Bug: Revert changes that impacts v3

[leandrodamascena]

Expected Behaviour

Some utilities should work without any extra dependency.

Current Behaviour

The V3 branch does not yet have nox installed to perform dependency testing and some of our changes have created some unexpected cross-dependency breaking changes.

Here is the list of breaking changes detected so far:

1 - We can't import BaseModel outside of TYPE_CHECKING because it will force to bring Pydantic at Runtime level even not using data_validation - https://github.com/aws-powertools/powertools-lambda-python/blob/v3/aws_lambda_powertools/event_handler/openapi/types.py#L7

2 - We can't import ConfigDict because we will force to bring Pydantic event not using data_validation - https://github.com/aws-powertools/powertools-lambda-python/blob/v3/aws_lambda_powertools/event_handler/openapi/constants.py#L1

3 - We need to restore get_header.. and get_querystring.. functions removed by this PR - https://github.com/aws-powertools/powertools-lambda-python/pull/4606/files

4 - We need to add more e2e tests.

Code snippet

No code snippet

Possible Solution

No response

Steps to Reproduce

Install and run

Powertools for AWS Lambda (Python) version

v3

AWS Lambda function runtime

3.11

Packaging format used

PyPi

Debugging logs

No response

[ericbn]

@leandrodamascena, regarding point 1, as far as I can see, code has always been importing BaseModel outside TYPE_CHECKING. This makes sense, because otherwise it would break Pydantic.

Looks like for both points 1 and 2, Pydantic-specific code will need to be under some runtime check depending on each case. For example instead of:

        if isinstance(res, BaseModel):
            return _model_dump(
                res,
                by_alias=True,
                exclude_unset=exclude_unset,
                exclude_defaults=exclude_defaults,
                exclude_none=exclude_none,
            )

we could use a duck-typing-"Look Before You Leap" (LBYL) style like:

        if hasattr(res, "model_dump"):
            return _model_dump(
                res,
                by_alias=True,
                exclude_unset=exclude_unset,
                exclude_defaults=exclude_defaults,
                exclude_none=exclude_none,
            )

or an arguably preferred (because under the hood, hasattr is actually checking for an AttributeError exception being raised by getattr) "Easier to Ask Forgiveness Than Permission" (EAFP) style like:

        try:
            return _model_dump(
                res,
                by_alias=True,
                exclude_unset=exclude_unset,
                exclude_defaults=exclude_defaults,
                exclude_none=exclude_none,
            )
        except AttributeError:
            ...

I'd say it goes on depending on each code that should only optionally work with Pydantic, only if it's available.