Omit constraints referencing dict if a conflicting TypedDict constraint exists
#19225
Conversation
This comment has been minimized.
This comment has been minimized.
|
Diff from mypy_primer, showing the effect of this PR on open source code: core (https://github.com/home-assistant/core)
+ homeassistant/core.py:2342: error: Unused "type: ignore" comment [unused-ignore]
+ homeassistant/scripts/benchmark/__init__.py:144: error: Unused "type: ignore" comment [unused-ignore]
+ homeassistant/scripts/benchmark/__init__.py:144: error: Argument 2 to "async_fire" of "EventBus" has incompatible type "dict[str, object]"; expected "EventStateChangedData | None" [arg-type]
+ homeassistant/scripts/benchmark/__init__.py:144: note: Error code "arg-type" not covered by "type: ignore" comment
+ homeassistant/scripts/benchmark/__init__.py:177: error: Unused "type: ignore" comment [unused-ignore]
+ homeassistant/scripts/benchmark/__init__.py:177: error: Argument 2 to "async_fire" of "EventBus" has incompatible type "dict[str, object]"; expected "EventStateChangedData | None" [arg-type]
+ homeassistant/scripts/benchmark/__init__.py:177: note: Error code "arg-type" not covered by "type: ignore" comment
+ homeassistant/scripts/benchmark/__init__.py:215: error: Unused "type: ignore" comment [unused-ignore]
+ homeassistant/scripts/benchmark/__init__.py:215: error: Argument 2 to "async_fire" of "EventBus" has incompatible type "dict[str, object]"; expected "EventStateChangedData | None" [arg-type]
+ homeassistant/scripts/benchmark/__init__.py:215: note: Error code "arg-type" not covered by "type: ignore" comment
|
Do you think there's any neat way to handle prioritization in such a general way that it solves this too? (i.e. marking constraints as higher quality than others) I did some basic version of that in #18958 and I would be happy if there's a method that generalizes to solve both this issue and that issue. |
There was a problem hiding this comment.
Never mind I thought more about my comment and any nice generalizations I can think of would rely on the dict argument being inferred as a TypedDict. I think this makes sense.
It looks like literals don't work in a similar fashion either, should we preemptively add support?
a_lit = A[Literal[1]]()
f(a_lit, 1)
Fixes #19201.
When inferring callable constraints, excludes constraints with
dicttarget when other constraints haveTypedDicttype. The right way to do this would be to allow backtracking in meet and join (so thatdictcan be promoted toTypedDictif possible), but that seems to be way more difficult and needs significant refactoring first.This should be safe as we never rely solely on solutions, such inference is followed by actually checking all arguments, so any incompatibilities will still be reported.