Discussion: new operators for numerical computation

[Travis Oliphant]

> We are planning to use Numeric (or its future equivalent) heavily for
> our software. One thing we would strongly object to is making 
> matrix operators the default. The vast majority of our manipulations 
> (on images and spectra) involve elementwise operations. This would be
> consistent with what most astronomers would expect (and how IDL--currently
> the most commonly used array-based language in the astronomical community 
> by far--handles it).

I also take this view.  Definining * to be matrixwise should be handled by
subclassing the generic array (which will be much easier to do
without breaking compatibility with compiled extension modules in the
future Numeric).
> 
> Use of parentheses for matrix operators that Konrad Hinsen proposed
> appears to be the best solution I've seen so far. 

I agree.  I'm glad Konrad is still with actively with us.  His ideas were
also pivotal in getting Numeric into its current useful state.  I do like
Tim Hochberg's suggestion of (|) as matsolve.

The only question is can the (*) be implemented in the current grammar?  

I don't think we should start introducing [*] {*} variants.  That would
likely weaken our case in the PEP.  

I like:

(*)  outer product
(.)  inner product
(|)  matrix solve
(**) matrix power  
     (^) is better but only if ^ is power (which is not likely!)

I'm not sure we should blow our chance to get some useful operators into
Python by over-reaching...  Also, unless you've used Numeric with Python
for a while, it's easy to over-value the use of the infix notation.
This really bugged me when I first started using Numeric Python as
well, but it doesn't bother me anymore.  The syntax does not have
to be as bad as Huaiyu demonstrates.  

-Travis