Python Decorators

I’ve now been working in Python for more than a year and we have been doing some pretty crazy stuff, especially around decorators and authorization/permissions implementations.

Admittedly, Python decorators are one of the bits of the language that can be best defined as ‘magic’ and certainly remains puzzling for someone like me, used to Java strict and (I still believe, much safer) compile-time type-checking.

Anyway, sparked by this very useful article on decorators, I have decided to further explore the topic of class decorators, that in the original article was just mentioned in passing, and found a few twists – the most surprising of which has certainly been finding out that the call sequence, and the type of objects being associated with type names depends on whether the decorator annotation is followed by a parameter list.

It is worth noting that python Mocks use decorators pretty heavily, and it’s easy to see why, once you browse the code below: it’s heavily commented and should give enough of an idea of what goes on.

"""
Code to try out Python decorators.

Original idea from:
http://www.brianholdefehr.com/decorators-and-functional-python
"""

class DecoratorClass(object):

    def __init__(self, klass=None, **kwargs):
        """This gets called every time the @DecoratorClass annotation
        is encountered.

        @param klass: the type of the class being decorated, it will be 
                passed in if no parameters are passed in the annotation
        @type klass: type
        @param kwargs: a dictionary containing any of the named 
                parameters passed in the decorator's declaration
        @type kwargs: dict
        """

        # If the decorator is declared without parameters, 
        # the class type will be passed in
        self._name = 'default'
        self._klass = None
        if klass:
            print 'No-args decorator for: ', klass.__name__
            self._klass = klass
        else:
            print 'Args declared: ', kwargs
            self._name = kwargs.get('name')

    def __call__(self, *args, **kwargs):
        """This gets invoked every time a decorated class gets created, 
        with the actual arguments in what would appear to look like a 
        constructor call: they do not have to (and in fact, won't) match 
        the actual formal argument list of the
        constructor of the decorated class (which may not even get invoked).

        If the decorator is declared with one or more arguments 
        (see Different class below), then the first item in ``args`` will be the
        **type** of the class being decorated (which
        will not have been passed in to the __init__()).

        This gets invoked immediately after self#__init__() if the decorator is 
        declared with one or more arguments, or when the "constructor" is invoked.
        """
        print 'Decorator __call__ invoked with: ', args, kwargs
        if self._klass:
            print 'Decorated instance: ', self._klass.__name__
            return self._klass(self._name)
        else:
            print 'Setting up klass'
            self._klass = args[0]
            # Here we inject a 'class-level' static method
            self._klass.get_name = self.get_name
        return self._klass

    def get_name(self):
        return self._name


@DecoratorClass
class Decorated(object):
    def __init__(self, name):
        print '>>>>> I am being decorated! <<<<<'         self.name = name     def call_me(self, *args):         print 'I\'m being called: ', self.name         print 'These are my args: ', args # Notice how here a 'type' name (Decorated) has been completely  # 'hijacked' to point to a specific # instance of a DecoratorClass object (this provides 'closure') print 'what is Decorated here?', Decorated # >> what is Decorated here? 

@DecoratorClass(name='another')
class Different(object):
    def __init__(self):
        # Notice how here we are using a 'static' method 
        # that has been injected by the decorator
        self.name = Different.get_name()

    def method(self, *args):
        """This is the same implementation as Decorated#call_me(), 
        just to show how they behave differently
        """
        print 'I\'m being called: ', self.name
        print 'These are my args: ', args

    def __call__(self, *args, **kwargs):
        return 'Different called with: ', args, kwargs

# Here, Different is whatever DecoratorClass#__call__()  returned: 
# this happens to be what one would expect it to be (a Different class 
# type) but that's only because the code makes it so

print 'what is Different here? ', Different
# >> what is Different here?  

print 'Calling Decorated class constructor'
# Note how the params being passed here have no relation with the 
# constructor argument list and in fact, the name of the 'decorated' 
# class is not even given here, but in the decorator

# Due to the 'magic' of decorators, DecoratorClass#__call__() is
# instead invoked here,
# on the instance that was created at declaration
deco = Decorated(123, "Hello", foo='foo', baz='baz')

# Again, now ``deco`` here happens to be what one would expect it to 
# be (an instance of the Decorated class) but solely because the code 
# makes it so - it could have been anything, really

deco.call_me(1, 2, 3, 1)
# >> I'm being called:  default
# >> These are my args:  (1, 2, 3, 1)

print 'Creating now a Different class:'
# Here the call does actually invoke the constructor (__init__()) 
# for the Different class, and the returned object is again what one 
# would expect it to be (an instance of Different):
diff = Different()

diff.method('quart', 'naught')
# >> I'm being called:  another
# >> These are my args:  ('quart', 'naught')

# Obviously, as the Different#__call__() method is defined, we can call it too:

print diff(1, 2, 3, value='val')
# >> ('Different called with: ', (1, 2, 3), {'value': 'val'})

# If we now create an entirely different instance of a Different object,
# it will still have the same 'static' method injected by the decorator:
another_diff = Different()

print 'My name is:', another_diff.name
# >> My name is: another
Advertisements

One thought on “Python Decorators

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s