non-self-return-type (PYI034)

Derived from the flake8-pyi linter.

What it does

Checks for methods that are annotated with a fixed return type, which should instead be returning self.

Why is this bad?

If methods like __new__ or __enter__ are annotated with a fixed return type, and the class is subclassed, type checkers will not be able to infer the correct return type.

For example:

class Shape:
    def set_scale(self, scale: float) -> Shape:
        self.scale = scale
        return self

class Circle(Shape):
    def set_radius(self, radius: float) -> Circle:
        self.radius = radius
        return self

# This returns `Shape`, not `Circle`.
Circle().set_scale(0.5)

# Thus, this expression is invalid, as `Shape` has no attribute `set_radius`.
Circle().set_scale(0.5).set_radius(2.7)

Specifically, this check enforces that the return type of the following methods is Self:

1. In-place binary operations, like __iadd__, __imul__, etc.
2. __new__, __enter__, and __aenter__, if those methods return the class name.
3. __iter__ methods that return Iterator, despite the class inheriting directly from Iterator.
4. __aiter__ methods that return AsyncIterator, despite the class inheriting directly from AsyncIterator.

Example

class Foo:
    def __new__(cls, *args: Any, **kwargs: Any) -> Bad:
        ...

    def __enter__(self) -> Bad:
        ...

    async def __aenter__(self) -> Bad:
        ...

    def __iadd__(self, other: Bad) -> Bad:
        ...

Use instead:

from typing_extensions import Self


class Foo:
    def __new__(cls, *args: Any, **kwargs: Any) -> Self:
        ...

    def __enter__(self) -> Self:
        ...

    async def __aenter__(self) -> Self:
        ...

    def __iadd__(self, other: Bad) -> Self:
        ...

References

• PEP 673