自定义模式上的 Pydantic 序列化

Question

我正在使用 pydantic v2，并且有一个 FutureAwareDatetime 的自定义架构。

class FutureAwareDatetime:
    """A datetime that requires timezone info and must be in the future."""

    @classmethod
    def __get_pydantic_core_schema__(
        cls,
        source: type[Any],
        handler: GetCoreSchemaHandler,
    ) -> CoreSchema:
        if cls is source:
            # used directly as a type
            return datetime_schema(
                now_op="future",
                tz_constraint="aware",
            )
        else:
            schema = handler(source)
            _check_annotated_type(schema["type"], "datetime", cls.__name__)
            schema["now_op"] = "future"
            schema["tz_constraint"] = "aware"
            return schema

    def __repr__(self) -> str:
        return "FutureAwareDatetime"

虽然这工作正常，但当我使用例如序列化数据时，我希望能够使用 get

dt.isoformat(timespec="seconds")

model_dump()

我不想使用

field_serializer

装饰器，而是直接在类中使用它。

谢谢你。

Answer 1

搜索了一段时间后发现

datetime_schema()

有说法

serialization

所以我添加了一个

PlainSerializer

。

我的代码看起来像

class FutureAwareDatetime:
    """A datetime that requires timezone info and must be in the future."""

    @classmethod
    def __get_pydantic_core_schema__(
        cls,
        source: type[Any],
        handler: GetCoreSchemaHandler,
    ) -> CoreSchema:
        serialize_to_iso8601 = PlainSerializer(
            func=lambda dt: dt.isoformat(timespec="seconds"),
            return_type=str,
        )
        if cls is source:
            # used directly as a type
            return datetime_schema(
                now_op="future",
                tz_constraint="aware",
                serialization=serialize_to_iso8601,
            )
        else:
            schema = handler(source)
            _check_annotated_type(schema["type"], "datetime", cls.__name__)
            return datetime_schema(
                now_op="future",
                tz_constraint="aware",
                serialization=serialize_to_iso8601,
            )