[Numpy-discussion] Ready for review: PyArrayNeighIterObject, an iterator to iterate over a neighborhood in arbitrary arrays

[Charles R Harris]

On Sat, Jun 20, 2009 at 8:15 AM, David Cournapeau<cournape at gmail.com> wrote:
> On Thu, Jun 18, 2009 at 2:05 PM, Charles R
> Harris<charlesr.harris at gmail.com> wrote:
>
>>
>> I don't know. I think it is cleaner to depend on 1.4.0
>
> Oh, definitely - I certainly prefer this solution as well.
>
>> but there
>> might be other considerations such as release schedules, both for
>> numpy/scipy and perhaps some of the major linux distros
>> (Fedora/Ubuntu). You are probably a better judge of that than I, but
>> we should discuss it along with setting some goals for the next
>> release.
>
> I don't think there are many new features for numpy 1.4.0. Most of the
> changes (in term of commits) are the ones I did to put everything in
> separate files for multiarray/ufunc, but that has 0 consequence for
> users (hopefully :) ).
>
> There is the usual set of bug-fixes. One thing I want to check is
> python 2.7 compatibility: there are some massive distutils changes, I
> have already detected and reported a couple of regressions, but I
> should test more thoroughly (especially on windows).
>
>  I know that Travis and Robert pushed some stuff in a branch for date
> handling, I don't know if they intend to push this soon in the trunk
> or not ?
>

Since we are extending the API and bumping up the API number, I wonder
if we should change the python dependency to 2.6? If there are
otherwise no major improvements in the pipe besides bug fixes now
might be the time to start laying the foundations for Python 3.0
compatibility.

We could make a bugfix release of 1.3.0 for those folks still using
python 2.4-2.5, which is probably quite a few. If we want to go that
way we should probably make an ongoing effort to backport fixes as we
go because it's such a pain to sort them all out later.

Chuck