Description
(Opening a new issue for this as discussed in #12390.)
Live demo: https://mypy-play.net/?mypy=latest&python=3.10&flags=strict&gist=d3e1a06857e4c6f1c698b7fa89e22f40
Example code:
import typing as t
class A:
@t.overload
def pop(self) -> int: ...
@t.overload
def pop(self, d: int = ...) -> int: ...
def pop(self, d: int = ...) -> int: ...
class B(A):
@t.overload
def pop(self) -> int: ...
@t.overload
def pop(self, d: int) -> int: ...
def pop(self, d: int = ...) -> int: ...
Current mypy output:
main.py:14: error: Signature of "pop" incompatible with supertype "A"
main.py:14: note: Superclass:
main.py:14: note: @overload
main.py:14: note: def pop(self) -> int
main.py:14: note: @overload
main.py:14: note: def pop(self, d: int = ...) -> int
main.py:14: note: Subclass:
main.py:14: note: @overload
main.py:14: note: def pop(self) -> int
main.py:14: note: @overload
main.py:14: note: def pop(self, d: int) -> int
Found 1 error in 1 file (checked 1 source file)
It's somewhat helpful that mypy quotes the code in this error output, but it would be really helpful if mypy's output helped us zero in on exactly where the problem is, e.g.:
main.py:14: error: Signature of "pop" incompatible with supertype "A" in second overload
main.py:14: note: Superclass:
main.py:14: note: @overload
main.py:14: note: def pop(self) -> int
main.py:14: note: @overload
main.py:14: note: def pop(self, d: int = ...) -> int
^^^^^
main.py:14: note: Subclass:
main.py:14: note: @overload
main.py:14: note: def pop(self) -> int
main.py:14: note: @overload
main.py:14: note: def pop(self, d: int) -> int
Found 1 error in 1 file (checked 1 source file)
Note the addition of "in second overload" in the error message, as well as the all-important ^
arrows pointing to exactly where the supertype differs (namely, d
should be a keyword argument). Without the arrows, it's harder to see in the output that the only difference is the = ...
in the code for the superclass.
This is exacerbated when method signatures are more complex than this, e.g. MutableMapping.pop()
, where the current output makes it even harder to spot where the problem is compared to this minimized example (see #12390 (comment)). It is also exacerbated by mypy ignoring the signature in the actual implementation, where d
correctly appears as a keyword argument in the subclass. Finally, it can be exacerbated when upgrading to a mypy version that includes typeshed changes like this one: python/typeshed@92438dc#diff-b12c4be0f8e39f89957bbe7be6b6327185e18b69ea80a0f62c0746bb26f4a58eL478