How to call the original method when it is monkey-patched?

Question

I have a class in the main project I don't want to change.

class A():
    def __init__(self, firstname, lastname):
        self.firstname = firstname
        self.lastname = lastname

    def name(self):
        # this method could be much more complex
        return self.lastname.upper()

I'm trying to build a plugin mechansim. So far so good, I have an extension point like this:

if __name__ == '__main__':
    ''' The main project has an extension point that allows me to do'''
    # for each class extension such as AExtended:
    A.name = AExtended.name
    ''' After the extensions are loaded, some behaviours may be changed'''
    a = A("John", "Doe")
    print(a.name())

A plugin can be written like this:

class AExtended(A):
    ''' This is an extension I provide through a plugin mechanism
    '''
    def name(self):
        return self.firstname + ' ' + self.lastname.upper()

This all works very well. I now get "John DOE".

My problem is that the original name() method can be quite complex. In other words, I can't afford to call self.lastname.upper() in the AExtended. I'd like to call the "super" method, which does not exist any more, because it has been overwritten.

How can I change my code, in order to achieve something like this:

class AExtended(A):
    def name(self):
        # I'm looking for a way to call the base implementation that was in A.name()
        return self.firstname + ' ' + parent.name()

Thanks for your help!

Edit: Some explanations of what I try to do.

• I want the plugin to patch the behaviour of A. I can't afford to change existing consumers of A
• There are many classes like A that could be changed, I'd like plugins to have full control and responsibility
• It's true AExtended does not have to inherit from A, but it was an easy way to access self.firstname. I have no problem following a different design pattern if it can help.

I have a workaround, but it's not very elegant and hard to generalize

class AExtended(A):
    def name(self):
        # I'm looking for a way to call the base implementation that was in A.name()
        return self.firstname + ' ' + self.parentname()
#in main
A.parentname = A.name
A.name = AExtended.name

Answer 1

This is what we call a 'decorator' pattern. Replace the original reassignment of name to have it call a function instead, which takes the original. It then returns a new function.

def name_decorator(method):
    def decorate_name(self=None):
        return stuff + method(self)
    return decorate_name
A.name = name_decorator(A.name)

Later, calling A.name will call decorate_name with self as the current instance and method will be available to it which points to the function at the time of the reassignment.

Answer 2

Here's a full example of what I was hinting at. Feel free to yell at me and have me merge my answers, or downvote one or whatever, just easier to provide an alternative as a new answer. I'll let the code do the talking instead of poorly explaining it. :)

## Some shared class that is used all over the place and needs to be patched.

class A(object):
    def __init__(self):
        self.firstname = 'Bob'

    # Print my first name.
    def name(self):
        return self.firstname

    # Use this to allow patching arbitrary methods...
    @classmethod
    def patch(cls, func_name):
        def patch_by_name(new_func):
            old_func = getattr(cls, func_name)
            def patched_func(self):
                return new_func(self, old_func)
            setattr(cls, func_name, patched_func)
        return patch_by_name

## Some other area of the code where you want to throw in a patch

class PatchedA(A):  # doesn't need to subclass, but comes in handy sometimes
    @A.patch('name')
    def name(self, orig_func):
        return 'I am ' + orig_func(self) + 'McWizwaz'

print 'Who are you, A class?'
print A().name()  # prints 'I am Bob McWizwaz'

Answer 3

class ABase(object):
    def name(self):
        pass

class A(object):
    pass

class AExtension(ABase):
    def name(self):
        return ABase.name(self)

A.name = AExtension.name

Answer 4

Standing on the shoulders of giants (@spencer), here's a bit more generic sample. It lets one patch a class Original without touching its source code. In this version, parameters fly all over the place, and a reference to an arbitrary context is given to the new method.

class Original:
    def f( self, a, b = None ):
        print( "f", a, b )
        self.g( b )

    def g( self, b ):
        print( "g", b )

class Patcher:
    def prepare_class( self, clazz ):
        @classmethod
        def on_class_patcher( cls, func_name, context ):
            def patch_by_name( new_func) :
                old_func = getattr( cls, func_name )
                def patched_func( self, *args, **kwargs ):
                    return new_func( self, old_func, context, *args, **kwargs )
                setattr( cls, func_name, patched_func )
            return patch_by_name

        setattr( clazz, "patch", on_class_patcher )

class Another:
    def log( self, level, info ):
        print( level, info )

Now, let's patch something:

obj = Original()
obj.f( 1, b = "hello" )

p = Patcher()
p.prepare_class( clazz = Original )

logger = Another()

@Original.patch( "f", context = logger  )
def new_f( self, old_f, context, a, b ):
    print( "new_f", a, b )
    context.log( "zzz", a )
    old_f( self, a, b )

obj = Original()
obj.f( 1, b = "hello" )

And the output is:

f 1 hello
g hello
new_f 1 hello
zzz 1
f 1 hello
g hello

Answer 5

One option that might not always be the best in a language like Python is to use the @override decorator from this non-standard package. But this is a viable option only if your two functions work on different types or different number of arguments. Other than that, there's not much you can do, besides renaming your function.