Integers and booleans can often be used interchangeably (in fact, bool is an int subclass in builtin python). Created_ArrayLikeIntOrBool to reflect this compatibility.
Rename _ScalarGeneric to _ScalarGenericDT; add a new _ScalarGeneric TypeVar which is only bound to np.generic.
np.choose() can only accept array-like objects consisting of integers (or booleans); reflect this in its annotation.
np.argpartition() will return an np.integer if a scalar is passed; not an np.ndarray.
Further Notes
It seems np.partition() might actually bugged in NumPy 1.17.3.
it's default value for axis is set to -1 instead of None. Consequently, it will raise an AxisError if a scalar is passed unless one explicitly passes axis=None to the function.
>>> import numpy as np
>>> np.partition(0, 0, axis=None)
array([0])
>>> np.partition(0, 0)
Traceback (most recent call last):
...
AxisError: axis -1 is out of bounds for array of dimension 0
person142
commented
4 years ago
Follow-up on https://github.com/numpy/numpy-stubs/pull/71#issuecomment-619435429.
https://github.com/numpy/numpy-stubs/pull/67 Fixes & Updates
boolis anintsubclass in builtin python). Created_ArrayLikeIntOrBoolto reflect this compatibility._ScalarGenericto_ScalarGenericDT; add a new_ScalarGenericTypeVar which is only bound tonp.generic.np.choose()can only accept array-like objects consisting of integers (or booleans); reflect this in its annotation.np.argpartition()will return annp.integerif a scalar is passed; not annp.ndarray.Further Notes
np.partition()might actually bugged in NumPy1.17.3. it's default value foraxisis set to-1instead ofNone. Consequently, it will raise anAxisErrorif a scalar is passed unless one explicitly passesaxis=Noneto the function.