respect field_serializer when using serialize_as_any=True
#1835
Conversation
| // FIXME: should `exclude_if` be applied to extra fields too? | ||
| if serialization_exclude_if(field.serialization_exclude_if.as_ref(), value)? { | ||
| return Ok(None); | ||
| } |
There was a problem hiding this comment.
This looks like a possible bug which maybe needs changing in 2.13 rather than backporting?
There was a problem hiding this comment.
Are you talking about the FIXME comment?
There was a problem hiding this comment.
I think it makes sense to support it, regarding backporting I have the same opinion as in #1835 (comment).
There was a problem hiding this comment.
| let output_key = field.get_key_json(key_str, &field_extra); | ||
| map.serialize_entry(&output_key, &s)?; | ||
| } else if self.mode == FieldsMode::TypedDictAllow { | ||
| // FIXME: why is `extra_serializer` not used here when `serialize_as_any` is not set? |
There was a problem hiding this comment.
This is probably a bug we should repro and fix in 2.12?
There was a problem hiding this comment.
Here's a MRE:
class Model(BaseModel, extra='allow'):
a: int
__pydantic_extra__: dict[str, int]
m = Model(a=1, extra=1)
m.a = 'not_an_int'
m.model_dump_json() # PydanticSerializationUnexpectedValue warning raised
m.a = 1
m.extra = 'not_an_int'
m.model_dump_json() # No warning raisedI created pydantic/pydantic#12385. We could backport to 2.12, but this may introduce warnings that will break CI, which is unideal for a patch release (and this wasn't reported by any user yet).
CodSpeed Performance ReportMerging #1835 will not alter performanceComparing Summary
|
Co-authored-by: Victorien <65306057+Viicos@users.noreply.github.com>
2 participants
Change Summary
Fixes pydantic/pydantic#12379
Related issue number
This adjusts logic in
GeneralFieldsSerializerso that the type inference application ofSerializeAsAnyis applied after field serializers are applied. This is more consistent withSerializeAsAnyannotation on a field that also has afield_serializer; thefield_serializerwould still be applied in that case too.Checklist
pydantic-core(except for expected changes)