Colon, None, slice(None) in numpy array indexers

Question

As for a = np.arange(24).reshape(2,3,4)

a[0,:,1] or a[0,slice(None),1] outputs array([1, 5, 9])

while a[0,None,1] gives array([[4, 5, 6, 7]])

Could sb explain the latter?

Answer 1

Using a raw None (not in slice) is the same thing as using np.newaxis, of which it is but an alias.

In your case:

• a[0,None,1] is like a[0,np.newaxis,1], hence the output
• whereas slice(None) is like "slice nothing", which is why a[0,:,1] is the same as a[0,slice(None),1]. See numpy's Indexing doc.

Answer 2

a[0,None,1] is the same as a[0, 1] but with an extra axis in the result.

The newaxis object can be used in all slicing operations to create an axis of length one. :const: newaxis is an alias for ‘None’, and ‘None’ can be used in place of this with the same result.

So a[0,None,1] is the same as a[0,np.newaxis,1]

In this case, where None is placed is not of relevance, but every None adds a new axis.

>>> a[0,None, 1]
array([[4, 5, 6, 7]])
>>> a[None,None,0,1]
array([[[4, 5, 6, 7]]])
>>> a[0,np.newaxis,1]
array([[4, 5, 6, 7]])