Add conformance tests for basic generic spec

[hauntsaninja]

No description provided.

[erictraut]

This gets into constraint-solving behavior which is underspecified, but I'm still surprised to see that mypy generates an Any in this case. That's not what I would expect, and it seems inconsistent with mypy's handling of non-constrained TypeVars.

[erictraut]

It may also be useful to add tests for accessing a TypeVar as a runtime object — to make sure that this isn't disallowed.

assert_type(T.__name__, str)
assert_type(T.__bound__, Any | None)
T.other # Type error

runtime_typevar: TypeVar = T # OK

[hauntsaninja]

@erictraut I don't think we expect any "derives from protocol but doesn't implement all methods" errors here, or at least I don't see any.

On the following file pyright 1.1.345 has no complaints:

from typing import TypeVar
from collections.abc import Iterator
T = TypeVar('T')
class MyIter1(Iterator[T]): ...
MyIter1()

mypy, pyre and pytype are all fine with the class definition, but treat it as an abstract class and error when you attempt to instantiate.

I can still add the missing methods if you want, but wanted to make sure I wasn't missing something / we should add this to the list of things to exactly specify.

[erictraut]

but treat it as an abstract class and error when you attempt to instantiate

Yeah, that makes sense. After reviewing the text in the Explicitly Declaring Implementation section, I agree that pyright should emit those errors only if it's marked @final. I'll make that change.

[hauntsaninja]

Okay, this is mergeable. Remaining items:

• Add conformance tests for basic generic spec #1553 (comment) I agree mypy's behaviour here is surprising. I'll look into changing it and report back
• Add conformance tests for basic generic spec #1553 (comment) These proved controversial, I'll follow up separately on this one