LGTM

Maybe we should extract the part related to plugin installation into a separate PR, but doesn't seem critical to me

Ok, so in an attempt to get this merged, I've moved the pydantic.mypy plugin changes to a separate PR#14081. This makes it easier to review which changes are causing what, as several reviewers had questions about what exactly is needed and what relates to what.

For this PR, the goal is simply to update tomypy 1.14.1. When doing so, we're getting 2 similar errors:

fastapi\encoders.py:244: error: Argument 1 to "asdict" has incompatible type "DataclassInstance | type[DataclassInstance]"; expected "DataclassInstance"  [arg-type]

What happens here is that first, we're callingdataclasses.is_dataclass(obj).

Due tothe typing of this method in theirdataclasses.pyi,mypy will go and typeobj asDataclassInstance | type[DataclassInstance]. This then doesn't align with the signature of the followingdataclasses.asdict(obj) statement, which expects aDataclassInstance.

So, these are the possible fixes for this:

• add an ignore statement
• add atype() call before callingis_dataclass, as@tamird suggests. This isn't actually necessary, because the implementation ofdataclasses.is_dataclass will calltype() when necessary. However, by doing this, we trickmypy into assuming thatobj is NOT atype[DataclassInstance] and then it won't error on the followingasdict call
• inbetween the originalis_dataclass check and theasdict call, add the following:assert not isinstance(obj, type). This again will tellmypy not to assume theobj is a type. To me, this is the most direct/clear solution, but I think it's a matter of taste.

Will run this by Tiangolo to get his opinion.

Nice, thanks! 🚀

And thanks@svlandeg for the analysis, discussion and work on this! 🙌