On Oct 27, 2014 1:25 PM, "Todd" <toddrjen@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.