Fix dataclasses.Field to use _MISSING_TYPE

[sirosen]

default, default_factory, and kw_only are all attributes of Field objects which may have a value of MISSING. This is likely to arise when examining Field objects and checking attributes.

For example,

 [f for f in fields(obj) if f.default is MISSING]

The types for the attributes should therefore be ... | _MISSING_TYPE
This adjustment is only applied to the attributes of Field, not the arguments to functions and methods.

---

I believe this is the correct resolution for python/mypy#12124

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

pydantic (https://github.com/samuelcolvin/pydantic)
+ pydantic/dataclasses.py:173: error: Incompatible types in assignment (expression has type "Union[_DefaultFactory[Any], _MISSING_TYPE]", variable has type "Optional[Callable[[], Any]]")  [assignment]

[hauntsaninja]

Thanks, but we previously rejected this change in #5900 .
As discussed over there, the main issue is that type checkers will not recognise if x is dataclasses.MISSING as narrowing the type

[hauntsaninja]

Actually, given that this has come up a couple times, I would accept a PR that uses the enum trick as described in: #5900 (comment) , if you'd be willing to make one :-)

[sirosen]

Actually, given that this has come up a couple times, I would accept a PR that uses the enum trick as described in: #5900 (comment) , if you'd be willing to make one :-)

Sorry for not seeing that there had been an earlier PR! Yes, I'll read a bit more and then try that workaround in a PR.

The enum trick seems good to me. It will make type narrowing work, and is valid for Literal.