Reputation: 4537
I have a list of slices
and use them to index a numpy
array.
arr = np.arange(25).reshape(5, 5)
# 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]])
slice_list = list(map(lambda i: slice(i, i+2), [1, 2]))
# [slice(1, 3, None), slice(2, 4, None)]
print(arr[slice_list])
# == arr[1:3, 2:4]
# [[ 7 8]
# [12 13]]
This works fine but it breaks if I have fewer slices than the number of dimensions of the array I want to index.
arr3d = arr[np.newaxis, :, :] # dims: [1, 5, 5]
arr3d[:, slice_list]
# IndexError: only integers, slices (`:`), ellipsis (`...`),(`None`)
# numpy.newaxis and integer or boolean arrays are valid indices
The following examples work however:
arr3d[:, slice_list[0], slice_list[1]]
arr3d[[slice(None)] + slice_list]
arr3d[:, [[1], [2]], [2, 3]]
Is there a way I can use a list of slices to index an array with more dimensions. I want to do things like:
arr[..., slice_list]
arr[..., slice_list, :]
arr[:, slice_list, :]
without thinking about the dimensions of the array and figuring out how many [slice(None)]*X
I have to pad on either side of my slice_list
.
Upvotes: 3
Views: 3244
Reputation: 59741
You can do that using tuples of slices and ellipsis objects. Just put all the elements to you want to use for indexing into a tuple and use it as index:
import numpy as np
arr = np.arange(24).reshape(2, 3, 4)
print(arr)
# [[[ 0 1 2 3]
# [ 4 5 6 7]
# [ 8 9 10 11]]
#
# [[12 13 14 15]
# [16 17 18 19]
# [20 21 22 23]]]
slice_tup = tuple(map(lambda i: slice(i, i+2), [1, 2]))
print(slice_tup)
# (slice(1, 3, None), slice(2, 4, None))
print(arr[slice_tup])
# [[[20 21 22 23]]]
# arr[..., slice_list]
print(arr[(Ellipsis, *slice_tup)])
# [[[ 6 7]
# [10 11]]
#
# [[18 19]
# [22 23]]]
# arr[..., slice_list, :]
print(arr[(Ellipsis, *slice_tup, slice(None))])
# [[[20 21 22 23]]]
# arr[:, slice_list, :]
print(arr[(slice(None), *slice_tup, slice(None))])
# IndexError: too many indices for array
Upvotes: 3