BUG: where keyword argument is ignored in _array_ufunc__/__array_function__ if it is not an instance of numpy.ndarray #23219
Description
Describe the issue:
If the where keyword argument is a ndarray duck type that implements __array_ufunc__ and __array_function__, I would've expected its implementation of those methods to be used. Instead, it appears that where is just ignored if it is not an instance of numpy.ndarray.
Reproduce the code example:
import dataclasses
import numpy as np
@dataclasses.dataclass
class DuckArray(np.lib.mixins.NDArrayOperatorsMixin):
value: np.ndarray
def __array_ufunc__(self, ufunc, method, *inputs, **kwargs):
print("in DuckArray.__array_ufunc__")
inputs = [inp.value if isinstance(inp, DuckArray) else inp for inp in inputs]
kwargs = {k: kwargs[k].value if isinstance(kwargs[k], DuckArray) else kwargs[k] for k in kwargs}
return DuckArray(getattr(ufunc, method)(*inputs, **kwargs))
def __array_function__(self, func, types, args, kwargs):
print("in DuckArray.__array_function_")
args = [arg.value if isinstance(arg, DuckArray) else arg for arg in args]
kwargs = {k: kwargs[k].value if isinstance(kwargs[k], DuckArray) else kwargs[k] for k in kwargs}
return DuckArray(func(*args, **kwargs))
def test_ufunc():
a = np.random.random(5)
b = DuckArray(a > 0.5)
c = np.negative(a, where=b) # doesn't print "in DuckArray.__array_ufunc__" :(
print(type(c)) # prints <class 'numpy.ndarray'>
print(c)
def test_function():
a = np.random.random(5)
b = DuckArray(a > 0.5)
c = np.sum(a, where=b) # doesn't print "in DuckArray.__array_function_" :(
print(type(c)) # prints <class 'numpy.float64'>
print(c)Error message:
No response
Runtime information:
1.24.1
3.11.1 (tags/v3.11.1:a7a450f, Dec 6 2022, 19:58:39) [MSC v.1934 64 bit (AMD64)]
print(numpy.show_runtime())
WARNING: `threadpoolctl` not found in system! Install it by `pip install threadpoolctl`. Once installed, try `np.show_runtime` again for more detailed build information
[{'simd_extensions': {'baseline': ['SSE', 'SSE2', 'SSE3'],
'found': ['SSSE3',
'SSE41',
'POPCNT',
'SSE42',
'AVX',
'F16C',
'FMA3',
'AVX2'],
'not_found': ['AVX512F',
'AVX512CD',
'AVX512_SKX',
'AVX512_CLX',
'AVX512_CNL',
'AVX512_ICL']}}]
None
Context for the issue:
I am working on an array duck type that tries to propagate uncertainties similar to astropy.uncertainty.Distribution. I would like to make sure that uncertainties propagate properly in the case where the where keyword argument is uncertain, but the inputs are exact.