python中类属性与实例属性总结

2012-09-11 10:33311人阅读评论(0)收藏举报

stackoverflow上面的相关讨论

http://stackoverflow.com/questions/2923579/python-class-attribute

http://stackoverflow.com/questions/1944625/what-is-the-relationship-between-getattr-and-getattr

1. 类属性

为在类定义时直接指定的属性(不是在__init__方法中)

[python]
view plaincopyprint?

class Test:
class_attribute1="attr-value"

class Test:
class_attribute1="attr-value"

2. 实例属性

在__init__方法中添加的属性, 在其他位置添加也是可以的, 实际是通过setattr内置函数(调用__setattr__)完成, 另外也可以直接修改__dict__属性手动添加

[python]
view plaincopyprint?

t=Test()
setattr(t,attr_name,attr_value) #等价于 t.attr_name=attr_value

t.__dict__[attr_name]=attr_value

t=Test()
setattr(t,attr_name,attr_value) #等价于 t.attr_name=attr_value
t.__dict__[attr_name]=attr_value

3. 通过重写__setattr__控制属性的添加

该方法在setattr(t,attr_name,attr_value)或者t.attr_name=attr_value时被调用,通过重写该方法可以达到控制属性添加的功能

[python]
view plaincopyprint?

class Test:
def __setattr__(self,name,value):

if name=="key":

self.__dict__[name]=value

class Test:
def __setattr__(self,name,value):
if name=="key":
self.__dict__[name]=value

这样, 在通过setattr(t,attr_name,attr_value)或者t.attr_name=attr_value添加新属性时, 就只能添加名称为key的属性了,

实际上这种控制是不完整的, 因为可以通过t.__dict__[attr_name]=attr_value的方式直接添加!



4. 属性的访问顺序

"instance.attr_name"访问实例属性时, 首先在instance.__dict__中查找, 如果找到返回对应值,否则在

instance.__class__.__dict__中查找, 也就是在类属性中查找, 如果找到, 返回对应值, 否则产生attributeError异常

[python]
view plaincopyprint?

class Test:
derived_val=1
def __init__(self):

pass

t1=Test()

print "t1.derived_val=",t1.derived_val
#t1.derived_val= 1
print "Test.derived_val=",Test.derived_val
#Test.derived_val=1

print "t1.__dict__=",t1.__dict__
#{}
print "Test.__dict__=",Test.__dict__
#Test.__dict__= {'derived_val': 1,...}

print "t1.__class__.__dict__=",t1.__class__.__dict__
#Test.__dict__= {'derived_val': 1,...}

Test.derived_val+=1
#类属性derived_val+1=2
t1.derived_val+=1 #t1并没有属性derived_val, 根据访问顺序, 将访问类属性derived_val, 结果为 t1.derived_val=Test.derived_val+1=3

print "t1.derived_val=",t1.derived_val
#t1.derived_val=3
print "Test.derived_val=",Test.derived_val
#Test.derived_val=2

print "t1.__dict__=",t1.__dict__
#t1.__dict__={'derived_val':3}
print "Test.__dict__=",Test.__dict__
#Test.__dict__= {'derived_val': 2, ...}

print "t1.__class__.__dict__=",t1.__class__.__dict__
#Test.__dict__= {'derived_val': 2, ...}

class Test:
derived_val=1
def __init__(self):
pass

t1=Test()

print "t1.derived_val=",t1.derived_val #t1.derived_val= 1
print "Test.derived_val=",Test.derived_val #Test.derived_val=1

print "t1.__dict__=",t1.__dict__ #{}
print "Test.__dict__=",Test.__dict__ #Test.__dict__= {'derived_val': 1,...}
print "t1.__class__.__dict__=",t1.__class__.__dict__ #Test.__dict__= {'derived_val': 1,...}

Test.derived_val+=1 #类属性derived_val+1=2
t1.derived_val+=1 #t1并没有属性derived_val, 根据访问顺序, 将访问类属性derived_val, 结果为 t1.derived_val=Test.derived_val+1=3

print "t1.derived_val=",t1.derived_val  #t1.derived_val=3
print "Test.derived_val=",Test.derived_val #Test.derived_val=2

print "t1.__dict__=",t1.__dict__ #t1.__dict__={'derived_val':3}
print "Test.__dict__=",Test.__dict__ #Test.__dict__= {'derived_val': 2, ...}
print "t1.__class__.__dict__=",t1.__class__.__dict__ #Test.__dict__= {'derived_val': 2, ...}

实际上没有找到属性对应的行为与类的__getattr__和__getattribute__方法有关

5. __getattr__和__getattribute__的区别

另请参考: http://stackoverflow.com/questions/4295678/understanding-the-difference-between-getattr-and-getattribute
如果在类中定义__getattr__方法, 该方法会在搜寻属性失败时调用, lookup attribute成功不会调用该方法

[python]
view plaincopyprint?

class A(object):

def __getattr__(self, name):

return "I pretend I have an attribute called '%s'" % name

a = A()
print a.foo # there's no attribute 'foo', but no AttributeError currs, prints "I pretend I have an attribute called 'foo'"

class A(object):
def __getattr__(self, name):
return "I pretend I have an attribute called '%s'" % name

a = A()
print a.foo # there's no attribute 'foo', but no AttributeError currs, prints "I pretend I have an attribute called 'foo'"

如果在类中定义__getattribute__方法, 该方法在获取instance的任何属性(方法也是属性!)时调用, 因为__getattribute__需要

调用super类(见"__getattribute__方法的无限循环"), 故只能使用在new-style类中(这个逻辑是否正确, 希望指出)

[python]
view plaincopyprint?

class TestClass(object):

def __init__(self):

pass
def __getattr__(self,attr_name):

print "attribute '%s' is not defined!(in function TestClass.__getattr__)" %attr_name

return "not defined!"

def __getattribute__(self,attr_name):

print "in function '__getattribute__'"

attr_val=super(TestClass,self).__getattribute__(attr_name)

return attr_val
#TestClass类定义了两个方法__getattr__ 和__getattribute__方法, 下面几个例子说明两者区别

tc=TestClass() #tc!
tc.a #a
tc.a=1 #add attribute 'a' to tc

tc.a #b
tc.__getattr__('a') #c

tc.__getattribute__('a')
#d

class TestClass(object):
def __init__(self):
pass
def __getattr__(self,attr_name):
print "attribute '%s' is not defined!(in function TestClass.__getattr__)" %attr_name
return "not defined!"
def __getattribute__(self,attr_name):
print "in function '__getattribute__'"
attr_val=super(TestClass,self).__getattribute__(attr_name)
return attr_val
#TestClass类定义了两个方法__getattr__ 和__getattribute__方法, 下面几个例子说明两者区别
tc=TestClass() #tc!
tc.a  #a
tc.a=1 #add attribute 'a' to tc
tc.a #b
tc.__getattr__('a') #c
tc.__getattribute__('a') #d

************************************************************分析*************************************************************

a. 访问属性'a',得到结果为:

in function '__getattribute__'

attribute 'a' is not defined!(in function TestClass.__getattr__)

'not defined!'

首先调用了'__getattribute__'方法,因为该方法在属性访问时一定会被调用

接着发现属性a尚未定义, 应当抛出AttributeError异常, 但是发现TestClass中定义了'__getattr__'方法, 该方法

类似AttributeError的异常处理方法, 所以python内部机制调用'__getattr__'方法

b. 这次访问属性'a', 得到结果为:

in function '__getattribute__'

1

因为'a'已经定义, 故而不再调用'__getattr__'方法, 但'__getattribute__'仍然执行

c. 显示的调用方法'__getattr__', 得到结果:

in function '__getattribute__'

attribute 'a' is not defined!(in function TestClass.__getattr__)

'not defined!'

'__getattr__'同样是'tc'的属性, 同样也会调用'__getattribute__'

注意, 属性'a'并不是不存在, 只不过'__getattr__'方法这样输出而已!

d. 调用'_getattribute__', 得到结果:

in function '__getattribute__'

in function '__getattribute__'

1

该方法只需要了解'__getattribute__'本身也是'tc'的属性, 就一目了然了

************************************************************************************************************************************

6. __getattribute__方法的无限循环

[python]
view plaincopyprint?

class TestClass(object):

def __init__(self):

pass

def __getattribute__(self,attr_name):

print "in function '__getattribute__'"

attr_val=self.__dict__[attr_name]

return attr_val

class TestClass(object):
def __init__(self):
pass

def __getattribute__(self,attr_name):
print "in function '__getattribute__'"
attr_val=self.__dict__[attr_name]
return attr_val

调用'__getattribute__'方法会陷入循环嵌套调用

根据5中测试, 我们知道在访问self.__dict__时, 会调用'__getattribute__', 这样就出现了循环嵌套调用

因此显然不能够在'__getattribute__'方法中调用self的任何属性方法, 那么既然连'__dict__'都

不能够访问, 该怎样得到attr_name对应的值那?

解决方法是使用super函数显示调用父类的'__getattribute__'方法, 这里一个问题:
object的__getattribute__是如何实现的？

7. 危险的__getattribute__的使用时机

使用'__getattribute__'存在无限循环的风险, 但是如果需要在一个较低层次控制属性的访问, 可以使用它

下面是一个例子

[python]
view plaincopyprint?

import random
#I think this is a good example to show the power you can get from overriding '__getattribute__'

class TestClass(object):

def __init__(self,stu1,*args):
#at least one student!
self.students=[stu1]

self.students.extend(args)

def __getattribute__(self,attr_name):

if attr_name=="random_student":

students=super(TestClass,self).__getattribute__("students")

pos=random.randint(0,len(students)-1)

return students[pos]
else:
return super(TestClass,self).__getattribute__(attr_name)

import random
#I think this is a good example to show the power you can get from overriding '__getattribute__'
class TestClass(object):
def __init__(self,stu1,*args): #at least one student!
self.students=[stu1]
self.students.extend(args)

def __getattribute__(self,attr_name):
if attr_name=="random_student":
students=super(TestClass,self).__getattribute__("students")
pos=random.randint(0,len(students)-1)
return students[pos]
else:
return super(TestClass,self).__getattribute__(attr_name)

该例子中, 使用'__getattribute__'方法添加了一个TestClass本不存在的属性'random_student'

通过该属性, 可以随机的访问students中的某一个student