[BioPython] NumPy changes ready to check in

[Bruce Southey]

Brad Chapman wrote:
> Hi Peter;
>
>   
>> Wow - welcome back!
>>     
>
> Thanks. Glad to be able to help; great to see you around.
>
>   
>> Its certainly the C/C++ bits that make life complicated.
>>
>> Have you looked at the patch on Bug 2251 at all?
>> http://bugzilla.open-bio.org/show_bug.cgi?id=2251
>>     
>
> I did; sorry, forgot to mention that in my last e-mail. I wasn't
> exactly sure why this wasn't checked in then as the thread dies a
> bit. It was a great help for getting started.
>
> I think the idea there is good, but might be a little more
> complicated than needed. My thoughts were that we need to manage a
> transition to NumPy rather than support both Numeric and NumPy
> concurrently. Numeric hasn't been updated since 2005 and the
> compatibility modules for transitioning seem solid, so most people
> should have moved by now or need some encouragement to do so. We are
> probably being more annoying to up to date users by keeping Numeric
> support.
>
> The way it is implemented in my recent changes are that Numeric users will
> not get broken builds, but will rather not get the C extensions
> compiled unless they have NumPy. So the users who absolutely need to
> install NumPy are those still using Numeric and using the C/C++
> extensions.
>
>   
>> Michiel and I had chatted about doing one more release with Numeric
>> only - there are a few things I'd like to try and get in first like
>> ambiguous translation (Bug 2530) but perhaps we should just do
>> Biopython 1.48 pretty much as is.
>> http://bugzilla.open-bio.org/show_bug.cgi?id=2530
>>
>> I'm happy to try and put together the release this week - but will
>> need a hand for the windows builds.
>>
>> If that pans out, then we could start the Numeric to numpy transition in CVS.
>>     
>
> That sounds fine to me. Once the new release is out I can check
> things in and we can find out if there are any issues.
>
> Brad
>   
Hi,
This is excellent work!

I agree that at least one more version of BioPython should be released. 
This version would allow backporting of any fixes if necessary.

The transition should require NumPy 1.2 or at least NumPy 1.1 due to 
NumPy API changes. On that note (referring to the related thread), 
NumPy's API has to change to support Python3K (due to Python API 
changes) which limit the porting of BioPython  to Python3K (ignoring the 
significant porting effort for BioPython). Also that probably only NumPy 
will be ported to Python3K.

Also I would like to point out this thread on the future directions of 
NumPy that should make NumPy easier to support:
http://projects.scipy.org/pipermail/numpy-discussion/2008-August/036909.html


Bruce