[Numpy-discussion] I want to help with a numpy python 3.1.x port

[Charles R Harris]

On Fri, Sep 18, 2009 at 10:19 AM, René Dudfield <renesd at gmail.com> wrote:

> On Fri, Sep 18, 2009 at 4:29 PM, Charles R Harris
> <charlesr.harris at gmail.com> wrote:
> >
> >
> > On Fri, Sep 18, 2009 at 9:09 AM, René Dudfield <renesd at gmail.com> wrote:
> >>
> >> On Fri, Sep 18, 2009 at 4:05 PM, Charles R Harris
> >> <charlesr.harris at gmail.com> wrote:
> >> >
> >> >
> >> > On Fri, Sep 18, 2009 at 8:52 AM, René Dudfield <renesd at gmail.com>
> wrote:
> >> >>
> >> >> one more thing...
> >> >>
> >> >> there's also notes about porting to py3k here:
> >> >>    http://wiki.python.org/moin/cporting
> >> >> and here:
> >> >>    http://wiki.python.org/moin/PortingExtensionModulesToPy3k
> >> >>
> >> >> Which are better than the python.org docs for cporting.  That's
> >> >> probably a pretty good page to store notes about porting as we go.
> >> >>
> >> >>
> >> >
> >> > Thanks! Numpy defines a lot of extension python types, so that is what
> I
> >> > got
> >> > started on, using NPY_PY3K as the flag. Numpy also exports the numeric
> >> > operations (don't ask) and I think that needs to be changed so it
> looks
> >> > like
> >> > a reqular c++ in c class with getters and setters, which will make
> >> > things a
> >> > bit easier with the major changes that happened there.
> >> >
> >> > IIRC, there is an include file that provides the old conversions
> between
> >> > python numeric types and c types. Did you use that?
> >>
> >> no, I don't know about that.
> >>
> >> >
> >> > We could give you commit privileges for this work, or we could work
> >> > offline
> >> > with git, i.e., you could use git svn and I would pull from you to
> make
> >> > the
> >> > commits.
> >>
> >> if that works for you, that sounds good.  Should I clone from trunk,
> >> or is it going to be in a separate branch?
> >>
> >
> > If things work like we want, the changes will have to end up in trunk.
> > That's not to say you can't work in a git branch. One concern I have is
> > pulling between git repos that have been rebased to svn, as I understand
> it
> > this can lead to problems. OTOH, I think things should work if both git
> > repos are rebased to the same svn revision. Hmm, it might be best to work
> in
> > a branch that is kept rebased to the master which is kept in sync with
> svn.
> > Maybe someone who knows a bit more about git can weigh in here.
> >
> > Chuck
> >
>
>
> Well, for now I can just send patches with a svn diff... if you'd be
> so kind as to apply them after a review :)
>
> Integrating working changes into trunk incrementally seems like a good
> idea (not know numpy dev process very well).
>
>
Numpy relies on nose for testing. I know that there is a py3k branch for
nose but it doesn't look very active and I don't know its current state. Do
you know anything about that?

Chuck
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20090918/fd317acd/attachment.html>