1

I have a 3d array in numpy that I would like to index on two of these axes. For one of the axes I am using a simple integer index, whereas on the other axis I use a list indexing. If I apply the indexes separately, the result is different from when I apply them within one pair of brackets:

import numpy as np
a = np.ones((10, 20, 30))

print(a.shape)
> (10, 20, 30)

# the unexpected result:
print(a[3,:,[1,2]].shape)
> (2, 20)

# the two expected results:
print(a[3][:,[1,2]].shape)
> (20, 2)

print(a[:,:,[1,2]][3].shape)
> (20, 2)

Could you please explain to me why this is the expected/desired behaviour? For me, this looks inconsistent.

Milla Well
  • 3,193
  • 3
  • 35
  • 50
  • This is case of mixed basic and advanced indexing, documented in https://docs.scipy.org/doc/numpy-1.13.0/reference/arrays.indexing.html#combining-advanced-and-basic-indexing. It's also been discussed in previous SO questions. In the 'unexpected' case, the sliced dimension has been placed at the end of the result. – hpaulj Apr 27 '18 at 16:26

0 Answers0