Do not warn about shadowed fields if they are not redefined in a child class

[chan-vince]

Change Summary

Fix for issue #9107.

Adds another early exit condition when evaluating whether to log a warning message during detection for shadowed fields.

In the case where a field is defined in a parent class, but it has not been defined at all in a child class, it is technically not a shadowed field, and so shouldn't be warned as such.

Note this is very different from the case where a child class does redefine a field but with a narrower type or even defined as the same type but with a different default. Conceptually this is probably ok, but checking for that is quite complex and this PR does not attempt to try. So this is about checking if a field is defined or not defined - if it is, regardless of type or default value, the warning message is still logged.

Example of a case where a warning is currently logged, but is now no longer logged:

class MyMixin:
    my_field: bool = False

class MyModel(BaseModel, MyMixin):
    pass

Example of a case where a warning is still logged, and it is important to do so:

class MyMixin:
    my_field: bool = False

class MyModel(BaseModel, MyMixin):
    my_field: bool = True

Related issue number

fix #9107

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

Selected Reviewer: @adriangb

[codspeed-hq]

CodSpeed Performance Report

Merging #9111 will not alter performance

_{Comparing chan-vince:bug/9107-improve-warning-on-parent-inherited-field (f42b70a) with main (e3b4633)}

Summary

✅ 10 untouched benchmarks

[chan-vince]

please review

[sydney-runkle]

@chan-vince,

Looks great. Could you add a test for each of those above cases, showcasing that a warning is raised for one and not the other?

[sydney-runkle]

Why is this necessary?

[chan-vince]

Sorry, I misunderstood the simplefilter() docs - I have adjusted it now.

Essentially follows this example from the docs, to ensure that all warnings are captured, then run the code, then check the recorded warnings list is empty.

[sydney-runkle]

Ahh I see, looks good.

[sydney-runkle]

Looks good overall, just one comment!

[sydney-runkle]

Thanks for adding the test!