Discussion: new operators for numerical computation

[Huaiyu Zhu]

On Thu, 20 Jul 2000 18:18:10 -0500, Travis Oliphant <olipt at mayo.edu> wrote:
>
>These are my sentiments too.  Huaiyu, it's obvious you've spent a lot of
>time in Matlab but not so obvious that you've considered the design of
>NumPy very much.  There was A LOT of discussion that wen't in to NumPy to
>make it more generic than 2-D Matlab (I wouldn't be using it if it
>weren't).  That's why you meet resistance when you seem to want to change
>syntax to make just 2-D Matrix notation easier.  

That is a very good point.  Would you like to summarise some of the design
issues here?  I would be especially interested in seeing some tensor
operations that's candidate for infix notation, like kron, and how they are
currently done.  Some example codes that uses the expressions would be most
helpful, imho.

>I think that 6+5 operators will never fly.  I think we'll be lucky if we
>get four into Python itself.  I suppose if you want to maintain a
>separate patched-interpreter than you can do whatever you feel best about,
>though.. ;-)

Well, as I said, that's not really a proposal.  It's just one way to list
the options so as to be as compatible with everyone as possible.  If we
agree that's the case we can start to suggest what's the best migration
path.  I would be glad to see MatPy and NumPy merge in the future, as long
as both interfaces remain available, as both make sense to a lot of people.

Huaiyu