Numpy.array - wrong dtype for 'string enum'

When creating a numpy.array of a class that inherits from both str and Enum the dtype is dtype='<UX where X is the length of the string. the data saved is X chars from the string representation of the enum.

I would expect the minimal value to be object, so the Enum is preserved. Problem is of course solved when passing object as dtype

Reproducing code example:

>>> import numpy as np
>>> from enum import Enum
>>> class StrEnum(str, Enum):
...    A = 'a'
...    B = 'bb'
...
>>> np.array([StrEnum.A])
array(['S'], dtype='<U1')
>>> np.array([StrEnum.B])
array(['St'], dtype='<U2')

Numpy/Python version information:

1.18.0 3.7.5 (default, Nov 1 2019, 02:16:23) [Clang 11.0.0 (clang-1100.0.33.8)]

[alonme]

Not very familiar with the C-API but will be happy to get my hands dirty. Ill try and take a look.

[seberg]

We currently also coerce subclasses of types that NumPy knows about to store them into arrays. I agree with it being bad behaviour, in the long run, I would like to deprecate it.