如何修改Pydantic默认映射，实现嵌套Schema的ORM对象转换？

背景与需求 #

现有SQLAlchemy ORM模型：

class Building(Base):
    id = Column(Integer, primary_key=True)
    name = Column(String)
    street = Column(String)
    city = Column(String)
    post_code = Column(String)

原Pydantic Schema直接平映射ORM字段可正常运行：

class BuildingOutSchema(BaseModel):
    id: int
    name: str
    street: str
    city: str
    post_code: str

    class Config:
        orm_mode = True

# 调用示例
company = db.query(Company).get(1)
return CompanyOutSchema.from_orm(company)

期望将地址相关字段嵌套为AddressSchema，且无需手动循环映射（该Schema会作为列表字段用于其他模型）：

class AddressSchema(BaseModel):
    street: str
    city: str
    post_code: str

class BuildingOutSchema(BaseModel):
    id: int
    name: str
    address: AddressSchema

    class Config:
        orm_mode = True

# 期望直接生效的调用
company = db.query(Company).get(1)
return CompanyOutSchema.from_orm(company)

可行解决方案 #

方案1：Pydantic v2 计算字段（推荐） #

利用Pydantic v2的computed_field特性，自动从ORM字段组装嵌套Schema：

from pydantic import BaseModel, computed_field

class AddressSchema(BaseModel):
    street: str
    city: str
    post_code: str

class BuildingOutSchema(BaseModel):
    id: int
    name: str

    class Config:
        orm_mode = True

    @computed_field
    @property
    def address(self) -> AddressSchema:
        return AddressSchema(
            street=self.street,
            city=self.city,
            post_code=self.post_code
        )

此方案无需额外配置，from_orm会自动识别ORM模型的平级字段，通过计算字段生成嵌套的address对象，列表场景下直接用list[BuildingOutSchema].from_orm(query_result)即可批量转换。

方案2：Validator字段填充（兼容v1/v2） #

通过@validator在验证阶段自动构建嵌套字段：

from pydantic import BaseModel, validator

class AddressSchema(BaseModel):
    street: str
    city: str
    post_code: str

class BuildingOutSchema(BaseModel):
    id: int
    name: str
    address: AddressSchema

    class Config:
        orm_mode = True

    @validator('address', pre=True, always=True)
    def build_address(cls, v, values):
        # 从平级字段中提取值组装AddressSchema
        return AddressSchema(
            street=values.get('street'),
            city=values.get('city'),
            post_code=values.get('post_code')
        )

pre=True确保在字段验证前执行，always=True保证即使没有传入address也会触发，完美适配ORM模型的平级字段结构。

方案3：重写from_orm方法 #

直接控制ORM到Schema的转换逻辑，适合复杂场景：

class AddressSchema(BaseModel):
    street: str
    city: str
    post_code: str

class BuildingOutSchema(BaseModel):
    id: int
    name: str
    address: AddressSchema

    class Config:
        orm_mode = True

    @classmethod
    def from_orm(cls, obj):
        return cls(
            id=obj.id,
            name=obj.name,
            address=AddressSchema(
                street=obj.street,
                city=obj.city,
                post_code=obj.post_code
            )
        )

重写后from_orm会直接按自定义逻辑转换ORM对象，列表场景下同样无需手动循环，Pydantic会自动迭代处理。

内容的提问来源于stack exchange，提问作者Ruuza