But that looks like a variable display issue (a different bug?)

octave:17> a= zeros(2)
a =

   0   0
   0   0

octave:18> as = sparse (a)
as =

Compressed Column Sparse (rows = 2, cols = 2, nnz = 0 [0%])


octave:19> aa = full(as)
aa =

   0   0
   0   0

octave:20> as(1)
ans = Compressed Column Sparse (rows = 1, cols = 1, nnz = 0 [0%])

octave:21> as(1)+1
ans = 1
octave:22> 1+ as(1)
ans = 1
octave:23> whos as
Variables visible from the current scope:

variables in scope: top scope

  Attr   Name        Size                     Bytes  Class
  ====   ====        ====                     =====  =====
   s     as          2x2                         40  double

Dmitri.
--

Not sure why this wasn't reviewed four years ago, but this is still a real problem.  Minimum example below.

octave:1> a = 0
a = 0
octave:2> b = 2
b = 2
octave:3> a .^ b
ans = 0
octave:4> as = sparse (a)
as = Compressed Column Sparse (rows = 1, cols = 1, nnz = 0 [0%])

octave:5> bs = sparse (b)
bs = Compressed Column Sparse (rows = 1, cols = 1, nnz = 1 [100%])

  (1, 1) -> 2
octave:6> as .^ bs
ans = Compressed Column Sparse (rows = 1, cols = 1, nnz = 1 [100%])

  (1, 1) -> 1

I'm attaching a patch to fix the functions in sparse-xpow.cc.  I'll need to also add tests.  I'd be glad to see any improvements in either the algorithm or the use of templates for this function.

(file #48517)

I think the attached function does the right indexing when there is no need to convert to complex.  A proper replacement for the elem_xpow function in sparse-xpow.cc also needs to handle the complex case.  And since the indexing should be the same for the real and complex cases, maybe it should be done with a template rather than duplicating the code.


(file #48511)

The element-by-element sparse exponentiation functions miss the case when the LHS is 0 and the RHS is non-zero and the result should be zero.  In the following, I believe the results should agree for both sparse and full, and that the sparse result is wrong:

a = sparse (4, 6);
b = sparse (4, 6);
a(2,2) = 1;
b(3,3) = 2;
a(3,4) = 3; b(2,4) = 4;
a(2,5) = 5; b(3,5) = 6;
a(4,6) = 7; b(4,6) = 8;
full (a .^ b) =>

         1         1         1         1         1         1
         1         1         1         1         1         1
         1         1         1         1         1         1
         1         1         1         1         1   5764801

full (a) .^ full (b) =>

         1         1         1         1         1         1
         1         1         1         0         1         1
         1         1         0         1         0         1
         1         1         1         1         1   5764801