python
Wrong type inference with "yield from" from an Iterator class
Bug Report
When using "yield from" for an iterable instance the type inference if wrong. Normal for-loop iteration over the same iterable does work though.
To Reproduce
from typing import Iterator, Self
class goes_to_11(Iterator[int]):
def __init__(self):
self.elevens = iter((11,))
def __next__(self) -> int:
return next(self.elevens)
def __iter__(self) -> Self:
return self
def bunch_of_11() -> Iterator[tuple[int, ...]]:
yield from goes_to_11()
for integer in goes_to_11():
yield integer
for value in bunch_of_11():
print(type(value), value)
Expected Behavior
Mypy should complain about both the "yield from" line and the "yield integer" line, but for the former it doesn't.
Actual Behavior
mypy --strict buggy.py
buggy.py:19: error: Incompatible types in "yield" (actual type "int", expected type "tuple[int, ...]") [misc]
Found 1 error in 1 file (checked 1 source file)
Your Environment
Tested on Linux Python 3.12.3 using mypy wheel mypy-1.11.0+dev.b88fdbd32fe0a45d40531a6504317aa3fd48489e-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
As a reference point, pyright does warn about this (output taken from a reduce version of the above):
(tools_venv3) ➜ m python3 -m pyright buggy.py
/home/arni/m/buggy.py
/home/arni/m/buggy.py:16:16 - error: Return type of generator function must be compatible with "Generator[int, Any, Any]"
"Generator[int, Unknown, Unknown]" is incompatible with "Generator[tuple[int, ...], None, None]"
Type parameter "_YieldT_co@Generator" is covariant, but "int" is not a subtype of "tuple[int, ...]"
"int" is incompatible with "tuple[int, ...]" (reportReturnType)
1 error, 0 warnings, 0 informations
The reduced example:
from typing import Generator, Iterator, Self
class goes_to_11(Iterator[int]):
def __init__(self) -> None:
self.elevens = iter((11,))
def __next__(self) -> int:
return next(self.elevens)
def __iter__(self) -> Self:
return self
def bunch_of_11() -> Generator[tuple[int, ...]]:
yield from goes_to_11()
for value in bunch_of_11():
print(type(value), value)
It seems, the error lays in the get_generator_yield_type function which is not able to get the yielding type for such a constellation accuratley and assumes Any
