[Numpy-discussion] Toward release 1.0 of NumPy

[Sven Schreiber]

[Sorry for the late reaction, I was on vacation.]

Tim Hochberg schrieb:

>>
> Here's my best guess as to what is going on:
>    1. There is a relatively large group of people who use Kronecker
> product as Alan does (probably the matrix as opposed to tensor math
> folks). I'm guessing it's a large group since they manage to write the
> definitions at both mathworld and planetmath.

Yes.

>    2. kron was meant to implement this.

That's what I thought, anyway.

>    2.5 People who need the other meaning of kron can just use outer, so
> no real conflict.
>    3. The implementation was either inappropriately generalized or it
> was assumed that all inputs would be matrices (and hence rank-2).
> 
> Assuming 3. is correct, and I'd like to hear from people if they think
> that the behaviour in the non rank-2 cases is sensible, the next
> question is whether the behaviour in the rank-2 cases makes sense. It
> seem to, but I'm not a user of kron. If both of the preceeding are true,
> it seems like a complete fix entails the following two things:
>    1. Forbid arguments that are not rank-2. This allows all matrices,
> which is really the main target here I think.
>    2. Fix the return type issue. I have a fix for this ready to commit,
> but I want to figure out the first part as well.
> 

Both 1 and 2 sound very good to me as a user.

So, should I still submit a new ticket about kron, or is it already
being fixed?

Greetings,
Sven