[PYTHON MATRIX-SIG] Re: NumPy in Python 1.4

[Jim Hugunin]

>  > I don't have an argument as to why this was chosen over the old
version,
>  > but note that you can do:
>  > 
>  > 	zmin = min(ravel(zz))
>  > 
>  > I am in the process of updating the tutorial to NumPy to document
these
>  > changes from the early alpha releases.  

The last time I checked, there was no min function that works as you're
requesting.  the following will work: minimum.reduce(zz.flat).  The problem
here is that min is a builtin python function with a different meaning from
the expected one for arrays.

Notice that while reshape is no longer a method on array objects, but
rather a function, a new data member has been added called flat which will
give a 1d flat indexed version of any array.

> More documentation on NumPy would be most desirable.  Also, I find the
> distribution format of the 1.0a2 to be more difficult to work with
> than the ones that were for 1.3.  It was easier to use when it was
> distributed as a direct patch to the main Python source tree.  I say
> this b/c the (maybe just my) goal is to add the numerical extension to
> the Python I am using for general system use, not merely to build a
> new python with the numerical extension.

There are two ways of building NumPy1.0a5.  One of those (and by far the
easiest) is directly in the Python source tree.  This should be as easy to
work with as the previous releases.

> I understand.  OTOH, when a package is in "alpha" for extremely long
> periods of time, people are driven to just use it as is, since waiting
> is not practical for all users, if the wait will be measured in
> months.  I know--I am the author of a package which has been in
> "alpha" distribution status for 3+ years.  :-).  Drives even the most
> stability conscious of would-be users into alpha testers.

Well, I don't mind "early adopters", but on the other hand, I make no
apologies for the fact that you will have to change skads of code when
switching to later versions.  I've probably had to change far more code
(and for each new revision) than any other users, so I'm a bit short on
sympathy here.

OTOH - The point of the current release is that it should be stable.  I'm
finally happy enough with how things work to be willing to worry about
minor issues like backwards compatibilty.

-Jim

=================
MATRIX-SIG  - SIG on Matrix Math for Python

send messages to: matrix-sig@python.org
administrivia to: matrix-sig-request@python.org
=================