Allow case where __pydantic_extra__ is None, even if extra='allow'

[sydney-runkle]

Fix pydantic/pydantic#9043. The issue reported has been simplified in the example below:

For the following setup:

from pydantic import BaseModel, ConfigDict, Field, TypeAdapter


class Base(BaseModel):
    a: int = Field(default=1)
    
    model_config = ConfigDict(extra='allow')


class Sub(Base):
    b: str = Field(default="2")

    model_config = ConfigDict(extra='forbid')


sub = Sub()
ta = TypeAdapter(Base)

Before this fix:

result = ta.dump_python(sub)
"""
/Users/programming/pydantic_work/pydantic/pydantic/type_adapter.py:356: UserWarning: Pydantic serializer warnings:
  Expected `general-fields` but got `tuple` - serialized value may not be as expected
  return self.serializer.to_python(
"""
print(result)
#> ({'a': 1, 'b': '2'}, None)

After this fix:

result = ta.dump_python(sub)
print(result)
#> {'a': 1}

Selected Reviewer: @adriangb

[sydney-runkle]

I can either add a test here, or in the existing serialize as any PR (which currently has a failing test bc of this bug), or in pydantic itself.

[sydney-runkle]

Please review

[codspeed-hq]

CodSpeed Performance Report

Merging #1236 will improve performances by 43.9%

_{Comparing fix_extra_ser (3fd8dae) with main (abff9ba)}

Summary

⚡ 6 improvements
✅ 142 untouched benchmarks

🆕 4 new benchmarks

Benchmarks breakdown

	Benchmark	main	fix_extra_ser	Change
⚡	test_complete_core_json	2.6 ms	2.1 ms	+22.02%
⚡	test_core_json_fs	736.6 µs	542.6 µs	+35.76%
⚡	test_dict_of_ints_core_json	13.3 ms	10.1 ms	+31.55%
🆕	test_enum_int_core	N/A	20.5 µs	N/A
🆕	test_enum_int_python	N/A	39.1 µs	N/A
🆕	test_enum_str_core	N/A	20.9 µs	N/A
🆕	test_enum_str_python	N/A	38.5 µs	N/A
⚡	test_list_of_ints_core_json	5.1 ms	4 ms	+28.74%
⚡	test_set_of_ints_core_json	5.9 ms	4.8 ms	+22.55%
⚡	test_set_of_ints_core_json_duplicates	3.6 ms	2.5 ms	+43.9%

[adriangb]

A test here would be nice

[davidhewitt]

I'm not entirely sure on the justification, maybe a test is helpful to explain why?

I couldn't resist some Rust golfing...

[sydney-runkle]

Alright, odd issue, the current test passes, but if you use extra_behavior='allow' instead of config=core_schema.CoreConfig(extra_fields_behavior='allow'), then things fail...

[sydney-runkle]

Alright, odd issue, the current test passes, but if you use extra_behavior='allow' instead of config=core_schema.CoreConfig(extra_fields_behavior='allow'), then things fail...

I think worth addressing this in a different issue, I just discovered this while trying to write the test here