sparse indexing issue

[David Bateman]

John W. Eaton wrote:
> On 18-Dec-2007, David Bateman wrote:
> 
> | The attached patch fixes all the bugs reported in this thread. These
> | bugs are either only recent (Change from 9th Oct to speed up the sparse
> | indexing) or only became obvious due to recent changes (cf
> | sparse_auto_mutate) for the sparse(42)([1,1]) case. I added test cases
> | for them so that they shouldn't creep back.
> 
> Thanks, I applied the patch and checked in the change.
> 
> | *** ./test/build_sparse_tests.sh.orig9	2007-12-18 21:34:24.452169877 +0100
> | --- ./test/build_sparse_tests.sh	2007-12-18 21:08:44.963865073 +0100
> | ***************
> | *** 811,816 ****
> | --- 811,823 ----
> |   %!assert(as((size(as,1):-1:1),:),sparse(af((size(af,1):-1:1),:)))
> |   %!assert(as(:,(size(as,2):-1:1)),sparse(af(:,(size(af,2):-1:1))))
> |   
> | + %% Indexing tests
> | + %!assert(full(as([1,1],:)), af([1,1],:))
> | + %!assert(full(as(:,[1,1])), af(:,[1,1]))
> | + %!test
> | + %! [i,j,v] = find (as);
> | + %! assert (as(i(1),j(1))([1,1]), sparse([v(1),v(1)]))
> 
> Maybe we should also test the indexing against a value constructed
> without indexing, like
> 
>   as = sparse (42);
>   assert (as([1,1]), sparse([42,42]))
> 
> to avoid the possibility that the code for indexing full matrices has
> the same bug?
> 
> jwe


Sure if you like.. Do you want a patch? In any case the reason I did it
like the above is that "as" can be real or complex and so the above
implicitly treats both cases with the manner this script works..

D.