[Biopython-dev] NumPy

[Peter]

On Thu, Aug 28, 2008 at 1:12 AM, Michiel de Hoon <mjldehoon at yahoo.com> wrote:
> Hi everybody,
>
> Previously we discussed on this mailing list whether Biopython should adopt
> the "new" Numerical Python (aka NumPy, currently at version 1.1.1)
> instead of the "old" Numerical Python (version 24.2). My objections against
> NumPy were that its documentation is not freely available, it doesn't
> compile cleanly on all platforms, and some other scientific and
> computational biology libraries use the old Numerical Python.
>
> Last week, the NumPy documentation did become freely available.
> Compilation of NumPy is still not perfect on all platforms (e.g. on Cygwin
> it may fail), however recently I have also noticed that compilation of the
> "old" Numerical Python may fail on modern systems. As far as I can tell,
> MMTK and PyMOL are (still?) based on the "old" Numerical Python, but
> Matplotlib now relies on the "new" Numerical Python.
>
> In my opinion, the balance is now tilting in favor of the new NumPy,
> and we should consider transitioning Biopython to the new NumPy.
> Does anybody have a different opinion? If not, I suggest we bring this
> up on the user mailing list.

I agree that we should be aiming to transition from Numeric to NumPy.

One question is will Numpy 1.2 be binary compatible with Numpy 1.1
(for our compiled C extensions).  This isn't true for the Numpy 1.2 beta
releases thus far, but it sounds like they are going to try:
http://projects.scipy.org/pipermail/numpy-discussion/2008-August/036508.html
http://projects.scipy.org/pipermail/numpy-discussion/2008-August/036909.html

Probably in the short term we should just target NumPy 1.1

Peter