Skip to content

Instantly share code, notes, and snippets.

@rpgoldman
Last active May 2, 2024 16:48
Show Gist options
  • Save rpgoldman/91b1d0b20e09a4b5f27dbf3d395a8217 to your computer and use it in GitHub Desktop.
Save rpgoldman/91b1d0b20e09a4b5f27dbf3d395a8217 to your computer and use it in GitHub Desktop.
Python file showing issue with (spurious?) overlap in overloads
from typing import overload, Literal, List, Tuple, Any
class Example:
pass
class Prediction:
pass
class TestMypy:
@overload
def __call__(
self,
program,
return_all_scores: Literal[True] = ...,
return_outputs: Literal[True] = ...,
) -> Tuple[float, List[Tuple[Example, Prediction, Any]], List[float]]:
...
@overload
def __call__(
self,
program,
return_all_scores: Literal[True] = ...,
return_outputs: Literal[False] = ...,
) -> Tuple[float, List[float]]:
...
@overload
def __call__(
self,
program,
return_all_scores: Literal[False] = ...,
return_outputs: Literal[True] = ...,
) -> Tuple[float, List[Tuple[Example, Prediction, Any]]]:
...
@overload
def __call__(
self,
program,
return_all_scores: Literal[False] = ...,
return_outputs: Literal[False] = ...,
) -> float:
...
def __call__(
self,
program,
return_all_scores: bool = False,
return_outputs: bool = False,
):
if return_all_scores:
if return_outputs:
return 0.1, [(Example(), Prediction(), 0.1)], [0.1]
else:
return 0.1, [0.1]
elif return_outputs:
return 0.1, [(Example(), Prediction(), 0.1)]
else:
return 0.1
$ poetry run mypy dspy/evaluate/mypy_issue.py
dspy/evaluate/mypy_issue.py:13: error: Overloaded function signatures 1 and 2 overlap with incompatible return types [overload-overlap]
dspy/evaluate/mypy_issue.py:13: error: Overloaded function signatures 1 and 3 overlap with incompatible return types [overload-overlap]
dspy/evaluate/mypy_issue.py:13: error: Overloaded function signatures 1 and 4 overlap with incompatible return types [overload-overlap]
dspy/evaluate/mypy_issue.py:22: error: Overloaded function signatures 2 and 3 overlap with incompatible return types [overload-overlap]
dspy/evaluate/mypy_issue.py:22: error: Overloaded function signatures 2 and 4 overlap with incompatible return types [overload-overlap]
dspy/evaluate/mypy_issue.py:31: error: Overloaded function signatures 3 and 4 overlap with incompatible return types [overload-overlap]
Found 6 errors in 1 file (checked 1 source file)
$ poetry run mypy dspy/evaluate/mypy_issue.py
dspy/evaluate/mypy_issue.py:13: error: Overloaded function signatures 1 and 2 overlap with incompatible return types [overload-overlap]
dspy/evaluate/mypy_issue.py:13: error: Overloaded function signatures 1 and 3 overlap with incompatible return types [overload-overlap]
dspy/evaluate/mypy_issue.py:13: error: Overloaded function signatures 1 and 4 overlap with incompatible return types [overload-overlap]
dspy/evaluate/mypy_issue.py:23: error: Overloaded function signatures 2 and 3 overlap with incompatible return types [overload-overlap]
dspy/evaluate/mypy_issue.py:23: error: Overloaded function signatures 2 and 4 overlap with incompatible return types [overload-overlap]
dspy/evaluate/mypy_issue.py:33: error: Overloaded function signatures 3 and 4 overlap with incompatible return types [overload-overlap]
Found 6 errors in 1 file (checked 1 source file)
from typing import overload, Literal, List, Tuple, Any
class Example:
pass
class Prediction:
pass
class TestMypy:
@overload
def __call__(
self,
program,
foo: int = ...,
return_all_scores: Literal[True] = ...,
return_outputs: Literal[True] = ...,
) -> Tuple[float, List[Tuple[Example, Prediction, Any]], List[float]]:
...
@overload
def __call__(
self,
program,
foo: int = ...,
return_all_scores: Literal[True] = ...,
return_outputs: Literal[False] = ...,
) -> Tuple[float, List[float]]:
...
@overload
def __call__(
self,
program,
foo: int = ...,
return_all_scores: Literal[False] = ...,
return_outputs: Literal[True] = ...,
) -> Tuple[float, List[Tuple[Example, Prediction, Any]]]:
...
@overload
def __call__(
self,
program,
foo: int = ...,
return_all_scores: Literal[False] = ...,
return_outputs: Literal[False] = ...,
) -> float:
...
def __call__(
self,
program,
foo: int = 5,
return_all_scores: bool = False,
return_outputs: bool = False,
):
if return_all_scores:
if return_outputs:
return foo, [(Example(), Prediction(), 0.1)], [0.1]
else:
return foo, [0.1]
elif return_outputs:
return foo, [(Example(), Prediction(), 0.1)]
else:
return foo
@rpgoldman
Copy link
Author

@Dr-Irv Yes, I just checked, and when I drop the = ... in my full code, mypy errors out because there are non-defaulted parameters following defaulted parameters. So I can't omit that (unless there's something I am missing). I will make a new version of the gist

@rpgoldman
Copy link
Author

The additional two files show what goes wrong if I omit the = ....

Note that this argument is not required, but it is determined by the method form's default arguments. So this still looks like a bug to me.

@Dr-Irv
Copy link

Dr-Irv commented May 2, 2024

These overloads worked:

    @overload
    def __call__(
        self,
        program,
        foo: float = ...,
        *,
        return_all_scores: Literal[True],
        return_outputs: Literal[True],
    ) -> Tuple[float, List[Tuple[Example, Prediction, Any]], List[float]]: ...

    @overload
    def __call__(
        self,
        program,
        foo: float = ...,
        *,
        return_all_scores: Literal[True],
        return_outputs: Literal[False],
    ) -> Tuple[float, List[float]]: ...

    @overload
    def __call__(
        self,
        program,
        foo: float = ...,
        *,
        return_all_scores: Literal[False],
        return_outputs: Literal[True],
    ) -> Tuple[float, List[Tuple[Example, Prediction, Any]]]: ...

    @overload
    def __call__(
        self,
        program,
        foo: float = ...,
        *,
        return_all_scores: Literal[False],
        return_outputs: Literal[False],
    ) -> float: ...

@rpgoldman
Copy link
Author

Thank you so much! I am very grateful. I made the corresponding changes to the real code from which this MWE was taken and all is well now.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment