np.where returns tuple instead of array

Contradictory to the documentation, np.where returns a tupel with the new array instead of only the array.

Reproducing code example:

import numpy as np

a =  np.arange(10)
np.where(a < 5)
# (array([0, 1, 2, 3, 4]),)

Numpy/Python version information:

1.17.2 3.7.4 and also 1.18.4 3.7.7

[eric-wieser]

It’s unfortunate that where was built this way. The simplest mental model is “never use where(arr), always use nonzero(arr)”. If you follow that rule, no one reading your codebase has to work out what these docs mean in the first place!

[rossbar]

I’m going to close this due to inactivity - PRs with proposed wording improvements are of course still welcome!