How to correctly handle initMethodSelfParamTypeVar errors

[runfalk]

Since pyright was made stricter in a recent release I have encountered initMethodSelfParamTypeVar for my FrozenMap implementation. It currently looks something like this:

import collections.abc
import itertools
import typing as t

K = t.TypeVar("K")
V = t.TypeVar("V")


class FrozenMap(t.Generic[K, V], collections.abc.Mapping[K, V]):
    """An immutable dict implementation"""

    __slots__ = ("_dict", "_hash")
    _dict: dict[K, V]
    _hash: int | None

    @t.overload
    def __init__(self) -> None: ...

    @t.overload
    def __init__(self, mapping: collections.abc.Mapping[K, V]) -> None: ...

    @t.overload
    def __init__(self, it: collections.abc.Iterable[tuple[K, V]]) -> None: ...

    @t.overload
    def __init__(self: "FrozenMap[str, V]", **kwargs: V) -> None: ...

    def __init__(self, *args, **kwargs) -> None:  # type: ignore
        self._dict = dict(*args, **kwargs)
        self._hash = None

    ...

def __init__(self: "FrozenMap[str, V]", **kwargs: V) -> None: ... is no longer allowed, but unless I keep it in my K will be Unknown. Even though this now raises an error, it still resolves the correct type of str for the key. What is the proper way to handle this going forwards? If I add an explicit ignore, will it likely break in the future?

I'm using strict mode.

[erictraut]

The typing spec was recently updated to clarify that class-scoped type variables cannot be used in the self parameter of an __init__ method, so pyright's error is correct here.

You can use method-scoped type variables in this case, so that's the approach I'd recommend. Create a new type variable (say, T), and change your last overload signature to be:

    @t.overload
    def __init__(self: "FrozenMap[str, T]", **kwargs: T) -> None: ...

[runfalk]

Perfect, I had misinterpreted the change as if self couldn't be type hinted. This does exactly what I want. Thanks!

[erictraut]

If you're interested in the spec, here is a link.