Update to mypy 1.15 #3201
Update to mypy 1.15 #3201
Conversation
In python/typeshed#13372, the type of the stdlib's `socket.getaddrinfo` was updated to add a union member for cases where Python was compiled with `--disable-ipv6`. This causes custom resolvers that use the stdlib function to fail mypy This is probably going to (mypy) break everyone's custom `HostnameResolver`, since the parent type is changing
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #3201 +/- ##
====================================================
- Coverage 100.00000% 99.95744% -0.04256%
====================================================
Files 124 124
Lines 18792 18796 +4
Branches 1268 1268
====================================================
- Hits 18792 18788 -4
- Misses 0 8 +8
|
|
I guess this PR should be more generally "update to mypy 1.15"... And major typing changes are fine, type hints are optional/not at runtime anyways. |
CoolCat467
added
dependencies
|
OK, I'll take a look at the other typing changes needed for 1.15 as well |
|
looks like we need to replace a lot of |
There was a problem hiding this comment.
I actually just was looking at doing this and thinking through what I would prefer, but I actually prefer the choices you made here (e.g. pushing up the CoroutineType change rather than littering code with assert isinstance(task.coro, CoroutineType))
A couple comments, but most are very nitpicky so feel free to ignore.
| additional_dependencies: | ||
| - tomli |
There was a problem hiding this comment.
Without this change, pre-commit was not picking up the config settings in pyproject.toml. I'm curious how others haven't been running into this.
There was a problem hiding this comment.
Well I have Python 3.13 locally and so maybe it uses tomllib (3.11+) instead? Wouldn't be surprised if it's just a variant of that for everyone ... or they just don't run pre-commit.
There was a problem hiding this comment.
Yeah, I'm using 3.10 in the venv I do trio work in.
There was a problem hiding this comment.
I've been meaning to add this for quite some time, thanks for catching this!
| class RunSync(Generic[RetT]): # type: ignore[explicit-any] | ||
| fn: Callable[..., RetT] # type: ignore[explicit-any] |
| match that of the stdlib `socket.getaddrinfo`, which was updated in mypy 1.15 (via | ||
| a typeshed update) to include the possibility of ``tuple[int, bytes]`` for the | ||
| ``sockaddr`` field of the result. This happens in situations where Python was compiled | ||
| with ``--disable-ipv6``. |
There was a problem hiding this comment.
It's probably worth calling out the improved typing around to_thread/from_thread functions too.
|
|
||
|
|
||
| # Decorator to mark methods public. This does nothing by itself, but | ||
| # trio/_tools/gen_exports.py looks for it. | ||
| # Explicit "Any" is not allowed | ||
| def _public(fn: F) -> F: # type: ignore[misc] | ||
| def _public(fn: Callable[P, T]) -> Callable[P, T]: |
There was a problem hiding this comment.
This should not be ParamSpec. There's a subtle difference between this and a callable-bound TypeVar - ParamSpec can't capture an overload since the return type might change, but the typevar form preserves it, so it would be passed through unchanged.
In python/typeshed#13372, the type of the stdlib's
socket.getaddrinfowas updated to add a union member for cases where Python was compiled with--disable-ipv6. This causes custom resolvers that use the stdlib function to fail mypyThis is probably going to (mypy) break everyone's custom
HostnameResolver, since the parent type is changing