编写一个类装饰器，使所有方法都应用装饰器

Question

我正在尝试写一个类装饰器，这个装饰器可以对类里面的所有方法都应用另一个装饰器：

import inspect


def decorate_func(func):
    def wrapper(*args, **kwargs):
        print "before"
        ret = func(*args, **kwargs)
        print "after"
        return ret
    for attr in "__module__", "__name__", "__doc__":
        setattr(wrapper, attr, getattr(func, attr))
    return wrapper


def decorate_class(cls):
    for name, meth in inspect.getmembers(cls, inspect.ismethod):
        setattr(cls, name, decorate_func(meth))
    return cls


@decorate_class
class MyClass(object):

    def __init__(self):
        self.a = 10
        print "__init__"

    def foo(self):
        print self.a

    @staticmethod
    def baz():
        print "baz"

    @classmethod
    def bar(cls):
        print "bar"


obj = MyClass()
obj.foo()
obj.baz()
MyClass.baz()
obj.bar()
MyClass.bar()

这个方法差不多可以用，但@classmethod需要特别处理：

$ python test.py
before
__init__
after
before
10
after
baz
baz
before
Traceback (most recent call last):
  File "test.py", line 44, in <module>
    obj.bar()
  File "test.py", line 7, in wrapper
    ret = func(*args, **kwargs)
TypeError: bar() takes exactly 1 argument (2 given)

有没有什么好的办法来解决这个问题呢？我查看了被@classmethod装饰的方法，但我没有找到什么能把它们和其他“类型”的方法区分开的东西。

更新

这里是完整的解决方案，记录一下（使用描述符来很好地处理@staticmethod和@classmethod，还有aix的技巧来区分@classmethod和普通方法）：

import inspect


class DecoratedMethod(object):

    def __init__(self, func):
        self.func = func

    def __get__(self, obj, cls=None):
        def wrapper(*args, **kwargs):
            print "before"
            ret = self.func(obj, *args, **kwargs)
            print "after"
            return ret
        for attr in "__module__", "__name__", "__doc__":
            setattr(wrapper, attr, getattr(self.func, attr))
        return wrapper


class DecoratedClassMethod(object):

    def __init__(self, func):
        self.func = func

    def __get__(self, obj, cls=None):
        def wrapper(*args, **kwargs):
            print "before"
            ret = self.func(*args, **kwargs)
            print "after"
            return ret
        for attr in "__module__", "__name__", "__doc__":
            setattr(wrapper, attr, getattr(self.func, attr))
        return wrapper


def decorate_class(cls):
    for name, meth in inspect.getmembers(cls):
        if inspect.ismethod(meth):
            if inspect.isclass(meth.im_self):
                # meth is a classmethod
                setattr(cls, name, DecoratedClassMethod(meth))
            else:
                # meth is a regular method
                setattr(cls, name, DecoratedMethod(meth))
        elif inspect.isfunction(meth):
            # meth is a staticmethod
            setattr(cls, name, DecoratedClassMethod(meth))
    return cls


@decorate_class
class MyClass(object):

    def __init__(self):
        self.a = 10
        print "__init__"

    def foo(self):
        print self.a

    @staticmethod
    def baz():
        print "baz"

    @classmethod
    def bar(cls):
        print "bar"


obj = MyClass()
obj.foo()
obj.baz()
MyClass.baz()
obj.bar()
MyClass.bar()

Answer 1

上面的回答不直接适用于python3。根据其他一些很棒的回答，我想出了以下解决方案：

import inspect
import types
import networkx as nx


def override_methods(cls):
    for name, meth in inspect.getmembers(cls):
        if name in cls.methods_to_override:
            setattr(cls, name, cls.DecorateMethod(meth))
    return cls


@override_methods
class DiGraph(nx.DiGraph):

    methods_to_override = ("add_node", "remove_edge", "add_edge")

    class DecorateMethod:

        def __init__(self, func):
            self.func = func

        def __get__(self, obj, cls=None):
            def wrapper(*args, **kwargs):
                ret = self.func(obj, *args, **kwargs)
                obj._dirty = True  # This is the attribute I want to update
                return ret
            return wrapper

    def __init__(self):
        super().__init__()
        self._dirty = True

现在，每当调用元组 methods_to_override 中的方法时，都会设置一个“脏标志”。当然，其他任何东西也可以放在这里。并不一定要在需要重写方法的类中包含 DecorateMethod 类。不过，由于 DecorateMethod 使用了特定的类属性，我更倾向于将其作为类属性来处理。

Answer 2

（评论太长了）

我擅自为你的解决方案增加了一个功能，可以指定哪些方法需要被装饰：

def class_decorator(*method_names):

    def wrapper(cls):

        for name, meth in inspect.getmembers(cls):
            if name in method_names or len(method_names) == 0:
                if inspect.ismethod(meth):
                    if inspect.isclass(meth.im_self):
                        # meth is a classmethod
                        setattr(cls, name, VerifyTokenMethod(meth))
                    else:
                        # meth is a regular method
                        setattr(cls, name, VerifyTokenMethod(meth))
                elif inspect.isfunction(meth):
                    # meth is a staticmethod
                    setattr(cls, name, VerifyTokenMethod(meth))

        return cls

    return wrapper

使用方法：

@class_decorator('some_method')
class Foo(object):

    def some_method(self):
        print 'I am decorated'

    def another_method(self):
        print 'I am NOT decorated'

Answer 3

你可以用 inspect.isclass(meth.im_self) 这个代码来判断 meth 是否是一个类方法。

def decorate_class(cls):
    for name, meth in inspect.getmembers(cls, inspect.ismethod):
        if inspect.isclass(meth.im_self):
          print '%s is a class method' % name
          # TODO
        ...
    return cls