我该使用描述符类还是属性工厂？

Question

假设我有一个类，我想要创建很多类似“属性风格”的属性。例如：

class Foo(object):
    @property
    def foo(self):
        return self._foo

    @foo.setter
    def foo(self, val):
        self.update()
        self._foo = val

    @property
    def bar(self):
        return self._bar

    @bar.setter
    def bar(self, val):
        self.update()
        self._bar = val

    @property
    def baz(self):
        return self._baz

    @baz.setter
    def baz(self, val):
        self.update()
        self._baz = val

    def update(self):
        print "Updating..."

显然，这里有很多重复的部分，如果能把这些重复的内容提取出来就好了。有一种方法是写一个函数，让它返回一个属性：

def create_property(var):
    def _setter(self, val):
        self.update()
        setattr(self, '_' + var, val)

    def _getter(self):
        return getattr(self, '_' + var)

    return property(_getter, _setter)

class Foo(object):
    foo = create_property("foo")
    bar = create_property("bar")
    baz = create_property("baz")

    def update(self):
        print "Updating..."

另外，我也可以写一个描述符类：

class UpdateOnChange(object):
    def __init__(self):
        self.data = WeakKeyDictionary()

    def __get__(self, instance, owner):
        try:
            return self.data[instance]
        except KeyError:
            raise AttributeError("attribute not set")

    def __set__(self, instance, value):
        instance.update()
        self.data[instance] = value

class Foo(object):
    foo = UpdateOnChange()
    bar = UpdateOnChange()
    baz = UpdateOnChange()

    def update(self):
        print "Updating"

那么，哪种方法最好、最快、最符合Python的风格呢？

Answer 1

把描述符的概念和一个元类结合起来，就像我在评论里提到的那样。

class UpdateOnChange(object):
    def __init__(self, name):
        self.name = name
        self._name = '_' + name

    def __get__(self, instance, owner):
        try:
            return getattr(instance, self._name)
        except AttributeError:
            raise AttributeError('%s has no attribute %s.' % (instance, self.name))

    def __set__(self, instance, value):
        # Assume first assignment is initialization and does not require update.
        if not hasattr(instance, self._name):
            setattr(instance, self._name, value)
        else:
            instance.update()
            setattr(instance, self._name, value)

class UpdateOnChangeMeta(type):

     def __new__(cls, name, bases, attrs):
         for attr_name in attrs:
             if attrs[attr_name] == UpdateOnChange:
                 attrs[attr_name] = UpdateOnChange(attr_name)
         return type.__new__(cls, name, bases, attrs)

class Foo(object):

    __metaclass__ = UpdateOnChangeMeta
    foo = UpdateOnChange
    bar = UpdateOnChange
    baz = UpdateOnChange

    def update(self):
        print "Updating"