make sure marker colors also accept np.array, fixes #8750

[joergdietrich]

PR Summary

Setting markerface/edgecolor with an np.array leads to a ValueError with numpy-1.13. See #8750 As suggested in the issue thread I replaced the comparison with np.array_equal(). I modified the unit test for marker styles to test a range of color specifications.

PR Checklist

• [ x] Has Pytest style unit tests
• [ x] Code is PEP 8 compliant
• [ N/A] New features are documented, with examples if plot related
• [ N/A] Documentation is sphinx and numpydoc compliant
• [ N/A] Added an entry to doc/users/whats_new.rst if major new feature
• [ N/A] Documented in doc/api/api_changes.rst if API changed in a backward-incompatible way

First PR to matplotlib. Hope this is fine.

[joergdietrich]

Ah well, it seems np.array_equal doesn't work with numpy-1.7.1.

In [6]: equal('auto', 'k')
Out[6]: NotImplemented

Back to the drawing board.

[afvincent]

Huh. It is weird because one can find these kind of functions in the 1.5 reference guide. See for example np.equal, np.array_equal or np.array_equiv.

Edit:
For the record

In [15]: np.__version__
Out[15]: '1.11.1'  # so I guess the issue is not Numpy 1.7 ^^

In [16]: np.equal('auto', 'k')
Out[16]: NotImplemented

In [17]: np.array_equal('auto', 'k')
Out[17]: False

In [18]: np.array_equiv('auto', 'k')
Out[18]: False

[tacaswell]

Internally colors should be 4-tuples for actual colors or strings for 'special' values (like 'none', 'edge', or 'auto'). I think @anntzer made great progress on this actually being true. Instead of trying to run this through numpy, just checking isinstance(v, six.text_types) should be enough.

[joergdietrich]

@afvincent Ok, so the problem is that in numpy-1.7.1 array_equal uses

return bool(equal(a1,a2).all())

which trips on NotImplemented but starting with 1.8 it is

return bool(asarray(a1 == a2).all())

[dstansby]

How about if not np.all(self._markerfacecolor == fc) instead? (ie. if at least one markerfacecolor is differend to fc)

[dstansby]

Or if np.any(self._markerfacecolor != fc) would be closer to the original and do the same

[joergdietrich]

Thanks, @dstansby! Not sure what the failing codecov/patch check wants to tell me.

[dstansby]

I'm not entirely sure why codecov/patch is failing either, all the changed lines seem to be tested. Thanks for the patch!

[anntzer]

thanks!