Fix indexing on ma.mvoid to be consistent with void; fixes #6724
#6755
Conversation
ma.mvoid to be consistent with void; fixed #6724ma.mvoid to be consistent with void; fixes #6724
574d209 to
5d69826
Compare
…y.void Make indexing on numpy.ma.mvoid consistent with indexing on numpy.void. Does not change non-broken behaviour. Fixes numpy#6724. Sometimes, indexing ma.mvoid results in a non-scalar mask. For example, dimension increases if indexing with a multi-dimensional field. Previously, this led to a ValueError (truth value ambiguous). With this commit, indexing now returns an ma.masked_array so that there is no loss of information. Note that there is a precedence for returning from void to array. Z = zeros((2,), dtype="(2,)i2,(2,)i2"), then Z[0] is a void, but Z[0][0] and Z[0]["f1"] are array. This commit therefore implements behaviouk such that numpy.ma.mvoid is consistent with numpy.void. Also adds a related test.
5d69826 to
f36db77
Compare
|
Had to wrap my head around this one, but agree you are making it do the right thing. |
| @@ -5869,6 +5869,18 @@ def __getitem__(self, indx): | |||
|
|
|||
| """ | |||
| m = self._mask | |||
| if m[indx].size > 1: | |||
There was a problem hiding this comment.
It is possible to have subarrays of size 1, eg dtype(["A", "i2", (1,1,1)]) so this probably needs something better. m[indx].shape == () was an idea, but that won't work for object types. Probably you just have to check isinstance(m[indx], np.ndarray).
There was a problem hiding this comment.
@ahaldane True, my reasoning was that the status quo does not fail for subarrays of size 1. So there is a trade-off between changing existing behaviour, or being consistent with new behaviour.
There was a problem hiding this comment.
Now that I've tested I see your point.
But what if we get really crazy with a subarray of size 1 of structured dtype:
>>> a = np.ma.array(np.ones(2, dtype=[('A', 'i4,i4', (1,1,1)), ('B', 'i4')]),
mask=[([[[(0, 0)]]], 0), ([[[(0, 0)]]], 0)])
>>> a[0]['A']
masked
There was a problem hiding this comment.
It appears that crazier dtypes do not get assigned to the mask properly anyway:
In [234]: Z = ma.zeros(2, dtype([("A", "(2,2)i1,(2,2)i1", (2,2))]))
In [237]: print(Z.dtype.descr)
[('A', [('f0', '|i1', (2, 2)), ('f1', '|i1', (2, 2))], (2, 2))]
In [238]: print(Z.mask.dtype.descr)
[('A', '|b1', (2, 2))]
That's bound to lead to problems somewhere as well...
|
@gerritholl I'm going to trust you and @mhvk on this one unless @ahaldane wants to do the honors. |
|
If I find time tonight I will think again about my alternative vs. ahaldanes alternative, I might not have time before tomorrow night though. |
|
I haven't had a chance to test it yet because of build issues! But I will soon. |
|
Closing in favor of #6763, this can be reopened if discussion warrants it. |
Repair indexing on
ma.mvoidso that it is consistent withvoid. Changes only behaviour currently broken. Fixes #6724.Currently, or a standard
ndarray, givenZ = zeros((2,), dtype="(2,)i2,(2,)i2"),Z[0]isvoidbutZ[0][0]andZ[0]["f1"]arendarray. However, trying the same withma.mvoidresults inValueError. With this commit, indexing ama.mvoidwith a similar dtype results inma.masked_array.