Updates type signature for Field() constructor

[bjmc]

Replaces float in lt/gt/le/ge, arguments with the appropriate Protocol from annotated_types.

Change Summary

Related to #9472 updates the type signatures in Field() to be less restrictive, using the appropriate protocols from annotated_types instead of strict float or int types. This should be a nonbreaking change.

Related issue number

fixes #9472

Checklist

• The pull request title is a good summary of the changes - it will be used in the changelog
• Unit tests for the changes exist
• Tests pass on CI
• Documentation reflects the changes where applicable
• My PR is ready to review, please add a comment including the phrase "please review" to assign reviewers

[codspeed-hq]

CodSpeed Performance Report

Merging #9484 will not alter performance

_{Comparing bjmc:comparison-protocols (2673881) with main (28b36d9)}

Summary

✅ 13 untouched benchmarks

[bjmc]

The only test failures look to be with an older version of mypy. Maybe it doesn't fully recognize the protocols from annotated_types?

[sydney-runkle]

@bjmc,

Thanks, this looks great. Let's consult with @dmontagu re the test failures.

[sydney-runkle]

For these, let's actually keep the original type (float and int) and add an additional union member as the annotated_types type. Ex:
multiple_of: int | annotated_types.MultipleOf | None

[sydney-runkle]

@bjmc,

Thinking about this more, let's also add some tests accordingly, verifying that MultipleOf works in this case (it might not, due to expectations on the pydantic-core end of things.

[dmontagu]

@bjmc the SupportsX look like protocols that should work (though as @sydney-runkle mentioned, we should confirm that they actually do work).

But the MinLen, MaxLen, and MultipleOf seem to be dataclasses and not protocols, right? (Am I missing something?) Did you check if instances of these classes actually work? (I'd expect that they won't without further changes, but it's been a while since I looked at these bits of code.)

[bjmc]

@dmontagu Good shout, I'll revert those.

[bjmc]

I was overgeneralizing from the metadata_lookup here but you're absolutely right those are dataclasses to store the constraints, not Protocols to describe input types. I've reverted them now.

I suppose to be extremely pedantic, min/max lengths should be typed Annotated[int, Ge(0)] rather than just int but maybe we don't care about that.

[bjmc]

@sydney-runkle - Confirming this fixes my issue, using the same example.py from #9472 :

bjmc@host:~/Sandbox/pydantic$ git checkout main 
Switched to branch 'main'
Your branch is up to date with 'origin/main'.
bjmc@host:~/Sandbox/pydantic$ pyright example.py 
/home/bjmc/Sandbox/pydantic/example.py
  /home/bjmc/Sandbox/pydantic/example.py:6:44 - error: Argument of type "timedelta" cannot be assigned to parameter "gt" of type "float | None" in function "Field"
    Type "timedelta" is incompatible with type "float | None"
      "timedelta" is incompatible with "float"
      "timedelta" is incompatible with "None" (reportArgumentType)
1 error, 0 warnings, 0 informations 
bjmc@host:~/Sandbox/pydantic$ git checkout comparison-protocols 
Switched to branch 'comparison-protocols'
Your branch is up to date with 'origin/comparison-protocols'.
bjmc@host:~/Sandbox/pydantic$ pyright example.py 
0 errors, 0 warnings, 0 informations 

[sydney-runkle]

@bjmc,

Nice, the new changes look good. I think we do still want to add some explicit tests for the new protocols, but I'm not sure what the best way to do that is. I'll consult with @dmontagu and get back to you soon!

[sydney-runkle]

@bjmc,

I spoke with @adriangb, our resident annotated types guru, and we've decided that this PR is good to go.

Awesome work here, thanks for your contribution!!