two_d = np.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]])
first = np.array((True, True, False, False, False))
second = np.array((False, False, False, True, True))
Now, when I enter:
two_d[first, second]
I get:
array([3,9])
which doesn't make a whole lot of sense to me. Can anybody explain that simply?
When given multiple boolean arrays to index with, NumPy pairs up the indices of the True values. The first true value in first in paired with the first true value in second, and so on. NumPy then fetches the elements at each of these (x, y) indices.
This means that two_d[first, second] is equivalent to:
two_d[[0, 1], [3, 4]]
In other words you're retrieving the values at index (0, 3) and index (1, 4); 3 and 9. Note that if the two arrays had different numbers of true values an error would be raised!
The documents on advanced indexing mention this behaviour briefly and suggest np.ix_ as a 'less surprising' alternative:
Combining multiple Boolean indexing arrays or a Boolean with an integer indexing array can best be understood with the
obj.nonzero()analogy. The functionix_also supports boolean arrays and will work without any surprises.
Hence you may be looking for:
>>> two_d[np.ix_(first, second)]
array([[3, 4],
[8, 9]])
np.ix_ to get the desired result.Check the documentation on boolean indexing.
two_d[first, second] is the same as
two_d[first.nonzero(), second.nonzero()], where:
>>> first.nonzero()
(array([0, 1]),)
>>> second.nonzero()
(array([3, 4]),)
Used as indices, this will select 3 and 9 because
>>> two_d[0,3]
3
>>> two_d[1,4]
9
and
>>> two_d[[0,1],[3,4]]
array([3, 9])
Also mildy related: NumPy indexing using List?
