I was learning boolean indexing in numpy and came across this. How is the indexing below not producing a Index Error as for axis 0 as there are only two blocks?
x = np.arange(30).reshape(2, 3, 5)
x
array([[[ 0, 1, 2, 3, 4],
[ 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14]],
[[15, 16, 17, 18, 19],
[20, 21, 22, 23, 24],
[25, 26, 27, 28, 29]]])
x[[[True, True, False], [False, True, True]]]
array([[ 0, 1, 2, 3, 4],
[ 5, 6, 7, 8, 9],
[20, 21, 22, 23, 24],
[25, 26, 27, 28, 29]])
You are performing boolean array indexing, which is fine.
You would have an indexing error with:
# first dimension second dimension
x[[True, True, False], [False, True, True]]
# IndexError: boolean index did not match indexed array along dimension 0;
# dimension is 2 but corresponding boolean dimension is 3
However, in your case, you have an extra set of brackets, which makes it index only the first dimension, using an array:
# [ first dimension ]
# [ second dimension], [ second dimension]
x[[[True, True, False], [False, True, True]]]
This means, that using a 2x3 array, you request in the second dimension, for the first "row" [True, True, False], and for the second "row" [False, True, True].
Since your array shape matches the first two dimensions, this is valid, and more or less equivalent to:
np.concatenate([x[0][[True, True, False]],
x[1][[False, True, True]]])