[Numpy-discussion] the path forward

Keith Goodman kwgoodman@gmail....
Tue Apr 29 19:40:01 CDT 2008


On Tue, Apr 29, 2008 at 5:18 PM, Alan G Isaac <aisaac@american.edu> wrote:
> On Tue, 29 Apr 2008, Keith Goodman apparently wrote:
>  > I hope that changing x[0,:] is considered a major change since it will
>  > break most any package based on matrices (mine). And so
>  > I hope that such a change wouldn't show up, if at all,
>  > until 2.0.
>
>  What if the extant matrix class would continue to be
>  available for awhile as "oldmatrix", which you could then
>  import as "matrix" if you wished.  Would that meet your
>  needs?  (I am simply assuming this would be feasible,
>  without being sure how a lot of the special casing for
>  matrices has been done.)

I never import matrix directly (is that what you are suggesting?). I
usually create it with M.ones, M.rand, x * y, etc., where M is

import numpy.matlib as M

If a big change is made to matrix behavior could it be accessed in 1.X
from the __future__ and, if successful, switched to the "present" in
2.X? I, obviously, have no idea what would be involved to make that
happen.

In my use, changing x[0] is not a big deal, but changing x[0,:] is.


More information about the Numpy-discussion mailing list