Fixing "GCXS matmul => slice leads to incorrect results"

[EuGig]

It fixes #607. The solution simply inverts the order of indices and of data before returning them in _dot_csr_csr.
(I did not write any test, since I'm not implementing a new feature. Let me know if I have to do that.)

[hameerabbasi]

Could you perhaps add a regression test, I.e. a test that would trigger the bug before and now doesn't.

[codecov]

Codecov Report

Merging #611 (b8e1818) into master (7c0c962) will increase coverage by 0.02%.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #611      +/-   ##
==========================================
+ Coverage   91.95%   91.97%   +0.02%     
==========================================
  Files          20       20              
  Lines        3330     3340      +10     
==========================================
+ Hits         3062     3072      +10     
  Misses        268      268              

[EuGig]

Please ignore the last commit. I got an unexpected _utills.py.swp file. I will get rid of it. For the regression test I've thought of something like that:

Code

# tests for good slicing
assert np.array_equal(x[0].data, y[0])

# inverting indices and data to cause slicing error
x.indices = x.indices[::-1]
x.data = x.data[::-1]
assert not np.array_equal(x[0].data, y[0])

where x is the sparse version of a tensor product and y is the dense one.

[hameerabbasi]

It seems like a lot of existing tests are failing -- Would it be possible for you to look into that?

Failing tests.

[EuGig]

Yes, of course, already on that.

[EuGig]

Hi @hameerabbasi, is it normal to have failing tests without any modification? I got 178 tests failed after simply cloning the repo.

[hameerabbasi]

That's quite odd -- I'll look into it in depth this weekend, but certainly not normal.

[hameerabbasi]

Hi @hameerabbasi, is it normal to have failing tests without any modification? I got 178 tests failed after simply cloning the repo.

Did you do an editable pip install before running the tests? If not, that's normal.

[EuGig]

Oh yeah sorry, I fixed that. Now tests are passing.

[EuGig]

Actually, pre-commit.ci is failing. I don't know why!

[hameerabbasi]

Actually, pre-commit.ci is failing. I don't know why!

That can be ignored for now.

[hameerabbasi]

Thanks for the fix, @EuGig!