Dynamically generating object methods in Python

I have recently been working on an object-oriented wrapper for the Python bindings to my [libxtract][1] feature extraction library. I used SWIG to auto-generate the bindings, and SWIG generally does a great job. The problem is that the functions it generates follow the original C code quite precisely, and aren’t very ‘Pythonic’. I wanted to create a nice Python class to hide some of the nastiness of the auto-generated functions, but became slightly daunted by having to hand-code every method of the new Xtract class with 50 very similar methods! Instead I decided to take advantage of Python’s support for [lexical closures][2] to auto-generate the methods from the libxtract function descriptors using a method factory. The concept works like this (I think this is a fairly common idiom in Python):

#!/usr/bin/python

class foo:
    '''Dummy class'''
    def __init__(self):
        for y in range(3):
            doc = "Returns the sum of a number and %.2f" % y
            name = 'add%d' % y
            adder = self.adder_factory(y)
            setattr(self, name, adder)
            # Set the docstring
            self.__dict__[name].__doc__ = doc

    @staticmethod
    def adder_factory(y):
        "Method for generating adder functions using a lexical closure"
        def adder(x):
            return x + y
        return adder

a = foo()
x = 3

print a.add0(x) # prints 3
print a.add1(x) # prints 4
print a.add2(x) # prints 5