[Python] pyarrow.array segfaults when passed masked array with shrunken mask

[asfimport]

>>> import numpy as np, pyarrow as pa                               
>>> pa.array(np.ma.array([0]))

I think the issue might be here because when the masked array has a shrunken mask, it's values is False rather than the boolean mask array.

        if isinstance(values, np.ma.MaskedArray):
            if mask is not None:
                raise ValueError("Cannot pass a numpy masked array and "
                                 "specify a mask at the same time")
            else:
                mask = values.mask
                values = values.data

Reporter: Daniel Nugent / @nugend
Assignee: Daniel Nugent / @nugend

PRs and other links:

• GitHub Pull Request #6612

_{Note: This issue was originally created as ARROW-8105. Please see the migration documentation for further details.}

[asfimport]

Wes McKinney / @wesm:
I don't think the numpy.ma path has seen a lot of use (because numpy.ma is not widely used) so it doesn't surprise me that there are some issues like this

[asfimport]

Daniel Nugent / @nugend:
Put up a PR. Independent of the big: if masked arrays are not widely used, does that mean they should be avoided?

[asfimport]

Wes McKinney / @wesm:
I think it's fine for us to "unbox" them into Arrow format. I wouldn't recommend building applications based on them nowadays since the level of support / compatibility in other projects is low.

[asfimport]

Joris Van den Bossche / @jorisvandenbossche:
Issue resolved by pull request 6612
#6612