[Numpy-discussion] argmax & py2exe
jmiller at stsci.edu
Fri Aug 6 13:55:12 EDT 2004
I was on leave this week so I'm getting to this later than usual, but
Perry and I discussed it this morning. We concluded that this is the
MODEL numarray problem report because:
1. You reported the problem.
2. You suggested a fix even I could implement in a half hour or less.
3. You declared the problem a non-problem before I could waste a single
solitary second on it. Well, except this one anyway.
Everyone take note because I think Tim is on to something. We need more
bug reports exactly like this one...
On Thu, 2004-08-05 at 16:39, Tim Hochberg wrote:
> Tim Hochberg wrote:
> Please ignore this. The problem went away when I deleted numerics pyc
> files and started over. It's a mysterious phenomena, that I ran into
> once before with py2exe and stale pyc files. Very strange, but no time
> to dig into it now since things appear to be working again. Sorry for
> the confusion.
> Although getting rid of those circular imports would still probably be
> good style.
> > I just ran into a little snag using py2exe with CVS numarray: argmax
> > breaks in py2exe packaged programs although it works fine when run
> > normally. My believe is that py2exe is getting confused by the trick
> > of using 'import numeric' inside the function argmax, which I believe
> > is there to avoid circular imports. In any event, generic.argmax and
> > numarraycore.NumArray.argmax mysteriously ending up calling each other
> > recursively until the recursion limit is exceeded.
> > This is almost certainly a problem with py2exe, not with numarray, but
> > since circular imports are sort of bad form and since numarray side of
> > things is almost certainly easier to to fix, I though I'd try to
> > persuade you, the numarray developers, to reorganize argmax so the
> > circular imports go away.
> > I think that this is mostly a matter of moving the
> > numeric.argmax/argmin code into NumArray.argmax/argmin and then
> > numeric.argmax/argmin can just call a.argmax/a.argmin on the array a
> > that it is passed./
> > -tim
> > -------------------------------------------------------
> > This SF.Net email is sponsored by OSTG. Have you noticed the changes on
> > Linux.com, ITManagersJournal and NewsForge in the past few weeks? Now,
> > one more big change to announce. We are now OSTG- Open Source Technology
> > Group. Come see the changes on the new OSTG site. www.ostg.com
> > _______________________________________________
> > Numpy-discussion mailing list
> > Numpy-discussion at lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/numpy-discussion
> This SF.Net email is sponsored by OSTG. Have you noticed the changes on
> Linux.com, ITManagersJournal and NewsForge in the past few weeks? Now,
> one more big change to announce. We are now OSTG- Open Source Technology
> Group. Come see the changes on the new OSTG site. www.ostg.com
> Numpy-discussion mailing list
> Numpy-discussion at lists.sourceforge.net
More information about the NumPy-Discussion