Python 数据类 - dataclass 的作用与不足

https://docs.python.org/zh-cn/3/library/dataclasses.html
https://peps.python.org/pep-0526/
https://peps.python.org/pep-0557/

dataclass 简单示例

from dataclasses import dataclass@dataclass
class User:name: strage: intif __name__ == '__main__':response_json = {'name': 'python', 'age': 32}user = User(**response_json)print(user.name, user.age)  # python 32

不用使用 dataclass 也可以

class User:def __init__(self, name: str, age: int):self.name = nameself.age = ageif __name__ == '__main__':response_json = {'name': 'python', 'age': 32}user = User(**response_json)print(user.name, user.age)

User 显得稍显麻烦，不过这里很简单，倒也无所谓。

dataclass 优势

1. dataclass 比直接使用 class 书写更简洁, 看起来一目了然
2. dataclass 自动实现实用方法，比如 __repr__
3. Pycharm 的完全代码提示

dataclass 的不足

不支持嵌套

from dataclasses import dataclass@dataclass
class Info:sex: strphone: straddress: str = None@dataclass
class User:name: strage: intinfo: Info = Noneif __name__ == '__main__':response_json = {'name': 'python','age': 32,'info': {'sex': '','phone': '','address': 'https://python.org/'},}user = User(**response_json)# 显然这里我们期望的不是 dictassert isinstance(user.info, dict)

不支持扩展字段

from dataclasses import dataclass@dataclass
class User:name: strage: intif __name__ == '__main__':response_json = {'name': 'python','age': 32,'address': 'https://python.org/'}# TypeError: __init__() got an unexpected keyword argument 'address'user = User(**response_json)

使用 datclass 库弥补这些不足

https://github.com/foyoux/datclass

pip install datclass

datclass 简单示例

from dataclasses import dataclassfrom datclass import DatClass@dataclass
class Info(DatClass):sex: strphone: straddress: str = None@dataclass
class User(DatClass):name: strage: intinfo: Info = Noneif __name__ == '__main__':response_json = {'name': 'python','age': 32,'info': {'sex': '','phone': '','address': 'https://python.org/','field': 'extra'},}user = User(**response_json)assert isinstance(user.info, Info)assert isinstance(user.info, Info)print(user.info.address)  # https://python.org/

datclass 附带一个命令行命令用以自动生成 DatClass → datclass