惰性类属性装饰器

我有一个 Django 模型，需要对自定义用户模型进行一些处理。

我无法在类加载时使用该模型的类，因为类的加载顺序未知。

所以我需要在运行时添加一些类属性，目前我将它们添加到__init__ or __new__ like:

def __new__(cls, *args, **kwargs):
    # hack to avoid INSTALLED_APPS initialization conflicts.
    # get_user_model() can't be called from this module at class loading time,
    # so some class attributes must be added later.
    # Metaclasses could me more appropiate but I don't want to override
    # dango's metaclasses.
    if not hasattr(cls, '_reverse_field_name_to_user'):
        cls._find_reverse_field_name_to_user()
    return Group.__new__(cls, *args, **kwargs)

它有效，但看起来很糟糕，所以我考虑过使用类似的东西@lazyclassproperty对于这些属性。

我找到了几个@classproperty and @lazyproperty装饰器，但不是一个同时适用于两者的装饰器，而且我不知道如何自己编写一个。

Question：我如何编写这样的装饰器？或者建议我当前愚蠢的实现的另一种更清洁的替代方案。

Pyramid 框架有一个非常好的装饰器，称为reify，但它只适用于实例级别，而您需要类级别，所以让我们稍微修改一下

class class_reify(object):
    def __init__(self, wrapped):
        self.wrapped = wrapped
        try:
            self.__doc__ = wrapped.__doc__
        except: # pragma: no cover
            pass

    # original sets the attributes on the instance
    # def __get__(self, inst, objtype=None):
    #    if inst is None:
    #        return self
    #    val = self.wrapped(inst)
    #    setattr(inst, self.wrapped.__name__, val)
    #    return val

    # ignore the instance, and just set them on the class
    # if called on a class, inst is None and objtype is the class
    # if called on an instance, inst is the instance, and objtype 
    # the class
    def __get__(self, inst, objtype=None):
        # ask the value from the wrapped object, giving it
        # our class
        val = self.wrapped(objtype)

        # and set the attribute directly to the class, thereby
        # avoiding the descriptor to be called multiple times
        setattr(objtype, self.wrapped.__name__, val)

        # and return the calculated value
        return val

class Test(object):
    @class_reify
    def foo(cls):
        print("foo called for class", cls)
        return 42

print(Test.foo)
print(Test.foo)

运行程序并打印

foo called for class <class '__main__.Test'>
42
42