pythonのdecorator@

decoratorはpythonの特性です.
decoratorは関数であってもクラスであってもよい.
 
decorator@の用途:
1)パラメータチェック,呼び出し統計など,関数の開始前に初期作業を行う.
2)関数の終了後にいくつかの終了作業を行う.
3)C++テンプレートのような機能.
 
decoratorが関数の場合、

def decorator(fun):
    def call(*args, **kwargs):
        print "do something before fun!"
        if len(args) < 1:
            print "args is less"
        else:
            fun(args[0])
        print "do something after fun!"
    return call

@decorator
def hello(a):
    print "hello"

# hello(*args, **kwargs)
# is equivalent to
# decorator(hello)(*args, **kwargs)
# see the example for more detail

print "hello() is called:"
hello()

print "-----------------------------"

print "hello(1) is called:"
hello(1)


# decorator chain

def chain0(fun):
    def call(*args0, **kwargs0):
        print "chain0"
        fun(*args0, **kwargs0)
    return call


def chain1(fun):
    def call(*args1, **kwargs1):
        print "chain1"
        fun(*args1, **kwargs1)
    return call

@chain0
@chain1
def chain2(*args2, **kwargs2):
    print "chain2"

# attention!!
# chain2(*args0, **kwargs0)
# is equivalent to chain0(chain1(chain2))(*args0, **kwargs0)

# it is not chain2(*args2, **kwargs2)
# it is not chain0(chain1(chain2(*args0, **kwargs0)))
# it is not chain0(chain1(chain2(*args2, **kwargs2)))

chain2()

 
実行結果:
hello() is called:
do something before fun!
args is less
do something after fun!
-----------------------------
hello(1) is called:
do something before fun!
hello
do something after fun!
chain0
chain1
chain2
 
具体的な原因は注釈を見れば分かります.
 
decoratorがクラスの場合、
 

class decorator(object):
    def __init__(self, func):
        print func
        self.func = func
        print '__init__'

    def __call__(self, *args, **kwargs):
        print 'obj = %s, cls = %s' % (self.obj, self.cls)
        return self.func.__call__(*args, **kwargs)

    def __get__(self, obj, cls):
        self.obj = obj 
        self.cls = cls 
        print '__get__'
        return self.__call__

class foo(object):
    @decorator
    def hello(self):
        print 'hello'

    @decorator
    def world(self):
        print 'world'


tmp = foo()
print '-------------------------'
tmp.hello(tmp)
tmp.world(tmp)

 
 
 
実行結果:
 

__init__

__init__
-------------------------
__get__
obj = <__main__.foo object at 0x2b978f03a710>, cls =
hello
__get__
obj = <__main__.foo object at 0x2b978f03a710>, cls =
world
 
fooインスタンスを作成するとき、
fooの関数に対応するクラスdecoratorがあれば、
クラスdecoratorもそれぞれ対応する初期化を行います.
__get__その後、関数を返し、前の関数decoratorと同じになります.