[dataclasses plugin] Support kw_only=True

[tgallant]

Description

Fixes #10865

If all fields in a dataclass are keyword only then mypy should not raise errors regarding the order of attributes with default values. See the linked issue for a more detailed description.

This PR makes the relevant changes to handle that.

Test Plan

I've added tests to check-dataclasses.test to cover the new changes.

[cdce8p]

Came across an interesting use case of KW_ONLY with inheritance, that should probably be supported as well (if not already done).

from dataclasses import dataclass, KW_ONLY

@dataclass
class Base:
    x: str
    _: KW_ONLY
    y: int = 0
    w: int = 1

@dataclass
class D(Base):
    z: str
    a: str = "a"

>>> D("Hello", "World")
D(x='Hello', y=0, w=1, z='World', a='a')

That works because of keyword reordering.
https://docs.python.org/3.10/library/dataclasses.html#re-ordering-of-keyword-only-parameters-in-init

[hauntsaninja]

Thanks for working on this! Could we also test that __init__ has the appropriate signature (e.g. by instantiating the classes)

[tgallant]

Thanks for raising this case @cdce8p. I've added this example as a test case and updated the code to handle it. @hauntsaninja I also made sure to check the initialization in the test cases.

I was looking through the docs again and spotted this:

In a single dataclass, it is an error to specify more than one field whose type is KW_ONLY.

https://docs.python.org/3.10/library/dataclasses.html#dataclasses.KW_ONLY

It seemed like this would be a good thing to have a mypy check for. I added a check for this case in commit 75ef138

[hauntsaninja]

python/typeshed#5826 is now on mypy master, so if you merge you'll be able to pull that in.

It would be good to add tests of class instantiation that produce errors. Specifically, test cases should prove that the arguments are in fact keyword only and complain when passed positionally and that parameter reordering happens and we get type errors (i.e. make sure positional parameters have types that are distinct from each other and from kw only parameters)

[tgallant]

i rebased this branch with master and updated the KW_ONLY usage according to the typeshed change. i will follow up with some more test cases

[tgallant]

thanks for the helpful suggestions @hauntsaninja and @cdce8p. i've added test cases covering positional vs keyword arguments, reordering, and type errors. let me know if you think there are other cases that should be covered in this PR.

[hauntsaninja]

Thanks, this is great! :-)

[cdce8p]

Thanks @tgallant 🎉