Bug#387137: [Pkg-octave-devel] Bug#387137: spkron(a, b) not working when arguments are scalar

[Rafael Laboissiere]

* John W. Eaton <jwe at bevo.che.wisc.edu> [2006-09-12 11:38]:

> On 12-Sep-2006, Thomas Weber wrote:
> 
> | Hi, 
> | 
> | Am Dienstag, den 12.09.2006, 14:39 +0200 schrieb Nicolas Guilbert:
> | > Package: octave2.9
> | > Version: 2.9.8-1
> | > 
> | > Steps to reproduce:
> | > octave-2.9.8:1> spkron(2,3)
> | > error: octave_base_value::sparse_matrix_value(): wrong type argument `scalar'
> | > error: octave_base_value::sparse_matrix_value(): wrong type argument `scalar'
> | > 
> | > would expect
> | > 
> | > ans =  6
> | 
> | Try
> | 	spkron(sparse(2), sparse(3))
> | 
> | I don't consider this a bug. The overhead for sparse-implementations
> | just doesn't make sense for scalars.
> 
> Also, why are you using spkron directly?  I think it would be better
> to use kron instead.  It will handle both
> 
>   kron (2, 3)
>   kron (sparse (2), sparse (3))
> 
> Once Octave has proper dispatching, spkron may only be available via
> kron with sparse arguments rather than being directly callable.

I tend to agree with the both of you.  However the current on-line help for
spkron is misleading:

########################################################################
octave2.9:1> help spkron
 -- Function File:  spkron (A, B)
     Form the kronecker product of two sparse matrices. This is defined
     block by block as

          x = [a(i, j) b]

     For example,

          kron (1:4, ones (3, 1))
          =>  1  2  3  4
                    1  2  3  4
                    1  2  3  4
########################################################################

Through its example, one may think that spkron(2,3) should be valid.
Notice that in the example, full matrices are used (!:4 and ones(3,1)).
Shouldn't that be fixed?

-- 
Rafael