Conformance Tests directives_type_ignore should probably allow Mypy-like type ignores

[davidhalter]

The conformance test directives_type_ignore feels a bit weird since it disallows the Mypy syntax:

z: int = ""  # type: ignore[<some-error-code>]
# It has the following line where the violation should be suppressed:
z: int = ""  # type: ignore[additional_stuff]

The documentation says:

In some cases, linting tools or other comments may be needed on the same line as a type comment. In these cases, the type comment should be before other comments and linting markers:

# type: ignore # <comment or other marker>

I feel it is therefore not really clear if the conformance tests are correct or not and I would really like to avoid adding a flag to Zuban with the name --make-conformance-tests-pass. # type: ignore[misc] can also be found in numerous places in Typeshed.

I would just suggest allowing an error there.

---

There's also this line that should be ignored to the conformance tests:

y: int = ""  # type: ignore - additional stuff

where I'm not sure if it's better to allow an error there, but I don't have an opinion there really. It might be useful to generate an error that the - additional stuff is not conformant with the spec that says we should use an additional comment. Mypy for example has the error Invalid "type: ignore" comment [syntax]. I'm however totally fine if the conformance tests disallow this.

cc @erictraut since you wrote these tests AFAIK.

[erictraut]

I think the current tests accurately reflect the way the spec is worded currently. If you'd like to suggest a modification or clarification in the spec, you could propose a change and see what the community thinks.

My personal preference is that type checkers support # type: ignore as it's currently spec'ed and then provide their own variants if they want to allow for type-check-specific features like filters on specific error codes. That's the approach I took with pyright, and I think some other type checkers have followed the same approach. This approach has been difficult for mypy to implement because mypy relies on the Python parser and its special-cased support for # type: ignore comments. The mypy maintainers are considering switching to a separate parser, which would eliminate this limitation.