[SciPy-user] Numerical Achievements in SciPy

[Anne Archibald]

On 02/05/07, David Cournapeau <david at ar.media.kyoto-u.ac.jp> wrote:

> If you want to share anything non trivial, you will need distutils, I
> think, or something as complicated. I don't know much about R, but it
> looks like it uses autoconf when you have some code written in C or
> Fortran (and compared to autoconf, almost anything is a pleasant
> experience, including debugging someone's else perl code). Now, someone
> could create something  like a system to create a skeleton project, with
> distutils setup.py already written, etc... To make things a bit more
> automated. A page could also be set up om the wiki for instructions how
> to do it for someone new to scipy, etc..
>
> But generally, when you want to share some code which is a bit more than
> one python file, it will require some work, and I don't see how it can
> be really different than what Robert described.

I made a distutils package for a library with a few python files the
other day. I'd never done it before and it took me about ten minutes.
It's really not hard. You write one file, setup.py, you make sure you
have an __init__.py containing __all__, and it just works.

If you have C files, it becomes more complicated - but getting C files
to compile on any computer but your own basically requires something
like distutils.

I really don't think there's much of a gap between projects small
enough to go in the wiki and projects for which distutils is worth the
trouble.

Anne