Python 的类中,所有以双下划线__包起来的方法,叫魔术方法,魔术方法在类或对象的某些事件发出后可以自动执行,让类具有神奇的魔力,比如常见的构造方法__new__、初始化方法__init__、析构方法__del__,今天来聊一聊__new__的妙用,主要分享以下几点:
- __new__ 和 __init__ 的区别
- 应用1:改变内置的不可变类型
- 应用2:实现一个单例
- 应用3:客户端缓存
- 应用4:不同文件不同的解密方法
- 应用5:Metaclasses
__new__ 和 __init__ 的区别
1、调用时机不同:new 是真正创建实例的方法,init 用于实例的初始化,new 先于 init 运行。
2、返回值不同,new 返回一个类的实例,而 init 不返回任何信息。
3、new 是 class 的方法,而 init 是对象的方法。
示例代码:
- class A:
- def __new__(cls, *args, **kwargs):
- print("new", cls, args, kwargs)
- return super().__new__(cls)
- def __init__(self, *args, **kwargs):
- print("init", self, args, kwargs)
- def how_object_construction_works():
- x = A(1, 2, 3, x=4)
- print(x)
- print("===================")
- x = A.__new__(A, 1, 2, 3, x=4)
- if isinstance(x, A):
- type(x).__init__(x, 1, 2, 3, x=4)
- print(x)
- if __name__ == "__main__":
- how_object_construction_works()
上述代码定义了一个类 A,在调用 A(1, 2, 3, x=4) 时先执行 new,再执行 init,等价于:
- x = A.__new__(A, 1, 2, 3, x=4)
- if isinstance(x, A):
- type(x).__init__(x, 1, 2, 3, x=4)
代码的运行结果如下:
- new <class '__main__.A'> (1, 2, 3) {'x': 4}
- init <__main__.A object at 0x7fccaec97610> (1, 2, 3) {'x': 4}
- <__main__.A object at 0x7fccaec97610>
- ===================
- new <class '__main__.A'> (1, 2, 3) {'x': 4}
- init <__main__.A object at 0x7fccaec97310> (1, 2, 3) {'x': 4}
- <__main__.A object at 0x7fccaec97310>
new 的主要作用就是让程序员可以自定义类的创建行为,以下是其主要应用场景:
应用1:改变内置的不可变类型
我们知道,元组是不可变类型,但是我们继承 tuple ,然后可以在 new 中,对其元组的元素进行修改,因为 new 返回之前,元组还不是元组,这在 init 函数中是无法实现的。比如说,实现一个大写的元组,代码如下:
- class UppercaseTuple(tuple):
- def __new__(cls, iterable):
- upper_iterable = (s.upper() for s in iterable)
- return super().__new__(cls, upper_iterable)
- # 以下代码会报错,初始化时是无法修改的
- # def __init__(self, iterable):
- # print(f'init {iterable}')
- # for i, arg in enumerate(iterable):
- # self[i] = arg.upper()
- if __name__ == '__main__':
- print("UPPERCASE TUPLE EXAMPLE")
- print(UppercaseTuple(["hello", "world"]))
- # UPPERCASE TUPLE EXAMPLE
- # ('HELLO', 'WORLD')
应用2:实现一个单例
- class Singleton:
- _instance = None
- def __new__(cls, *args, **kwargs):
- if cls._instance is None:
- cls._instance = super().__new__(cls, *args, **kwargs)
- return cls._instance
- if __name__ == "__main__":
- print("SINGLETON EXAMPLE")
- x = Singleton()
- y = Singleton()
- print(f"{x is y=}")
- # SINGLETON EXAMPLE
- # x is y=True
应用3:客户端缓存
当客户端的创建成本比较高时,比如读取文件或者数据库,可以采用以下方法,同一个客户端属于同一个实例,节省创建对象的成本,这本质就是多例模式。
- class Client:
- _loaded = {}
- _db_file = "file.db"
- def __new__(cls, client_id):
- if (client := cls._loaded.get(client_id)) is not None:
- print(f"returning existing client {client_id} from cache")
- return client
- client = super().__new__(cls)
- cls._loaded[client_id] = client
- client._init_from_file(client_id, cls._db_file)
- return client
- def _init_from_file(self, client_id, file):
- # lookup client in file and read properties
- print(f"reading client {client_id} data from file, db, etc.")
- name = ...
- email = ...
- self.name = name
- self.email = email
- self.id = client_id
- if __name__ == '__main__':
- print("CLIENT CACHE EXAMPLE")
- x = Client(0)
- y = Client(0)
- print(f"{x is y=}")
- z = Client(1)
- # CLIENT CACHE EXAMPLE
- # reading client 0 data from file, db, etc.
- # returning existing client 0 from cache
- # x is y=True
- # reading client 1 data from file, db, etc.
应用4:不同文件不同的解密方法
先在脚本所在目录创建三个文件:plaintext_hello.txt、rot13_hello.txt、otp_hello.txt,程序会根据不同的文件选择不同的解密算法
- import codecs
- import itertools
- class EncryptedFile:
- _registry = {} # 'rot13' -> ROT13Text
- def __init_subclass__(cls, prefix, **kwargs):
- super().__init_subclass__(**kwargs)
- cls._registry[prefix] = cls
- def __new__(cls, path: str, key=None):
- prefix, sep, suffix = path.partition(":///")
- if sep:
- file = suffix
- else:
- file = prefix
- prefix = "file"
- subclass = cls._registry[prefix]
- obj = object.__new__(subclass)
- obj.file = file
- obj.key = key
- return obj
- def read(self) -> str:
- raise NotImplementedError
- class Plaintext(EncryptedFile, prefix="file"):
- def read(self):
- with open(self.file, "r") as f:
- return f.read()
- class ROT13Text(EncryptedFile, prefix="rot13"):
- def read(self):
- with open(self.file, "r") as f:
- text = f.read()
- return codecs.decode(text, "rot_13")
- class OneTimePadXorText(EncryptedFile, prefix="otp"):
- def __init__(self, path, key):
- if isinstance(self.key, str):
- self.key = self.key.encode()
- def xor_bytes_with_key(self, b: bytes) -> bytes:
- return bytes(b1 ^ b2 for b1, b2 in zip(b, itertools.cycle(self.key)))
- def read(self):
- with open(self.file, "rb") as f:
- btext = f.read()
- text = self.xor_bytes_with_key(btext).decode()
- return text
- if __name__ == "__main__":
- print("ENCRYPTED FILE EXAMPLE")
- print(EncryptedFile("plaintext_hello.txt").read())
- print(EncryptedFile("rot13:///rot13_hello.txt").read())
- print(EncryptedFile("otp:///otp_hello.txt", key="1234").read())
- # ENCRYPTED FILE EXAMPLE
- # plaintext_hello.txt
- # ebg13_uryyb.gkg
- # ^FCkYW_X^GLE
应用5:Metaclasses
metaclass 可以像装饰器那样定制和修改继承它的子类,前文Python黑魔法之metaclass
本文转载自微信公众号「Python七号」,可以通过以下二维码关注。转载本文请联系Python七号公众号。