The partition and argpartition functions currently raise when the kth indices are supplied as a uint64 array-like.

The functions in question check that the indice-type can safelly be cast to intp, which of course will fail in the case of uint64. I'm not quite sure how the likes of __getitem__ and take do manage to deal with uint64, but there it seems to work just fine.

Example

In [1]: import numpy as np

In [2]: ar = np.arange(10)

In [3]: ar.partition(np.uint64(5))
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-5-fee31ee69481> in <module>
----> 1 ar.partition(np.uint64(5))

TypeError: Partition index must be integer

NumPy/Python version information:

1.22.0.dev0+997.g8e4e40f60
3.9.6 | packaged by conda-forge | (default, Jul 11 2021, 03:36:15) 
[Clang 11.1.0 ]