Problems with Enums: iteration, __new__

[ilinum]

Consider the following perfectly valid python code:

from enum import IntEnum, Enum

class Colors(IntEnum):
    RED = 0
    BLUE = 1

class Colors2(int, Enum):  # E: Definition of "__new__" in base class "int" is incompatible with definition in base class "Enum"
    GREEN = 0
    PURPLE = 1

for color in Colors:  # E: Type[Colors] has no attribute "__iter__"
    print(color.name)

for color in Colors2:  # E: Type[Colors2] has no attribute "__iter__"
    print(color.name)

This probably has to do with support for metaclasses. Here is the output from mypy and python.

$ mypy e.py
e.py:7: error: Definition of "__new__" in base class "int" is incompatible with definition in base class "Enum"
e.py:11: error: Iterable expected
e.py:11: error: Type[Colors] has no attribute "__iter__"
e.py:14: error: Iterable expected
e.py:14: error: Type[Colors2] has no attribute "__iter__"
$ mypy e.py --py2
e.py:7: error: Definition of "__new__" in base class "int" is incompatible with definition in base class "Enum"
e.py:11: error: Iterable expected
e.py:11: error: Type[Colors] has no attribute "__iter__"
e.py:14: error: Iterable expected
e.py:14: error: Type[Colors2] has no attribute "__iter__"
$ python3 e.py
RED
BLUE
GREEN
PURPLE
$ python e.py
RED
BLUE
GREEN
PURPLE

[gvanrossum]

The __new__ problem is probably separate; the lack of iteration is because IntEnum doesn't have the right metaclass (EnumMeta), and trying to add it gives a different error:

__tmp__.py:8: error: Inconsistent metaclass structure for 'Colors2'

[ilevkivskyi]

This seems to be partially related to #3210, and as I mentioned there, this may be solved easily by structural subtyping. However, the __new__ part looks harder to fix.

[ilinum]

The __new__ issue seems to be addressed here: python/typeshed#1464

(Haven't had time to test it to make sure though)

[JelleZijlstra]

Oh right, this is a near-duplicate of the typeshed issue I opened. I think there may be more regressions here since the last mypy release, so I'll look into this more (hopefully tonight).

[elazarg]

I believe this is a duplicate of #2824

[elazarg]

More general: python/typeshed#1595

[emmatyping]

Closing as a duplicate of #2824