[SciPy-User] Documenting numpy vs. scipy fft

[Todd]

On Oct 27, 2014 1:25 PM, "Todd" <toddrjen at gmail.com> wrote:
>
> As everyone here knows, both numpy and scipy have their own fft modules.
I have looked at previous discussion on the mailing list, and it looks like
scipy's is the recommended one.
>
> However, I can't find any mention of this in the numpy and scipy
documentation.  As far as I can see, neither numpy nor scipy even
acknowledge in the documentation that there is an fft module in the other,
not to mention provide any comparison or guidance to users on which they
would want to use.
>
> If I didn't just miss it, might this be a good thing to have the
differences between the two modules explained somewhere, and give users
some hint as to which they would use (either in general or in specific
situations)?

It seems they is as lot if discussion on how to move forward with overlap
between numpy and scipy.

However, in the short term there is still the issue of communicating to
users how the modules relate to each other, how they differ, and when users
should use one or the other.

I see four possible solutions (besides simply deprecating one of the
modules), although there may be more.  These are not mutually exclusive, in
fact they are probably complimentary:

1. Add documentation to the individual functions.  This could be as simple
as just a link to the corresponding function in the other project, although
ideally it would have at least a sentence summarizing the major differences.

2. Add a section in one of the modules root documentation explaining in
some detail the differences, with a link to that page in documentation for
the other module.

3. A set of documentation on the scipy.org website explaining all the
differences.  This could even be automatically generated from 2.

4. A more general website comparing similar tools in the scientific python
software stack.  For this case, it could also cover other fft
implementations like pyFFTW.  However, it would also contain comparisons of
other types of projects that seem superficially similar and where users
just getting started might have trouble deciding what use, such as blaze
vs. Pandas, the various plotting libraries, or the numerous physical units
libraries.  This could be either a conventional website, a wiki, or a blog
with an index of links to the latest comparisons.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20141029/8944c9bd/attachment.html>