Re: goals for 3.1

[David Bateman]

John W. Eaton wrote:
> On 19-Feb-2008, David Bateman wrote:
> 
> | John W. Eaton wrote:
> | >    1. Objects:
> | >         -- Merge object-branch
> | >         -- Implementation of superiorto/inferiorto
> | >         -- Operator overloading vs. constant folding
> | >   
> | I suppose that now that this code is in Octave the use of dispatch
> | should be discouraged, within the core of Octave at least. Though
> | dispatch itself is useful so I'd hate to see it go away. The only code
> | within Octave itself that uses dispatch is within the DLD-FUNCTIONS code
> | sparse.cc, splu.cc, spchol.cc, spdet.cc, spfind.cc and spqr.cc. That is
> | entirely the sparse code. However, for matlab compatibility the sparse
> | type is of the class "double/bool" and there is no indication of its
> | sparse nature within the class itself, and so the new object code can
> | dispatch to a sparse type basedon its class.
> | 
> | Therefore, should all of the sparse dispatch code be merged with the
> | corresponding functions in the same manner as spkron recently was? Or
> | are we happy to leave the dispatched sparse functions as is?
> 
> I think these should be merged with the corresponding full functions
> if possible, so the dispatch on various types of double objects
> (real, complex, sparse) is handled inside the DEFUN itself, or by an
> octave_value member function.
> 

The attached patch gets rid of some of them. An open question is whether
atan2(full,sparse) should return a full or sparse matrix (I chose
sparse) as I don't have access to matalab at the moment.

Regards
David

patch7555.bz2
Description: BZip2 compressed data