BUG: numpy.ma: Fix ma.in1d

[tirthasheshpatel]

Fixes #19877

Fixed ma.in1d and add tests cases from gh-19877.

I just rewrote the function to use numpy.in1d instead of reinventing the wheel for masked arrays. To me, this seems more maintainable. I don't know if there is any reason to handle masked arrays differently. If so, sorry; I'll close this PR.

[mattip]

Hi @tirthasheshpatel we discussed this at the last triage meeting and would like to review/merge. Could you rebase off main in order to pick up any CI changes that might have happened since this was first submitted?

[tirthasheshpatel]

we discussed this at the last triage meeting and would like to review/merge.

Great, thanks!

Could you rebase off main in order to pick up any CI changes that might have happened since this was first submitted?

Done.

[mattip]

Seems to make sense to me. @WarrenWeckesser want to take a look?

[stefanv]

Shouldn't the output array be masked by the intersection of the mask of ar1 and ar2?

[WarrenWeckesser]

For this function, the mask of the output array is determined by the mask of the first argument ar1 only. This is not a pairwise operation; the mask of the second argument actually has no effect on the output.

[WarrenWeckesser]

You already have ar1.mask in m, so here and in the next line, use m instead of ar1.mask.

[WarrenWeckesser]

... and if you prefer the readability of ar1.mask, you could rename m to ar1_mask.

[WarrenWeckesser]

It looks like this changed the behavior of ma.in1d with multidimensional array arguments. In the main development branch, this works:

In [6]: ma.in1d([[0, 1, 2, 3], [4, 5, 6, 7]], [0, 2, 3])
Out[6]: 
masked_array(data=[ True, False,  True,  True, False, False, False, False],
             mask=False,
       fill_value=True)

But in this branch, we get an exception:

In [2]: ma.in1d([[0, 1, 2, 3], [4, 5, 6, 7]], [0, 2, 3])
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-2-a44d06ca9c57> in <cell line: 1>()
----> 1 ma.in1d([[0, 1, 2, 3], [4, 5, 6, 7]], [0, 2, 3])

~/py3.10.1/lib/python3.10/site-packages/numpy/ma/extras.py in in1d(ar1, ar2, assume_unique, invert)
   1208         res[~m] = out
   1209     else:
-> 1210         res[:] = out
   1211     return res
   1212 

~/py3.10.1/lib/python3.10/site-packages/numpy/ma/core.py in __setitem__(self, indx, value)
   3375         if _mask is nomask:
   3376             # Set the data, then the mask
-> 3377             _data[indx] = dval
   3378             if mval is not nomask:
   3379                 _mask = self._mask = make_mask_none(self.shape, _dtype)

ValueError: could not broadcast input array from shape (8,) into shape (2,4)