
     9179  *         if isinstance(shape, int):
     9180@@ -16877,7 +17488,7 @@
     9181   __pyx_t_2 = __Pyx_PySequence_GetSlice(__pyx_v_shape, 0, PY_SSIZE_T_MAX); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3990; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9182   __Pyx_GOTREF(__pyx_t_2);
     9183   __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3990; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9184-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     9185+  __Pyx_GOTREF(__pyx_t_3);
     9186   PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_2);
     9187   __Pyx_GIVEREF(__pyx_t_2);
     9188   __pyx_t_2 = 0;
     9189@@ -16894,9 +17505,6 @@
     9190  *         # Create a matrix of independent standard normally distributed random
     9191  *         # numbers. The matrix has rows with the same length as mean and as
     9192  */
     9193-  if (unlikely(((PyObject *)__pyx_v_final_shape) == Py_None)) {
     9194-    PyErr_Format(PyExc_AttributeError, "'NoneType' object has no attribute '%s'", "append"); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3991; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9195-  }
     9196   __pyx_t_2 = PyObject_GetAttr(__pyx_v_mean, __pyx_n_s__shape); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3991; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9197   __Pyx_GOTREF(__pyx_t_2);
     9198   __pyx_t_3 = __Pyx_GetItemInt(__pyx_t_2, 0, sizeof(long), PyInt_FromLong); if (!__pyx_t_3) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3991; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9199@@ -16912,7 +17520,7 @@
     9200  *         x.shape = (np.multiply.reduce(final_shape[0:len(final_shape)-1]),
     9201  *                    mean.shape[0])
     9202  */
     9203-  __pyx_t_3 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__standard_normal); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3995; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9204+  __pyx_t_3 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__standard_normal); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3995; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9205   __Pyx_GOTREF(__pyx_t_3);
     9206   __pyx_t_2 = __Pyx_GetName(__pyx_m, __pyx_n_s__np); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3995; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9207   __Pyx_GOTREF(__pyx_t_2);
     9208@@ -16923,7 +17531,7 @@
     9209   __Pyx_GOTREF(__pyx_t_2);
     9210   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     9211   __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3995; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9212-  __Pyx_GOTREF(((PyObject *)__pyx_t_1));
     9213+  __Pyx_GOTREF(__pyx_t_1);
     9214   __Pyx_INCREF(((PyObject *)__pyx_v_final_shape));
     9215   PyTuple_SET_ITEM(__pyx_t_1, 0, ((PyObject *)__pyx_v_final_shape));
     9216   __Pyx_GIVEREF(((PyObject *)__pyx_v_final_shape));
     9217@@ -16932,7 +17540,7 @@
     9218   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     9219   __Pyx_DECREF(((PyObject *)__pyx_t_1)); __pyx_t_1 = 0;
     9220   __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3995; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9221-  __Pyx_GOTREF(((PyObject *)__pyx_t_1));
     9222+  __Pyx_GOTREF(__pyx_t_1);
     9223   PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_t_9);
     9224   __Pyx_GIVEREF(__pyx_t_9);
     9225   __pyx_t_9 = 0;
     9226@@ -16958,14 +17566,11 @@
     9227   __pyx_t_9 = PyObject_GetAttr(__pyx_t_1, __pyx_n_s__reduce); if (unlikely(!__pyx_t_9)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3996; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9228   __Pyx_GOTREF(__pyx_t_9);
     9229   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     9230-  if (unlikely(((PyObject *)__pyx_v_final_shape) == Py_None)) {
     9231-    PyErr_SetString(PyExc_TypeError, "object of type 'NoneType' has no len()"); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3996; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9232-  }
     9233-  __pyx_t_5 = PyList_GET_SIZE(((PyObject *)__pyx_v_final_shape));
     9234+  __pyx_t_5 = PyList_GET_SIZE(((PyObject *)__pyx_v_final_shape)); if (unlikely(__pyx_t_5 == -1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3996; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9235   __pyx_t_1 = __Pyx_PySequence_GetSlice(((PyObject *)__pyx_v_final_shape), 0, (__pyx_t_5 - 1)); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3996; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9236   __Pyx_GOTREF(((PyObject *)__pyx_t_1));
     9237   __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3996; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9238-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     9239+  __Pyx_GOTREF(__pyx_t_3);
     9240   PyTuple_SET_ITEM(__pyx_t_3, 0, ((PyObject *)__pyx_t_1));
     9241   __Pyx_GIVEREF(((PyObject *)__pyx_t_1));
     9242   __pyx_t_1 = 0;
     9243@@ -16987,7 +17592,7 @@
     9244   __Pyx_GOTREF(__pyx_t_9);
     9245   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     9246   __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3996; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9247-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     9248+  __Pyx_GOTREF(__pyx_t_3);
     9249   PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_1);
     9250   __Pyx_GIVEREF(__pyx_t_1);
     9251   PyTuple_SET_ITEM(__pyx_t_3, 1, __pyx_t_9);
     9252@@ -17013,14 +17618,18 @@
     9253  *         (u,s,v) = svd(cov)
     9254  */
     9255   __pyx_t_3 = PyList_New(1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4005; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9256-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     9257+  __Pyx_GOTREF(__pyx_t_3);
     9258   __Pyx_INCREF(((PyObject *)__pyx_n_s__svd));
     9259   PyList_SET_ITEM(__pyx_t_3, 0, ((PyObject *)__pyx_n_s__svd));
     9260   __Pyx_GIVEREF(((PyObject *)__pyx_n_s__svd));
     9261   __pyx_t_9 = __Pyx_Import(((PyObject *)__pyx_n_s_166), ((PyObject *)__pyx_t_3), -1); if (unlikely(!__pyx_t_9)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4005; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9262   __Pyx_GOTREF(__pyx_t_9);
     9263   __Pyx_DECREF(((PyObject *)__pyx_t_3)); __pyx_t_3 = 0;
     9264-  __pyx_t_3 = PyObject_GetAttr(__pyx_t_9, __pyx_n_s__svd); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4005; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9265+  __pyx_t_3 = PyObject_GetAttr(__pyx_t_9, __pyx_n_s__svd);
     9266+  if (__pyx_t_3 == NULL) {
     9267+    if (PyErr_ExceptionMatches(PyExc_AttributeError)) __Pyx_RaiseImportError(__pyx_n_s__svd);
     9268+    if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4005; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9269+  }
     9270   __Pyx_GOTREF(__pyx_t_3);
     9271   __Pyx_INCREF(__pyx_t_3);
     9272   __pyx_v_svd = __pyx_t_3;
     9273@@ -17035,7 +17644,7 @@
     9274  *         # The rows of x now have the correct covariance but mean 0. Add
     9275  */
     9276   __pyx_t_9 = PyTuple_New(1); if (unlikely(!__pyx_t_9)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4007; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9277-  __Pyx_GOTREF(((PyObject *)__pyx_t_9));
     9278+  __Pyx_GOTREF(__pyx_t_9);
     9279   __Pyx_INCREF(__pyx_v_cov);
     9280   PyTuple_SET_ITEM(__pyx_t_9, 0, __pyx_v_cov);
     9281   __Pyx_GIVEREF(__pyx_v_cov);
     9282@@ -17044,21 +17653,22 @@
     9283   __Pyx_DECREF(((PyObject *)__pyx_t_9)); __pyx_t_9 = 0;
     9284   if ((likely(PyTuple_CheckExact(__pyx_t_3))) || (PyList_CheckExact(__pyx_t_3))) {
     9285     PyObject* sequence = __pyx_t_3;
     9286+    #if CYTHON_COMPILING_IN_CPYTHON
     9287+    Py_ssize_t size = Py_SIZE(sequence);
     9288+    #else
     9289+    Py_ssize_t size = PySequence_Size(sequence);
     9290+    #endif
     9291+    if (unlikely(size != 3)) {
     9292+      if (size > 3) __Pyx_RaiseTooManyValuesError(3);
     9293+      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
     9294+      {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4007; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9295+    }
     9296+    #if CYTHON_COMPILING_IN_CPYTHON
     9297     if (likely(PyTuple_CheckExact(sequence))) {
     9298-      if (unlikely(PyTuple_GET_SIZE(sequence) != 3)) {
     9299-        if (PyTuple_GET_SIZE(sequence) > 3) __Pyx_RaiseTooManyValuesError(3);
     9300-        else __Pyx_RaiseNeedMoreValuesError(PyTuple_GET_SIZE(sequence));
     9301-        {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4007; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9302-      }
     9303       __pyx_t_9 = PyTuple_GET_ITEM(sequence, 0);
     9304       __pyx_t_1 = PyTuple_GET_ITEM(sequence, 1);
     9305       __pyx_t_2 = PyTuple_GET_ITEM(sequence, 2);
     9306     } else {
     9307-      if (unlikely(PyList_GET_SIZE(sequence) != 3)) {
     9308-        if (PyList_GET_SIZE(sequence) > 3) __Pyx_RaiseTooManyValuesError(3);
     9309-        else __Pyx_RaiseNeedMoreValuesError(PyList_GET_SIZE(sequence));
     9310-        {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4007; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9311-      }
     9312       __pyx_t_9 = PyList_GET_ITEM(sequence, 0);
     9313       __pyx_t_1 = PyList_GET_ITEM(sequence, 1);
     9314       __pyx_t_2 = PyList_GET_ITEM(sequence, 2);
     9315@@ -17066,28 +17676,35 @@
     9316     __Pyx_INCREF(__pyx_t_9);
     9317     __Pyx_INCREF(__pyx_t_1);
     9318     __Pyx_INCREF(__pyx_t_2);
     9319+    #else
     9320+    __pyx_t_9 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_9)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4007; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9321+    __pyx_t_1 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4007; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9322+    __pyx_t_2 = PySequence_ITEM(sequence, 2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4007; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9323+    #endif
     9324     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     9325-  } else {
     9326+  } else
     9327+  {
     9328     Py_ssize_t index = -1;
     9329     __pyx_t_10 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_10)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4007; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9330     __Pyx_GOTREF(__pyx_t_10);
     9331     __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     9332     __pyx_t_11 = Py_TYPE(__pyx_t_10)->tp_iternext;
     9333-    index = 0; __pyx_t_9 = __pyx_t_11(__pyx_t_10); if (unlikely(!__pyx_t_9)) goto __pyx_L11_unpacking_failed;
     9334+    index = 0; __pyx_t_9 = __pyx_t_11(__pyx_t_10); if (unlikely(!__pyx_t_9)) goto __pyx_L8_unpacking_failed;
     9335     __Pyx_GOTREF(__pyx_t_9);
     9336-    index = 1; __pyx_t_1 = __pyx_t_11(__pyx_t_10); if (unlikely(!__pyx_t_1)) goto __pyx_L11_unpacking_failed;
     9337+    index = 1; __pyx_t_1 = __pyx_t_11(__pyx_t_10); if (unlikely(!__pyx_t_1)) goto __pyx_L8_unpacking_failed;
     9338     __Pyx_GOTREF(__pyx_t_1);
     9339-    index = 2; __pyx_t_2 = __pyx_t_11(__pyx_t_10); if (unlikely(!__pyx_t_2)) goto __pyx_L11_unpacking_failed;
     9340+    index = 2; __pyx_t_2 = __pyx_t_11(__pyx_t_10); if (unlikely(!__pyx_t_2)) goto __pyx_L8_unpacking_failed;
     9341     __Pyx_GOTREF(__pyx_t_2);
     9342     if (__Pyx_IternextUnpackEndCheck(__pyx_t_11(__pyx_t_10), 3) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4007; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9343+    __pyx_t_11 = NULL;
     9344     __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
     9345-    goto __pyx_L12_unpacking_done;
     9346-    __pyx_L11_unpacking_failed:;
     9347+    goto __pyx_L9_unpacking_done;
     9348+    __pyx_L8_unpacking_failed:;
     9349     __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
     9350-    if (PyErr_Occurred() && PyErr_ExceptionMatches(PyExc_StopIteration)) PyErr_Clear();
     9351-    if (!PyErr_Occurred()) __Pyx_RaiseNeedMoreValuesError(index);
     9352+    __pyx_t_11 = NULL;
     9353+    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
     9354     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4007; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9355-    __pyx_L12_unpacking_done:;
     9356+    __pyx_L9_unpacking_done:;
     9357   }
     9358   __pyx_v_u = __pyx_t_9;
     9359   __pyx_t_9 = 0;
     9360@@ -17114,7 +17731,7 @@
     9361   __Pyx_GOTREF(__pyx_t_1);
     9362   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     9363   __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4008; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9364-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     9365+  __Pyx_GOTREF(__pyx_t_3);
     9366   __Pyx_INCREF(__pyx_v_s);
     9367   PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_v_s);
     9368   __Pyx_GIVEREF(__pyx_v_s);
     9369@@ -17126,7 +17743,7 @@
     9370   __Pyx_GOTREF(__pyx_t_3);
     9371   __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
     9372   __pyx_t_9 = PyTuple_New(2); if (unlikely(!__pyx_t_9)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4008; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9373-  __Pyx_GOTREF(((PyObject *)__pyx_t_9));
     9374+  __Pyx_GOTREF(__pyx_t_9);
     9375   PyTuple_SET_ITEM(__pyx_t_9, 0, __pyx_t_3);
     9376   __Pyx_GIVEREF(__pyx_t_3);
     9377   __Pyx_INCREF(__pyx_v_v);
     9378@@ -17154,7 +17771,7 @@
     9379   __Pyx_GOTREF(__pyx_t_9);
     9380   __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
     9381   __pyx_t_3 = PyTuple_New(3); if (unlikely(!__pyx_t_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4011; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9382-  __Pyx_GOTREF(((PyObject *)__pyx_t_3));
     9383+  __Pyx_GOTREF(__pyx_t_3);
     9384   __Pyx_INCREF(__pyx_v_mean);
     9385   PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_v_mean);
     9386   __Pyx_GIVEREF(__pyx_v_mean);
     9387@@ -17177,9 +17794,6 @@
     9388  *         return x
     9389  *
     9390  */
     9391-  if (unlikely(((PyObject *)__pyx_v_final_shape) == Py_None)) {
     9392-    PyErr_SetString(PyExc_TypeError, "'NoneType' object is not iterable"); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4012; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9393-  }
     9394   __pyx_t_2 = ((PyObject *)PyList_AsTuple(__pyx_v_final_shape)); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4012; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9395   __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     9396   if (PyObject_SetAttr(__pyx_v_x, __pyx_n_s__shape, ((PyObject *)__pyx_t_2)) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4012; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9397@@ -17222,50 +17836,32 @@
     9398   return __pyx_r;
     9399 }
     9400 
     9401-/* "mtrand.pyx":4015
     9402+/* Python wrapper */
     9403+static PyObject *__pyx_pw_6mtrand_11RandomState_97multinomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     9404+static char __pyx_doc_6mtrand_11RandomState_96multinomial[] = "\n        multinomial(n, pvals, size=None)\n\n        Draw samples from a multinomial distribution.\n\n        The multinomial distribution is a multivariate generalisation of the\n        binomial distribution.  Take an experiment with one of ``p``\n        possible outcomes.  An example of such an experiment is throwing a dice,\n        where the outcome can be 1 through 6.  Each sample drawn from the\n        distribution represents `n` such experiments.  Its values,\n        ``X_i = [X_0, X_1, ..., X_p]``, represent the number of times the outcome\n        was ``i``.\n\n        Parameters\n        ----------\n        n : int\n            Number of experiments.\n        pvals : sequence of floats, length p\n            Probabilities of each of the ``p`` different outcomes.  These\n            should sum to 1 (however, the last element is always assumed to\n            account for the remaining probability, as long as\n            ``sum(pvals[:-1]) <= 1)``.\n        size : tuple of ints\n            Given a `size` of ``(M, N, K)``, then ``M*N*K`` samples are drawn,\n            and the output shape becomes ``(M, N, K, p)``, since each sample\n            has shape ``(p,)``.\n\n        Examples\n        --------\n        Throw a dice 20 times:\n\n        >>> np.random.multinomial(20, [1/6.]*6, size=1)\n        array([[4, 1, 7, 5, 2, 1]])\n\n        It landed 4 times on 1, once on 2, etc.\n\n        Now, throw the dice 20 times, and 20 times again:\n\n        >>> np.random.multinomial(20, [1/6.]*6, size=2)\n        array([[3, 4, 3, 3, 4, 3],\n               [2, 4, 3, 4, 0, 7]])\n\n        For the first run, we threw 3 times 1, 4 times 2, etc.  For the second,\n        we threw 2 times 1, 4 times 2, etc.\n\n        A loaded dice is more likely to land on number 6:\n\n        >>> np.random.multinomial(100, [1/7.]*5)\n        array([13, 16, 13, 16, 42])\n\n        ";
     9405+static PyObject *__pyx_pw_6mtrand_11RandomState_97multinomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     9406+  npy_intp __pyx_v_n;
     9407+  PyObject *__pyx_v_pvals = 0;
     9408+  PyObject *__pyx_v_size = 0;
     9409+  PyObject *__pyx_r = 0;
     9410+  __Pyx_RefNannyDeclarations
     9411+  __Pyx_RefNannySetupContext("multinomial (wrapper)", 0);
     9412+  {
     9413+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__n,&__pyx_n_s__pvals,&__pyx_n_s__size,0};
     9414+    PyObject* values[3] = {0,0,0};
     9415+
     9416+    /* "mtrand.pyx":4015
     9417  *         return x
     9418  *
     9419  *     def multinomial(self, npy_intp n, object pvals, size=None):             # <<<<<<<<<<<<<<
     9420  *         """
     9421  *         multinomial(n, pvals, size=None)
     9422  */
     9423-
     9424-static PyObject *__pyx_pf_6mtrand_11RandomState_48multinomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     9425-static char __pyx_doc_6mtrand_11RandomState_48multinomial[] = "\n        multinomial(n, pvals, size=None)\n\n        Draw samples from a multinomial distribution.\n\n        The multinomial distribution is a multivariate generalisation of the\n        binomial distribution.  Take an experiment with one of ``p``\n        possible outcomes.  An example of such an experiment is throwing a dice,\n        where the outcome can be 1 through 6.  Each sample drawn from the\n        distribution represents `n` such experiments.  Its values,\n        ``X_i = [X_0, X_1, ..., X_p]``, represent the number of times the outcome\n        was ``i``.\n\n        Parameters\n        ----------\n        n : int\n            Number of experiments.\n        pvals : sequence of floats, length p\n            Probabilities of each of the ``p`` different outcomes.  These\n            should sum to 1 (however, the last element is always assumed to\n            account for the remaining probability, as long as\n            ``sum(pvals[:-1]) <= 1)``.\n        size : tuple of ints\n            Given a `size` of ``(M, N, K)``, then ``M*N*K`` samples are drawn,\n            and the output shape becomes ``(M, N, K, p)``, since each sample\n            has shape ``(p,)``.\n\n        Examples\n        --------\n        Throw a dice 20 times:\n\n        >>> np.random.multinomial(20, [1/6.]*6, size=1)\n        array([[4, 1, 7, 5, 2, 1]])\n\n        It landed 4 times on 1, once on 2, etc.\n\n        Now, throw the dice 20 times, and 20 times again:\n\n        >>> np.random.multinomial(20, [1/6.]*6, size=2)\n        array([[3, 4, 3, 3, 4, 3],\n               [2, 4, 3, 4, 0, 7]])\n\n        For the first run, we threw 3 times 1, 4 times 2, etc.  For the second,\n        we threw 2 times 1, 4 times 2, etc.\n\n        A loaded dice is more likely to land on number 6:\n\n        >>> np.random.multinomial(100, [1/7.]*5)\n        array([13, 16, 13, 16, 42])\n\n        ";
     9426-static PyObject *__pyx_pf_6mtrand_11RandomState_48multinomial(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     9427-  npy_intp __pyx_v_n;
     9428-  PyObject *__pyx_v_pvals = 0;
     9429-  PyObject *__pyx_v_size = 0;
     9430-  npy_intp __pyx_v_d;
     9431-  PyArrayObject *arrayObject_parr = 0;
     9432-  PyArrayObject *arrayObject_mnarr = 0;
     9433-  double *__pyx_v_pix;
     9434-  long *__pyx_v_mnix;
     9435-  npy_intp __pyx_v_i;
     9436-  npy_intp __pyx_v_j;
     9437-  npy_intp __pyx_v_dn;
     9438-  double __pyx_v_Sum;
     9439-  PyObject *__pyx_v_shape = NULL;
     9440-  PyObject *__pyx_v_multin = NULL;
     9441-  PyObject *__pyx_r = NULL;
     9442-  __Pyx_RefNannyDeclarations
     9443-  Py_ssize_t __pyx_t_1;
     9444-  PyObject *__pyx_t_2 = NULL;
     9445-  int __pyx_t_3;
     9446-  PyObject *__pyx_t_4 = NULL;
     9447-  PyObject *__pyx_t_5 = NULL;
     9448-  long __pyx_t_6;
     9449-  int __pyx_lineno = 0;
     9450-  const char *__pyx_filename = NULL;
     9451-  int __pyx_clineno = 0;
     9452-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__n,&__pyx_n_s__pvals,&__pyx_n_s__size,0};
     9453-  __Pyx_RefNannySetupContext("multinomial");
     9454-  {
     9455-    PyObject* values[3] = {0,0,0};
     9456     values[2] = ((PyObject *)Py_None);
     9457     if (unlikely(__pyx_kwds)) {
     9458       Py_ssize_t kw_args;
     9459-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     9460+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     9461+      switch (pos_args) {
     9462         case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
     9463         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     9464         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     9465@@ -17273,14 +17869,12 @@
     9466         default: goto __pyx_L5_argtuple_error;
     9467       }
     9468       kw_args = PyDict_Size(__pyx_kwds);
     9469-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     9470+      switch (pos_args) {
     9471         case  0:
     9472-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__n);
     9473-        if (likely(values[0])) kw_args--;
     9474+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__n)) != 0)) kw_args--;
     9475         else goto __pyx_L5_argtuple_error;
     9476         case  1:
     9477-        values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__pvals);
     9478-        if (likely(values[1])) kw_args--;
     9479+        if (likely((values[1] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__pvals)) != 0)) kw_args--;
     9480         else {
     9481           __Pyx_RaiseArgtupleInvalid("multinomial", 0, 2, 3, 1); {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4015; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     9482         }
     9483@@ -17291,7 +17885,7 @@
     9484         }
     9485       }
     9486       if (unlikely(kw_args > 0)) {
     9487-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "multinomial") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4015; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     9488+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "multinomial") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4015; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     9489       }
     9490     } else {
     9491       switch (PyTuple_GET_SIZE(__pyx_args)) {
     9492@@ -17314,6 +17908,35 @@
     9493   __Pyx_RefNannyFinishContext();
     9494   return NULL;
     9495   __pyx_L4_argument_unpacking_done:;
     9496+  __pyx_r = __pyx_pf_6mtrand_11RandomState_96multinomial(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_n, __pyx_v_pvals, __pyx_v_size);
     9497+  __Pyx_RefNannyFinishContext();
     9498+  return __pyx_r;
     9499+}
     9500+
     9501+static PyObject *__pyx_pf_6mtrand_11RandomState_96multinomial(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, npy_intp __pyx_v_n, PyObject *__pyx_v_pvals, PyObject *__pyx_v_size) {
     9502+  npy_intp __pyx_v_d;
     9503+  PyArrayObject *arrayObject_parr = 0;
     9504+  PyArrayObject *arrayObject_mnarr = 0;
     9505+  double *__pyx_v_pix;
     9506+  long *__pyx_v_mnix;
     9507+  npy_intp __pyx_v_i;
     9508+  npy_intp __pyx_v_j;
     9509+  npy_intp __pyx_v_dn;
     9510+  double __pyx_v_Sum;
     9511+  PyObject *__pyx_v_shape = NULL;
     9512+  PyObject *__pyx_v_multin = NULL;
     9513+  PyObject *__pyx_r = NULL;
     9514+  __Pyx_RefNannyDeclarations
     9515+  Py_ssize_t __pyx_t_1;
     9516+  PyObject *__pyx_t_2 = NULL;
     9517+  int __pyx_t_3;
     9518+  PyObject *__pyx_t_4 = NULL;
     9519+  PyObject *__pyx_t_5 = NULL;
     9520+  long __pyx_t_6;
     9521+  int __pyx_lineno = 0;
     9522+  const char *__pyx_filename = NULL;
     9523+  int __pyx_clineno = 0;
     9524+  __Pyx_RefNannySetupContext("multinomial", 0);
     9525 
     9526   /* "mtrand.pyx":4074
     9527  *         cdef double Sum
     9528@@ -17369,9 +17992,9 @@
     9529     __Pyx_Raise(__pyx_t_2, 0, 0, 0);
     9530     __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     9531     {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4079; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9532-    goto __pyx_L6;
     9533+    goto __pyx_L3;
     9534   }
     9535-  __pyx_L6:;
     9536+  __pyx_L3:;
     9537 
     9538   /* "mtrand.pyx":4081
     9539  *             raise ValueError("sum(pvals[:-1]) > 1.0")
     9540@@ -17393,13 +18016,13 @@
     9541     __pyx_t_2 = __Pyx_PyInt_to_py_npy_intp(__pyx_v_d); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4082; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9542     __Pyx_GOTREF(__pyx_t_2);
     9543     __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4082; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9544-    __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     9545+    __Pyx_GOTREF(__pyx_t_4);
     9546     PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     9547     __Pyx_GIVEREF(__pyx_t_2);
     9548     __pyx_t_2 = 0;
     9549     __pyx_v_shape = ((PyObject *)__pyx_t_4);
     9550     __pyx_t_4 = 0;
     9551-    goto __pyx_L7;
     9552+    goto __pyx_L4;
     9553   }
     9554 
     9555   /* "mtrand.pyx":4083
     9556@@ -17422,7 +18045,7 @@
     9557     __pyx_t_4 = __Pyx_PyInt_to_py_npy_intp(__pyx_v_d); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4084; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9558     __Pyx_GOTREF(__pyx_t_4);
     9559     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4084; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9560-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     9561+    __Pyx_GOTREF(__pyx_t_2);
     9562     __Pyx_INCREF(__pyx_v_size);
     9563     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     9564     __Pyx_GIVEREF(__pyx_v_size);
     9565@@ -17431,7 +18054,7 @@
     9566     __pyx_t_4 = 0;
     9567     __pyx_v_shape = ((PyObject *)__pyx_t_2);
     9568     __pyx_t_2 = 0;
     9569-    goto __pyx_L7;
     9570+    goto __pyx_L4;
     9571   }
     9572   /*else*/ {
     9573 
     9574@@ -17445,7 +18068,7 @@
     9575     __pyx_t_2 = __Pyx_PyInt_to_py_npy_intp(__pyx_v_d); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4086; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9576     __Pyx_GOTREF(__pyx_t_2);
     9577     __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4086; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9578-    __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     9579+    __Pyx_GOTREF(__pyx_t_4);
     9580     PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     9581     __Pyx_GIVEREF(__pyx_t_2);
     9582     __pyx_t_2 = 0;
     9583@@ -17455,7 +18078,7 @@
     9584     __pyx_v_shape = __pyx_t_2;
     9585     __pyx_t_2 = 0;
     9586   }
     9587-  __pyx_L7:;
     9588+  __pyx_L4:;
     9589 
     9590   /* "mtrand.pyx":4088
     9591  *             shape = size + (d,)
     9592@@ -17470,7 +18093,7 @@
     9593   __Pyx_GOTREF(__pyx_t_4);
     9594   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     9595   __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4088; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9596-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     9597+  __Pyx_GOTREF(__pyx_t_2);
     9598   __Pyx_INCREF(__pyx_v_shape);
     9599   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_shape);
     9600   __Pyx_GIVEREF(__pyx_v_shape);
     9601@@ -17562,7 +18185,7 @@
     9602         PyErr_Format(PyExc_ZeroDivisionError, "float division");
     9603         {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4096; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9604       }
     9605-      (__pyx_v_mnix[(__pyx_v_i + __pyx_v_j)]) = rk_binomial(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, __pyx_v_dn, ((__pyx_v_pix[__pyx_v_j]) / __pyx_v_Sum));
     9606+      (__pyx_v_mnix[(__pyx_v_i + __pyx_v_j)]) = rk_binomial(__pyx_v_self->internal_state, __pyx_v_dn, ((__pyx_v_pix[__pyx_v_j]) / __pyx_v_Sum));
     9607 
     9608       /* "mtrand.pyx":4097
     9609  *             for j from 0 <= j < d-1:
     9610@@ -17590,10 +18213,10 @@
     9611  *                 Sum = Sum - pix[j]
     9612  *             if dn > 0:
     9613  */
     9614-        goto __pyx_L11_break;
     9615-        goto __pyx_L12;
     9616+        goto __pyx_L8_break;
     9617+        goto __pyx_L9;
     9618       }
     9619-      __pyx_L12:;
     9620+      __pyx_L9:;
     9621 
     9622       /* "mtrand.pyx":4100
     9623  *                 if dn <= 0:
     9624@@ -17604,7 +18227,7 @@
     9625  */
     9626       __pyx_v_Sum = (__pyx_v_Sum - (__pyx_v_pix[__pyx_v_j]));
     9627     }
     9628-    __pyx_L11_break:;
     9629+    __pyx_L8_break:;
     9630 
     9631     /* "mtrand.pyx":4101
     9632  *                     break
     9633@@ -17624,9 +18247,9 @@
     9634  *             i = i + d
     9635  */
     9636       (__pyx_v_mnix[((__pyx_v_i + __pyx_v_d) - 1)]) = __pyx_v_dn;
     9637-      goto __pyx_L13;
     9638+      goto __pyx_L10;
     9639     }
     9640-    __pyx_L13:;
     9641+    __pyx_L10:;
     9642 
     9643     /* "mtrand.pyx":4104
     9644  *                 mnix[i+d-1] = dn
     9645@@ -17668,60 +18291,40 @@
     9646   return __pyx_r;
     9647 }
     9648 
     9649-/* "mtrand.pyx":4108
     9650+/* Python wrapper */
     9651+static PyObject *__pyx_pw_6mtrand_11RandomState_99dirichlet(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     9652+static char __pyx_doc_6mtrand_11RandomState_98dirichlet[] = "\n        dirichlet(alpha, size=None)\n\n        Draw samples from the Dirichlet distribution.\n\n        Draw `size` samples of dimension k from a Dirichlet distribution. A\n        Dirichlet-distributed random variable can be seen as a multivariate\n        generalization of a Beta distribution. Dirichlet pdf is the conjugate\n        prior of a multinomial in Bayesian inference.\n\n        Parameters\n        ----------\n        alpha : array\n            Parameter of the distribution (k dimension for sample of\n            dimension k).\n        size : array\n            Number of samples to draw.\n\n        Returns\n        -------\n        samples : ndarray,\n            The drawn samples, of shape (alpha.ndim, size).\n\n        Notes\n        -----\n        .. math:: X \\approx \\prod_{i=1}^{k}{x^{\\alpha_i-1}_i}\n\n        Uses the following property for computation: for each dimension,\n        draw a random sample y_i from a standard gamma generator of shape\n        `alpha_i`, then\n        :math:`X = \\frac{1}{\\sum_{i=1}^k{y_i}} (y_1, \\ldots, y_n)` is\n        Dirichlet distributed.\n\n        References\n        ----------\n        .. [1] David McKay, \"Information Theory, Inference and Learning\n               Algorithms,\" chapter 23,\n               http://www.inference.phy.cam.ac.uk/mackay/\n        .. [2] Wikipedia, \"Dirichlet distribution\",\n               http://en.wikipedia.org/wiki/Dirichlet_distribution\n\n        Examples\n        --------\n        Taking an example cited in Wikipedia, this distribution can be used if\n        one wanted to cut strings (each of initial length 1.0) into K pieces\n        with different lengths, where each piece had, on average, a designated\n        average length, but allowing some variation in the relative sizes of the\n        pieces.\n\n        >>> s = np.random.dirichlet((10, 5, 3), 20).transpose()\n\n        >>> plt.barh(range(20), s[0])\n        >>> plt.barh(range(20), s[1], left=s[0], color='g')""\n        >>> plt.barh(range(20), s[2], left=s[0]+s[1], color='r')\n        >>> plt.title(\"Lengths of Strings\")\n\n        ";
     9653+static PyObject *__pyx_pw_6mtrand_11RandomState_99dirichlet(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     9654+  PyObject *__pyx_v_alpha = 0;
     9655+  PyObject *__pyx_v_size = 0;
     9656+  PyObject *__pyx_r = 0;
     9657+  __Pyx_RefNannyDeclarations
     9658+  __Pyx_RefNannySetupContext("dirichlet (wrapper)", 0);
     9659+  {
     9660+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__alpha,&__pyx_n_s__size,0};
     9661+    PyObject* values[2] = {0,0};
     9662+
     9663+    /* "mtrand.pyx":4108
     9664  *         return multin
     9665  *
     9666  *     def dirichlet(self, object alpha, size=None):             # <<<<<<<<<<<<<<
     9667  *         """
     9668  *         dirichlet(alpha, size=None)
     9669  */
     9670-
     9671-static PyObject *__pyx_pf_6mtrand_11RandomState_49dirichlet(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
     9672-static char __pyx_doc_6mtrand_11RandomState_49dirichlet[] = "\n        dirichlet(alpha, size=None)\n\n        Draw samples from the Dirichlet distribution.\n\n        Draw `size` samples of dimension k from a Dirichlet distribution. A\n        Dirichlet-distributed random variable can be seen as a multivariate\n        generalization of a Beta distribution. Dirichlet pdf is the conjugate\n        prior of a multinomial in Bayesian inference.\n\n        Parameters\n        ----------\n        alpha : array\n            Parameter of the distribution (k dimension for sample of\n            dimension k).\n        size : array\n            Number of samples to draw.\n\n        Returns\n        -------\n        samples : ndarray,\n            The drawn samples, of shape (alpha.ndim, size).\n\n        Notes\n        -----\n        .. math:: X \\approx \\prod_{i=1}^{k}{x^{\\alpha_i-1}_i}\n\n        Uses the following property for computation: for each dimension,\n        draw a random sample y_i from a standard gamma generator of shape\n        `alpha_i`, then\n        :math:`X = \\frac{1}{\\sum_{i=1}^k{y_i}} (y_1, \\ldots, y_n)` is\n        Dirichlet distributed.\n\n        References\n        ----------\n        .. [1] David McKay, \"Information Theory, Inference and Learning\n               Algorithms,\" chapter 23,\n               http://www.inference.phy.cam.ac.uk/mackay/\n        .. [2] Wikipedia, \"Dirichlet distribution\",\n               http://en.wikipedia.org/wiki/Dirichlet_distribution\n\n        Examples\n        --------\n        Taking an example cited in Wikipedia, this distribution can be used if\n        one wanted to cut strings (each of initial length 1.0) into K pieces\n        with different lengths, where each piece had, on average, a designated\n        average length, but allowing some variation in the relative sizes of the\n        pieces.\n\n        >>> s = np.random.dirichlet((10, 5, 3), 20).transpose()\n\n        >>> plt.barh(range(20), s[0])\n        >>> plt.barh(range(20), s[1], left=s[0], color='g')""\n        >>> plt.barh(range(20), s[2], left=s[0]+s[1], color='r')\n        >>> plt.title(\"Lengths of Strings\")\n\n        ";
     9673-static PyObject *__pyx_pf_6mtrand_11RandomState_49dirichlet(PyObject *__pyx_v_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
     9674-  PyObject *__pyx_v_alpha = 0;
     9675-  PyObject *__pyx_v_size = 0;
     9676-  npy_intp __pyx_v_k;
     9677-  npy_intp __pyx_v_totsize;
     9678-  PyArrayObject *__pyx_v_alpha_arr = 0;
     9679-  PyArrayObject *__pyx_v_val_arr = 0;
     9680-  double *__pyx_v_alpha_data;
     9681-  double *__pyx_v_val_data;
     9682-  npy_intp __pyx_v_i;
     9683-  npy_intp __pyx_v_j;
     9684-  double __pyx_v_acc;
     9685-  double __pyx_v_invacc;
     9686-  PyObject *__pyx_v_shape = NULL;
     9687-  PyObject *__pyx_v_diric = NULL;
     9688-  PyObject *__pyx_r = NULL;
     9689-  __Pyx_RefNannyDeclarations
     9690-  Py_ssize_t __pyx_t_1;
     9691-  PyObject *__pyx_t_2 = NULL;
     9692-  int __pyx_t_3;
     9693-  PyObject *__pyx_t_4 = NULL;
     9694-  PyObject *__pyx_t_5 = NULL;
     9695-  npy_intp __pyx_t_6;
     9696-  int __pyx_lineno = 0;
     9697-  const char *__pyx_filename = NULL;
     9698-  int __pyx_clineno = 0;
     9699-  static PyObject **__pyx_pyargnames[] = {&__pyx_n_s__alpha,&__pyx_n_s__size,0};
     9700-  __Pyx_RefNannySetupContext("dirichlet");
     9701-  {
     9702-    PyObject* values[2] = {0,0};
     9703     values[1] = ((PyObject *)Py_None);
     9704     if (unlikely(__pyx_kwds)) {
     9705       Py_ssize_t kw_args;
     9706-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     9707+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
     9708+      switch (pos_args) {
     9709         case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
     9710         case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
     9711         case  0: break;
     9712         default: goto __pyx_L5_argtuple_error;
     9713       }
     9714       kw_args = PyDict_Size(__pyx_kwds);
     9715-      switch (PyTuple_GET_SIZE(__pyx_args)) {
     9716+      switch (pos_args) {
     9717         case  0:
     9718-        values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__alpha);
     9719-        if (likely(values[0])) kw_args--;
     9720+        if (likely((values[0] = PyDict_GetItem(__pyx_kwds, __pyx_n_s__alpha)) != 0)) kw_args--;
     9721         else goto __pyx_L5_argtuple_error;
     9722         case  1:
     9723         if (kw_args > 0) {
     9724@@ -17730,7 +18333,7 @@
     9725         }
     9726       }
     9727       if (unlikely(kw_args > 0)) {
     9728-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, PyTuple_GET_SIZE(__pyx_args), "dirichlet") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4108; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     9729+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "dirichlet") < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4108; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     9730       }
     9731     } else {
     9732       switch (PyTuple_GET_SIZE(__pyx_args)) {
     9733@@ -17751,6 +18354,36 @@
     9734   __Pyx_RefNannyFinishContext();
     9735   return NULL;
     9736   __pyx_L4_argument_unpacking_done:;
     9737+  __pyx_r = __pyx_pf_6mtrand_11RandomState_98dirichlet(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), __pyx_v_alpha, __pyx_v_size);
     9738+  __Pyx_RefNannyFinishContext();
     9739+  return __pyx_r;
     9740+}
     9741+
     9742+static PyObject *__pyx_pf_6mtrand_11RandomState_98dirichlet(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_alpha, PyObject *__pyx_v_size) {
     9743+  npy_intp __pyx_v_k;
     9744+  npy_intp __pyx_v_totsize;
     9745+  PyArrayObject *__pyx_v_alpha_arr = 0;
     9746+  PyArrayObject *__pyx_v_val_arr = 0;
     9747+  double *__pyx_v_alpha_data;
     9748+  double *__pyx_v_val_data;
     9749+  npy_intp __pyx_v_i;
     9750+  npy_intp __pyx_v_j;
     9751+  double __pyx_v_acc;
     9752+  double __pyx_v_invacc;
     9753+  PyObject *__pyx_v_shape = NULL;
     9754+  PyObject *__pyx_v_diric = NULL;
     9755+  PyObject *__pyx_r = NULL;
     9756+  __Pyx_RefNannyDeclarations
     9757+  Py_ssize_t __pyx_t_1;
     9758+  PyObject *__pyx_t_2 = NULL;
     9759+  int __pyx_t_3;
     9760+  PyObject *__pyx_t_4 = NULL;
     9761+  PyObject *__pyx_t_5 = NULL;
     9762+  npy_intp __pyx_t_6;
     9763+  int __pyx_lineno = 0;
     9764+  const char *__pyx_filename = NULL;
     9765+  int __pyx_clineno = 0;
     9766+  __Pyx_RefNannySetupContext("dirichlet", 0);
     9767 
     9768   /* "mtrand.pyx":4194
     9769  *         cdef double     acc, invacc
     9770@@ -17804,13 +18437,13 @@
     9771     __pyx_t_2 = __Pyx_PyInt_to_py_npy_intp(__pyx_v_k); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4199; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9772     __Pyx_GOTREF(__pyx_t_2);
     9773     __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4199; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9774-    __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     9775+    __Pyx_GOTREF(__pyx_t_4);
     9776     PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     9777     __Pyx_GIVEREF(__pyx_t_2);
     9778     __pyx_t_2 = 0;
     9779     __pyx_v_shape = ((PyObject *)__pyx_t_4);
     9780     __pyx_t_4 = 0;
     9781-    goto __pyx_L6;
     9782+    goto __pyx_L3;
     9783   }
     9784 
     9785   /* "mtrand.pyx":4200
     9786@@ -17833,7 +18466,7 @@
     9787     __pyx_t_4 = __Pyx_PyInt_to_py_npy_intp(__pyx_v_k); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4201; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9788     __Pyx_GOTREF(__pyx_t_4);
     9789     __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4201; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9790-    __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     9791+    __Pyx_GOTREF(__pyx_t_2);
     9792     __Pyx_INCREF(__pyx_v_size);
     9793     PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_size);
     9794     __Pyx_GIVEREF(__pyx_v_size);
     9795@@ -17842,7 +18475,7 @@
     9796     __pyx_t_4 = 0;
     9797     __pyx_v_shape = ((PyObject *)__pyx_t_2);
     9798     __pyx_t_2 = 0;
     9799-    goto __pyx_L6;
     9800+    goto __pyx_L3;
     9801   }
     9802   /*else*/ {
     9803 
     9804@@ -17856,7 +18489,7 @@
     9805     __pyx_t_2 = __Pyx_PyInt_to_py_npy_intp(__pyx_v_k); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4203; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9806     __Pyx_GOTREF(__pyx_t_2);
     9807     __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4203; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9808-    __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     9809+    __Pyx_GOTREF(__pyx_t_4);
     9810     PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
     9811     __Pyx_GIVEREF(__pyx_t_2);
     9812     __pyx_t_2 = 0;
     9813@@ -17866,7 +18499,7 @@
     9814     __pyx_v_shape = __pyx_t_2;
     9815     __pyx_t_2 = 0;
     9816   }
     9817-  __pyx_L6:;
     9818+  __pyx_L3:;
     9819 
     9820   /* "mtrand.pyx":4205
     9821  *             shape = size + (k,)
     9822@@ -17886,7 +18519,7 @@
     9823   __Pyx_GOTREF(__pyx_t_5);
     9824   __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
     9825   __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4205; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     9826-  __Pyx_GOTREF(((PyObject *)__pyx_t_2));
     9827+  __Pyx_GOTREF(__pyx_t_2);
     9828   __Pyx_INCREF(__pyx_v_shape);
     9829   PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_shape);
     9830   __Pyx_GIVEREF(__pyx_v_shape);
     9831@@ -17974,7 +18607,7 @@
     9832  *                 acc             = acc + val_data[i+j]
     9833  *             invacc  = 1/acc
     9834  */
     9835-      (__pyx_v_val_data[(__pyx_v_i + __pyx_v_j)]) = rk_standard_gamma(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state, (__pyx_v_alpha_data[__pyx_v_j]));
     9836+      (__pyx_v_val_data[(__pyx_v_i + __pyx_v_j)]) = rk_standard_gamma(__pyx_v_self->internal_state, (__pyx_v_alpha_data[__pyx_v_j]));
     9837 
     9838       /* "mtrand.pyx":4215
     9839  *             for j from 0 <= j < k:
     9840@@ -18059,6 +18692,18 @@
     9841   return __pyx_r;
     9842 }
     9843 
     9844+/* Python wrapper */
     9845+static PyObject *__pyx_pw_6mtrand_11RandomState_101shuffle(PyObject *__pyx_v_self, PyObject *__pyx_v_x); /*proto*/
     9846+static char __pyx_doc_6mtrand_11RandomState_100shuffle[] = "\n        shuffle(x)\n\n        Modify a sequence in-place by shuffling its contents.\n\n        Parameters\n        ----------\n        x : array_like\n            The array or list to be shuffled.\n\n        Returns\n        -------\n        None\n\n        Examples\n        --------\n        >>> arr = np.arange(10)\n        >>> np.random.shuffle(arr)\n        >>> arr\n        [1 7 5 2 9 4 3 6 0 8]\n\n        This function only shuffles the array along the first index of a\n        multi-dimensional array:\n\n        >>> arr = np.arange(9).reshape((3, 3))\n        >>> np.random.shuffle(arr)\n        >>> arr\n        array([[3, 4, 5],\n               [6, 7, 8],\n               [0, 1, 2]])\n\n        ";
     9847+static PyObject *__pyx_pw_6mtrand_11RandomState_101shuffle(PyObject *__pyx_v_self, PyObject *__pyx_v_x) {
     9848+  PyObject *__pyx_r = 0;
     9849+  __Pyx_RefNannyDeclarations
     9850+  __Pyx_RefNannySetupContext("shuffle (wrapper)", 0);
     9851+  __pyx_r = __pyx_pf_6mtrand_11RandomState_100shuffle(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), ((PyObject *)__pyx_v_x));
     9852+  __Pyx_RefNannyFinishContext();
     9853+  return __pyx_r;
     9854+}
     9855+
     9856 /* "mtrand.pyx":4224
     9857  *
     9858  *     # Shuffling and permutations:
     9859@@ -18067,9 +18712,7 @@
     9860  *         shuffle(x)
     9861  */
     9862 
     9863-static PyObject *__pyx_pf_6mtrand_11RandomState_50shuffle(PyObject *__pyx_v_self, PyObject *__pyx_v_x); /*proto*/
     9864-static char __pyx_doc_6mtrand_11RandomState_50shuffle[] = "\n        shuffle(x)\n\n        Modify a sequence in-place by shuffling its contents.\n\n        Parameters\n        ----------\n        x : array_like\n            The array or list to be shuffled.\n\n        Returns\n        -------\n        None\n\n        Examples\n        --------\n        >>> arr = np.arange(10)\n        >>> np.random.shuffle(arr)\n        >>> arr\n        [1 7 5 2 9 4 3 6 0 8]\n\n        This function only shuffles the array along the first index of a\n        multi-dimensional array:\n\n        >>> arr = np.arange(9).reshape((3, 3))\n        >>> np.random.shuffle(arr)\n        >>> arr\n        array([[3, 4, 5],\n               [6, 7, 8],\n               [0, 1, 2]])\n\n        ";
     9865-static PyObject *__pyx_pf_6mtrand_11RandomState_50shuffle(PyObject *__pyx_v_self, PyObject *__pyx_v_x) {
     9866+static PyObject *__pyx_pf_6mtrand_11RandomState_100shuffle(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_x) {
     9867   npy_intp __pyx_v_i;
     9868   npy_intp __pyx_v_j;
     9869   int __pyx_v_copy;
     9870@@ -18086,7 +18729,7 @@
     9871   int __pyx_lineno = 0;
     9872   const char *__pyx_filename = NULL;
     9873   int __pyx_clineno = 0;
     9874-  __Pyx_RefNannySetupContext("shuffle");
     9875+  __Pyx_RefNannySetupContext("shuffle", 0);
     9876 
     9877   /* "mtrand.pyx":4260
     9878  *         cdef int copy
     9879@@ -18119,17 +18762,17 @@
     9880  *         except:
     9881  *             j = 0
     9882  */
     9883-      __pyx_t_5 = __Pyx_GetItemInt(__pyx_v_x, 0, sizeof(long), PyInt_FromLong); if (!__pyx_t_5) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4262; __pyx_clineno = __LINE__; goto __pyx_L5_error;}
     9884+      __pyx_t_5 = __Pyx_GetItemInt(__pyx_v_x, 0, sizeof(long), PyInt_FromLong); if (!__pyx_t_5) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4262; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     9885       __Pyx_GOTREF(__pyx_t_5);
     9886-      __pyx_t_1 = PyObject_Length(__pyx_t_5); if (unlikely(__pyx_t_1 == -1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4262; __pyx_clineno = __LINE__; goto __pyx_L5_error;}
     9887+      __pyx_t_1 = PyObject_Length(__pyx_t_5); if (unlikely(__pyx_t_1 == -1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4262; __pyx_clineno = __LINE__; goto __pyx_L3_error;}
     9888       __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     9889       __pyx_v_j = __pyx_t_1;
     9890     }
     9891     __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
     9892     __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
     9893     __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
     9894-    goto __pyx_L12_try_end;
     9895-    __pyx_L5_error:;
     9896+    goto __pyx_L10_try_end;
     9897+    __pyx_L3_error:;
     9898     __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
     9899 
     9900     /* "mtrand.pyx":4263
     9901@@ -18141,7 +18784,7 @@
     9902  */
     9903     /*except:*/ {
     9904       __Pyx_AddTraceback("mtrand.RandomState.shuffle", __pyx_clineno, __pyx_lineno, __pyx_filename);
     9905-      if (__Pyx_GetException(&__pyx_t_5, &__pyx_t_6, &__pyx_t_7) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4263; __pyx_clineno = __LINE__; goto __pyx_L7_except_error;}
     9906+      if (__Pyx_GetException(&__pyx_t_5, &__pyx_t_6, &__pyx_t_7) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4263; __pyx_clineno = __LINE__; goto __pyx_L5_except_error;}
     9907       __Pyx_GOTREF(__pyx_t_5);
     9908       __Pyx_GOTREF(__pyx_t_6);
     9909       __Pyx_GOTREF(__pyx_t_7);
     9910@@ -18157,20 +18800,20 @@
     9911       __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
     9912       __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
     9913       __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
     9914-      goto __pyx_L6_exception_handled;
     9915+      goto __pyx_L4_exception_handled;
     9916     }
     9917-    __pyx_L7_except_error:;
     9918+    __pyx_L5_except_error:;
     9919     __Pyx_XGIVEREF(__pyx_t_2);
     9920     __Pyx_XGIVEREF(__pyx_t_3);
     9921     __Pyx_XGIVEREF(__pyx_t_4);
     9922     __Pyx_ExceptionReset(__pyx_t_2, __pyx_t_3, __pyx_t_4);
     9923     goto __pyx_L1_error;
     9924-    __pyx_L6_exception_handled:;
     9925+    __pyx_L4_exception_handled:;
     9926     __Pyx_XGIVEREF(__pyx_t_2);
     9927     __Pyx_XGIVEREF(__pyx_t_3);
     9928     __Pyx_XGIVEREF(__pyx_t_4);
     9929     __Pyx_ExceptionReset(__pyx_t_2, __pyx_t_3, __pyx_t_4);
     9930-    __pyx_L12_try_end:;
     9931+    __pyx_L10_try_end:;
     9932   }
     9933 
     9934   /* "mtrand.pyx":4266
     9935@@ -18201,7 +18844,7 @@
     9936  *                 x[i], x[j] = x[j], x[i]
     9937  *                 i = i - 1
     9938  */
     9939-      __pyx_v_j = rk_interval(__pyx_v_i, ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state);
     9940+      __pyx_v_j = rk_interval(__pyx_v_i, __pyx_v_self->internal_state);
     9941 
     9942       /* "mtrand.pyx":4270
     9943  *             while i > 0:
     9944@@ -18228,7 +18871,7 @@
     9945  */
     9946       __pyx_v_i = (__pyx_v_i - 1);
     9947     }
     9948-    goto __pyx_L15;
     9949+    goto __pyx_L13;
     9950   }
     9951   /*else*/ {
     9952 
     9953@@ -18275,7 +18918,7 @@
     9954  *                     x[i], x[j] = x[j].copy(), x[i].copy()
     9955  *                     i = i - 1
     9956  */
     9957-        __pyx_v_j = rk_interval(__pyx_v_i, ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state);
     9958+        __pyx_v_j = rk_interval(__pyx_v_i, __pyx_v_self->internal_state);
     9959 
     9960         /* "mtrand.pyx":4278
     9961  *                 while(i > 0):
     9962@@ -18314,7 +18957,7 @@
     9963  */
     9964         __pyx_v_i = (__pyx_v_i - 1);
     9965       }
     9966-      goto __pyx_L18;
     9967+      goto __pyx_L16;
     9968     }
     9969     /*else*/ {
     9970 
     9971@@ -18336,7 +18979,7 @@
     9972  *                     x[i], x[j] = x[j][:], x[i][:]
     9973  *                     i = i - 1
     9974  */
     9975-        __pyx_v_j = rk_interval(__pyx_v_i, ((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self)->internal_state);
     9976+        __pyx_v_j = rk_interval(__pyx_v_i, __pyx_v_self->internal_state);
     9977 
     9978         /* "mtrand.pyx":4283
     9979  *                 while(i > 0):
     9980@@ -18370,9 +19013,9 @@
     9981         __pyx_v_i = (__pyx_v_i - 1);
     9982       }
     9983     }
     9984-    __pyx_L18:;
     9985+    __pyx_L16:;
     9986   }
     9987-  __pyx_L15:;
     9988+  __pyx_L13:;
     9989 
     9990   __pyx_r = Py_None; __Pyx_INCREF(Py_None);
     9991   goto __pyx_L0;
     9992@@ -18388,6 +19031,18 @@
     9993   return __pyx_r;
     9994 }
     9995 
     9996+/* Python wrapper */
     9997+static PyObject *__pyx_pw_6mtrand_11RandomState_103permutation(PyObject *__pyx_v_self, PyObject *__pyx_v_x); /*proto*/
     9998+static char __pyx_doc_6mtrand_11RandomState_102permutation[] = "\n        permutation(x)\n\n        Randomly permute a sequence, or return a permuted range.\n\n        If `x` is a multi-dimensional array, it is only shuffled along its\n        first index.\n\n        Parameters\n        ----------\n        x : int or array_like\n            If `x` is an integer, randomly permute ``np.arange(x)``.\n            If `x` is an array, make a copy and shuffle the elements\n            randomly.\n\n        Returns\n        -------\n        out : ndarray\n            Permuted sequence or array range.\n\n        Examples\n        --------\n        >>> np.random.permutation(10)\n        array([1, 7, 4, 3, 0, 9, 2, 5, 8, 6])\n\n        >>> np.random.permutation([1, 4, 9, 12, 15])\n        array([15,  1,  9,  4, 12])\n\n        >>> arr = np.arange(9).reshape((3, 3))\n        >>> np.random.permutation(arr)\n        array([[6, 7, 8],\n               [0, 1, 2],\n               [3, 4, 5]])\n\n        ";
     9999+static PyObject *__pyx_pw_6mtrand_11RandomState_103permutation(PyObject *__pyx_v_self, PyObject *__pyx_v_x) {
     10000+  PyObject *__pyx_r = 0;
     10001+  __Pyx_RefNannyDeclarations
     10002+  __Pyx_RefNannySetupContext("permutation (wrapper)", 0);
     10003+  __pyx_r = __pyx_pf_6mtrand_11RandomState_102permutation(((struct __pyx_obj_6mtrand_RandomState *)__pyx_v_self), ((PyObject *)__pyx_v_x));
     10004+  __Pyx_RefNannyFinishContext();
     10005+  return __pyx_r;
     10006+}
     10007+
     10008 /* "mtrand.pyx":4286
     10009  *                     i = i - 1
     10010  *
     10011@@ -18396,9 +19051,7 @@
     10012  *         permutation(x)
     10013  */
     10014 
     10015-static PyObject *__pyx_pf_6mtrand_11RandomState_51permutation(PyObject *__pyx_v_self, PyObject *__pyx_v_x); /*proto*/
     10016-static char __pyx_doc_6mtrand_11RandomState_51permutation[] = "\n        permutation(x)\n\n        Randomly permute a sequence, or return a permuted range.\n\n        If `x` is a multi-dimensional array, it is only shuffled along its\n        first index.\n\n        Parameters\n        ----------\n        x : int or array_like\n            If `x` is an integer, randomly permute ``np.arange(x)``.\n            If `x` is an array, make a copy and shuffle the elements\n            randomly.\n\n        Returns\n        -------\n        out : ndarray\n            Permuted sequence or array range.\n\n        Examples\n        --------\n        >>> np.random.permutation(10)\n        array([1, 7, 4, 3, 0, 9, 2, 5, 8, 6])\n\n        >>> np.random.permutation([1, 4, 9, 12, 15])\n        array([15,  1,  9,  4, 12])\n\n        >>> arr = np.arange(9).reshape((3, 3))\n        >>> np.random.permutation(arr)\n        array([[6, 7, 8],\n               [0, 1, 2],\n               [3, 4, 5]])\n\n        ";
     10017-static PyObject *__pyx_pf_6mtrand_11RandomState_51permutation(PyObject *__pyx_v_self, PyObject *__pyx_v_x) {
     10018+static PyObject *__pyx_pf_6mtrand_11RandomState_102permutation(struct __pyx_obj_6mtrand_RandomState *__pyx_v_self, PyObject *__pyx_v_x) {
     10019   PyObject *__pyx_v_arr = NULL;
     10020   PyObject *__pyx_r = NULL;
     10021   __Pyx_RefNannyDeclarations
     10022@@ -18409,7 +19062,7 @@
     10023   int __pyx_lineno = 0;
     10024   const char *__pyx_filename = NULL;
     10025   int __pyx_clineno = 0;
     10026-  __Pyx_RefNannySetupContext("permutation");
     10027+  __Pyx_RefNannySetupContext("permutation", 0);
     10028 
     10029   /* "mtrand.pyx":4322
     10030  *
     10031@@ -18424,7 +19077,7 @@
     10032   __Pyx_GOTREF(__pyx_t_2);
     10033   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     10034   __pyx_t_1 = PyTuple_New(3); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4322; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10035-  __Pyx_GOTREF(((PyObject *)__pyx_t_1));
     10036+  __Pyx_GOTREF(__pyx_t_1);
     10037   __Pyx_INCREF(((PyObject *)((PyObject*)(&PyInt_Type))));
     10038   PyTuple_SET_ITEM(__pyx_t_1, 0, ((PyObject *)((PyObject*)(&PyInt_Type))));
     10039   __Pyx_GIVEREF(((PyObject *)((PyObject*)(&PyInt_Type))));
     10040@@ -18451,7 +19104,7 @@
     10041     __Pyx_GOTREF(__pyx_t_2);
     10042     __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     10043     __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4323; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10044-    __Pyx_GOTREF(((PyObject *)__pyx_t_1));
     10045+    __Pyx_GOTREF(__pyx_t_1);
     10046     __Pyx_INCREF(__pyx_v_x);
     10047     PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_v_x);
     10048     __Pyx_GIVEREF(__pyx_v_x);
     10049@@ -18461,7 +19114,7 @@
     10050     __Pyx_DECREF(((PyObject *)__pyx_t_1)); __pyx_t_1 = 0;
     10051     __pyx_v_arr = __pyx_t_4;
     10052     __pyx_t_4 = 0;
     10053-    goto __pyx_L5;
     10054+    goto __pyx_L3;
     10055   }
     10056   /*else*/ {
     10057 
     10058@@ -18478,7 +19131,7 @@
     10059     __Pyx_GOTREF(__pyx_t_1);
     10060     __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
     10061     __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4325; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10062-    __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     10063+    __Pyx_GOTREF(__pyx_t_4);
     10064     __Pyx_INCREF(__pyx_v_x);
     10065     PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_x);
     10066     __Pyx_GIVEREF(__pyx_v_x);
     10067@@ -18489,7 +19142,7 @@
     10068     __pyx_v_arr = __pyx_t_2;
     10069     __pyx_t_2 = 0;
     10070   }
     10071-  __pyx_L5:;
     10072+  __pyx_L3:;
     10073 
     10074   /* "mtrand.pyx":4326
     10075  *         else:
     10076@@ -18498,10 +19151,10 @@
     10077  *         return arr
     10078  *
     10079  */
     10080-  __pyx_t_2 = PyObject_GetAttr(__pyx_v_self, __pyx_n_s__shuffle); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4326; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10081+  __pyx_t_2 = PyObject_GetAttr(((PyObject *)__pyx_v_self), __pyx_n_s__shuffle); if (unlikely(!__pyx_t_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4326; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10082   __Pyx_GOTREF(__pyx_t_2);
     10083   __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4326; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10084-  __Pyx_GOTREF(((PyObject *)__pyx_t_4));
     10085+  __Pyx_GOTREF(__pyx_t_4);
     10086   __Pyx_INCREF(__pyx_v_arr);
     10087   PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_arr);
     10088   __Pyx_GIVEREF(__pyx_v_arr);
     10089@@ -18538,7 +19191,7 @@
     10090   return __pyx_r;
     10091 }
     10092 
     10093-static PyObject *__pyx_tp_new_6mtrand_RandomState(PyTypeObject *t, PyObject *a, PyObject *k) {
     10094+static PyObject *__pyx_tp_new_6mtrand_RandomState(PyTypeObject *t, CYTHON_UNUSED PyObject *a, CYTHON_UNUSED PyObject *k) {
     10095   PyObject *o = (*t->tp_alloc)(t, 0);
     10096   if (!o) return 0;
     10097   return o;
     10098@@ -18549,7 +19202,7 @@
     10099     PyObject *etype, *eval, *etb;
     10100     PyErr_Fetch(&etype, &eval, &etb);
     10101     ++Py_REFCNT(o);
     10102-    __pyx_pf_6mtrand_11RandomState_1__dealloc__(o);
     10103+    __pyx_pw_6mtrand_11RandomState_3__dealloc__(o);
     10104     if (PyErr_Occurred()) PyErr_WriteUnraisable(o);
     10105     --Py_REFCNT(o);
     10106     PyErr_Restore(etype, eval, etb);
     10107@@ -18558,56 +19211,56 @@
     10108 }
     10109 
     10110 static PyMethodDef __pyx_methods_6mtrand_RandomState[] = {
     10111-  {__Pyx_NAMESTR("seed"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_2seed, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_2seed)},
     10112-  {__Pyx_NAMESTR("get_state"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_3get_state, METH_NOARGS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_3get_state)},
     10113-  {__Pyx_NAMESTR("set_state"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_4set_state, METH_O, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_4set_state)},
     10114-  {__Pyx_NAMESTR("__getstate__"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_5__getstate__, METH_NOARGS, __Pyx_DOCSTR(0)},
     10115-  {__Pyx_NAMESTR("__setstate__"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_6__setstate__, METH_O, __Pyx_DOCSTR(0)},
     10116-  {__Pyx_NAMESTR("__reduce__"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_7__reduce__, METH_NOARGS, __Pyx_DOCSTR(0)},
     10117-  {__Pyx_NAMESTR("random_sample"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_8random_sample, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_8random_sample)},
     10118-  {__Pyx_NAMESTR("tomaxint"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_9tomaxint, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_9tomaxint)},
     10119-  {__Pyx_NAMESTR("randint"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_10randint, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_10randint)},
     10120-  {__Pyx_NAMESTR("bytes"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_11bytes, METH_O, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_11bytes)},
     10121-  {__Pyx_NAMESTR("uniform"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_12uniform, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_12uniform)},
     10122-  {__Pyx_NAMESTR("rand"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_13rand, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_13rand)},
     10123-  {__Pyx_NAMESTR("randn"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_14randn, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_14randn)},
     10124-  {__Pyx_NAMESTR("random_integers"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_15random_integers, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_15random_integers)},
     10125-  {__Pyx_NAMESTR("standard_normal"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_16standard_normal, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_16standard_normal)},
     10126-  {__Pyx_NAMESTR("normal"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_17normal, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_17normal)},
     10127-  {__Pyx_NAMESTR("beta"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_18beta, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_18beta)},
     10128-  {__Pyx_NAMESTR("exponential"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_19exponential, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_19exponential)},
     10129-  {__Pyx_NAMESTR("standard_exponential"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_20standard_exponential, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_20standard_exponential)},
     10130-  {__Pyx_NAMESTR("standard_gamma"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_21standard_gamma, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_21standard_gamma)},
     10131-  {__Pyx_NAMESTR("gamma"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_22gamma, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_22gamma)},
     10132-  {__Pyx_NAMESTR("f"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_23f, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_23f)},
     10133-  {__Pyx_NAMESTR("noncentral_f"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_24noncentral_f, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_24noncentral_f)},
     10134-  {__Pyx_NAMESTR("chisquare"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_25chisquare, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_25chisquare)},
     10135-  {__Pyx_NAMESTR("noncentral_chisquare"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_26noncentral_chisquare, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_26noncentral_chisquare)},
     10136-  {__Pyx_NAMESTR("standard_cauchy"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_27standard_cauchy, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_27standard_cauchy)},
     10137-  {__Pyx_NAMESTR("standard_t"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_28standard_t, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_28standard_t)},
     10138-  {__Pyx_NAMESTR("vonmises"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_29vonmises, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_29vonmises)},
     10139-  {__Pyx_NAMESTR("pareto"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_30pareto, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_30pareto)},
     10140-  {__Pyx_NAMESTR("weibull"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_31weibull, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_31weibull)},
     10141-  {__Pyx_NAMESTR("power"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_32power, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_32power)},
     10142-  {__Pyx_NAMESTR("laplace"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_33laplace, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_33laplace)},
     10143-  {__Pyx_NAMESTR("gumbel"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_34gumbel, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_34gumbel)},
     10144-  {__Pyx_NAMESTR("logistic"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_35logistic, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_35logistic)},
     10145-  {__Pyx_NAMESTR("lognormal"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_36lognormal, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_36lognormal)},
     10146-  {__Pyx_NAMESTR("rayleigh"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_37rayleigh, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_37rayleigh)},
     10147-  {__Pyx_NAMESTR("wald"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_38wald, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_38wald)},
     10148-  {__Pyx_NAMESTR("triangular"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_39triangular, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_39triangular)},
     10149-  {__Pyx_NAMESTR("binomial"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_40binomial, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_40binomial)},
     10150-  {__Pyx_NAMESTR("negative_binomial"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_41negative_binomial, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_41negative_binomial)},
     10151-  {__Pyx_NAMESTR("poisson"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_42poisson, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_42poisson)},
     10152-  {__Pyx_NAMESTR("zipf"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_43zipf, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_43zipf)},
     10153-  {__Pyx_NAMESTR("geometric"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_44geometric, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_44geometric)},
     10154-  {__Pyx_NAMESTR("hypergeometric"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_45hypergeometric, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_45hypergeometric)},
     10155-  {__Pyx_NAMESTR("logseries"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_46logseries, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_46logseries)},
     10156-  {__Pyx_NAMESTR("multivariate_normal"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_47multivariate_normal, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_47multivariate_normal)},
     10157-  {__Pyx_NAMESTR("multinomial"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_48multinomial, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_48multinomial)},
     10158-  {__Pyx_NAMESTR("dirichlet"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_49dirichlet, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_49dirichlet)},
     10159-  {__Pyx_NAMESTR("shuffle"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_50shuffle, METH_O, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_50shuffle)},
     10160-  {__Pyx_NAMESTR("permutation"), (PyCFunction)__pyx_pf_6mtrand_11RandomState_51permutation, METH_O, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_51permutation)},
     10161+  {__Pyx_NAMESTR("seed"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_5seed, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_4seed)},
     10162+  {__Pyx_NAMESTR("get_state"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_7get_state, METH_NOARGS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_6get_state)},
     10163+  {__Pyx_NAMESTR("set_state"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_9set_state, METH_O, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_8set_state)},
     10164+  {__Pyx_NAMESTR("__getstate__"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_11__getstate__, METH_NOARGS, __Pyx_DOCSTR(0)},
     10165+  {__Pyx_NAMESTR("__setstate__"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_13__setstate__, METH_O, __Pyx_DOCSTR(0)},
     10166+  {__Pyx_NAMESTR("__reduce__"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_15__reduce__, METH_NOARGS, __Pyx_DOCSTR(0)},
     10167+  {__Pyx_NAMESTR("random_sample"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_17random_sample, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_16random_sample)},
     10168+  {__Pyx_NAMESTR("tomaxint"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_19tomaxint, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_18tomaxint)},
     10169+  {__Pyx_NAMESTR("randint"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_21randint, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_20randint)},
     10170+  {__Pyx_NAMESTR("bytes"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_23bytes, METH_O, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_22bytes)},
     10171+  {__Pyx_NAMESTR("uniform"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_25uniform, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_24uniform)},
     10172+  {__Pyx_NAMESTR("rand"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_27rand, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_26rand)},
     10173+  {__Pyx_NAMESTR("randn"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_29randn, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_28randn)},
     10174+  {__Pyx_NAMESTR("random_integers"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_31random_integers, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_30random_integers)},
     10175+  {__Pyx_NAMESTR("standard_normal"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_33standard_normal, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_32standard_normal)},
     10176+  {__Pyx_NAMESTR("normal"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_35normal, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_34normal)},
     10177+  {__Pyx_NAMESTR("beta"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_37beta, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_36beta)},
     10178+  {__Pyx_NAMESTR("exponential"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_39exponential, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_38exponential)},
     10179+  {__Pyx_NAMESTR("standard_exponential"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_41standard_exponential, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_40standard_exponential)},
     10180+  {__Pyx_NAMESTR("standard_gamma"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_43standard_gamma, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_42standard_gamma)},
     10181+  {__Pyx_NAMESTR("gamma"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_45gamma, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_44gamma)},
     10182+  {__Pyx_NAMESTR("f"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_47f, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_46f)},
     10183+  {__Pyx_NAMESTR("noncentral_f"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_49noncentral_f, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_48noncentral_f)},
     10184+  {__Pyx_NAMESTR("chisquare"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_51chisquare, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_50chisquare)},
     10185+  {__Pyx_NAMESTR("noncentral_chisquare"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_53noncentral_chisquare, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_52noncentral_chisquare)},
     10186+  {__Pyx_NAMESTR("standard_cauchy"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_55standard_cauchy, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_54standard_cauchy)},
     10187+  {__Pyx_NAMESTR("standard_t"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_57standard_t, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_56standard_t)},
     10188+  {__Pyx_NAMESTR("vonmises"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_59vonmises, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_58vonmises)},
     10189+  {__Pyx_NAMESTR("pareto"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_61pareto, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_60pareto)},
     10190+  {__Pyx_NAMESTR("weibull"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_63weibull, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_62weibull)},
     10191+  {__Pyx_NAMESTR("power"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_65power, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_64power)},
     10192+  {__Pyx_NAMESTR("laplace"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_67laplace, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_66laplace)},
     10193+  {__Pyx_NAMESTR("gumbel"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_69gumbel, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_68gumbel)},
     10194+  {__Pyx_NAMESTR("logistic"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_71logistic, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_70logistic)},
     10195+  {__Pyx_NAMESTR("lognormal"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_73lognormal, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_72lognormal)},
     10196+  {__Pyx_NAMESTR("rayleigh"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_75rayleigh, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_74rayleigh)},
     10197+  {__Pyx_NAMESTR("wald"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_77wald, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_76wald)},
     10198+  {__Pyx_NAMESTR("triangular"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_79triangular, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_78triangular)},
     10199+  {__Pyx_NAMESTR("binomial"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_81binomial, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_80binomial)},
     10200+  {__Pyx_NAMESTR("negative_binomial"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_83negative_binomial, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_82negative_binomial)},
     10201+  {__Pyx_NAMESTR("poisson"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_85poisson, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_84poisson)},
     10202+  {__Pyx_NAMESTR("zipf"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_87zipf, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_86zipf)},
     10203+  {__Pyx_NAMESTR("geometric"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_89geometric, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_88geometric)},
     10204+  {__Pyx_NAMESTR("hypergeometric"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_91hypergeometric, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_90hypergeometric)},
     10205+  {__Pyx_NAMESTR("logseries"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_93logseries, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_92logseries)},
     10206+  {__Pyx_NAMESTR("multivariate_normal"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_95multivariate_normal, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_94multivariate_normal)},
     10207+  {__Pyx_NAMESTR("multinomial"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_97multinomial, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_96multinomial)},
     10208+  {__Pyx_NAMESTR("dirichlet"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_99dirichlet, METH_VARARGS|METH_KEYWORDS, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_98dirichlet)},
     10209+  {__Pyx_NAMESTR("shuffle"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_101shuffle, METH_O, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_100shuffle)},
     10210+  {__Pyx_NAMESTR("permutation"), (PyCFunction)__pyx_pw_6mtrand_11RandomState_103permutation, METH_O, __Pyx_DOCSTR(__pyx_doc_6mtrand_11RandomState_102permutation)},
     10211   {0, 0, 0, 0}
     10212 };
     10213 
     10214@@ -18749,7 +19402,7 @@
     10215   0, /*tp_descr_get*/
     10216   0, /*tp_descr_set*/
     10217   0, /*tp_dictoffset*/
     10218-  __pyx_pf_6mtrand_11RandomState___init__, /*tp_init*/
     10219+  __pyx_pw_6mtrand_11RandomState_1__init__, /*tp_init*/
     10220   0, /*tp_alloc*/
     10221   __pyx_tp_new_6mtrand_RandomState, /*tp_new*/
     10222   0, /*tp_free*/
     10223@@ -19035,7 +19688,7 @@
     10224 
     10225 static int __Pyx_InitCachedConstants(void) {
     10226   __Pyx_RefNannyDeclarations
     10227-  __Pyx_RefNannySetupContext("__Pyx_InitCachedConstants");
     10228+  __Pyx_RefNannySetupContext("__Pyx_InitCachedConstants", 0);
     10229 
     10230   /* "mtrand.pyx":184
     10231  *                                                 <void *>oa)
     10232@@ -19045,7 +19698,7 @@
     10233  *             oa_data = <double *>PyArray_MultiIter_DATA(multi, 1)
     10234  */
     10235   __pyx_k_tuple_2 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_2)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 184; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10236-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_2));
     10237+  __Pyx_GOTREF(__pyx_k_tuple_2);
     10238   __Pyx_INCREF(((PyObject *)__pyx_kp_s_1));
     10239   PyTuple_SET_ITEM(__pyx_k_tuple_2, 0, ((PyObject *)__pyx_kp_s_1));
     10240   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_1));
     10241@@ -19059,7 +19712,7 @@
     10242  *             oa_data = <double *>PyArray_MultiIter_DATA(multi, 1)
     10243  */
     10244   __pyx_k_tuple_3 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_3)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 233; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10245-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_3));
     10246+  __Pyx_GOTREF(__pyx_k_tuple_3);
     10247   __Pyx_INCREF(((PyObject *)__pyx_kp_s_1));
     10248   PyTuple_SET_ITEM(__pyx_k_tuple_3, 0, ((PyObject *)__pyx_kp_s_1));
     10249   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_1));
     10250@@ -19073,7 +19726,7 @@
     10251  *             oa_data = <double *>PyArray_MultiIter_DATA(multi, 1)
     10252  */
     10253   __pyx_k_tuple_4 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_4)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 288; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10254-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_4));
     10255+  __Pyx_GOTREF(__pyx_k_tuple_4);
     10256   __Pyx_INCREF(((PyObject *)__pyx_kp_s_1));
     10257   PyTuple_SET_ITEM(__pyx_k_tuple_4, 0, ((PyObject *)__pyx_kp_s_1));
     10258   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_1));
     10259@@ -19087,7 +19740,7 @@
     10260  *             on_data = <long *>PyArray_MultiIter_DATA(multi, 1)
     10261  */
     10262   __pyx_k_tuple_5 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_5)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 352; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10263-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_5));
     10264+  __Pyx_GOTREF(__pyx_k_tuple_5);
     10265   __Pyx_INCREF(((PyObject *)__pyx_kp_s_1));
     10266   PyTuple_SET_ITEM(__pyx_k_tuple_5, 0, ((PyObject *)__pyx_kp_s_1));
     10267   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_1));
     10268@@ -19101,7 +19754,7 @@
     10269  *             on_data = <double *>PyArray_MultiIter_DATA(multi, 1)
     10270  */
     10271   __pyx_k_tuple_6 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_6)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 401; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10272-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_6));
     10273+  __Pyx_GOTREF(__pyx_k_tuple_6);
     10274   __Pyx_INCREF(((PyObject *)__pyx_kp_s_1));
     10275   PyTuple_SET_ITEM(__pyx_k_tuple_6, 0, ((PyObject *)__pyx_kp_s_1));
     10276   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_1));
     10277@@ -19115,7 +19768,7 @@
     10278  *             on_data = <long *>PyArray_MultiIter_DATA(multi, 1)
     10279  */
     10280   __pyx_k_tuple_7 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_7)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 455; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10281-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_7));
     10282+  __Pyx_GOTREF(__pyx_k_tuple_7);
     10283   __Pyx_INCREF(((PyObject *)__pyx_kp_s_1));
     10284   PyTuple_SET_ITEM(__pyx_k_tuple_7, 0, ((PyObject *)__pyx_kp_s_1));
     10285   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_1));
     10286@@ -19129,7 +19782,7 @@
     10287  *             oa_data = <double *>PyArray_MultiIter_DATA(multi, 1)
     10288  */
     10289   __pyx_k_tuple_8 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_8)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 503; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10290-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_8));
     10291+  __Pyx_GOTREF(__pyx_k_tuple_8);
     10292   __Pyx_INCREF(((PyObject *)__pyx_kp_s_1));
     10293   PyTuple_SET_ITEM(__pyx_k_tuple_8, 0, ((PyObject *)__pyx_kp_s_1));
     10294   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_1));
     10295@@ -19143,7 +19796,7 @@
     10296  *         if len(state) == 3:
     10297  */
     10298   __pyx_k_tuple_10 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_10)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 689; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10299-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_10));
     10300+  __Pyx_GOTREF(__pyx_k_tuple_10);
     10301   __Pyx_INCREF(((PyObject *)__pyx_kp_s_9));
     10302   PyTuple_SET_ITEM(__pyx_k_tuple_10, 0, ((PyObject *)__pyx_kp_s_9));
     10303   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_9));
     10304@@ -19157,7 +19810,7 @@
     10305  *         self.internal_state.pos = pos
     10306  */
     10307   __pyx_k_tuple_12 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_12)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 702; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10308-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_12));
     10309+  __Pyx_GOTREF(__pyx_k_tuple_12);
     10310   __Pyx_INCREF(((PyObject *)__pyx_kp_s_11));
     10311   PyTuple_SET_ITEM(__pyx_k_tuple_12, 0, ((PyObject *)__pyx_kp_s_11));
     10312   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_11));
     10313@@ -19171,7 +19824,7 @@
     10314  *         diff = <unsigned long>hi - <unsigned long>lo - 1UL
     10315  */
     10316   __pyx_k_tuple_14 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_14)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 874; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10317-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_14));
     10318+  __Pyx_GOTREF(__pyx_k_tuple_14);
     10319   __Pyx_INCREF(((PyObject *)__pyx_kp_s_13));
     10320   PyTuple_SET_ITEM(__pyx_k_tuple_14, 0, ((PyObject *)__pyx_kp_s_13));
     10321   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_13));
     10322@@ -19185,7 +19838,7 @@
     10323  *
     10324  */
     10325   __pyx_k_tuple_20 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_20)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1301; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10326-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_20));
     10327+  __Pyx_GOTREF(__pyx_k_tuple_20);
     10328   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10329   PyTuple_SET_ITEM(__pyx_k_tuple_20, 0, ((PyObject *)__pyx_kp_s_19));
     10330   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10331@@ -19199,7 +19852,7 @@
     10332  *
     10333  */
     10334   __pyx_k_tuple_21 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_21)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1309; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10335-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_21));
     10336+  __Pyx_GOTREF(__pyx_k_tuple_21);
     10337   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10338   PyTuple_SET_ITEM(__pyx_k_tuple_21, 0, ((PyObject *)__pyx_kp_s_19));
     10339   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10340@@ -19213,7 +19866,7 @@
     10341  *                 raise ValueError("b <= 0")
     10342  */
     10343   __pyx_k_tuple_23 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_23)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1356; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10344-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_23));
     10345+  __Pyx_GOTREF(__pyx_k_tuple_23);
     10346   __Pyx_INCREF(((PyObject *)__pyx_kp_s_22));
     10347   PyTuple_SET_ITEM(__pyx_k_tuple_23, 0, ((PyObject *)__pyx_kp_s_22));
     10348   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_22));
     10349@@ -19227,7 +19880,7 @@
     10350  *
     10351  */
     10352   __pyx_k_tuple_25 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_25)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1358; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10353-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_25));
     10354+  __Pyx_GOTREF(__pyx_k_tuple_25);
     10355   __Pyx_INCREF(((PyObject *)__pyx_kp_s_24));
     10356   PyTuple_SET_ITEM(__pyx_k_tuple_25, 0, ((PyObject *)__pyx_kp_s_24));
     10357   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_24));
     10358@@ -19241,7 +19894,7 @@
     10359  *             raise ValueError("b <= 0")
     10360  */
     10361   __pyx_k_tuple_26 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_26)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1366; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10362-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_26));
     10363+  __Pyx_GOTREF(__pyx_k_tuple_26);
     10364   __Pyx_INCREF(((PyObject *)__pyx_kp_s_22));
     10365   PyTuple_SET_ITEM(__pyx_k_tuple_26, 0, ((PyObject *)__pyx_kp_s_22));
     10366   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_22));
     10367@@ -19255,7 +19908,7 @@
     10368  *
     10369  */
     10370   __pyx_k_tuple_27 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_27)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1368; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10371-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_27));
     10372+  __Pyx_GOTREF(__pyx_k_tuple_27);
     10373   __Pyx_INCREF(((PyObject *)__pyx_kp_s_24));
     10374   PyTuple_SET_ITEM(__pyx_k_tuple_27, 0, ((PyObject *)__pyx_kp_s_24));
     10375   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_24));
     10376@@ -19269,7 +19922,7 @@
     10377  *
     10378  */
     10379   __pyx_k_tuple_29 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_29)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1415; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10380-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_29));
     10381+  __Pyx_GOTREF(__pyx_k_tuple_29);
     10382   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10383   PyTuple_SET_ITEM(__pyx_k_tuple_29, 0, ((PyObject *)__pyx_kp_s_19));
     10384   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10385@@ -19283,7 +19936,7 @@
     10386  *
     10387  */
     10388   __pyx_k_tuple_30 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_30)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1422; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10389-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_30));
     10390+  __Pyx_GOTREF(__pyx_k_tuple_30);
     10391   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10392   PyTuple_SET_ITEM(__pyx_k_tuple_30, 0, ((PyObject *)__pyx_kp_s_19));
     10393   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10394@@ -19297,7 +19950,7 @@
     10395  *
     10396  */
     10397   __pyx_k_tuple_32 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_32)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1526; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10398-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_32));
     10399+  __Pyx_GOTREF(__pyx_k_tuple_32);
     10400   __Pyx_INCREF(((PyObject *)__pyx_kp_s_31));
     10401   PyTuple_SET_ITEM(__pyx_k_tuple_32, 0, ((PyObject *)__pyx_kp_s_31));
     10402   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_31));
     10403@@ -19311,7 +19964,7 @@
     10404  *
     10405  */
     10406   __pyx_k_tuple_33 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_33)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1532; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10407-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_33));
     10408+  __Pyx_GOTREF(__pyx_k_tuple_33);
     10409   __Pyx_INCREF(((PyObject *)__pyx_kp_s_31));
     10410   PyTuple_SET_ITEM(__pyx_k_tuple_33, 0, ((PyObject *)__pyx_kp_s_31));
     10411   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_31));
     10412@@ -19325,7 +19978,7 @@
     10413  *                 raise ValueError("scale <= 0")
     10414  */
     10415   __pyx_k_tuple_35 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_35)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1612; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10416-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_35));
     10417+  __Pyx_GOTREF(__pyx_k_tuple_35);
     10418   __Pyx_INCREF(((PyObject *)__pyx_kp_s_31));
     10419   PyTuple_SET_ITEM(__pyx_k_tuple_35, 0, ((PyObject *)__pyx_kp_s_31));
     10420   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_31));
     10421@@ -19339,7 +19992,7 @@
     10422  *
     10423  */
     10424   __pyx_k_tuple_36 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_36)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1614; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10425-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_36));
     10426+  __Pyx_GOTREF(__pyx_k_tuple_36);
     10427   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10428   PyTuple_SET_ITEM(__pyx_k_tuple_36, 0, ((PyObject *)__pyx_kp_s_19));
     10429   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10430@@ -19353,7 +20006,7 @@
     10431  *             raise ValueError("scale <= 0")
     10432  */
     10433   __pyx_k_tuple_37 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_37)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1621; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10434-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_37));
     10435+  __Pyx_GOTREF(__pyx_k_tuple_37);
     10436   __Pyx_INCREF(((PyObject *)__pyx_kp_s_31));
     10437   PyTuple_SET_ITEM(__pyx_k_tuple_37, 0, ((PyObject *)__pyx_kp_s_31));
     10438   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_31));
     10439@@ -19367,7 +20020,7 @@
     10440  *
     10441  */
     10442   __pyx_k_tuple_38 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_38)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1623; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10443-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_38));
     10444+  __Pyx_GOTREF(__pyx_k_tuple_38);
     10445   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10446   PyTuple_SET_ITEM(__pyx_k_tuple_38, 0, ((PyObject *)__pyx_kp_s_19));
     10447   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10448@@ -19381,7 +20034,7 @@
     10449  *                 raise ValueError("scale <= 0")
     10450  */
     10451   __pyx_k_tuple_39 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_39)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1714; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10452-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_39));
     10453+  __Pyx_GOTREF(__pyx_k_tuple_39);
     10454   __Pyx_INCREF(((PyObject *)__pyx_kp_s_31));
     10455   PyTuple_SET_ITEM(__pyx_k_tuple_39, 0, ((PyObject *)__pyx_kp_s_31));
     10456   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_31));
     10457@@ -19395,7 +20048,7 @@
     10458  *
     10459  */
     10460   __pyx_k_tuple_40 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_40)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1716; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10461-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_40));
     10462+  __Pyx_GOTREF(__pyx_k_tuple_40);
     10463   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10464   PyTuple_SET_ITEM(__pyx_k_tuple_40, 0, ((PyObject *)__pyx_kp_s_19));
     10465   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10466@@ -19409,7 +20062,7 @@
     10467  *             raise ValueError("dfden <= 0")
     10468  */
     10469   __pyx_k_tuple_42 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_42)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1724; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10470-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_42));
     10471+  __Pyx_GOTREF(__pyx_k_tuple_42);
     10472   __Pyx_INCREF(((PyObject *)__pyx_kp_s_41));
     10473   PyTuple_SET_ITEM(__pyx_k_tuple_42, 0, ((PyObject *)__pyx_kp_s_41));
     10474   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_41));
     10475@@ -19423,7 +20076,7 @@
     10476  *
     10477  */
     10478   __pyx_k_tuple_44 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_44)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1726; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10479-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_44));
     10480+  __Pyx_GOTREF(__pyx_k_tuple_44);
     10481   __Pyx_INCREF(((PyObject *)__pyx_kp_s_43));
     10482   PyTuple_SET_ITEM(__pyx_k_tuple_44, 0, ((PyObject *)__pyx_kp_s_43));
     10483   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_43));
     10484@@ -19437,7 +20090,7 @@
     10485  *                 raise ValueError("dfden <= 0")
     10486  */
     10487   __pyx_k_tuple_46 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_46)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1801; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10488-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_46));
     10489+  __Pyx_GOTREF(__pyx_k_tuple_46);
     10490   __Pyx_INCREF(((PyObject *)__pyx_kp_s_45));
     10491   PyTuple_SET_ITEM(__pyx_k_tuple_46, 0, ((PyObject *)__pyx_kp_s_45));
     10492   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_45));
     10493@@ -19451,7 +20104,7 @@
     10494  *                 raise ValueError("nonc < 0")
     10495  */
     10496   __pyx_k_tuple_47 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_47)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1803; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10497-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_47));
     10498+  __Pyx_GOTREF(__pyx_k_tuple_47);
     10499   __Pyx_INCREF(((PyObject *)__pyx_kp_s_43));
     10500   PyTuple_SET_ITEM(__pyx_k_tuple_47, 0, ((PyObject *)__pyx_kp_s_43));
     10501   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_43));
     10502@@ -19465,7 +20118,7 @@
     10503  *                                   fdfnum, fdfden, fnonc)
     10504  */
     10505   __pyx_k_tuple_49 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_49)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1805; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10506-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_49));
     10507+  __Pyx_GOTREF(__pyx_k_tuple_49);
     10508   __Pyx_INCREF(((PyObject *)__pyx_kp_s_48));
     10509   PyTuple_SET_ITEM(__pyx_k_tuple_49, 0, ((PyObject *)__pyx_kp_s_48));
     10510   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_48));
     10511@@ -19479,7 +20132,7 @@
     10512  *             raise ValueError("dfden <= 0")
     10513  */
     10514   __pyx_k_tuple_50 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_50)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1816; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10515-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_50));
     10516+  __Pyx_GOTREF(__pyx_k_tuple_50);
     10517   __Pyx_INCREF(((PyObject *)__pyx_kp_s_45));
     10518   PyTuple_SET_ITEM(__pyx_k_tuple_50, 0, ((PyObject *)__pyx_kp_s_45));
     10519   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_45));
     10520@@ -19493,7 +20146,7 @@
     10521  *             raise ValueError("nonc < 0")
     10522  */
     10523   __pyx_k_tuple_51 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_51)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1818; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10524-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_51));
     10525+  __Pyx_GOTREF(__pyx_k_tuple_51);
     10526   __Pyx_INCREF(((PyObject *)__pyx_kp_s_43));
     10527   PyTuple_SET_ITEM(__pyx_k_tuple_51, 0, ((PyObject *)__pyx_kp_s_43));
     10528   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_43));
     10529@@ -19507,7 +20160,7 @@
     10530  *             odfden, ononc)
     10531  */
     10532   __pyx_k_tuple_52 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_52)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1820; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10533-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_52));
     10534+  __Pyx_GOTREF(__pyx_k_tuple_52);
     10535   __Pyx_INCREF(((PyObject *)__pyx_kp_s_48));
     10536   PyTuple_SET_ITEM(__pyx_k_tuple_52, 0, ((PyObject *)__pyx_kp_s_48));
     10537   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_48));
     10538@@ -19521,7 +20174,7 @@
     10539  *
     10540  */
     10541   __pyx_k_tuple_54 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_54)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1892; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10542-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_54));
     10543+  __Pyx_GOTREF(__pyx_k_tuple_54);
     10544   __Pyx_INCREF(((PyObject *)__pyx_kp_s_53));
     10545   PyTuple_SET_ITEM(__pyx_k_tuple_54, 0, ((PyObject *)__pyx_kp_s_53));
     10546   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_53));
     10547@@ -19535,7 +20188,7 @@
     10548  *
     10549  */
     10550   __pyx_k_tuple_55 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_55)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1899; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10551-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_55));
     10552+  __Pyx_GOTREF(__pyx_k_tuple_55);
     10553   __Pyx_INCREF(((PyObject *)__pyx_kp_s_53));
     10554   PyTuple_SET_ITEM(__pyx_k_tuple_55, 0, ((PyObject *)__pyx_kp_s_53));
     10555   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_53));
     10556@@ -19549,7 +20202,7 @@
     10557  *                 raise ValueError("nonc <= 0")
     10558  */
     10559   __pyx_k_tuple_56 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_56)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1977; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10560-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_56));
     10561+  __Pyx_GOTREF(__pyx_k_tuple_56);
     10562   __Pyx_INCREF(((PyObject *)__pyx_kp_s_53));
     10563   PyTuple_SET_ITEM(__pyx_k_tuple_56, 0, ((PyObject *)__pyx_kp_s_53));
     10564   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_53));
     10565@@ -19563,7 +20216,7 @@
     10566  *                                   size, fdf, fnonc)
     10567  */
     10568   __pyx_k_tuple_58 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_58)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1979; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10569-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_58));
     10570+  __Pyx_GOTREF(__pyx_k_tuple_58);
     10571   __Pyx_INCREF(((PyObject *)__pyx_kp_s_57));
     10572   PyTuple_SET_ITEM(__pyx_k_tuple_58, 0, ((PyObject *)__pyx_kp_s_57));
     10573   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_57));
     10574@@ -19577,7 +20230,7 @@
     10575  *             raise ValueError("nonc < 0")
     10576  */
     10577   __pyx_k_tuple_60 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_60)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1988; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10578-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_60));
     10579+  __Pyx_GOTREF(__pyx_k_tuple_60);
     10580   __Pyx_INCREF(((PyObject *)__pyx_kp_s_59));
     10581   PyTuple_SET_ITEM(__pyx_k_tuple_60, 0, ((PyObject *)__pyx_kp_s_59));
     10582   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_59));
     10583@@ -19591,7 +20244,7 @@
     10584  *             odf, ononc)
     10585  */
     10586   __pyx_k_tuple_61 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_61)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1990; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10587-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_61));
     10588+  __Pyx_GOTREF(__pyx_k_tuple_61);
     10589   __Pyx_INCREF(((PyObject *)__pyx_kp_s_48));
     10590   PyTuple_SET_ITEM(__pyx_k_tuple_61, 0, ((PyObject *)__pyx_kp_s_48));
     10591   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_48));
     10592@@ -19605,7 +20258,7 @@
     10593  *
     10594  */
     10595   __pyx_k_tuple_62 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_62)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2146; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10596-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_62));
     10597+  __Pyx_GOTREF(__pyx_k_tuple_62);
     10598   __Pyx_INCREF(((PyObject *)__pyx_kp_s_53));
     10599   PyTuple_SET_ITEM(__pyx_k_tuple_62, 0, ((PyObject *)__pyx_kp_s_53));
     10600   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_53));
     10601@@ -19619,7 +20272,7 @@
     10602  *
     10603  */
     10604   __pyx_k_tuple_63 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_63)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2153; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10605-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_63));
     10606+  __Pyx_GOTREF(__pyx_k_tuple_63);
     10607   __Pyx_INCREF(((PyObject *)__pyx_kp_s_53));
     10608   PyTuple_SET_ITEM(__pyx_k_tuple_63, 0, ((PyObject *)__pyx_kp_s_53));
     10609   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_53));
     10610@@ -19633,7 +20286,7 @@
     10611  *
     10612  */
     10613   __pyx_k_tuple_65 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_65)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2239; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10614-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_65));
     10615+  __Pyx_GOTREF(__pyx_k_tuple_65);
     10616   __Pyx_INCREF(((PyObject *)__pyx_kp_s_64));
     10617   PyTuple_SET_ITEM(__pyx_k_tuple_65, 0, ((PyObject *)__pyx_kp_s_64));
     10618   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_64));
     10619@@ -19647,7 +20300,7 @@
     10620  *
     10621  */
     10622   __pyx_k_tuple_66 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_66)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2247; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10623-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_66));
     10624+  __Pyx_GOTREF(__pyx_k_tuple_66);
     10625   __Pyx_INCREF(((PyObject *)__pyx_kp_s_64));
     10626   PyTuple_SET_ITEM(__pyx_k_tuple_66, 0, ((PyObject *)__pyx_kp_s_64));
     10627   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_64));
     10628@@ -19661,7 +20314,7 @@
     10629  *
     10630  */
     10631   __pyx_k_tuple_67 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_67)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2336; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10632-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_67));
     10633+  __Pyx_GOTREF(__pyx_k_tuple_67);
     10634   __Pyx_INCREF(((PyObject *)__pyx_kp_s_22));
     10635   PyTuple_SET_ITEM(__pyx_k_tuple_67, 0, ((PyObject *)__pyx_kp_s_22));
     10636   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_22));
     10637@@ -19675,7 +20328,7 @@
     10638  *
     10639  */
     10640   __pyx_k_tuple_68 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_68)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2343; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10641-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_68));
     10642+  __Pyx_GOTREF(__pyx_k_tuple_68);
     10643   __Pyx_INCREF(((PyObject *)__pyx_kp_s_22));
     10644   PyTuple_SET_ITEM(__pyx_k_tuple_68, 0, ((PyObject *)__pyx_kp_s_22));
     10645   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_22));
     10646@@ -19689,7 +20342,7 @@
     10647  *
     10648  */
     10649   __pyx_k_tuple_69 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_69)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2436; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10650-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_69));
     10651+  __Pyx_GOTREF(__pyx_k_tuple_69);
     10652   __Pyx_INCREF(((PyObject *)__pyx_kp_s_22));
     10653   PyTuple_SET_ITEM(__pyx_k_tuple_69, 0, ((PyObject *)__pyx_kp_s_22));
     10654   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_22));
     10655@@ -19703,7 +20356,7 @@
     10656  *
     10657  */
     10658   __pyx_k_tuple_70 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_70)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2443; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10659-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_70));
     10660+  __Pyx_GOTREF(__pyx_k_tuple_70);
     10661   __Pyx_INCREF(((PyObject *)__pyx_kp_s_22));
     10662   PyTuple_SET_ITEM(__pyx_k_tuple_70, 0, ((PyObject *)__pyx_kp_s_22));
     10663   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_22));
     10664@@ -19717,7 +20370,7 @@
     10665  *
     10666  */
     10667   __pyx_k_tuple_71 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_71)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2545; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10668-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_71));
     10669+  __Pyx_GOTREF(__pyx_k_tuple_71);
     10670   __Pyx_INCREF(((PyObject *)__pyx_kp_s_22));
     10671   PyTuple_SET_ITEM(__pyx_k_tuple_71, 0, ((PyObject *)__pyx_kp_s_22));
     10672   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_22));
     10673@@ -19731,7 +20384,7 @@
     10674  *
     10675  */
     10676   __pyx_k_tuple_72 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_72)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2552; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10677-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_72));
     10678+  __Pyx_GOTREF(__pyx_k_tuple_72);
     10679   __Pyx_INCREF(((PyObject *)__pyx_kp_s_22));
     10680   PyTuple_SET_ITEM(__pyx_k_tuple_72, 0, ((PyObject *)__pyx_kp_s_22));
     10681   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_22));
     10682@@ -19745,7 +20398,7 @@
     10683  *
     10684  */
     10685   __pyx_k_tuple_75 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_75)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2635; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10686-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_75));
     10687+  __Pyx_GOTREF(__pyx_k_tuple_75);
     10688   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10689   PyTuple_SET_ITEM(__pyx_k_tuple_75, 0, ((PyObject *)__pyx_kp_s_19));
     10690   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10691@@ -19759,7 +20412,7 @@
     10692  *
     10693  */
     10694   __pyx_k_tuple_76 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_76)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2642; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10695-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_76));
     10696+  __Pyx_GOTREF(__pyx_k_tuple_76);
     10697   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10698   PyTuple_SET_ITEM(__pyx_k_tuple_76, 0, ((PyObject *)__pyx_kp_s_19));
     10699   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10700@@ -19773,7 +20426,7 @@
     10701  *
     10702  */
     10703   __pyx_k_tuple_79 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_79)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2766; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10704-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_79));
     10705+  __Pyx_GOTREF(__pyx_k_tuple_79);
     10706   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10707   PyTuple_SET_ITEM(__pyx_k_tuple_79, 0, ((PyObject *)__pyx_kp_s_19));
     10708   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10709@@ -19787,7 +20440,7 @@
     10710  *
     10711  */
     10712   __pyx_k_tuple_80 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_80)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2773; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10713-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_80));
     10714+  __Pyx_GOTREF(__pyx_k_tuple_80);
     10715   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10716   PyTuple_SET_ITEM(__pyx_k_tuple_80, 0, ((PyObject *)__pyx_kp_s_19));
     10717   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10718@@ -19801,7 +20454,7 @@
     10719  *
     10720  */
     10721   __pyx_k_tuple_83 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_83)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2854; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10722-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_83));
     10723+  __Pyx_GOTREF(__pyx_k_tuple_83);
     10724   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10725   PyTuple_SET_ITEM(__pyx_k_tuple_83, 0, ((PyObject *)__pyx_kp_s_19));
     10726   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10727@@ -19815,7 +20468,7 @@
     10728  *
     10729  */
     10730   __pyx_k_tuple_84 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_84)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2861; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10731-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_84));
     10732+  __Pyx_GOTREF(__pyx_k_tuple_84);
     10733   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10734   PyTuple_SET_ITEM(__pyx_k_tuple_84, 0, ((PyObject *)__pyx_kp_s_19));
     10735   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10736@@ -19829,7 +20482,7 @@
     10737  *
     10738  */
     10739   __pyx_k_tuple_88 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_88)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2984; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10740-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_88));
     10741+  __Pyx_GOTREF(__pyx_k_tuple_88);
     10742   __Pyx_INCREF(((PyObject *)__pyx_kp_s_87));
     10743   PyTuple_SET_ITEM(__pyx_k_tuple_88, 0, ((PyObject *)__pyx_kp_s_87));
     10744   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_87));
     10745@@ -19843,7 +20496,7 @@
     10746  *
     10747  */
     10748   __pyx_k_tuple_90 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_90)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 2992; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10749-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_90));
     10750+  __Pyx_GOTREF(__pyx_k_tuple_90);
     10751   __Pyx_INCREF(((PyObject *)__pyx_kp_s_89));
     10752   PyTuple_SET_ITEM(__pyx_k_tuple_90, 0, ((PyObject *)__pyx_kp_s_89));
     10753   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_89));
     10754@@ -19857,7 +20510,7 @@
     10755  *
     10756  */
     10757   __pyx_k_tuple_92 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_92)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3057; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10758-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_92));
     10759+  __Pyx_GOTREF(__pyx_k_tuple_92);
     10760   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10761   PyTuple_SET_ITEM(__pyx_k_tuple_92, 0, ((PyObject *)__pyx_kp_s_19));
     10762   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10763@@ -19871,7 +20524,7 @@
     10764  *
     10765  */
     10766   __pyx_k_tuple_94 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_94)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3064; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10767-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_94));
     10768+  __Pyx_GOTREF(__pyx_k_tuple_94);
     10769   __Pyx_INCREF(((PyObject *)__pyx_kp_s_93));
     10770   PyTuple_SET_ITEM(__pyx_k_tuple_94, 0, ((PyObject *)__pyx_kp_s_93));
     10771   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_93));
     10772@@ -19885,7 +20538,7 @@
     10773  *                 raise ValueError("scale <= 0")
     10774  */
     10775   __pyx_k_tuple_96 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_96)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3137; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10776-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_96));
     10777+  __Pyx_GOTREF(__pyx_k_tuple_96);
     10778   __Pyx_INCREF(((PyObject *)__pyx_kp_s_95));
     10779   PyTuple_SET_ITEM(__pyx_k_tuple_96, 0, ((PyObject *)__pyx_kp_s_95));
     10780   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_95));
     10781@@ -19899,7 +20552,7 @@
     10782  *
     10783  */
     10784   __pyx_k_tuple_97 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_97)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3139; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10785-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_97));
     10786+  __Pyx_GOTREF(__pyx_k_tuple_97);
     10787   __Pyx_INCREF(((PyObject *)__pyx_kp_s_19));
     10788   PyTuple_SET_ITEM(__pyx_k_tuple_97, 0, ((PyObject *)__pyx_kp_s_19));
     10789   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_19));
     10790@@ -19913,7 +20566,7 @@
     10791  *             raise ValueError("scale <= 0.0")
     10792  */
     10793   __pyx_k_tuple_99 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_99)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3146; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10794-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_99));
     10795+  __Pyx_GOTREF(__pyx_k_tuple_99);
     10796   __Pyx_INCREF(((PyObject *)__pyx_kp_s_98));
     10797   PyTuple_SET_ITEM(__pyx_k_tuple_99, 0, ((PyObject *)__pyx_kp_s_98));
     10798   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_98));
     10799@@ -19927,7 +20580,7 @@
     10800  *
     10801  */
     10802   __pyx_k_tuple_100 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_100)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3148; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10803-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_100));
     10804+  __Pyx_GOTREF(__pyx_k_tuple_100);
     10805   __Pyx_INCREF(((PyObject *)__pyx_kp_s_93));
     10806   PyTuple_SET_ITEM(__pyx_k_tuple_100, 0, ((PyObject *)__pyx_kp_s_93));
     10807   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_93));
     10808@@ -19941,7 +20594,7 @@
     10809  *                 raise ValueError("mode > right")
     10810  */
     10811   __pyx_k_tuple_102 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_102)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3218; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10812-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_102));
     10813+  __Pyx_GOTREF(__pyx_k_tuple_102);
     10814   __Pyx_INCREF(((PyObject *)__pyx_kp_s_101));
     10815   PyTuple_SET_ITEM(__pyx_k_tuple_102, 0, ((PyObject *)__pyx_kp_s_101));
     10816   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_101));
     10817@@ -19955,7 +20608,7 @@
     10818  *                 raise ValueError("left == right")
     10819  */
     10820   __pyx_k_tuple_104 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_104)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3220; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10821-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_104));
     10822+  __Pyx_GOTREF(__pyx_k_tuple_104);
     10823   __Pyx_INCREF(((PyObject *)__pyx_kp_s_103));
     10824   PyTuple_SET_ITEM(__pyx_k_tuple_104, 0, ((PyObject *)__pyx_kp_s_103));
     10825   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_103));
     10826@@ -19969,7 +20622,7 @@
     10827  *                                   fmode, fright)
     10828  */
     10829   __pyx_k_tuple_106 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_106)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3222; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10830-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_106));
     10831+  __Pyx_GOTREF(__pyx_k_tuple_106);
     10832   __Pyx_INCREF(((PyObject *)__pyx_kp_s_105));
     10833   PyTuple_SET_ITEM(__pyx_k_tuple_106, 0, ((PyObject *)__pyx_kp_s_105));
     10834   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_105));
     10835@@ -19983,7 +20636,7 @@
     10836  *             raise ValueError("mode > right")
     10837  */
     10838   __pyx_k_tuple_107 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_107)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3232; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10839-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_107));
     10840+  __Pyx_GOTREF(__pyx_k_tuple_107);
     10841   __Pyx_INCREF(((PyObject *)__pyx_kp_s_101));
     10842   PyTuple_SET_ITEM(__pyx_k_tuple_107, 0, ((PyObject *)__pyx_kp_s_101));
     10843   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_101));
     10844@@ -19997,7 +20650,7 @@
     10845  *             raise ValueError("left == right")
     10846  */
     10847   __pyx_k_tuple_108 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_108)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3234; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10848-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_108));
     10849+  __Pyx_GOTREF(__pyx_k_tuple_108);
     10850   __Pyx_INCREF(((PyObject *)__pyx_kp_s_103));
     10851   PyTuple_SET_ITEM(__pyx_k_tuple_108, 0, ((PyObject *)__pyx_kp_s_103));
     10852   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_103));
     10853@@ -20011,7 +20664,7 @@
     10854  *             omode, oright)
     10855  */
     10856   __pyx_k_tuple_109 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_109)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3236; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10857-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_109));
     10858+  __Pyx_GOTREF(__pyx_k_tuple_109);
     10859   __Pyx_INCREF(((PyObject *)__pyx_kp_s_105));
     10860   PyTuple_SET_ITEM(__pyx_k_tuple_109, 0, ((PyObject *)__pyx_kp_s_105));
     10861   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_105));
     10862@@ -20025,7 +20678,7 @@
     10863  *                 raise ValueError("p < 0")
     10864  */
     10865   __pyx_k_tuple_111 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_111)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3330; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10866-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_111));
     10867+  __Pyx_GOTREF(__pyx_k_tuple_111);
     10868   __Pyx_INCREF(((PyObject *)__pyx_kp_s_110));
     10869   PyTuple_SET_ITEM(__pyx_k_tuple_111, 0, ((PyObject *)__pyx_kp_s_110));
     10870   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_110));
     10871@@ -20039,7 +20692,7 @@
     10872  *                 raise ValueError("p > 1")
     10873  */
     10874   __pyx_k_tuple_113 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_113)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3332; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10875-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_113));
     10876+  __Pyx_GOTREF(__pyx_k_tuple_113);
     10877   __Pyx_INCREF(((PyObject *)__pyx_kp_s_112));
     10878   PyTuple_SET_ITEM(__pyx_k_tuple_113, 0, ((PyObject *)__pyx_kp_s_112));
     10879   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_112));
     10880@@ -20053,7 +20706,7 @@
     10881  *
     10882  */
     10883   __pyx_k_tuple_115 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_115)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3334; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10884-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_115));
     10885+  __Pyx_GOTREF(__pyx_k_tuple_115);
     10886   __Pyx_INCREF(((PyObject *)__pyx_kp_s_114));
     10887   PyTuple_SET_ITEM(__pyx_k_tuple_115, 0, ((PyObject *)__pyx_kp_s_114));
     10888   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_114));
     10889@@ -20067,7 +20720,7 @@
     10890  *             raise ValueError("p < 0")
     10891  */
     10892   __pyx_k_tuple_116 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_116)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3342; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10893-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_116));
     10894+  __Pyx_GOTREF(__pyx_k_tuple_116);
     10895   __Pyx_INCREF(((PyObject *)__pyx_kp_s_110));
     10896   PyTuple_SET_ITEM(__pyx_k_tuple_116, 0, ((PyObject *)__pyx_kp_s_110));
     10897   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_110));
     10898@@ -20081,7 +20734,7 @@
     10899  *             raise ValueError("p > 1")
     10900  */
     10901   __pyx_k_tuple_117 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_117)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3344; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10902-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_117));
     10903+  __Pyx_GOTREF(__pyx_k_tuple_117);
     10904   __Pyx_INCREF(((PyObject *)__pyx_kp_s_112));
     10905   PyTuple_SET_ITEM(__pyx_k_tuple_117, 0, ((PyObject *)__pyx_kp_s_112));
     10906   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_112));
     10907@@ -20095,7 +20748,7 @@
     10908  *
     10909  */
     10910   __pyx_k_tuple_118 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_118)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3346; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10911-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_118));
     10912+  __Pyx_GOTREF(__pyx_k_tuple_118);
     10913   __Pyx_INCREF(((PyObject *)__pyx_kp_s_114));
     10914   PyTuple_SET_ITEM(__pyx_k_tuple_118, 0, ((PyObject *)__pyx_kp_s_114));
     10915   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_114));
     10916@@ -20109,7 +20762,7 @@
     10917  *                 raise ValueError("p < 0")
     10918  */
     10919   __pyx_k_tuple_119 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_119)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3423; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10920-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_119));
     10921+  __Pyx_GOTREF(__pyx_k_tuple_119);
     10922   __Pyx_INCREF(((PyObject *)__pyx_kp_s_110));
     10923   PyTuple_SET_ITEM(__pyx_k_tuple_119, 0, ((PyObject *)__pyx_kp_s_110));
     10924   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_110));
     10925@@ -20123,7 +20776,7 @@
     10926  *                 raise ValueError("p > 1")
     10927  */
     10928   __pyx_k_tuple_120 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_120)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3425; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10929-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_120));
     10930+  __Pyx_GOTREF(__pyx_k_tuple_120);
     10931   __Pyx_INCREF(((PyObject *)__pyx_kp_s_112));
     10932   PyTuple_SET_ITEM(__pyx_k_tuple_120, 0, ((PyObject *)__pyx_kp_s_112));
     10933   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_112));
     10934@@ -20137,7 +20790,7 @@
     10935  *                                    size, fn, fp)
     10936  */
     10937   __pyx_k_tuple_121 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_121)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3427; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10938-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_121));
     10939+  __Pyx_GOTREF(__pyx_k_tuple_121);
     10940   __Pyx_INCREF(((PyObject *)__pyx_kp_s_114));
     10941   PyTuple_SET_ITEM(__pyx_k_tuple_121, 0, ((PyObject *)__pyx_kp_s_114));
     10942   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_114));
     10943@@ -20151,7 +20804,7 @@
     10944  *             raise ValueError("p < 0")
     10945  */
     10946   __pyx_k_tuple_122 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_122)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3436; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10947-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_122));
     10948+  __Pyx_GOTREF(__pyx_k_tuple_122);
     10949   __Pyx_INCREF(((PyObject *)__pyx_kp_s_110));
     10950   PyTuple_SET_ITEM(__pyx_k_tuple_122, 0, ((PyObject *)__pyx_kp_s_110));
     10951   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_110));
     10952@@ -20165,7 +20818,7 @@
     10953  *             raise ValueError("p > 1")
     10954  */
     10955   __pyx_k_tuple_123 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_123)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3438; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10956-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_123));
     10957+  __Pyx_GOTREF(__pyx_k_tuple_123);
     10958   __Pyx_INCREF(((PyObject *)__pyx_kp_s_112));
     10959   PyTuple_SET_ITEM(__pyx_k_tuple_123, 0, ((PyObject *)__pyx_kp_s_112));
     10960   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_112));
     10961@@ -20179,7 +20832,7 @@
     10962  *                             on, op)
     10963  */
     10964   __pyx_k_tuple_124 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_124)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3440; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10965-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_124));
     10966+  __Pyx_GOTREF(__pyx_k_tuple_124);
     10967   __Pyx_INCREF(((PyObject *)__pyx_kp_s_114));
     10968   PyTuple_SET_ITEM(__pyx_k_tuple_124, 0, ((PyObject *)__pyx_kp_s_114));
     10969   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_114));
     10970@@ -20193,7 +20846,7 @@
     10971  *                 raise ValueError("lam value too large")
     10972  */
     10973   __pyx_k_tuple_127 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_127)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3501; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10974-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_127));
     10975+  __Pyx_GOTREF(__pyx_k_tuple_127);
     10976   __Pyx_INCREF(((PyObject *)__pyx_kp_s_126));
     10977   PyTuple_SET_ITEM(__pyx_k_tuple_127, 0, ((PyObject *)__pyx_kp_s_126));
     10978   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_126));
     10979@@ -20207,7 +20860,7 @@
     10980  *
     10981  */
     10982   __pyx_k_tuple_129 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_129)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3503; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10983-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_129));
     10984+  __Pyx_GOTREF(__pyx_k_tuple_129);
     10985   __Pyx_INCREF(((PyObject *)__pyx_kp_s_128));
     10986   PyTuple_SET_ITEM(__pyx_k_tuple_129, 0, ((PyObject *)__pyx_kp_s_128));
     10987   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_128));
     10988@@ -20221,7 +20874,7 @@
     10989  *             raise ValueError("lam value too large.")
     10990  */
     10991   __pyx_k_tuple_130 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_130)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3510; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     10992-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_130));
     10993+  __Pyx_GOTREF(__pyx_k_tuple_130);
     10994   __Pyx_INCREF(((PyObject *)__pyx_kp_s_126));
     10995   PyTuple_SET_ITEM(__pyx_k_tuple_130, 0, ((PyObject *)__pyx_kp_s_126));
     10996   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_126));
     10997@@ -20235,7 +20888,7 @@
     10998  *
     10999  */
     11000   __pyx_k_tuple_132 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_132)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3512; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11001-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_132));
     11002+  __Pyx_GOTREF(__pyx_k_tuple_132);
     11003   __Pyx_INCREF(((PyObject *)__pyx_kp_s_131));
     11004   PyTuple_SET_ITEM(__pyx_k_tuple_132, 0, ((PyObject *)__pyx_kp_s_131));
     11005   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_131));
     11006@@ -20249,7 +20902,7 @@
     11007  *
     11008  */
     11009   __pyx_k_tuple_134 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_134)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3593; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11010-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_134));
     11011+  __Pyx_GOTREF(__pyx_k_tuple_134);
     11012   __Pyx_INCREF(((PyObject *)__pyx_kp_s_133));
     11013   PyTuple_SET_ITEM(__pyx_k_tuple_134, 0, ((PyObject *)__pyx_kp_s_133));
     11014   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_133));
     11015@@ -20263,7 +20916,7 @@
     11016  *
     11017  */
     11018   __pyx_k_tuple_135 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_135)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3600; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11019-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_135));
     11020+  __Pyx_GOTREF(__pyx_k_tuple_135);
     11021   __Pyx_INCREF(((PyObject *)__pyx_kp_s_133));
     11022   PyTuple_SET_ITEM(__pyx_k_tuple_135, 0, ((PyObject *)__pyx_kp_s_133));
     11023   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_133));
     11024@@ -20277,7 +20930,7 @@
     11025  *                 raise ValueError("p > 1.0")
     11026  */
     11027   __pyx_k_tuple_137 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_137)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3654; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11028-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_137));
     11029+  __Pyx_GOTREF(__pyx_k_tuple_137);
     11030   __Pyx_INCREF(((PyObject *)__pyx_kp_s_136));
     11031   PyTuple_SET_ITEM(__pyx_k_tuple_137, 0, ((PyObject *)__pyx_kp_s_136));
     11032   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_136));
     11033@@ -20291,7 +20944,7 @@
     11034  *
     11035  */
     11036   __pyx_k_tuple_139 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_139)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3656; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11037-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_139));
     11038+  __Pyx_GOTREF(__pyx_k_tuple_139);
     11039   __Pyx_INCREF(((PyObject *)__pyx_kp_s_138));
     11040   PyTuple_SET_ITEM(__pyx_k_tuple_139, 0, ((PyObject *)__pyx_kp_s_138));
     11041   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_138));
     11042@@ -20305,7 +20958,7 @@
     11043  *             raise ValueError("p > 1.0")
     11044  */
     11045   __pyx_k_tuple_140 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_140)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3664; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11046-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_140));
     11047+  __Pyx_GOTREF(__pyx_k_tuple_140);
     11048   __Pyx_INCREF(((PyObject *)__pyx_kp_s_136));
     11049   PyTuple_SET_ITEM(__pyx_k_tuple_140, 0, ((PyObject *)__pyx_kp_s_136));
     11050   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_136));
     11051@@ -20319,7 +20972,7 @@
     11052  *
     11053  */
     11054   __pyx_k_tuple_141 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_141)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3666; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11055-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_141));
     11056+  __Pyx_GOTREF(__pyx_k_tuple_141);
     11057   __Pyx_INCREF(((PyObject *)__pyx_kp_s_138));
     11058   PyTuple_SET_ITEM(__pyx_k_tuple_141, 0, ((PyObject *)__pyx_kp_s_138));
     11059   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_138));
     11060@@ -20333,7 +20986,7 @@
     11061  *                 raise ValueError("nbad < 1")
     11062  */
     11063   __pyx_k_tuple_143 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_143)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3761; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11064-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_143));
     11065+  __Pyx_GOTREF(__pyx_k_tuple_143);
     11066   __Pyx_INCREF(((PyObject *)__pyx_kp_s_142));
     11067   PyTuple_SET_ITEM(__pyx_k_tuple_143, 0, ((PyObject *)__pyx_kp_s_142));
     11068   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_142));
     11069@@ -20347,7 +21000,7 @@
     11070  *                 raise ValueError("nsample < 1")
     11071  */
     11072   __pyx_k_tuple_145 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_145)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3763; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11073-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_145));
     11074+  __Pyx_GOTREF(__pyx_k_tuple_145);
     11075   __Pyx_INCREF(((PyObject *)__pyx_kp_s_144));
     11076   PyTuple_SET_ITEM(__pyx_k_tuple_145, 0, ((PyObject *)__pyx_kp_s_144));
     11077   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_144));
     11078@@ -20361,7 +21014,7 @@
     11079  *                 raise ValueError("ngood + nbad < nsample")
     11080  */
     11081   __pyx_k_tuple_147 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_147)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3765; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11082-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_147));
     11083+  __Pyx_GOTREF(__pyx_k_tuple_147);
     11084   __Pyx_INCREF(((PyObject *)__pyx_kp_s_146));
     11085   PyTuple_SET_ITEM(__pyx_k_tuple_147, 0, ((PyObject *)__pyx_kp_s_146));
     11086   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_146));
     11087@@ -20375,7 +21028,7 @@
     11088  *                                     lngood, lnbad, lnsample)
     11089  */
     11090   __pyx_k_tuple_149 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_149)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3767; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11091-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_149));
     11092+  __Pyx_GOTREF(__pyx_k_tuple_149);
     11093   __Pyx_INCREF(((PyObject *)__pyx_kp_s_148));
     11094   PyTuple_SET_ITEM(__pyx_k_tuple_149, 0, ((PyObject *)__pyx_kp_s_148));
     11095   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_148));
     11096@@ -20389,7 +21042,7 @@
     11097  *             raise ValueError("nbad < 1")
     11098  */
     11099   __pyx_k_tuple_150 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_150)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3778; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11100-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_150));
     11101+  __Pyx_GOTREF(__pyx_k_tuple_150);
     11102   __Pyx_INCREF(((PyObject *)__pyx_kp_s_142));
     11103   PyTuple_SET_ITEM(__pyx_k_tuple_150, 0, ((PyObject *)__pyx_kp_s_142));
     11104   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_142));
     11105@@ -20403,7 +21056,7 @@
     11106  *             raise ValueError("nsample < 1")
     11107  */
     11108   __pyx_k_tuple_151 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_151)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3780; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11109-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_151));
     11110+  __Pyx_GOTREF(__pyx_k_tuple_151);
     11111   __Pyx_INCREF(((PyObject *)__pyx_kp_s_144));
     11112   PyTuple_SET_ITEM(__pyx_k_tuple_151, 0, ((PyObject *)__pyx_kp_s_144));
     11113   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_144));
     11114@@ -20417,7 +21070,7 @@
     11115  *             raise ValueError("ngood + nbad < nsample")
     11116  */
     11117   __pyx_k_tuple_152 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_152)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3782; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11118-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_152));
     11119+  __Pyx_GOTREF(__pyx_k_tuple_152);
     11120   __Pyx_INCREF(((PyObject *)__pyx_kp_s_146));
     11121   PyTuple_SET_ITEM(__pyx_k_tuple_152, 0, ((PyObject *)__pyx_kp_s_146));
     11122   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_146));
     11123@@ -20431,7 +21084,7 @@
     11124  *             ongood, onbad, onsample)
     11125  */
     11126   __pyx_k_tuple_153 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_153)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3784; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11127-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_153));
     11128+  __Pyx_GOTREF(__pyx_k_tuple_153);
     11129   __Pyx_INCREF(((PyObject *)__pyx_kp_s_148));
     11130   PyTuple_SET_ITEM(__pyx_k_tuple_153, 0, ((PyObject *)__pyx_kp_s_148));
     11131   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_148));
     11132@@ -20445,7 +21098,7 @@
     11133  *                 raise ValueError("p >= 1.0")
     11134  */
     11135   __pyx_k_tuple_155 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_155)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3868; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11136-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_155));
     11137+  __Pyx_GOTREF(__pyx_k_tuple_155);
     11138   __Pyx_INCREF(((PyObject *)__pyx_kp_s_154));
     11139   PyTuple_SET_ITEM(__pyx_k_tuple_155, 0, ((PyObject *)__pyx_kp_s_154));
     11140   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_154));
     11141@@ -20459,7 +21112,7 @@
     11142  *
     11143  */
     11144   __pyx_k_tuple_157 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_157)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3870; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11145-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_157));
     11146+  __Pyx_GOTREF(__pyx_k_tuple_157);
     11147   __Pyx_INCREF(((PyObject *)__pyx_kp_s_156));
     11148   PyTuple_SET_ITEM(__pyx_k_tuple_157, 0, ((PyObject *)__pyx_kp_s_156));
     11149   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_156));
     11150@@ -20473,7 +21126,7 @@
     11151  *             raise ValueError("p >= 1.0")
     11152  */
     11153   __pyx_k_tuple_158 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_158)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3877; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11154-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_158));
     11155+  __Pyx_GOTREF(__pyx_k_tuple_158);
     11156   __Pyx_INCREF(((PyObject *)__pyx_kp_s_154));
     11157   PyTuple_SET_ITEM(__pyx_k_tuple_158, 0, ((PyObject *)__pyx_kp_s_154));
     11158   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_154));
     11159@@ -20487,7 +21140,7 @@
     11160  *
     11161  */
     11162   __pyx_k_tuple_159 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_159)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3879; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11163-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_159));
     11164+  __Pyx_GOTREF(__pyx_k_tuple_159);
     11165   __Pyx_INCREF(((PyObject *)__pyx_kp_s_156));
     11166   PyTuple_SET_ITEM(__pyx_k_tuple_159, 0, ((PyObject *)__pyx_kp_s_156));
     11167   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_156));
     11168@@ -20501,7 +21154,7 @@
     11169  *                raise ValueError("cov must be 2 dimensional and square")
     11170  */
     11171   __pyx_k_tuple_161 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_161)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3982; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11172-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_161));
     11173+  __Pyx_GOTREF(__pyx_k_tuple_161);
     11174   __Pyx_INCREF(((PyObject *)__pyx_kp_s_160));
     11175   PyTuple_SET_ITEM(__pyx_k_tuple_161, 0, ((PyObject *)__pyx_kp_s_160));
     11176   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_160));
     11177@@ -20515,7 +21168,7 @@
     11178  *                raise ValueError("mean and cov must have same length")
     11179  */
     11180   __pyx_k_tuple_163 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_163)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3984; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11181-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_163));
     11182+  __Pyx_GOTREF(__pyx_k_tuple_163);
     11183   __Pyx_INCREF(((PyObject *)__pyx_kp_s_162));
     11184   PyTuple_SET_ITEM(__pyx_k_tuple_163, 0, ((PyObject *)__pyx_kp_s_162));
     11185   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_162));
     11186@@ -20529,7 +21182,7 @@
     11187  *         if isinstance(shape, int):
     11188  */
     11189   __pyx_k_tuple_165 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_165)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 3986; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11190-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_165));
     11191+  __Pyx_GOTREF(__pyx_k_tuple_165);
     11192   __Pyx_INCREF(((PyObject *)__pyx_kp_s_164));
     11193   PyTuple_SET_ITEM(__pyx_k_tuple_165, 0, ((PyObject *)__pyx_kp_s_164));
     11194   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_164));
     11195@@ -20543,7 +21196,7 @@
     11196  *         if size is None:
     11197  */
     11198   __pyx_k_tuple_168 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_168)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 4079; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11199-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_168));
     11200+  __Pyx_GOTREF(__pyx_k_tuple_168);
     11201   __Pyx_INCREF(((PyObject *)__pyx_kp_s_167));
     11202   PyTuple_SET_ITEM(__pyx_k_tuple_168, 0, ((PyObject *)__pyx_kp_s_167));
     11203   __Pyx_GIVEREF(((PyObject *)__pyx_kp_s_167));
     11204@@ -20557,13 +21210,13 @@
     11205  *     def __init__(self, seed=None):
     11206  */
     11207   __pyx_k_tuple_169 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_169)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 556; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11208-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_169));
     11209+  __Pyx_GOTREF(__pyx_k_tuple_169);
     11210   __Pyx_INCREF(((PyObject *)__pyx_n_s__l));
     11211   PyTuple_SET_ITEM(__pyx_k_tuple_169, 0, ((PyObject *)__pyx_n_s__l));
     11212   __Pyx_GIVEREF(((PyObject *)__pyx_n_s__l));
     11213   __Pyx_GIVEREF(((PyObject *)__pyx_k_tuple_169));
     11214   __pyx_k_tuple_170 = PyTuple_New(1); if (unlikely(!__pyx_k_tuple_170)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 556; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11215-  __Pyx_GOTREF(((PyObject *)__pyx_k_tuple_170));
     11216+  __Pyx_GOTREF(__pyx_k_tuple_170);
     11217   __Pyx_INCREF(((PyObject *)__pyx_n_s__l));
     11218   PyTuple_SET_ITEM(__pyx_k_tuple_170, 0, ((PyObject *)__pyx_n_s__l));
     11219   __Pyx_GIVEREF(((PyObject *)__pyx_n_s__l));
     11220@@ -20608,12 +21261,18 @@
     11221           Py_FatalError("failed to import 'refnanny' module");
     11222   }
     11223   #endif
     11224-  __Pyx_RefNannySetupContext("PyMODINIT_FUNC PyInit_mtrand(void)");
     11225+  __Pyx_RefNannySetupContext("PyMODINIT_FUNC PyInit_mtrand(void)", 0);
     11226   if ( __Pyx_check_binary_version() < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11227   __pyx_empty_tuple = PyTuple_New(0); if (unlikely(!__pyx_empty_tuple)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11228   __pyx_empty_bytes = PyBytes_FromStringAndSize("", 0); if (unlikely(!__pyx_empty_bytes)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11229-  #ifdef __pyx_binding_PyCFunctionType_USED
     11230-  if (__pyx_binding_PyCFunctionType_init() < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11231+  #ifdef __Pyx_CyFunction_USED
     11232+  if (__Pyx_CyFunction_init() < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11233+  #endif
     11234+  #ifdef __Pyx_FusedFunction_USED
     11235+  if (__pyx_FusedFunction_init() < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11236+  #endif
     11237+  #ifdef __Pyx_Generator_USED
     11238+  if (__pyx_Generator_init() < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11239   #endif
     11240   /*--- Library function declarations ---*/
     11241   /*--- Threads initialization code ---*/
     11242@@ -20624,16 +21283,15 @@
     11243   #endif
     11244   /*--- Module creation code ---*/
     11245   #if PY_MAJOR_VERSION < 3
     11246-  __pyx_m = Py_InitModule4(__Pyx_NAMESTR("mtrand"), __pyx_methods, 0, 0, PYTHON_API_VERSION);
     11247+  __pyx_m = Py_InitModule4(__Pyx_NAMESTR("mtrand"), __pyx_methods, 0, 0, PYTHON_API_VERSION); Py_XINCREF(__pyx_m);
     11248   #else
     11249   __pyx_m = PyModule_Create(&__pyx_moduledef);
     11250   #endif
     11251-  if (!__pyx_m) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;};
     11252-  #if PY_MAJOR_VERSION < 3
     11253-  Py_INCREF(__pyx_m);
     11254+  if (unlikely(!__pyx_m)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11255+  __pyx_b = PyImport_AddModule(__Pyx_NAMESTR(__Pyx_BUILTIN_MODULE_NAME)); if (unlikely(!__pyx_b)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11256+  #if CYTHON_COMPILING_IN_PYPY
     11257+  Py_INCREF(__pyx_b);
     11258   #endif
     11259-  __pyx_b = PyImport_AddModule(__Pyx_NAMESTR(__Pyx_BUILTIN_MODULE_NAME));
     11260-  if (!__pyx_b) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;};
     11261   if (__Pyx_SetAttrString(__pyx_m, "__builtins__", __pyx_b) < 0) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;};
     11262   /*--- Initialize various global constants etc. ---*/
     11263   if (unlikely(__Pyx_InitGlobals() < 0)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 1; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11264@@ -20716,7 +21374,7 @@
     11265   __Pyx_GOTREF(__pyx_t_4);
     11266   __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
     11267   __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) {__pyx_filename = __pyx_f[0]; __pyx_lineno = 556; __pyx_clineno = __LINE__; goto __pyx_L1_error;}
     11268-  __Pyx_GOTREF(((PyObject *)__pyx_t_1));
     11269+  __Pyx_GOTREF(__pyx_t_1);
     11270   PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_t_4);
     11271   __Pyx_GIVEREF(__pyx_t_4);
     11272   __pyx_t_4 = 0;
     11273@@ -21667,7 +22325,6 @@
     11274 }
     11275 
     11276 /* Runtime support code */
     11277-
     11278 #if CYTHON_REFNANNY
     11279 static __Pyx_RefNannyAPIStruct *__Pyx_RefNannyImportAPI(const char *modname) {
     11280     PyObject *m = NULL, *p = NULL;
     11281@@ -21700,9 +22357,9 @@
     11282 }
     11283 
     11284 static CYTHON_INLINE void __Pyx_ErrRestore(PyObject *type, PyObject *value, PyObject *tb) {
     11285+#if CYTHON_COMPILING_IN_CPYTHON
     11286     PyObject *tmp_type, *tmp_value, *tmp_tb;
     11287     PyThreadState *tstate = PyThreadState_GET();
     11288-
     11289     tmp_type = tstate->curexc_type;
     11290     tmp_value = tstate->curexc_value;
     11291     tmp_tb = tstate->curexc_traceback;
     11292@@ -21712,55 +22369,60 @@
     11293     Py_XDECREF(tmp_type);
     11294     Py_XDECREF(tmp_value);
     11295     Py_XDECREF(tmp_tb);
     11296+#else
     11297+    PyErr_Restore(type, value, tb);
     11298+#endif
     11299 }
     11300-
     11301 static CYTHON_INLINE void __Pyx_ErrFetch(PyObject **type, PyObject **value, PyObject **tb) {
     11302+#if CYTHON_COMPILING_IN_CPYTHON
     11303     PyThreadState *tstate = PyThreadState_GET();
     11304     *type = tstate->curexc_type;
     11305     *value = tstate->curexc_value;
     11306     *tb = tstate->curexc_traceback;
     11307-
     11308     tstate->curexc_type = 0;
     11309     tstate->curexc_value = 0;
     11310     tstate->curexc_traceback = 0;
     11311+#else
     11312+    PyErr_Fetch(type, value, tb);
     11313+#endif
     11314 }
     11315 
     11316-
     11317 #if PY_MAJOR_VERSION < 3
     11318-static void __Pyx_Raise(PyObject *type, PyObject *value, PyObject *tb, PyObject *cause) {
     11319-    /* cause is unused */
     11320+static void __Pyx_Raise(PyObject *type, PyObject *value, PyObject *tb,
     11321+                        CYTHON_UNUSED PyObject *cause) {
     11322     Py_XINCREF(type);
     11323-    Py_XINCREF(value);
     11324-    Py_XINCREF(tb);
     11325-    /* First, check the traceback argument, replacing None with NULL. */
     11326-    if (tb == Py_None) {
     11327-        Py_DECREF(tb);
     11328-        tb = 0;
     11329-    }
     11330-    else if (tb != NULL && !PyTraceBack_Check(tb)) {
     11331-        PyErr_SetString(PyExc_TypeError,
     11332-            "raise: arg 3 must be a traceback or None");
     11333-        goto raise_error;
     11334-    }
     11335-    /* Next, replace a missing value with None */
     11336-    if (value == NULL) {
     11337-        value = Py_None;
     11338+    if (!value || value == Py_None)
     11339+        value = NULL;
     11340+    else
     11341         Py_INCREF(value);
     11342+    if (!tb || tb == Py_None)
     11343+        tb = NULL;
     11344+    else {
     11345+        Py_INCREF(tb);
     11346+        if (!PyTraceBack_Check(tb)) {
     11347+            PyErr_SetString(PyExc_TypeError,
     11348+                "raise: arg 3 must be a traceback or None");
     11349+            goto raise_error;
     11350+        }
     11351     }
     11352     #if PY_VERSION_HEX < 0x02050000
     11353-    if (!PyClass_Check(type))
     11354+    if (PyClass_Check(type)) {
     11355     #else
     11356-    if (!PyType_Check(type))
     11357+    if (PyType_Check(type)) {
     11358     #endif
     11359-    {
     11360-        /* Raising an instance.  The value should be a dummy. */
     11361-        if (value != Py_None) {
     11362+#if CYTHON_COMPILING_IN_PYPY
     11363+        if (!value) {
     11364+            Py_INCREF(Py_None);
     11365+            value = Py_None;
     11366+        }
     11367+#endif
     11368+        PyErr_NormalizeException(&type, &value, &tb);
     11369+    } else {
     11370+        if (value) {
     11371             PyErr_SetString(PyExc_TypeError,
     11372                 "instance exception may not have a separate value");
     11373             goto raise_error;
     11374         }
     11375-        /* Normalize to raise <class>, <instance> */
     11376-        Py_DECREF(value);
     11377         value = type;
     11378         #if PY_VERSION_HEX < 0x02050000
     11379             if (PyInstance_Check(type)) {
     11380@@ -21783,7 +22445,6 @@
     11381             }
     11382         #endif
     11383     }
     11384-
     11385     __Pyx_ErrRestore(type, value, tb);
     11386     return;
     11387 raise_error:
     11388@@ -21792,10 +22453,9 @@
     11389     Py_XDECREF(tb);
     11390     return;
     11391 }
     11392-
     11393 #else /* Python 3+ */
     11394-
     11395 static void __Pyx_Raise(PyObject *type, PyObject *value, PyObject *tb, PyObject *cause) {
     11396+    PyObject* owned_instance = NULL;
     11397     if (tb == Py_None) {
     11398         tb = 0;
     11399     } else if (tb && !PyTraceBack_Check(tb)) {
     11400@@ -21805,7 +22465,6 @@
     11401     }
     11402     if (value == Py_None)
     11403         value = 0;
     11404-
     11405     if (PyExceptionInstance_Check(type)) {
     11406         if (value) {
     11407             PyErr_SetString(PyExc_TypeError,
     11408@@ -21814,13 +22473,36 @@
     11409         }
     11410         value = type;
     11411         type = (PyObject*) Py_TYPE(value);
     11412-    } else if (!PyExceptionClass_Check(type)) {
     11413+    } else if (PyExceptionClass_Check(type)) {
     11414+        PyObject *args;
     11415+        if (!value)
     11416+            args = PyTuple_New(0);
     11417+        else if (PyTuple_Check(value)) {
     11418+            Py_INCREF(value);
     11419+            args = value;
     11420+        }
     11421+        else
     11422+            args = PyTuple_Pack(1, value);
     11423+        if (!args)
     11424+            goto bad;
     11425+        owned_instance = PyEval_CallObject(type, args);
     11426+        Py_DECREF(args);
     11427+        if (!owned_instance)
     11428+            goto bad;
     11429+        value = owned_instance;
     11430+        if (!PyExceptionInstance_Check(value)) {
     11431+            PyErr_Format(PyExc_TypeError,
     11432+                         "calling %R should have returned an instance of "
     11433+                         "BaseException, not %R",
     11434+                         type, Py_TYPE(value));
     11435+            goto bad;
     11436+        }
     11437+    } else {
     11438         PyErr_SetString(PyExc_TypeError,
     11439             "raise: exception class must be a subclass of BaseException");
     11440         goto bad;
     11441     }
     11442-
     11443-    if (cause) {
     11444+    if (cause && cause != Py_None) {
     11445         PyObject *fixed_cause;
     11446         if (PyExceptionClass_Check(cause)) {
     11447             fixed_cause = PyObject_CallObject(cause, NULL);
     11448@@ -21837,14 +22519,9 @@
     11449                             "BaseException");
     11450             goto bad;
     11451         }
     11452-        if (!value) {
     11453-            value = PyObject_CallObject(type, NULL);
     11454-        }
     11455         PyException_SetCause(value, fixed_cause);
     11456     }
     11457-
     11458     PyErr_SetObject(type, value);
     11459-
     11460     if (tb) {
     11461         PyThreadState *tstate = PyThreadState_GET();
     11462         PyObject* tmp_tb = tstate->curexc_traceback;
     11463@@ -21854,8 +22531,8 @@
     11464             Py_XDECREF(tmp_tb);
     11465         }
     11466     }
     11467-
     11468 bad:
     11469+    Py_XDECREF(owned_instance);
     11470     return;
     11471 }
     11472 #endif
     11473@@ -21869,7 +22546,7 @@
     11474         "%s() got multiple values for keyword argument '%U'", func_name, kw_name);
     11475         #else
     11476         "%s() got multiple values for keyword argument '%s'", func_name,
     11477-        PyString_AS_STRING(kw_name));
     11478+        PyString_AsString(kw_name));
     11479         #endif
     11480 }
     11481 
     11482@@ -21885,55 +22562,77 @@
     11483     Py_ssize_t pos = 0;
     11484     PyObject*** name;
     11485     PyObject*** first_kw_arg = argnames + num_pos_args;
     11486-
     11487     while (PyDict_Next(kwds, &pos, &key, &value)) {
     11488         name = first_kw_arg;
     11489         while (*name && (**name != key)) name++;
     11490         if (*name) {
     11491             values[name-argnames] = value;
     11492-        } else {
     11493-            #if PY_MAJOR_VERSION < 3
     11494-            if (unlikely(!PyString_CheckExact(key)) && unlikely(!PyString_Check(key))) {
     11495-            #else
     11496-            if (unlikely(!PyUnicode_CheckExact(key)) && unlikely(!PyUnicode_Check(key))) {
     11497-            #endif
     11498-                goto invalid_keyword_type;
     11499-            } else {
     11500-                for (name = first_kw_arg; *name; name++) {
     11501-                    #if PY_MAJOR_VERSION >= 3
     11502-                    if (PyUnicode_GET_SIZE(**name) == PyUnicode_GET_SIZE(key) &&
     11503-                        PyUnicode_Compare(**name, key) == 0) break;
     11504-                    #else
     11505-                    if (PyString_GET_SIZE(**name) == PyString_GET_SIZE(key) &&
     11506-                        _PyString_Eq(**name, key)) break;
     11507-                    #endif
     11508-                }
     11509-                if (*name) {
     11510+            continue;
     11511+        }
     11512+        name = first_kw_arg;
     11513+        #if PY_MAJOR_VERSION < 3
     11514+        if (likely(PyString_CheckExact(key)) || likely(PyString_Check(key))) {
     11515+            while (*name) {
     11516+                if ((CYTHON_COMPILING_IN_PYPY || PyString_GET_SIZE(**name) == PyString_GET_SIZE(key))
     11517+                        && _PyString_Eq(**name, key)) {
     11518                     values[name-argnames] = value;
     11519-                } else {
     11520-                    /* unexpected keyword found */
     11521-                    for (name=argnames; name != first_kw_arg; name++) {
     11522-                        if (**name == key) goto arg_passed_twice;
     11523-                        #if PY_MAJOR_VERSION >= 3
     11524-                        if (PyUnicode_GET_SIZE(**name) == PyUnicode_GET_SIZE(key) &&
     11525-                            PyUnicode_Compare(**name, key) == 0) goto arg_passed_twice;
     11526-                        #else
     11527-                        if (PyString_GET_SIZE(**name) == PyString_GET_SIZE(key) &&
     11528-                            _PyString_Eq(**name, key)) goto arg_passed_twice;
     11529-                        #endif
     11530-                    }
     11531-                    if (kwds2) {
     11532-                        if (unlikely(PyDict_SetItem(kwds2, key, value))) goto bad;
     11533-                    } else {
     11534-                        goto invalid_keyword;
     11535+                    break;
     11536+                }
     11537+                name++;
     11538+            }
     11539+            if (*name) continue;
     11540+            else {
     11541+                PyObject*** argname = argnames;
     11542+                while (argname != first_kw_arg) {
     11543+                    if ((**argname == key) || (
     11544+                            (CYTHON_COMPILING_IN_PYPY || PyString_GET_SIZE(**argname) == PyString_GET_SIZE(key))
     11545+                             && _PyString_Eq(**argname, key))) {
     11546+                        goto arg_passed_twice;
     11547                     }
     11548+                    argname++;
     11549                 }
     11550             }
     11551+        } else
     11552+        #endif
     11553+        if (likely(PyUnicode_Check(key))) {
     11554+            while (*name) {
     11555+                int cmp = (**name == key) ? 0 :
     11556+                #if !CYTHON_COMPILING_IN_PYPY && PY_MAJOR_VERSION >= 3
     11557+                    (PyUnicode_GET_SIZE(**name) != PyUnicode_GET_SIZE(key)) ? 1 :
     11558+                #endif
     11559+                    PyUnicode_Compare(**name, key);
     11560+                if (cmp < 0 && unlikely(PyErr_Occurred())) goto bad;
     11561+                if (cmp == 0) {
     11562+                    values[name-argnames] = value;
     11563+                    break;
     11564+                }
     11565+                name++;
     11566+            }
     11567+            if (*name) continue;
     11568+            else {
     11569+                PyObject*** argname = argnames;
     11570+                while (argname != first_kw_arg) {
     11571+                    int cmp = (**argname == key) ? 0 :
     11572+                    #if !CYTHON_COMPILING_IN_PYPY && PY_MAJOR_VERSION >= 3
     11573+                        (PyUnicode_GET_SIZE(**argname) != PyUnicode_GET_SIZE(key)) ? 1 :
     11574+                    #endif
     11575+                        PyUnicode_Compare(**argname, key);
     11576+                    if (cmp < 0 && unlikely(PyErr_Occurred())) goto bad;
     11577+                    if (cmp == 0) goto arg_passed_twice;
     11578+                    argname++;
     11579+                }
     11580+            }
     11581+        } else
     11582+            goto invalid_keyword_type;
     11583+        if (kwds2) {
     11584+            if (unlikely(PyDict_SetItem(kwds2, key, value))) goto bad;
     11585+        } else {
     11586+            goto invalid_keyword;
     11587         }
     11588     }
     11589     return 0;
     11590 arg_passed_twice:
     11591-    __Pyx_RaiseDoubleKeywordsError(function_name, **name);
     11592+    __Pyx_RaiseDoubleKeywordsError(function_name, key);
     11593     goto bad;
     11594 invalid_keyword_type:
     11595     PyErr_Format(PyExc_TypeError,
     11596@@ -21961,7 +22660,6 @@
     11597 {
     11598     Py_ssize_t num_expected;
     11599     const char *more_or_less;
     11600-
     11601     if (num_found < num_min) {
     11602         num_expected = num_min;
     11603         more_or_less = "at least";
     11604@@ -21973,21 +22671,54 @@
     11605         more_or_less = "exactly";
     11606     }
     11607     PyErr_Format(PyExc_TypeError,
     11608-                 "%s() takes %s %"PY_FORMAT_SIZE_T"d positional argument%s (%"PY_FORMAT_SIZE_T"d given)",
     11609+                 "%s() takes %s %" CYTHON_FORMAT_SSIZE_T "d positional argument%s (%" CYTHON_FORMAT_SSIZE_T "d given)",
     11610                  func_name, more_or_less, num_expected,
     11611                  (num_expected == 1) ? "" : "s", num_found);
     11612 }
     11613 
     11614+static CYTHON_INLINE void __Pyx_RaiseTooManyValuesError(Py_ssize_t expected) {
     11615+    PyErr_Format(PyExc_ValueError,
     11616+                 "too many values to unpack (expected %" CYTHON_FORMAT_SSIZE_T "d)", expected);
     11617+}
     11618 
     11619 static CYTHON_INLINE void __Pyx_RaiseNeedMoreValuesError(Py_ssize_t index) {
     11620     PyErr_Format(PyExc_ValueError,
     11621-                 "need more than %"PY_FORMAT_SIZE_T"d value%s to unpack",
     11622+                 "need more than %" CYTHON_FORMAT_SSIZE_T "d value%s to unpack",
     11623                  index, (index == 1) ? "" : "s");
     11624 }
     11625 
     11626-static CYTHON_INLINE void __Pyx_RaiseTooManyValuesError(Py_ssize_t expected) {
     11627-    PyErr_Format(PyExc_ValueError,
     11628-                 "too many values to unpack (expected %"PY_FORMAT_SIZE_T"d)", expected);
     11629+static CYTHON_INLINE int __Pyx_IterFinish(void) {
     11630+#if CYTHON_COMPILING_IN_CPYTHON
     11631+    PyThreadState *tstate = PyThreadState_GET();
     11632+    PyObject* exc_type = tstate->curexc_type;
     11633+    if (unlikely(exc_type)) {
     11634+        if (likely(exc_type == PyExc_StopIteration) || PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration)) {
     11635+            PyObject *exc_value, *exc_tb;
     11636+            exc_value = tstate->curexc_value;
     11637+            exc_tb = tstate->curexc_traceback;
     11638+            tstate->curexc_type = 0;
     11639+            tstate->curexc_value = 0;
     11640+            tstate->curexc_traceback = 0;
     11641+            Py_DECREF(exc_type);
     11642+            Py_XDECREF(exc_value);
     11643+            Py_XDECREF(exc_tb);
     11644+            return 0;
     11645+        } else {
     11646+            return -1;
     11647+        }
     11648+    }
     11649+    return 0;
     11650+#else
     11651+    if (unlikely(PyErr_Occurred())) {
     11652+        if (likely(PyErr_ExceptionMatches(PyExc_StopIteration))) {
     11653+            PyErr_Clear();
     11654+            return 0;
     11655+        } else {
     11656+            return -1;
     11657+        }
     11658+    }
     11659+    return 0;
     11660+#endif
     11661 }
     11662 
     11663 static int __Pyx_IternextUnpackEndCheck(PyObject *retval, Py_ssize_t expected) {
     11664@@ -21995,19 +22726,15 @@
     11665         Py_DECREF(retval);
     11666         __Pyx_RaiseTooManyValuesError(expected);
     11667         return -1;
     11668-    } else if (PyErr_Occurred()) {
     11669-        if (likely(PyErr_ExceptionMatches(PyExc_StopIteration))) {
     11670-            PyErr_Clear();
     11671-            return 0;
     11672-        } else {
     11673-            return -1;
     11674-        }
     11675+    } else {
     11676+        return __Pyx_IterFinish();
     11677     }
     11678     return 0;
     11679 }
     11680 
     11681 static int __Pyx_GetException(PyObject **type, PyObject **value, PyObject **tb) {
     11682     PyObject *local_type, *local_value, *local_tb;
     11683+#if CYTHON_COMPILING_IN_CPYTHON
     11684     PyObject *tmp_type, *tmp_value, *tmp_tb;
     11685     PyThreadState *tstate = PyThreadState_GET();
     11686     local_type = tstate->curexc_type;
     11687@@ -22016,19 +22743,27 @@
     11688     tstate->curexc_type = 0;
     11689     tstate->curexc_value = 0;
     11690     tstate->curexc_traceback = 0;
     11691+#else
     11692+    PyErr_Fetch(&local_type, &local_value, &local_tb);
     11693+#endif
     11694     PyErr_NormalizeException(&local_type, &local_value, &local_tb);
     11695+#if CYTHON_COMPILING_IN_CPYTHON
     11696     if (unlikely(tstate->curexc_type))
     11697+#else
     11698+    if (unlikely(PyErr_Occurred()))
     11699+#endif
     11700         goto bad;
     11701     #if PY_MAJOR_VERSION >= 3
     11702     if (unlikely(PyException_SetTraceback(local_value, local_tb) < 0))
     11703         goto bad;
     11704     #endif
     11705-    *type = local_type;
     11706-    *value = local_value;
     11707-    *tb = local_tb;
     11708     Py_INCREF(local_type);
     11709     Py_INCREF(local_value);
     11710     Py_INCREF(local_tb);
     11711+    *type = local_type;
     11712+    *value = local_value;
     11713+    *tb = local_tb;
     11714+#if CYTHON_COMPILING_IN_CPYTHON
     11715     tmp_type = tstate->exc_type;
     11716     tmp_value = tstate->exc_value;
     11717     tmp_tb = tstate->exc_traceback;
     11718@@ -22036,10 +22771,13 @@
     11719     tstate->exc_value = local_value;
     11720     tstate->exc_traceback = local_tb;
     11721     /* Make sure tstate is in a consistent state when we XDECREF
     11722-       these objects (XDECREF may run arbitrary code). */
     11723+       these objects (DECREF may run arbitrary code). */
     11724     Py_XDECREF(tmp_type);
     11725     Py_XDECREF(tmp_value);
     11726     Py_XDECREF(tmp_tb);
     11727+#else
     11728+    PyErr_SetExcInfo(local_type, local_value, local_tb);
     11729+#endif
     11730     return 0;
     11731 bad:
     11732     *type = 0;
     11733@@ -22051,7 +22789,6 @@
     11734     return -1;
     11735 }
     11736 
     11737-
     11738 static CYTHON_INLINE int __Pyx_CheckKeywordStrings(
     11739     PyObject *kwdict,
     11740     const char* function_name,
     11741@@ -22059,13 +22796,17 @@
     11742 {
     11743     PyObject* key = 0;
     11744     Py_ssize_t pos = 0;
     11745+#if CPYTHON_COMPILING_IN_PYPY
     11746+    if (!kw_allowed && PyDict_Next(kwdict, &pos, &key, 0))
     11747+        goto invalid_keyword;
     11748+    return 1;
     11749+#else
     11750     while (PyDict_Next(kwdict, &pos, &key, 0)) {
     11751         #if PY_MAJOR_VERSION < 3
     11752         if (unlikely(!PyString_CheckExact(key)) && unlikely(!PyString_Check(key)))
     11753-        #else
     11754-        if (unlikely(!PyUnicode_CheckExact(key)) && unlikely(!PyUnicode_Check(key)))
     11755         #endif
     11756-            goto invalid_keyword_type;
     11757+            if (unlikely(!PyUnicode_Check(key)))
     11758+                goto invalid_keyword_type;
     11759     }
     11760     if ((!kw_allowed) && unlikely(key))
     11761         goto invalid_keyword;
     11762@@ -22074,6 +22815,7 @@
     11763     PyErr_Format(PyExc_TypeError,
     11764         "%s() keywords must be strings", function_name);
     11765     return 0;
     11766+#endif
     11767 invalid_keyword:
     11768     PyErr_Format(PyExc_TypeError,
     11769     #if PY_MAJOR_VERSION < 3
     11770@@ -22098,8 +22840,8 @@
     11771     return 0;
     11772 }
     11773 
     11774-
     11775 static CYTHON_INLINE void __Pyx_ExceptionSave(PyObject **type, PyObject **value, PyObject **tb) {
     11776+#if CYTHON_COMPILING_IN_CPYTHON
     11777     PyThreadState *tstate = PyThreadState_GET();
     11778     *type = tstate->exc_type;
     11779     *value = tstate->exc_value;
     11780@@ -22107,9 +22849,12 @@
     11781     Py_XINCREF(*type);
     11782     Py_XINCREF(*value);
     11783     Py_XINCREF(*tb);
     11784+#else
     11785+    PyErr_GetExcInfo(type, value, tb);
     11786+#endif
     11787 }
     11788-
     11789 static void __Pyx_ExceptionReset(PyObject *type, PyObject *value, PyObject *tb) {
     11790+#if CYTHON_COMPILING_IN_CPYTHON
     11791     PyObject *tmp_type, *tmp_value, *tmp_tb;
     11792     PyThreadState *tstate = PyThreadState_GET();
     11793     tmp_type = tstate->exc_type;
     11794@@ -22121,6 +22866,9 @@
     11795     Py_XDECREF(tmp_type);
     11796     Py_XDECREF(tmp_value);
     11797     Py_XDECREF(tmp_tb);
     11798+#else
     11799+    PyErr_SetExcInfo(type, value, tb);
     11800+#endif
     11801 }
     11802 
     11803 static PyObject *__Pyx_Import(PyObject *name, PyObject *from_list, long level) {
     11804@@ -22149,12 +22897,33 @@
     11805         goto bad;
     11806     #if PY_VERSION_HEX >= 0x02050000
     11807     {
     11808-        PyObject *py_level = PyInt_FromLong(level);
     11809-        if (!py_level)
     11810-            goto bad;
     11811-        module = PyObject_CallFunctionObjArgs(py_import,
     11812-            name, global_dict, empty_dict, list, py_level, NULL);
     11813-        Py_DECREF(py_level);
     11814+        #if PY_MAJOR_VERSION >= 3
     11815+        if (level == -1) {
     11816+            if (strchr(__Pyx_MODULE_NAME, '.')) {
     11817+                /* try package relative import first */
     11818+                PyObject *py_level = PyInt_FromLong(1);
     11819+                if (!py_level)
     11820+                    goto bad;
     11821+                module = PyObject_CallFunctionObjArgs(py_import,
     11822+                    name, global_dict, empty_dict, list, py_level, NULL);
     11823+                Py_DECREF(py_level);
     11824+                if (!module) {
     11825+                    if (!PyErr_ExceptionMatches(PyExc_ImportError))
     11826+                        goto bad;
     11827+                    PyErr_Clear();
     11828+                }
     11829+            }
     11830+            level = 0; /* try absolute import on failure */
     11831+        }
     11832+        #endif
     11833+        if (!module) {
     11834+            PyObject *py_level = PyInt_FromLong(level);
     11835+            if (!py_level)
     11836+                goto bad;
     11837+            module = PyObject_CallFunctionObjArgs(py_import,
     11838+                name, global_dict, empty_dict, list, py_level, NULL);
     11839+            Py_DECREF(py_level);
     11840+        }
     11841     }
     11842     #else
     11843     if (level>0) {
     11844@@ -22171,66 +22940,65 @@
     11845     return module;
     11846 }
     11847 
     11848-static CYTHON_INLINE int __Pyx_PyBytes_Equals(PyObject* s1, PyObject* s2, int equals) {
     11849-    if (s1 == s2) {   /* as done by PyObject_RichCompareBool(); also catches the (interned) empty string */
     11850-        return (equals == Py_EQ);
     11851-    } else if (PyBytes_CheckExact(s1) & PyBytes_CheckExact(s2)) {
     11852-        if (PyBytes_GET_SIZE(s1) != PyBytes_GET_SIZE(s2)) {
     11853-            return (equals == Py_NE);
     11854-        } else if (PyBytes_GET_SIZE(s1) == 1) {
     11855-            if (equals == Py_EQ)
     11856-                return (PyBytes_AS_STRING(s1)[0] == PyBytes_AS_STRING(s2)[0]);
     11857-            else
     11858-                return (PyBytes_AS_STRING(s1)[0] != PyBytes_AS_STRING(s2)[0]);
     11859-        } else {
     11860-            int result = memcmp(PyBytes_AS_STRING(s1), PyBytes_AS_STRING(s2), (size_t)PyBytes_GET_SIZE(s1));
     11861-            return (equals == Py_EQ) ? (result == 0) : (result != 0);
     11862+static CYTHON_INLINE npy_intp __Pyx_PyInt_from_py_npy_intp(PyObject* x) {
     11863+    const npy_intp neg_one = (npy_intp)-1, const_zero = (npy_intp)0;
     11864+    const int is_unsigned = const_zero < neg_one;
     11865+    if (sizeof(npy_intp) == sizeof(char)) {
     11866+        if (is_unsigned)
     11867+            return (npy_intp)__Pyx_PyInt_AsUnsignedChar(x);
     11868+        else
     11869+            return (npy_intp)__Pyx_PyInt_AsSignedChar(x);
     11870+    } else if (sizeof(npy_intp) == sizeof(short)) {
     11871+        if (is_unsigned)
     11872+            return (npy_intp)__Pyx_PyInt_AsUnsignedShort(x);
     11873+        else
     11874+            return (npy_intp)__Pyx_PyInt_AsSignedShort(x);
     11875+    } else if (sizeof(npy_intp) == sizeof(int)) {
     11876+        if (is_unsigned)
     11877+            return (npy_intp)__Pyx_PyInt_AsUnsignedInt(x);
     11878+        else
     11879+            return (npy_intp)__Pyx_PyInt_AsSignedInt(x);
     11880+    } else if (sizeof(npy_intp) == sizeof(long)) {
     11881+        if (is_unsigned)
     11882+            return (npy_intp)__Pyx_PyInt_AsUnsignedLong(x);
     11883+        else
     11884+            return (npy_intp)__Pyx_PyInt_AsSignedLong(x);
     11885+    } else if (sizeof(npy_intp) == sizeof(PY_LONG_LONG)) {
     11886+        if (is_unsigned)
     11887+            return (npy_intp)__Pyx_PyInt_AsUnsignedLongLong(x);
     11888+        else
     11889+            return (npy_intp)__Pyx_PyInt_AsSignedLongLong(x);
     11890+    }  else {
     11891+        npy_intp val;
     11892+        PyObject *v = __Pyx_PyNumber_Int(x);
     11893+        #if PY_VERSION_HEX < 0x03000000
     11894+        if (likely(v) && !PyLong_Check(v)) {
     11895+            PyObject *tmp = v;
     11896+            v = PyNumber_Long(tmp);
     11897+            Py_DECREF(tmp);
     11898         }
     11899-    } else if ((s1 == Py_None) & PyBytes_CheckExact(s2)) {
     11900-        return (equals == Py_NE);
     11901-    } else if ((s2 == Py_None) & PyBytes_CheckExact(s1)) {
     11902-        return (equals == Py_NE);
     11903-    } else {
     11904-        int result;
     11905-        PyObject* py_result = PyObject_RichCompare(s1, s2, equals);
     11906-        if (!py_result)
     11907-            return -1;
     11908-        result = __Pyx_PyObject_IsTrue(py_result);
     11909-        Py_DECREF(py_result);
     11910-        return result;
     11911+        #endif
     11912+        if (likely(v)) {
     11913+            int one = 1; int is_little = (int)*(unsigned char *)&one;
     11914+            unsigned char *bytes = (unsigned char *)&val;
     11915+            int ret = _PyLong_AsByteArray((PyLongObject *)v,
     11916+                                          bytes, sizeof(val),
     11917+                                          is_little, !is_unsigned);
     11918+            Py_DECREF(v);
     11919+            if (likely(!ret))
     11920+                return val;
     11921+        }
     11922+        return (npy_intp)-1;
     11923     }
     11924 }
     11925 
     11926-static CYTHON_INLINE int __Pyx_PyUnicode_Equals(PyObject* s1, PyObject* s2, int equals) {
     11927-    if (s1 == s2) {   /* as done by PyObject_RichCompareBool(); also catches the (interned) empty string */
     11928-        return (equals == Py_EQ);
     11929-    } else if (PyUnicode_CheckExact(s1) & PyUnicode_CheckExact(s2)) {
     11930-        if (PyUnicode_GET_SIZE(s1) != PyUnicode_GET_SIZE(s2)) {
     11931-            return (equals == Py_NE);
     11932-        } else if (PyUnicode_GET_SIZE(s1) == 1) {
     11933-            if (equals == Py_EQ)
     11934-                return (PyUnicode_AS_UNICODE(s1)[0] == PyUnicode_AS_UNICODE(s2)[0]);
     11935-            else
     11936-                return (PyUnicode_AS_UNICODE(s1)[0] != PyUnicode_AS_UNICODE(s2)[0]);
     11937-        } else {
     11938-            int result = PyUnicode_Compare(s1, s2);
     11939-            if ((result == -1) && unlikely(PyErr_Occurred()))
     11940-                return -1;
     11941-            return (equals == Py_EQ) ? (result == 0) : (result != 0);
     11942-        }
     11943-    } else if ((s1 == Py_None) & PyUnicode_CheckExact(s2)) {
     11944-        return (equals == Py_NE);
     11945-    } else if ((s2 == Py_None) & PyUnicode_CheckExact(s1)) {
     11946-        return (equals == Py_NE);
     11947-    } else {
     11948-        int result;
     11949-        PyObject* py_result = PyObject_RichCompare(s1, s2, equals);
     11950-        if (!py_result)
     11951-            return -1;
     11952-        result = __Pyx_PyObject_IsTrue(py_result);
     11953-        Py_DECREF(py_result);
     11954-        return result;
     11955-    }
     11956+static CYTHON_INLINE void __Pyx_RaiseImportError(PyObject *name) {
     11957+#if PY_MAJOR_VERSION < 3
     11958+    PyErr_Format(PyExc_ImportError, "cannot import name %.230s",
     11959+                 PyString_AsString(name));
     11960+#else
     11961+    PyErr_Format(PyExc_ImportError, "cannot import name %S", name);
     11962+#endif
     11963 }
     11964 
     11965 static CYTHON_INLINE PyObject *__Pyx_PyInt_to_py_npy_intp(npy_intp val) {
     11966@@ -22658,58 +23426,6 @@
     11967     }
     11968 }
     11969 
     11970-static CYTHON_INLINE npy_intp __Pyx_PyInt_from_py_npy_intp(PyObject* x) {
     11971-    const npy_intp neg_one = (npy_intp)-1, const_zero = (npy_intp)0;
     11972-    const int is_unsigned = const_zero < neg_one;
     11973-    if (sizeof(npy_intp) == sizeof(char)) {
     11974-        if (is_unsigned)
     11975-            return (npy_intp)__Pyx_PyInt_AsUnsignedChar(x);
     11976-        else
     11977-            return (npy_intp)__Pyx_PyInt_AsSignedChar(x);
     11978-    } else if (sizeof(npy_intp) == sizeof(short)) {
     11979-        if (is_unsigned)
     11980-            return (npy_intp)__Pyx_PyInt_AsUnsignedShort(x);
     11981-        else
     11982-            return (npy_intp)__Pyx_PyInt_AsSignedShort(x);
     11983-    } else if (sizeof(npy_intp) == sizeof(int)) {
     11984-        if (is_unsigned)
     11985-            return (npy_intp)__Pyx_PyInt_AsUnsignedInt(x);
     11986-        else
     11987-            return (npy_intp)__Pyx_PyInt_AsSignedInt(x);
     11988-    } else if (sizeof(npy_intp) == sizeof(long)) {
     11989-        if (is_unsigned)
     11990-            return (npy_intp)__Pyx_PyInt_AsUnsignedLong(x);
     11991-        else
     11992-            return (npy_intp)__Pyx_PyInt_AsSignedLong(x);
     11993-    } else if (sizeof(npy_intp) == sizeof(PY_LONG_LONG)) {
     11994-        if (is_unsigned)
     11995-            return (npy_intp)__Pyx_PyInt_AsUnsignedLongLong(x);
     11996-        else
     11997-            return (npy_intp)__Pyx_PyInt_AsSignedLongLong(x);
     11998-    }  else {
     11999-        npy_intp val;
     12000-        PyObject *v = __Pyx_PyNumber_Int(x);
     12001-        #if PY_VERSION_HEX < 0x03000000
     12002-        if (likely(v) && !PyLong_Check(v)) {
     12003-            PyObject *tmp = v;
     12004-            v = PyNumber_Long(tmp);
     12005-            Py_DECREF(tmp);
     12006-        }
     12007-        #endif
     12008-        if (likely(v)) {
     12009-            int one = 1; int is_little = (int)*(unsigned char *)&one;
     12010-            unsigned char *bytes = (unsigned char *)&val;
     12011-            int ret = _PyLong_AsByteArray((PyLongObject *)v,
     12012-                                          bytes, sizeof(val),
     12013-                                          is_little, !is_unsigned);
     12014-            Py_DECREF(v);
     12015-            if (likely(!ret))
     12016-                return val;
     12017-        }
     12018-        return (npy_intp)-1;
     12019-    }
     12020-}
     12021-
     12022 static int __Pyx_check_binary_version(void) {
     12023     char ctversion[4], rtversion[4];
     12024     PyOS_snprintf(ctversion, 4, "%d.%d", PY_MAJOR_VERSION, PY_MINOR_VERSION);
     12025@@ -22729,6 +23445,23 @@
     12026     return 0;
     12027 }
     12028 
     12029+#ifndef __PYX_HAVE_RT_ImportModule
     12030+#define __PYX_HAVE_RT_ImportModule
     12031+static PyObject *__Pyx_ImportModule(const char *name) {
     12032+    PyObject *py_name = 0;
     12033+    PyObject *py_module = 0;
     12034+    py_name = __Pyx_PyIdentifier_FromString(name);
     12035+    if (!py_name)
     12036+        goto bad;
     12037+    py_module = PyImport_Import(py_name);
     12038+    Py_DECREF(py_name);
     12039+    return py_module;
     12040+bad:
     12041+    Py_XDECREF(py_name);
     12042+    return 0;
     12043+}
     12044+#endif
     12045+
     12046 #ifndef __PYX_HAVE_RT_ImportType
     12047 #define __PYX_HAVE_RT_ImportType
     12048 static PyTypeObject *__Pyx_ImportType(const char *module_name, const char *class_name,
     12049@@ -22738,15 +23471,10 @@
     12050     PyObject *result = 0;
     12051     PyObject *py_name = 0;
     12052     char warning[200];
     12053-
     12054     py_module = __Pyx_ImportModule(module_name);
     12055     if (!py_module)
     12056         goto bad;
     12057-    #if PY_MAJOR_VERSION < 3
     12058-    py_name = PyString_FromString(class_name);
     12059-    #else
     12060-    py_name = PyUnicode_FromString(class_name);
     12061-    #endif
     12062+    py_name = __Pyx_PyIdentifier_FromString(class_name);
     12063     if (!py_name)
     12064         goto bad;
     12065     result = PyObject_GetAttr(py_module, py_name);
     12066@@ -22762,7 +23490,7 @@
     12067             module_name, class_name);
     12068         goto bad;
     12069     }
     12070-    if (!strict && ((PyTypeObject *)result)->tp_basicsize > (Py_ssize_t)size) {
     12071+    if (!strict && (size_t)((PyTypeObject *)result)->tp_basicsize > size) {
     12072         PyOS_snprintf(warning, sizeof(warning),
     12073             "%s.%s size changed, may indicate binary incompatibility",
     12074             module_name, class_name);
     12075@@ -22772,7 +23500,7 @@
     12076         if (PyErr_WarnEx(NULL, warning, 0) < 0) goto bad;
     12077         #endif
     12078     }
     12079-    else if (((PyTypeObject *)result)->tp_basicsize != (Py_ssize_t)size) {
     12080+    else if ((size_t)((PyTypeObject *)result)->tp_basicsize != size) {
     12081         PyErr_Format(PyExc_ValueError,
     12082             "%s.%s has the wrong size, try recompiling",
     12083             module_name, class_name);
     12084@@ -22786,51 +23514,105 @@
     12085 }
     12086 #endif
     12087 
     12088-#ifndef __PYX_HAVE_RT_ImportModule
     12089-#define __PYX_HAVE_RT_ImportModule
     12090-static PyObject *__Pyx_ImportModule(const char *name) {
     12091-    PyObject *py_name = 0;
     12092-    PyObject *py_module = 0;
     12093-
     12094-    #if PY_MAJOR_VERSION < 3
     12095-    py_name = PyString_FromString(name);
     12096-    #else
     12097-    py_name = PyUnicode_FromString(name);
     12098-    #endif
     12099-    if (!py_name)
     12100-        goto bad;
     12101-    py_module = PyImport_Import(py_name);
     12102-    Py_DECREF(py_name);
     12103-    return py_module;
     12104-bad:
     12105-    Py_XDECREF(py_name);
     12106-    return 0;
     12107+static int __pyx_bisect_code_objects(__Pyx_CodeObjectCacheEntry* entries, int count, int code_line) {
     12108+    int start = 0, mid = 0, end = count - 1;
     12109+    if (end >= 0 && code_line > entries[end].code_line) {
     12110+        return count;
     12111+    }
     12112+    while (start < end) {
     12113+        mid = (start + end) / 2;
     12114+        if (code_line < entries[mid].code_line) {
     12115+            end = mid;
     12116+        } else if (code_line > entries[mid].code_line) {
     12117+             start = mid + 1;
     12118+        } else {
     12119+            return mid;
     12120+        }
     12121+    }
     12122+    if (code_line <= entries[mid].code_line) {
     12123+        return mid;
     12124+    } else {
     12125+        return mid + 1;
     12126+    }
     12127+}
     12128+static PyCodeObject *__pyx_find_code_object(int code_line) {
     12129+    PyCodeObject* code_object;
     12130+    int pos;
     12131+    if (unlikely(!code_line) || unlikely(!__pyx_code_cache.entries)) {
     12132+        return NULL;
     12133+    }
     12134+    pos = __pyx_bisect_code_objects(__pyx_code_cache.entries, __pyx_code_cache.count, code_line);
     12135+    if (unlikely(pos >= __pyx_code_cache.count) || unlikely(__pyx_code_cache.entries[pos].code_line != code_line)) {
     12136+        return NULL;
     12137+    }
     12138+    code_object = __pyx_code_cache.entries[pos].code_object;
     12139+    Py_INCREF(code_object);
     12140+    return code_object;
     12141+}
     12142+static void __pyx_insert_code_object(int code_line, PyCodeObject* code_object) {
     12143+    int pos, i;
     12144+    __Pyx_CodeObjectCacheEntry* entries = __pyx_code_cache.entries;
     12145+    if (unlikely(!code_line)) {
     12146+        return;
     12147+    }
     12148+    if (unlikely(!entries)) {
     12149+        entries = (__Pyx_CodeObjectCacheEntry*)PyMem_Malloc(64*sizeof(__Pyx_CodeObjectCacheEntry));
     12150+        if (likely(entries)) {
     12151+            __pyx_code_cache.entries = entries;
     12152+            __pyx_code_cache.max_count = 64;
     12153+            __pyx_code_cache.count = 1;
     12154+            entries[0].code_line = code_line;
     12155+            entries[0].code_object = code_object;
     12156+            Py_INCREF(code_object);
     12157+        }
     12158+        return;
     12159+    }
     12160+    pos = __pyx_bisect_code_objects(__pyx_code_cache.entries, __pyx_code_cache.count, code_line);
     12161+    if ((pos < __pyx_code_cache.count) && unlikely(__pyx_code_cache.entries[pos].code_line == code_line)) {
     12162+        PyCodeObject* tmp = entries[pos].code_object;
     12163+        entries[pos].code_object = code_object;
     12164+        Py_DECREF(tmp);
     12165+        return;
     12166+    }
     12167+    if (__pyx_code_cache.count == __pyx_code_cache.max_count) {
     12168+        int new_max = __pyx_code_cache.max_count + 64;
     12169+        entries = (__Pyx_CodeObjectCacheEntry*)PyMem_Realloc(
     12170+            __pyx_code_cache.entries, new_max*sizeof(__Pyx_CodeObjectCacheEntry));
     12171+        if (unlikely(!entries)) {
     12172+            return;
     12173+        }
     12174+        __pyx_code_cache.entries = entries;
     12175+        __pyx_code_cache.max_count = new_max;
     12176+    }
     12177+    for (i=__pyx_code_cache.count; i>pos; i--) {
     12178+        entries[i] = entries[i-1];
     12179+    }
     12180+    entries[pos].code_line = code_line;
     12181+    entries[pos].code_object = code_object;
     12182+    __pyx_code_cache.count++;
     12183+    Py_INCREF(code_object);
     12184 }
     12185-#endif
     12186 
     12187 #include "compile.h"
     12188 #include "frameobject.h"
     12189 #include "traceback.h"
     12190-
     12191-static void __Pyx_AddTraceback(const char *funcname, int __pyx_clineno,
     12192-                               int __pyx_lineno, const char *__pyx_filename) {
     12193+static PyCodeObject* __Pyx_CreateCodeObjectForTraceback(
     12194+            const char *funcname, int c_line,
     12195+            int py_line, const char *filename) {
     12196+    PyCodeObject *py_code = 0;
     12197     PyObject *py_srcfile = 0;
     12198     PyObject *py_funcname = 0;
     12199-    PyObject *py_globals = 0;
     12200-    PyCodeObject *py_code = 0;
     12201-    PyFrameObject *py_frame = 0;
     12202-
     12203     #if PY_MAJOR_VERSION < 3
     12204-    py_srcfile = PyString_FromString(__pyx_filename);
     12205+    py_srcfile = PyString_FromString(filename);
     12206     #else
     12207-    py_srcfile = PyUnicode_FromString(__pyx_filename);
     12208+    py_srcfile = PyUnicode_FromString(filename);
     12209     #endif
     12210     if (!py_srcfile) goto bad;
     12211-    if (__pyx_clineno) {
     12212+    if (c_line) {
     12213         #if PY_MAJOR_VERSION < 3
     12214-        py_funcname = PyString_FromFormat( "%s (%s:%d)", funcname, __pyx_cfilenm, __pyx_clineno);
     12215+        py_funcname = PyString_FromFormat( "%s (%s:%d)", funcname, __pyx_cfilenm, c_line);
     12216         #else
     12217-        py_funcname = PyUnicode_FromFormat( "%s (%s:%d)", funcname, __pyx_cfilenm, __pyx_clineno);
     12218+        py_funcname = PyUnicode_FromFormat( "%s (%s:%d)", funcname, __pyx_cfilenm, c_line);
     12219         #endif
     12220     }
     12221     else {
     12222@@ -22841,28 +23623,45 @@
     12223         #endif
     12224     }
     12225     if (!py_funcname) goto bad;
     12226-    py_globals = PyModule_GetDict(__pyx_m);
     12227-    if (!py_globals) goto bad;
     12228-    py_code = PyCode_New(
     12229+    py_code = __Pyx_PyCode_New(
     12230         0,            /*int argcount,*/
     12231-        #if PY_MAJOR_VERSION >= 3
     12232         0,            /*int kwonlyargcount,*/
     12233-        #endif
     12234         0,            /*int nlocals,*/
     12235         0,            /*int stacksize,*/
     12236         0,            /*int flags,*/
     12237         __pyx_empty_bytes, /*PyObject *code,*/
     12238-        __pyx_empty_tuple,  /*PyObject *consts,*/
     12239-        __pyx_empty_tuple,  /*PyObject *names,*/
     12240-        __pyx_empty_tuple,  /*PyObject *varnames,*/
     12241-        __pyx_empty_tuple,  /*PyObject *freevars,*/
     12242-        __pyx_empty_tuple,  /*PyObject *cellvars,*/
     12243+        __pyx_empty_tuple, /*PyObject *consts,*/
     12244+        __pyx_empty_tuple, /*PyObject *names,*/
     12245+        __pyx_empty_tuple, /*PyObject *varnames,*/
     12246+        __pyx_empty_tuple, /*PyObject *freevars,*/
     12247+        __pyx_empty_tuple, /*PyObject *cellvars,*/
     12248         py_srcfile,   /*PyObject *filename,*/
     12249         py_funcname,  /*PyObject *name,*/
     12250-        __pyx_lineno,   /*int firstlineno,*/
     12251+        py_line,      /*int firstlineno,*/
     12252         __pyx_empty_bytes  /*PyObject *lnotab*/
     12253     );
     12254-    if (!py_code) goto bad;
     12255+    Py_DECREF(py_srcfile);
     12256+    Py_DECREF(py_funcname);
     12257+    return py_code;
     12258+bad:
     12259+    Py_XDECREF(py_srcfile);
     12260+    Py_XDECREF(py_funcname);
     12261+    return NULL;
     12262+}
     12263+static void __Pyx_AddTraceback(const char *funcname, int c_line,
     12264+                               int py_line, const char *filename) {
     12265+    PyCodeObject *py_code = 0;
     12266+    PyObject *py_globals = 0;
     12267+    PyFrameObject *py_frame = 0;
     12268+    py_code = __pyx_find_code_object(c_line ? c_line : py_line);
     12269+    if (!py_code) {
     12270+        py_code = __Pyx_CreateCodeObjectForTraceback(
     12271+            funcname, c_line, py_line, filename);
     12272+        if (!py_code) goto bad;
     12273+        __pyx_insert_code_object(c_line ? c_line : py_line, py_code);
     12274+    }
     12275+    py_globals = PyModule_GetDict(__pyx_m);
     12276+    if (!py_globals) goto bad;
     12277     py_frame = PyFrame_New(
     12278         PyThreadState_GET(), /*PyThreadState *tstate,*/
     12279         py_code,             /*PyCodeObject *code,*/
     12280@@ -22870,11 +23669,9 @@
     12281         0                    /*PyObject *locals*/
     12282     );
     12283     if (!py_frame) goto bad;
     12284-    py_frame->f_lineno = __pyx_lineno;
     12285+    py_frame->f_lineno = py_line;
     12286     PyTraceBack_Here(py_frame);
     12287 bad:
     12288-    Py_XDECREF(py_srcfile);
     12289-    Py_XDECREF(py_funcname);
     12290     Py_XDECREF(py_code);
     12291     Py_XDECREF(py_frame);
     12292 }
     12293@@ -22909,6 +23706,7 @@
     12294     return 0;
     12295 }
     12296 
     12297+
     12298 /* Type Conversion Functions */
     12299 
     12300 static CYTHON_INLINE int __Pyx_PyObject_IsTrue(PyObject* x) {

  • new file dports/python/py-numpy/files/patch-python33-shape.diff 

    diff --git a/dports/python/py-numpy/files/patch-python33-shape.diff b/dports/python/py-numpy/files/patch-python33-shape.diff
new file mode 100644
    - +  
     1--- numpy/core/tests/test_multiarray.py
     2+++ numpy/core/tests/test_multiarray.py
     3@@ -11,10 +11,19 @@
     4 
     5 from numpy.compat import asbytes, getexception, strchar
     6 
     7 from test_print import in_foreign_locale
     8 
     9+if sys.version_info[:2] > (3, 2):
     10+    # In Python 3.3 the representation of empty shape, strides and suboffsets
     11+    # is an empty tuple instead of None.
     12+    # http://docs.python.org/dev/whatsnew/3.3.html#api-changes
     13+    EMPTY = ()
     14+else:   
     15+    EMPTY = None
     16+
     17+
     18 class TestFlags(TestCase):
     19     def setUp(self):
     20         self.a = arange(10)
     21 
     22     def test_writeable(self):
     23@@ -2160,31 +2169,31 @@
     24             y = memoryview(x)
     25             assert_equal(y.format, 'i')
     26             assert_equal(y.shape, (5,))
     27             assert_equal(y.ndim, 1)
     28             assert_equal(y.strides, (4,))
     29-            assert_equal(y.suboffsets, None)
     30+            assert_equal(y.suboffsets, EMPTY)
     31             assert_equal(y.itemsize, 4)
     32 
     33         def test_export_simple_nd(self):
     34             x = np.array([[1,2],[3,4]], dtype=np.float64)
     35             y = memoryview(x)
     36             assert_equal(y.format, 'd')
     37             assert_equal(y.shape, (2, 2))
     38             assert_equal(y.ndim, 2)
     39             assert_equal(y.strides, (16, 8))
     40-            assert_equal(y.suboffsets, None)
     41+            assert_equal(y.suboffsets, EMPTY)
     42             assert_equal(y.itemsize, 8)
     43 
     44         def test_export_discontiguous(self):
     45             x = np.zeros((3,3,3), dtype=np.float32)[:,0,:]
     46             y = memoryview(x)
     47             assert_equal(y.format, 'f')
     48             assert_equal(y.shape, (3, 3))
     49             assert_equal(y.ndim, 2)
     50             assert_equal(y.strides, (36, 4))
     51-            assert_equal(y.suboffsets, None)
     52+            assert_equal(y.suboffsets, EMPTY)
     53             assert_equal(y.itemsize, 4)
     54 
     55         def test_export_record(self):
     56             dt = [('a', 'b'),
     57                   ('b', 'h'),
     58@@ -2212,11 +2221,11 @@
     59                            asbytes('aaaa'), 'bbbb', asbytes('   '), True, 1.0)],
     60                          dtype=dt)
     61             y = memoryview(x)
     62             assert_equal(y.shape, (1,))
     63             assert_equal(y.ndim, 1)
     64-            assert_equal(y.suboffsets, None)
     65+            assert_equal(y.suboffsets, EMPTY)
     66 
     67             sz = sum([dtype(b).itemsize for a, b in dt])
     68             if dtype('l').itemsize == 4:
     69                 assert_equal(y.format, 'T{b:a:=h:b:i:c:l:d:^q:dx:B:e:@H:f:=I:g:L:h:^Q:hx:=f:i:d:j:^g:k:=Zf:ix:Zd:jx:^Zg:kx:4s:l:=4w:m:3x:n:?:o:@e:p:}')
     70             else:
     71@@ -2226,14 +2235,14 @@
     72 
     73         def test_export_subarray(self):
     74             x = np.array(([[1,2],[3,4]],), dtype=[('a', ('i', (2,2)))])
     75             y = memoryview(x)
     76             assert_equal(y.format, 'T{(2,2)i:a:}')
     77-            assert_equal(y.shape, None)
     78+            assert_equal(y.shape, EMPTY)
     79             assert_equal(y.ndim, 0)
     80-            assert_equal(y.strides, None)
     81-            assert_equal(y.suboffsets, None)
     82+            assert_equal(y.strides, EMPTY)
     83+            assert_equal(y.suboffsets, EMPTY)
     84             assert_equal(y.itemsize, 16)
     85 
     86         def test_export_endian(self):
     87             x = np.array([1,2,3], dtype='>i')
     88             y = memoryview(x)

  • new file dports/python/py-numpy/files/patch-python33-unicode.diff 

    diff --git a/dports/python/py-numpy/files/patch-python33-unicode.diff b/dports/python/py-numpy/files/patch-python33-unicode.diff
new file mode 100644
    - +  
     1--- numpy/core/src/multiarray/scalarapi.c
     2+++ numpy/core/src/multiarray/scalarapi.c
     3@@ -650,10 +650,39 @@
     4              * so round up to nearest multiple
     5              */
     6             itemsize = (((itemsize - 1) >> 2) + 1) << 2;
     7         }
     8     }
     9+#if PY_VERSION_HEX >= 0x03030000
     10+    if (type_num == NPY_UNICODE) {
     11+        PyObject *u, *args;
     12+        int byteorder;
     13+
     14+#if NPY_BYTE_ORDER == NPY_LITTLE_ENDIAN
     15+        byteorder = -1;
     16+#elif NPY_BYTE_ORDER == NPY_BIG_ENDIAN
     17+        byteorder = +1;
     18+#else
     19+        #error Endianness undefined ?
     20+#endif
     21+        if (swap) byteorder *= -1;
     22+
     23+        u = PyUnicode_DecodeUTF32(data, itemsize, NULL, &byteorder);
     24+        if (u == NULL) {
     25+            return NULL;
     26+        }
     27+        args = Py_BuildValue("(O)", u);
     28+        if (args == NULL) {
     29+            Py_DECREF(u);
     30+            return NULL;
     31+        }
     32+        obj = type->tp_new(type, args, NULL);
     33+        Py_DECREF(u);
     34+        Py_DECREF(args);
     35+        return obj;
     36+    }
     37+#endif
     38     if (type->tp_itemsize != 0) {
     39         /* String type */
     40         obj = type->tp_alloc(type, itemsize);
     41     }
     42     else {
     43@@ -686,10 +715,11 @@
     44             ((PyStringObject *)obj)->ob_sstate = SSTATE_NOT_INTERNED;
     45 #endif
     46             memcpy(destptr, data, itemsize);
     47             return obj;
     48         }
     49+#if PY_VERSION_HEX < 0x03030000
     50         else if (type_num == PyArray_UNICODE) {
     51             /* tp_alloc inherited from Python PyBaseObject_Type */
     52             PyUnicodeObject *uni = (PyUnicodeObject*)obj;
     53             size_t length = itemsize >> 2;
     54             Py_UNICODE *dst;
     55@@ -757,10 +787,11 @@
     56             uni->str[length] = 0;
     57             uni->length = length;
     58 #endif
     59             return obj;
     60         }
     61+#endif /* PY_VERSION_HEX < 0x03030000 */
     62         else {
     63             PyVoidScalarObject *vobj = (PyVoidScalarObject *)obj;
     64             vobj->base = NULL;
     65             vobj->descr = descr;
     66             Py_INCREF(descr);
     67--- numpy/core/src/multiarray/scalartypes.c.src
     68+++ numpy/core/src/multiarray/scalartypes.c.src
     69@@ -2321,11 +2321,15 @@
     70     Py_DECREF(typecode);
     71 #if @default@ == 0
     72     *((npy_@name@ *)dest) = *((npy_@name@ *)src);
     73 #elif @default@ == 1 /* unicode and strings */
     74     if (itemsize == 0) { /* unicode */
     75+#if PY_VERSION_HEX >= 0x03030000
     76+        itemsize = PyUnicode_GetLength(robj) * PyUnicode_KIND(robj);
     77+#else
     78         itemsize = ((PyUnicodeObject *)robj)->length * sizeof(Py_UNICODE);
     79+#endif
     80     }
     81     memcpy(dest, src, itemsize);
     82     /* @default@ == 2 won't get here */
     83 #endif
     84     Py_DECREF(robj);
     85--- numpy/core/tests/test_unicode.py
     86+++ numpy/core/tests/test_unicode.py
     87@@ -24,14 +24,16 @@
     88     def buffer_length(arr):
     89         if isinstance(arr, ndarray):
     90             return len(arr.data)
     91         return len(buffer(arr))
     92 
     93+# In both cases below we need to make sure that the byte swapped value (as
     94+# UCS4) is still a valid unicode:
     95 # Value that can be represented in UCS2 interpreters
     96-ucs2_value = u'\uFFFF'
     97+ucs2_value = u'\u0900'
     98 # Value that cannot be represented in UCS2 interpreters (but can in UCS4)
     99-ucs4_value = u'\U0010FFFF'
     100+ucs4_value = u'\U00100900'
     101 
     102 
     103 ############################################################
     104 #    Creation tests
     105 ############################################################