Stub methods for abstract tzinfo to allow usage with mypy 1.0.0 #126
Conversation
|
Hm. Kind of a weird choice on |
Fixes pganssle#125. Ported from: python/typeshed@de0b366 Signed-off-by: Anders Kaseorg <[email protected]>
| @@ -30,6 +30,9 @@ class ZoneInfo(tzinfo): | |||
| ) -> _T: ... | |||
| @classmethod | |||
| def clear_cache(cls, *, only_keys: Iterable[str] = ...) -> None: ... | |||
| def tzname(self, __dt: datetime | None) -> str | None: ... | |||
| def utcoffset(self, __dt: datetime | None) -> timedelta | None: ... | |||
There was a problem hiding this comment.
I am not too concerned with this since it's maybe better taken up with typeshed, but I am not sure that these are actually the right type signatures for these functions.
They can take a time as well, and they never return None if you pass them a datetime, so the correct signatures are probably:
@overload
def utcoffset(self, __dt: datetime) -> timedelta: ...
@overload
def utcoffset(self, __dt: Union[time, None]) -> Union[timedelta, None]: ...
I created a PR upstream in typeshed for discussion: python/typeshed#9862
There was a problem hiding this comment.
I continue to request that this be merged as-is, because the purpose of a backport is to optimize compatibility with the released upstream version, not to pursue incompatible improvements. Obviously you’re welcome to continue pursuing those improvements upstream, and then we can backport them here in the future. But there’s no reason to block this current backport on that.
|
The typeshed and mypy issues aren’t going to be straightforward to resolve. Would you be open to merging and releasing this now, so that the library can be used with mypy ≥ 1.0.0, and we can later update it for future typeshed and mypy fixes as appropriate? |
Fixes #125. Ported from python/typeshed@de0b366.