mypy treats Enum's __call__ as calling __init__
Dieses Issue hat noch niemand übernommen.
- Vorherrschende Sprache
- Python
- Sterne
- 20.6k
- Forks
- 3.3k
- PR-Merge-Kennzahlen
- PR-Kennzahlen ausstehend
Beschreibung
Bug Report
The standard library's enum.Enum has a class method† __call__ that can take a member of the enumeration, or a member's value, and returns that member in either case.
It is also possible to override the enumeration's __init__. The Python docs show an example that uses this to attach additional data to the members.
If this is done, mypy reports an error when calling the class, expecting the call to follow the signature of __init__.
To Reproduce
This is a stripped-down example of the one from the Python docs. It runs successfully (the asserts pass).
from enum import Enum
class Planet(Enum):
MERCURY = (3.303e+23, 2.4397e6)
def __init__(self, mass: float, radius: float) -> None:
self.mass = mass
self.radius = radius
mercury = Planet(Planet.MERCURY)
assert mercury is Planet.MERCURY
mercury = Planet((3.303e+23, 2.4397e6))
assert mercury is Planet.MERCURY
Expected Behavior
One possibility, of course, would be for mypy to produce no errors. But this is a somewhat odd situation, so failing that, there should be a way‡ to signal to mypy that this is what's happening.
I couldn't find any such way, short of # type: ignore. Overriding and annotating __call__ didn't work:
@classmethod
def __call__(cls, value_or_member: Planet | tuple[float, float]) -> Planet:
return type(Enum).__call__(cls, value_or_member)
This definition type-checks fine, but it doesn't affect the previous errors. (It's worth noting that this doesn't seem to actually override anything; the calls bypass it and go for the metaclass.)
It is possibly to go the other way and adapt the code to suit mypy, by making the use of __call__ explicit. This type-checks fine:
mercury = Planet.__call__(Planet.MERCURY)
assert mercury is Planet.MERCURY
mercury = Planet.__call__((3.303e+23, 2.4397e6))
assert mercury is Planet.MERCURY
But that seems backwards to me.
Searching for existing issues or Q&As didn't turn up anything that looked like it matched. There are some issues related to the functional API (e.g. #10469, a dupe of #6037), which also uses __call__, but mypy specifically supports that. There's also #15024, but that appears to be about __call__ on an instance.
Actual Behavior
main.py:10: error: Missing positional argument "radius" in call to "Planet" [call-arg]
main.py:10: error: Argument 1 to "Planet" has incompatible type "Planet"; expected "float" [arg-type]
main.py:13: error: Missing positional argument "radius" in call to "Planet" [call-arg]
main.py:13: error: Argument 1 to "Planet" has incompatible type "tuple[float, float]"; expected "float" [arg-type]
Found 4 errors in 1 file (checked 1 source file)
Your Environment
- Mypy version used: 0.982 locally, latest on playground
- Mypy command-line flags and config options: none
- Python version used: 3.11.6 locally, 3.12 on playground
† Technically it's implemented on the metaclass rather than by @classmethod.
‡ Preferably something obvious, to be PEP 20-compliant. 😉 ("There should be one—and preferably only one—obvious way to do it. Although that way may not be obvious at first unless you're Dutch." Maybe the problem is that I'm not Dutch?)
Beitragsleitfaden
Erste Schritte
- Lies das ganze Issue und danach den Beitragsleitfaden des Projekts.
- Schreib ins Issue, dass du es übernimmst — das erspart doppelte Arbeit.
- Forke das Repository und arbeite in einem Branch.
- Öffne einen Pull Request, der die Issue-Nummer nennt.
Rechercherichtung
Beginne mit dem verlinkten Playground-Reproducer und den gemeldeten Enum-Aufrufen, die ein überschriebenes init betreffen. Verfolge, wie mypy die aufrufbare Signatur für eine Enum-Klasse ableitet, und vergleiche sie mit dem im Issue beschriebenen Verhalten von call der Metaklasse. Die Aufgabe ist erledigt, wenn der Reproducer ohne ignore eine Typprüfung besteht und dabei die bestehende Prüfung der init-Argumente erhalten bleibt.
Vom Indexierungsmodell aus dem Issue-Text verfasst.
Bewertung
- Tech-Stack
- python
- Bereich
- compilers, devtools
- Issue-Typ
- Bug
- Schwierigkeit
- 4/5
- Geschätzter Aufwand
- 3-5 Tage
- Aktivitätsstatus
- Veraltet
- Klarheit
- Größtenteils klar
- Anfängerfreundlichkeit
- 45/100