Automatically decorating every instance method in

问题:

I want to apply the same decorator to every method in a given class, other than those that start and end with __.

It seems to me it should be doable using a class decorator. Are there any pitfalls to be aware of?

Ideally, I'd also like to be able to:

1. disable this mechanism for some methods by marking them with a special decorator
2. enable this mechanism for subclasses as well
3. enable this mechanism even for methods that are added to this class in runtime

[Note: I'm using Python 3.2, so I'm fine if this relies on features added recently.]

Here's my attempt:

_methods_to_skip = {}

def apply(decorator):
  def apply_decorator(cls):
    for method_name, method in get_all_instance_methods(cls):
      if (cls, method) in _methods_to_skip:
        continue
      if method_name[:2] == `__` and method_name[-2:] == `__`:
        continue
      cls.method_name = decorator(method)
  return apply_decorator

def dont_decorate(method):
  _methods_to_skip.add((get_class_from_method(method), method))
  return method

Here are things I have problems with:

• how to implement get_all_instance_methods function
• not sure if my cls.method_name = decorator(method) line is correct
• how to do the same to any methods added to a class in runtime
• how to apply this to subclasses
• how to implement get_class_from_method

回答1:

I think this is better done with a metaclass, in order to handle both runtime and subclass method decoration. I don't see an elegant way to handle subclasses automatically with a class decorator.

from types import FunctionType

# check if an object should be decorated
def do_decorate(attr, value):
    return ('__' not in attr and
            isinstance(value, FunctionType) and
            getattr(value, 'decorate', True))

# decorate all instance methods (unless excluded) with the same decorator
def decorate_all(decorator):
    class DecorateAll(type):
        def __new__(cls, name, bases, dct):
            for attr, value in dct.iteritems():
                if do_decorate(attr, value):
                    dct[attr] = decorator(value)
            return super(DecorateAll, cls).__new__(cls, name, bases, dct)
        def __setattr__(self, attr, value):
            if do_decorate(attr, value):
                value = decorator(value)
            super(DecorateAll, self).__setattr__(attr, value)
    return DecorateAll

# decorator to exclude methods
def dont_decorate(f):
    f.decorate = False
    return f

And an example of its use (Python 2, but trivially modified for Python 3):

def printer(f):
    print f
    return f

class Foo(object):
    __metaclass__ = decorate_all(printer)
    def bar(self):
        pass
    @dont_decorate
    def baz(self):
        pass
    @classmethod
    def test(self):
        pass
# prints
# <function bar at 0x04EB59B0>

class AnotherName(Foo):
    def blah(self):
        pass
# prints
# <function blah at 0x04EB5930>

Foo.qux = lambda: 1
# prints
# <function <lambda> at 0x04EB57F0>

回答2:

You could do this (not sure if this is the most elegant way though):

def get_all_instance_methods(x):
    return filter(callable, map(lambda d: getattr(x, d), dir(x)))

As for the cls.method_name, you will have to use getattr(cls, method_name).