Closed
Description
>>> arr = LArray([0.1, 0.2], Axis([False, True], 'gender'))
>>> key = np.array([True, False, True, True])
>>> arr[key]
ValueError: boolean key with a different shape ((4,)) than array ((2,))
>>> arr.points[key]
ValueError: boolean key with a different shape ((4,)) than array ((2,))
Oddly, using list keys seems to work (as long as we only target a single dimension)
>>> arr[[True, False, True, True]]
gender True False True True
0.2 0.1 0.2 0.2
>>> arr.points[[True, False, True, True]]
gender True False True True
0.2 0.1 0.2 0.2
Using an LArray key does not seem to help either:
>>> arr[LArray(key, Axis(6, 'gender'))]
ValueError: gender[0, 2, 3] is not a valid label for any axis
This is a "regression" in 0.29. It is only a partial regression because it was known to be broken before. It's just more broken now. 😄 See liam2/liam2#296.