如何用Pydantic V2的model_validate从ORM非模型属性生成计算字段？

问题背景 #

使用FastAPI时，接口返回SQLAlchemy ORM实例，通过model_validate转换为定义了response_model的Pydantic模型。需要基于ORM实例中未在Pydantic模型里声明的属性生成计算字段，但直接结合__pydantic_extra__与extra='allow'的方案在处理ORM实例时失效，当前动态修改ORM实例的实现方式不够优雅。

根本原因 #

当用model_validate处理ORM实例时，from_attributes=True仅让Pydantic提取模型显式定义的字段，extra='allow'默认仅对字典格式的输入生效，不会自动捕获ORM实例的所有额外属性，导致__pydantic_extra__为空。

简洁解决方案 #

方案一：前置验证器转ORM为字典，激活extra机制 #

通过前置验证器将ORM实例转换为包含所有属性的字典，让extra='allow'正常捕获额外属性：

from typing import Any
from pydantic import BaseModel, ConfigDict, computed_field, model_validator


class Foo:
    def __init__(self):
        self.original_thing = "foo"


class CleanSolution1(BaseModel):
    model_config = ConfigDict(from_attributes=True, extra='allow')

    @model_validator(mode="before")
    @classmethod
    def convert_orm_to_dict(cls, values: Any) -> Any:
        # 实际场景可替换为判断SQLAlchemy基类，比如 isinstance(values, Base)
        if isinstance(values, Foo):
            return vars(values)
        return values

    @computed_field
    @property
    def computed_thing(self) -> str:
        return self.__pydantic_extra__["original_thing"] + "_computed"


# 测试用例
model = CleanSolution1(original_thing="bar")
print(f'CleanSolution1 (original_thing="bar") worked: {model.computed_thing == "bar_computed"}')

model_orm = CleanSolution1.model_validate(Foo())
print(f'CleanSolution1 model_validate(Foo()) worked: {model_orm.computed_thing == "foo_computed"}')

方案二：私有属性存储原始ORM实例，直接访问计算 #

通过Pydantic私有属性保留原始ORM实例，绕开extra限制直接访问所需属性：

from typing import Any
from pydantic import BaseModel, ConfigDict, computed_field, model_validator, PrivateAttr


class Foo:
    def __init__(self):
        self.original_thing = "foo"


class CleanSolution2(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    
    # 私有属性存储原始ORM实例
    _orm_instance: Any = PrivateAttr()

    @model_validator(mode="after")
    def capture_orm_instance(self) -> "CleanSolution2":
        # 从Pydantic验证器上下文获取原始ORM实例
        self._orm_instance = self.__pydantic_validator__.values.obj
        return self

    @computed_field
    @property
    def computed_thing(self) -> str:
        return self._orm_instance.original_thing + "_computed"


# 测试用例
model_orm = CleanSolution2.model_validate(Foo())
print(f'CleanSolution2 model_validate(Foo()) worked: {model_orm.computed_thing == "foo_computed"}')

内容的提问来源于stack exchange，提问作者jsnow