Use Pydantic model_dump in serializer

[baumgold]

Attempts to fix #754 by using the Pydantic serializer when serializing a model via the CLI

[kschwab]

+@hramezani

@baumgold this is a good idea, thanks for the PR. We just need to make sure to maintain backwards compatibility. Can you add a bool kwarg (use_serializers?) to the serialize method with the default set to False. Then you can just make dumped_model an empty dictionary if flag is not set and update dumped_field to:

dumped_field = dumped_model.get(field_name, model_default)

You will also need to add some tests.

[baumgold]

@kschwab - I've enhanced the PR as you suggested to maintain backwards compatibility.

[hramezani]

Thanks @baumgold for updating PR.

I think we still need

• @kschwab review
• Test
• Documentatin

[baumgold]

Apologies for the slow follow-up on this — thanks for your patience!

Here is a summary of what has been added since the initial review:

• Backwards compatibility: The feature is opt-in via a use_serializers: bool = False parameter on CliApp.serialize(), so existing behaviour is completely unchanged by default.
• Tests: Added four new test cases covering field_serializer transforms with use_serializers=True, the default-False behaviour, pydantic dataclasses, and propagation through nested subfields.
• Docs: Added a paragraph and code example under the Serializing Arguments section of docs/index.md showing the opt-in use_serializers=True behaviour.

@kschwab — would you be able to take another look when you get a chance? Happy to make any further changes you suggest.

[hramezani]

thanks @baumgold for updating your PR.

Could you please add a test with BaseSettings? or change one of the existing tests and use BaseSettings
Also, a test with PlainSerializer/WrapSerializer from Annotated (the original issue's pattern) would be great