1

I have the following 3D array of shape

In [159]: arr = np.arange(60).reshape(3, 4, 5)

And I'm trying to do advanced indexing to extract a sub-array like:

# behaves as expected
In [160]: arr[[1, 2], :, 1].shape
Out[160]: (2, 4)

In the following case, I'd expect the result to be of shape (4, 2). 

# unintended behaviour
In [161]: arr[1, :, [1, 2]].shape
Out[161]: (2, 4)

Since we do __getitem__ call along first dimension that dimension would be gone. Along the second axis, we slice everything so it should be 4 and along the last axis it should be 2. So, we should get the resulting sub-array of shape (4, 2) but we get a shape of (2, 4) instead. Why is this ambiguity? How should I interpret the result?

kmario23
  • 57,311
  • 13
  • 161
  • 150
  • When mixing advanced and basic indexing (list and slice), the slice dimension may be tacked on to the end. It's discussed in the docs, and some other SO questions. You may need to transpose the result. – hpaulj Dec 30 '17 at 16:45

0 Answers0