incorrect inferred type accessing typed dict with static keys #19961
Description
Bug Report
mypy does not infer the correct type of a dictionary when constructing a subset of a TypedDict using a constant subset of keys.
To Reproduce
https://mypy-play.net/?mypy=latest&python=3.12&gist=82fc1eb24628e1ba2aed8294b81e2912
from typing import TypedDict, reveal_type
class MyDict(TypedDict):
key_0: bool
key_1: int
key_2: bool
limits: MyDict = {
'key_0': True,
'key_1': 1,
'key_2': False
}
def test_func() -> dict[str, bool]:
result = {
label: limits.get(label, False)
for label in ('key_0', 'key_2')
}
reveal_type(limits.get('key_0', False))
reveal_type(result)
return result
test_func()Expected Behavior
In the example above, I expect mypy to accept the return type since both key_0 and key_2 are annotated as bool
Actual Behavior
mypy complains about the return type with
Incompatible return value type (got "dict[str, object]", expected "dict[str, bool]") [return-value]
After mentioning it in my team chat @LefterisJP noticed that this version works https://mypy-play.net/?mypy=latest&python=3.12&gist=63bece9a183493b170f6f55527fbd7d3
from typing import TypedDict, reveal_type, Literal
class MyDict(TypedDict):
key_0: bool
key_1: int
key_2: bool
limits: MyDict = {
'key_0': True,
'key_1': 1,
'key_2': False
}
def test_func() -> dict[str, bool]:
keys: tuple[Literal['key_0', 'key_2'], ...] = ('key_0', 'key_2')
result: dict[str, bool] = {
label: limits.get(label, False)
for label in keys
}
reveal_type(limits.get('key_0', False))
reveal_type(result)
return result
test_func()In this case, because the keys are explicitly typed as Literals, mypy infers the correct type.
This suggests that mypy is not analyzing constant tuples deeply enough to recognize the accessed keys.
Your Environment
- Mypy version used: mypy 1.18.2 (compiled: yes)
- Mypy command-line flags: None
- Mypy configuration options from
mypy.ini(and other config files): I can reproduce it in an environment without any extra configuration. - Python version used: 3.11.13