EGSimulation

[ben@…]

By using a simulation of firm sizes (using a lognormal distribution) and specified geographic regions, standard deviations and employee head count, we can compute the critical regions for the Ellison & Glaeser statistic. In the process, it also calculates herfindahl values and provides critical regions.

[ryandesign (Ryan Carsten Schmidt)]

The license should be listed as "public-domain" (no spaces in license names).

The line "extract.mkdir no" can be removed; that's the default.

There is a github portgroup that could be used, but for this port it might not actually help simplify anything.

[ben@…]

I've added the file "Portfile.2" which addresses Ryan's comments and diff.diff, which is a difference file from the two port files (someone at some point told me MacPorts preferred those).

[mf2k (Frank Schima)]

This port does not properly use the unified python portgroup.

1. Change categories to categories-append to not overwrite the python category
2. Change depends_lib to depends_lib-append to not overwrite required dependencies
3. The dependencies themselves must be written like port:py${python.version}-numpy
4. The entire depends_lib-append lines must be placed inside a if {$subport != $name} { block
5. Did the listed maintainer agree to maintain this port?

[mf2k (Frank Schima)]

Also, the port name should be "py-EGSimulation".

[ben@…]

macsforever2000 -

I've updated the port file to account for your suggestions: 2 and 3 (I've uploaded a new port file, portfile 3, as well as a diff -- a diff to the original). I did not do the others for the following reasons:

1) This script isn't a python module. In fact it has nothing to do with python other than it is written in it. What it is is a command line tool for economist that goes with a paper we just submitted to JUE studying firm agglomeration. Generally when you look through the macports categories, the category they are in depends on what they do, not what they are written in.

4) I've tested it with a if {$subport != $name} { block and it results in the dependancies not being installed when they are not already installed

5) The maintainer is me, but that address is associated to my university address.

For the same reason as 1, I have not renamed it to py-EGSimulation.

[mf2k (Frank Schima)]

That's fine. In that case, you need to add variants for different python versions so people can choose which python version they want. I'd suggest at least python26 and python27 variants and maybe python32 if it works with the python 3 series. Also, you can no longer use ${python.version} and ${python.default_version } because they rely on a chosen pyXY subport, which you don't have.

[mf2k (Frank Schima)]

See the ROOT port for an example among many others.

[ben@…]

According to your portgroup documentation here:

​http://guide.macports.org/chunked/reference.portgroup.html

the portgroup for python is loading port:pythonXY by default. From Portfile.3 and Portfile.4, I've been using append as you suggest.

I've added 26 support, though for performance reasons they really should use 27.

Note: Latest version is Portfile.4 ; there is also a diff file diff.2.diff which is the difference between the original submission and Portfile.4

[mf2k (Frank Schima)]

Unless you name the port "py-EGSimulation" and use subports, there is no way to choose py26. So this really cannot work. Have you tried it? I've never seen the unified python portgroup used in this way before. Unfortunately it is not fully documented yet.

Also, you do not need to attach diffs. This is a new port and only the latest portfile matters.

[ben@…]

I don't want them to install a py26 version. As I said, for performance reasons that doesn't make sense. For people doing real work with this tool, they will set the iteration count up at 100000 or so, every little bit counts and py27 is the fastest version (I'm not saying python 2.7 is faster, in general, or anything. Specifically: this tool with our particular packages, i.e. numpy, scipy etc is fastest under python 2.7). (I know because we built an enormous critical value table with this tool)

Yes, I've tested every portfile I've attached. Portfile.5 is really what I want (which I removed the 26 reference). I thought you were insisting that I have py26 support. If I misunderstood you, I'm sorry.

[ben@…]

I haven't heard any comments on portfile.5; does that mean everyone is ok with it? If not please let me know what I need to do to make it so.

[nerdling (Jeremy Lavergne)]

Portfile.5​ still seems written like a python module.

You should remove the PortGroup and simply use depends_run-append port:python27 (whichever depends_* are appropriate for your needs).

[pixilla (Bradley Giesbrecht)]

Replying to snc@…:

Portfile.5​ still seems written like a python module.

You should remove the PortGroup and simply use depends_run-append port:python27 (whichever depends_* are appropriate for your needs).

The ports use setup.py and not configure/make. Isn't it nice to use the python PortGroup in this case?

[ryandesign (Ryan Carsten Schmidt)]

There's nothing wrong with using the python portgroup for a program, rather than a module. And there's no need for the author to provide variants for different python versions, especially when the author (who is also the author of the software in this case) says python27 is the best version to use. That's the version we'd like MacPorts users to be using anyway.

[ryandesign (Ryan Carsten Schmidt)]

Thanks for this port! I've committed it in r99349 with these changes:

• made use of github portgroup
• added required "platforms" variable
• removed "python.default_version 27" since that's the default
• fixed livecheck
• rearranged some lines