描述符(__get__和__set__和__delete__)
一、描述符
- 描述符是什么：描述符本质就是一个新式类，在这个新式类中，至少实现了__get__()，set()，delete()中的一个，这也被称为描述符协议
- __get__()：调用一个属性时，触发
- __set__()：为一个属性赋值时，触发
- __delete__()：采用del删除属性时，触发
- 定义一个描述符
class Foo: # 在python3中Foo是新式类，它实现了__get__()，__set__()，__delete__()中的一个三种方法的一个，这个类就被称作一个描述符
def __get__(self, instance, owner):
pass
def __set__(self, instance, value):
pass
def __delete__(self, instance):
pass
二、描述符的作用
- 描述符是干什么的：描述符的作用是用来代理另外一个类的属性的，必须把描述符定义成这个类的类属性，不能定义到构造函数中
class Foo:
def __get__(self, instance, owner):
print('触发get')
def __set__(self, instance, value):
print('触发set')
def __delete__(self, instance):
print('触发delete')
f1 = Foo()
- 包含这三个方法的新式类称为描述符，由这个类产生的实例进行属性的调用/赋值/删除，并不会触发这三个方法
f1.name = 'randy'
f1.name
del f1.name
2.1 何时，何地，会触发这三个方法的执行
class Str:
"""描述符Str"""
def __get__(self, instance, owner):
print('Str调用')
def __set__(self, instance, value):
print('Str设置...')
def __delete__(self, instance):
print('Str删除...')
class Int:
"""描述符Int"""
def __get__(self, instance, owner):
print('Int调用')
def __set__(self, instance, value):
print('Int设置...')
def __delete__(self, instance):
print('Int删除...')
class People:
name = Str()
age = Int()
def __init__(self, name, age): # name被Str类代理，age被Int类代理
self.name = name
self.age = age
# 何地？：定义成另外一个类的类属性
# 何时？：且看下列演示
p1 = People('alex', 18)
Str设置...
Int设置...
- 描述符Str的使用
p1.name
p1.name = 'randy'
del p1.name
Str调用
Str设置...
Str删除...
- 描述符Int的使用
p1.age
p1.age = 18
del p1.age
Int调用
Int设置...
Int删除...
- 我们来瞅瞅到底发生了什么
print(p1.__dict__)
print(People.__dict__)
{}
{'__module__': '__main__', 'name': <__main__.Str object at 0x107a86940>, 'age': <__main__.Int object at 0x107a863c8>, '__init__': <function People.__init__ at 0x107ba2ae8>, '__dict__': <attribute '__dict__' of 'People' objects>, '__weakref__': <attribute '__weakref__' of 'People' objects>, '__doc__': None}
- 补充
print(type(p1) == People) # type(obj)其实是查看obj是由哪个类实例化来的
print(type(p1).__dict__ == People.__dict__)
True
True
三、两种描述符
3.1 数据描述符
- 至少实现了__get__()和__set__()
class Foo:
def __set__(self, instance, value):
print('set')
def __get__(self, instance, owner):
print('get')
3.2 非数据描述符
- 没有实现__set__()
class Foo:
def __get__(self, instance, owner):
print('get')
四、描述符注意事项
-
描述符本身应该定义成新式类，被代理的类也应该是新式类
-
必须把描述符定义成这个类的类属性，不能为定义到构造函数中
-
要严格遵循该优先级，优先级由高到底分别是
1.类属性
2.数据描述符
3.实例属性
4.非数据描述符
5.找不到的属性触发__getattr__()
五、使用描述符
- 众所周知，python是弱类型语言，即参数的赋值没有类型限制，下面我们通过描述符机制来实现类型限制功能
5.1 牛刀小试
class Str:
def __init__(self, name):
self.name = name
def __get__(self, instance, owner):
print('get--->', instance, owner)
return instance.__dict__[self.name]
def __set__(self, instance, value):
print('set--->', instance, value)
instance.__dict__[self.name] = value
def __delete__(self, instance):
print('delete--->', instance)
instance.__dict__.pop(self.name)
class People:
name = Str('name')
def __init__(self, name, age, salary):
self.name = name # 触发__set__
self.age = age
self.salary = salary
p1 = People('randy', 18, 3231.3)
set---> <__main__.People object at 0x107a86198> randy
- 调用
print(p1.__dict__)
print(p1.name)
{'name': 'randy', 'age': 18, 'salary': 3231.3}
get---> <__main__.People object at 0x107a86198> <class '__main__.People'>
randy
- 赋值
print(p1.__dict__)
p1.name = 'randysun'
print(p1.__dict__)
{'name': 'randy', 'age': 18, 'salary': 3231.3}
set---> <__main__.People object at 0x107a86198> randysun
{'name': 'randysun', 'age': 18, 'salary': 3231.3}
- 删除
print(p1.__dict__)
del p1.name
print(p1.__dict__)
{'name': 'randysun', 'age': 18, 'salary': 3231.3}
delete---> <__main__.People object at 0x107a86198>
{'age': 18, 'salary': 3231.3}
5.2 拔刀相助
class Str:
def __init__(self, name):
self.name = name
def __get__(self, instance, owner):
print('get--->', instance, owner)
return instance.__dict__[self.name]
def __set__(self, instance, value):
print('set--->', instance, value)
instance.__dict__[self.name] = value
def __delete__(self, instance):
print('delete--->', instance)
instance.__dict__.pop(self.name)
class People:
name = Str('name')
def __init__(self, name, age, salary):
self.name = name
self.age = age
self.salary = salary
# 疑问：如果我用类名去操作属性呢
try:
People.name # 报错，错误的根源在于类去操作属性时，会把None传给instance
except Exception as e:
print(e)
get---> None <class '__main__.People'>
'NoneType' object has no attribute '__dict__'
- 修订__get__方法
class Str:
def __init__(self, name):
self.name = name
def __get__(self, instance, owner):
print('get--->', instance, owner)
if instance is None:
return self
return instance.__dict__[self.name]
def __set__(self, instance, value):
print('set--->', instance, value)
instance.__dict__[self.name] = value
def __delete__(self, instance):
print('delete--->', instance)
instance.__dict__.pop(self.name)
class People:
name = Str('name')
def __init__(self, name, age, salary):
self.name = name
self.age = age
self.salary = salary
print(People.name) # 完美，解决
get---> None <class '__main__.People'>
<__main__.Str object at 0x107a86da0>
5.3 磨刀霍霍
class Str:
def __init__(self, name, expected_type):
self.name = name
self.expected_type = expected_type
def __get__(self, instance, owner):
print('get--->', instance, owner)
if instance is None:
return self
return instance.__dict__[self.name]
def __set__(self, instance, value):
print('set--->', instance, value)
if not isinstance(value, self.expected_type): # 如果不是期望的类型，则抛出异常
raise TypeError('Expected %s' % str(self.expected_type))
instance.__dict__[self.name] = value
def __delete__(self, instance):
print('delete--->', instance)
instance.__dict__.pop(self.name)
class People:
name = Str('name', str) # 新增类型限制str
def __init__(self, name, age, salary):
self.name = name
self.age = age
self.salary = salary
try:
p1 = People(123, 18, 3333.3) # 传入的name因不是字符串类型而抛出异常
except Exception as e:
print(e)
set---> <__main__.People object at 0x1084cd940> 123
Expected <class 'str'>
5.4 大刀阔斧
class Typed:
def __init__(self, name, expected_type):
self.name = name
self.expected_type = expected_type
def __get__(self, instance, owner):
print('get--->', instance, owner)
if instance is None:
return self
return instance.__dict__[self.name]
def __set__(self, instance, value):
print('set--->', instance, value)
if not isinstance(value, self.expected_type):
raise TypeError('Expected %s' % str(self.expected_type))
instance.__dict__[self.name] = value
def __delete__(self, instance):
print('delete--->', instance)
instance.__dict__.pop(self.name)
class People:
name = Typed('name', str)
age = Typed('name', int)
salary = Typed('name', float)
def __init__(self, name, age, salary):
self.name = name
self.age = age
self.salary = salary
try:
p1 = People(123, 18, 3333.3)
except Exception as e:
print(e)
set---> <__main__.People object at 0x1082c7908> 123
Expected <class 'str'>
try:
p1 = People('randy', '18', 3333.3)
except Exception as e:
print(e)
set---> <__main__.People object at 0x1078dd438> randy
set---> <__main__.People object at 0x1078dd438> 18
Expected <class 'int'>
p1 = People('randy', 18, 3333.3)
set---> <__main__.People object at 0x1081b3da0> randy
set---> <__main__.People object at 0x1081b3da0> 18
set---> <__main__.People object at 0x1081b3da0> 3333.3
- 大刀阔斧之后我们已然能实现功能了，但是问题是，如果我们的类有很多属性，你仍然采用在定义一堆类属性的方式去实现，low，这时候我需要教你一招：独孤九剑
5.4.1 类的装饰器:无参
def decorate(cls):
print('类的装饰器开始运行啦------>')
return cls
@decorate # 无参：People = decorate(People)
class People:
def __init__(self, name, age, salary):
self.name = name
self.age = age
self.salary = salary
p1 = People('randy', 18, 3333.3)
类的装饰器开始运行啦------>
5.4.2 类的装饰器:有参
def typeassert(**kwargs):
def decorate(cls):
print('类的装饰器开始运行啦------>', kwargs)
return cls
return decorate
@typeassert(
name=str, age=int, salary=float
) # 有参：1.运行typeassert(...)返回结果是decorate，此时参数都传给kwargs 2.People=decorate(People)
class People:
def __init__(self, name, age, salary):
self.name = name
self.age = age
self.salary = salary
p1 = People('randy', 18, 3333.3)
类的装饰器开始运行啦------> {'name': <class 'str'>, 'age': <class 'int'>, 'salary': <class 'float'>}
5.5 刀光剑影
class Typed:
def __init__(self, name, expected_type):
self.name = name
self.expected_type = expected_type
def __get__(self, instance, owner):
print('get--->', instance, owner)
if instance is None:
return self
return instance.__dict__[self.name]
def __set__(self, instance, value):
print('set--->', instance, value)
if not isinstance(value, self.expected_type):
raise TypeError('Expected %s' % str(self.expected_type))
instance.__dict__[self.name] = value
def __delete__(self, instance):
print('delete--->', instance)
instance.__dict__.pop(self.name)
def typeassert(**kwargs):
def decorate(cls):
print('类的装饰器开始运行啦------>', kwargs)
for name, expected_type in kwargs.items():
setattr(cls, name, Typed(name, expected_type))
return cls
return decorate
@typeassert(
name=str, age=int, salary=float
) # 有参：1.运行typeassert(...)返回结果是decorate，此时参数都传给kwargs 2.People=decorate(People)
class People:
def __init__(self, name, age, salary):
self.name = name
self.age = age
self.salary = salary
print(People.__dict__)
p1 = People('randy', 18, 3333.3)
类的装饰器开始运行啦------> {'name': <class 'str'>, 'age': <class 'int'>, 'salary': <class 'float'>}
{'__module__': '__main__', '__init__': <function People.__init__ at 0x10797a400>, '__dict__': <attribute '__dict__' of 'People' objects>, '__weakref__': <attribute '__weakref__' of 'People' objects>, '__doc__': None, 'name': <__main__.Typed object at 0x1080b2a58>, 'age': <__main__.Typed object at 0x1080b2ef0>, 'salary': <__main__.Typed object at 0x1080b2c18>}
set---> <__main__.People object at 0x1080b22e8> randy
set---> <__main__.People object at 0x1080b22e8> 18
set---> <__main__.People object at 0x1080b22e8> 3333.3
六、描述符总结
- 描述符是可以实现大部分python类特性中的底层魔法，包括@classmethod，@staticmethd，@property甚至是__slots__属性
- 描述父是很多高级库和框架的重要工具之一，描述符通常是使用到装饰器或者元类的大型框架中的一个组件.
七、自定制@property
- 利用描述符原理完成一个自定制@property，实现延迟计算（本质就是把一个函数属性利用装饰器原理做成一个描述符：类的属性字典中函数名为key，value为描述符类产生的对象）
7.1 property回顾
class Room:
def __init__(self, name, width, length):
self.name = name
self.width = width
self.length = length
@property
def area(self):
return self.width * self.length
r1 = Room('alex', 1, 1)
print(r1.area)
1
7.2 自定制property
class Lazyproperty:
def __init__(self, func):
self.func = func
def __get__(self, instance, owner):
print('这是我们自己定制的静态属性，r1.area实际是要执行r1.area()')
if instance is None:
return self
return self.func(instance) # 此时你应该明白，到底是谁在为你做自动传递self的事情
class Room:
def __init__(self, name, width, length):
self.name = name
self.width = width
self.length = length
@Lazyproperty # area=Lazyproperty(area) 相当于定义了一个类属性,即描述符
def area(self):
return self.width * self.length
r1 = Room('alex', 1, 1)
print(r1.area)
这是我们自己定制的静态属性，r1.area实际是要执行r1.area()
1
7.3 实现延迟计算功能
class Lazyproperty:
def __init__(self, func):
self.func = func
def __get__(self, instance, owner):
print('这是我们自己定制的静态属性，r1.area实际是要执行r1.area()')
if instance is None:
return self
else:
print('--->')
value = self.func(instance)
setattr(instance, self.func.__name__, value) # 计算一次就缓存到实例的属性字典中
return value
class Room:
def __init__(self, name, width, length):
self.name = name
self.width = width
self.length = length
@Lazyproperty # area=Lazyproperty(area) 相当于'定义了一个类属性,即描述符'
def area(self):
return self.width * self.length
r1 = Room('alex', 1, 1)
print(r1.area) # 先从自己的属性字典找,没有再去类的中找,然后出发了area的__get__方法
这是我们自己定制的静态属性，r1.area实际是要执行r1.area()
--->
1
print(r1.area) # 先从自己的属性字典找,找到了,是上次计算的结果,这样就不用每执行一次都去计算
1
八、打破延迟计算
- 一个小的改动，延迟计算的美梦就破碎了
class Lazyproperty:
def __init__(self, func):
self.func = func
def __get__(self, instance, owner):
print('这是我们自己定制的静态属性，r1.area实际是要执行r1.area()')
if instance is None:
return self
else:
value = self.func(instance)
instance.__dict__[self.func.__name__] = value
return value
# return self.func(instance) # 此时你应该明白,到底是谁在为你做自动传递self的事情
def __set__(self, instance, value):
print('hahahahahah')
class Room:
def __init__(self, name, width, length):
self.name = name
self.width = width
self.length = length
@Lazyproperty # area=Lazyproperty(area) 相当于定义了一个类属性,即描述符
def area(self):
return self.width * self.length
print(Room.__dict__)
{'__module__': '__main__', '__init__': <function Room.__init__ at 0x107d53620>, 'area': <__main__.Lazyproperty object at 0x107ba3860>, '__dict__': <attribute '__dict__' of 'Room' objects>, '__weakref__': <attribute '__weakref__' of 'Room' objects>, '__doc__': None}
r1 = Room('alex', 1, 1)
print(r1.area)
print(r1.area)
print(r1.area)
这是我们自己定制的静态属性，r1.area实际是要执行r1.area()
1
这是我们自己定制的静态属性，r1.area实际是要执行r1.area()
1
这是我们自己定制的静态属性，r1.area实际是要执行r1.area()
1
print(
r1.area
) #缓存功能失效,每次都去找描述符了,为何,因为描述符实现了set方法,它由非数据描述符变成了数据描述符,数据描述符比实例属性有更高的优先级,因而所有的属性操作都去找描述符了
这是我们自己定制的静态属性，r1.area实际是要执行r1.area()
1
九、自定制@classmethod
class ClassMethod:
def __init__(self, func):
self.func = func
def __get__(
self, instance,
owner): #类来调用,instance为None,owner为类本身,实例来调用,instance为实例,owner为类本身,
def feedback():
print('在这里可以加功能啊...')
return self.func(owner)
return feedback
class People:
name = 'randy'
@ClassMethod # say_hi=ClassMethod(say_hi)
def say_hi(cls):
print('你好啊,帅哥 %s' % cls.name)
People.say_hi()
p1 = People()
在这里可以加功能啊...
你好啊,帅哥 randy
p1.say_hi()
在这里可以加功能啊...
你好啊,帅哥 randy
- 疑问,类方法如果有参数呢,好说,好说
class ClassMethod:
def __init__(self, func):
self.func = func
def __get__(self, instance, owner
): # 类来调用,instance为None,owner为类本身,实例来调用,instance为实例,owner为类本身,
def feedback(*args, **kwargs):
print('在这里可以加功能啊...')
return self.func(owner, *args, **kwargs)
return feedback
class People:
name = 'randy'
@ClassMethod # say_hi=ClassMethod(say_hi)
def say_hi(cls, msg):
print('你好啊,帅哥 %s %s' % (cls.name, msg))
People.say_hi('你是那偷心的贼')
p1 = People()
在这里可以加功能啊...
你好啊,帅哥 randy 你是那偷心的贼
p1.say_hi('你是那偷心的贼')
在这里可以加功能啊...
你好啊,帅哥 randy 你是那偷心的贼
一十、自定制@staticmethod
class StaticMethod:
def __init__(self, func):
self.func = func
def __get__(
self, instance,
owner): # 类来调用，instance为None，owner为类本身，实例来调用，instance为实例，owner为类本身
def feedback(*args, **kwargs):
print('在这里可以加功能啊...')
return self.func(*args, **kwargs)
return feedback
class People:
@StaticMethod # say_hi = StaticMethod(say_hi)
def say_hi(x, y, z):
print('------>', x, y, z)
People.say_hi(1, 2, 3)
p1 = People()
在这里可以加功能啊...
------> 1 2 3
p1.say_hi(4, 5, 6)
在这里可以加功能啊...
------> 4 5 6