Pydantic v2移除json_loads/json_dumps配置，求替代方案指引

1. 单模型调用时自定义处理 #

直接在model_dump_json()（序列化）或model_validate_json()（反序列化）方法中传入对应参数，替代原配置项：

• 自定义序列化（替代原json_dumps）：

  from pydantic import BaseModel
  import json
  from datetime import datetime
  
  def custom_encoder(obj):
      if isinstance(obj, datetime):
          return obj.isoformat() + 'Z'
      raise TypeError(f"Object of type {obj.__class__.__name__} is not JSON serializable")
  
  class Item(BaseModel):
      created_at: datetime
  
  item = Item(created_at=datetime.now())
  json_str = item.model_dump_json(encoder=custom_encoder)
  

• 自定义反序列化（替代原json_loads）：

  def custom_decoder(dct):
      if 'created_at' in dct:
          dct['created_at'] = datetime.fromisoformat(dct['created_at'].rstrip('Z'))
      return dct
  
  item = Item.model_validate_json(json_str, decoder=custom_decoder)
  

2. 全局统一配置 #

如果需要让所有模型共享自定义逻辑，可以封装一个基类，把参数固化到方法中：

from pydantic import BaseModel

class CustomBaseModel(BaseModel):
    @classmethod
    def validate_json(cls, json_str: str):
        return cls.model_validate_json(json_str, decoder=custom_decoder)

    def dump_json(self):
        return self.model_dump_json(encoder=custom_encoder)

# 后续业务模型直接继承该基类
class Item(CustomBaseModel):
    created_at: datetime

3. 单个字段级别的自定义处理 #

针对特定字段的JSON序列化需求，结合Json类型字段的serialize和deserialize参数实现：

from pydantic import BaseModel, Json
from typing import Any

def custom_json_serialize(value: Any) -> str:
    return json.dumps(value, default=custom_encoder)

def custom_json_deserialize(value: str) -> Any:
    return json.loads(value, object_hook=custom_decoder)

class Item(BaseModel):
    data: Json[Any] = Json(..., serialize=custom_json_serialize, deserialize=custom_json_deserialize)

内容的提问来源于stack exchange，提问作者user2465039