🐛 compatibility issues with ArrayAPI and SparseAPI Protocols in `namedarray` about xarray HOT 2 OPEN

Note that on python 3.12 and higher this will actually start to work because it will not use hasattr anymore (which, by calling getattr, evaluates properties / descriptors).

See also #8605

from xarray.

Isn't this a bug in sparse?

import numpy as np
import numpy.array_api as xp

import sparse

a_np = np.array(['a', 'b']).astype(object)
a_sparse = sparse.COO(a_np)
print(a_np.real) # works
print(a_sparse.real) # ValueError: invalid literal for int() with base 10: 'a'

a_xp = xp.asarray(a_np) # TypeError: The array_api namespace does not support the dtype 'object'

is_duck_array looks like this:

By using _arrayfunction_or_api you will guarantee that mypy finds places where namedarray uses more methods than the requirement and this happens quite often as there are many places where we assume (incorrectly?) numpy arrays are used.

from xarray.