It’s all good- Decorating Python like Martha Stewart


You can get by in Python with basic constructs but...

  • you might get bored
  • be confused by other’s code
  • want more power

Function Review

A function is an instance ot type function

>>> def spam():
...     "A function"
...     print 'eggs'
>>> spam
<function 0x2342342>
>>> callable(spam)
>>> spam()

Functions have attributes

>>> spam.func_name
>>> spam.__docstring__
"A function"

A function knows about itself

>>> def foo2():
...     print "NAME", foo2.func_name

A function can have attributes assigned:

>>> def foo3():
...     print "STUFF", foo3.stuff
>>> foo3.stuff = "Data"
>>> foo3()

Function Definition

def func_name(arg1, arg2=value, *args, **kwargs):
    """ docstring """
    # implementation

Function Gotcha

When a function is created, the named/default parameters are defined when the function is created

def named_param(a, foo=[]):
    if not poo:

print named_param.func_defaults

print named_param.func_defaults
([1, ])

Lists and dicts are mutable. When you modify them you don’t create a new list (or dict). Strings and ints are immutable

Parameters are evaluated when the def they belong to is imported

Don’t default to mutable types.

def named_param(a, foo=None):
    foo = foo or []
    if not foo:

*args and **kwargs


  • *args is a tuple of parameter values.
  • **kwargs is a dictionary of key/values
def param_func(a, b=2, c=5):
    print [x for x in [a, b, c]]

The ‘*’ before args flattens the tuple of parameters values.

def param_func(a, *args):
    print [x for x in [args]]
    # TODO  check I got this right

def kwargs_func(a, **kwargs):
    print [x for x in [kwargs]]
    # TODO  check I got this right

def param_func(a, b='b', *args, **kwargs):
    print [x for x in [a, b, args, kwargs]]


  • PEP 227 and came out in Python 2.1
  • Don’t be afraid of them
  • In Python a function can return a new function. The inner function a closuse and any variable it accesses that are defined outside of that function are free variables.
def add_x(x):
    def adder(num):
        # we have read acces to x
        return x + num # x is a free variable here
    return adder

sadd_5 = add_x(5)
add_5 # doctest: + ELLIPSESS
<function add at 0x12324ewe>
print add_5(10)

Nested functions only have write access to global and local scope.

x = 3
def outer():
    x = 4 # now local
    y = 2
    def inner():
        global x
        x = 5 #
    print x
    inner() # only changes the local inside the function
    print x
print outer()
print x # since global the global value

Python 3.x has a non-local keyword that replaces the global in Python 2.x


  • PEPS 318, 3129, implemented in Python 2.4

  • allow you to

    • modify arguments
    • modify function
    • modify results
# count how many times a function is called
call_count = 0
def count(func):
    def wrapper(*args, **kwargs):
        global call_count
        call_count += 1
        return func(*args, **kwargs)
    return wrapper

def hello():
    print 'invoked hello'
>>> hello = count(hello) # invoking count with the argument being the hello object
>>> hello()
>>> print call_count
>>> 1
>>> hello()
>>> print call_count
>>> 2
# Decorator Shortcut
def hello():
    return 'hello'

Better decorator:

def count2(func):
    # TODO - show this one out

Decorator Template

def decorator(function_to_decorate):
    def wrapper(*args, **kwargs):
        # do something before invoation
        result = func_to_decorate(*args, **kwargs)

        # do something after
        return result
    # update wrapper.__doc__ and .func_name
    # or functools.wraps
    return wrapper
# class as a decorator
class decorator_class(object):
    def __init__(self, function):
        self.function = function
    def __call__(self, *arg, **kwargs):
        result  = self.function(*arg, **kwargs):
        # do stuff to result
        return result

def hello():
    return 'hello'


Anything that is callable can be used to create a decorator

# using a class instance as a decorator
# instead of using __call__ use __init__ and then instantiate the class before using it.
deco = Decorator()

def hello():
    return 'hello'

# You can modify deco later! This is UBER powerful!


Not the same as “Class Decorators”. See PEP 3129

Paramaterized decorators

  • need 2 closures
def limit(length):
    def decorator(function):
        def wrapper(*args, **kwargs):
            result = function(*args, **kwargs)
            return result[:length]
        return wrapper
    return decorator

@limit(5) #notice parens
def echo(foo):
    return foo

# usage

#syntactical sugar for
echo = limit(5)

Warning: Function attributes get mangled in decorators

  • I’ve run into this - when you wrap a function a decorator the attributes get lost
  • Docstring kills me
  • Do this:
def limit(length):
    def decorator(function)
        def wrapper(*args, **kwargs):
            result = function(*args, **kwargs)
            result = result[:length]
        return wrapper
        wrapper.__docstring__ = function.__docstring__
    return decorator

You can also use functools to deal with this issue, but it’s not as clear a read

import functools
def limit(length):
    def decorator(function)
        def wrapper(*args, **kwargs):
            result = function(*args, **kwargs)
            result = result[:length]
        return wrapper
        wrapper.__docstring__ = function.__docstring__
    return decorator

Uses for decorators

  • caching

    • I wrote a cache decorator that uses Raymond Hettinger’s LRU cache code.
  • monkey patching stfio

  • jsonify

  • logging time in function call

  • change cwd

  • timeout a function call

What if I want to tweak decorator paramers at runtime?

What if I made a mistake in a param and want to change values?

  • Use class instance decorator

  • Tweak wrapper attributes

  • Use context manager

  • or...

    • Since a decorator is just a class you can invoke it at runtime. Like this:
# TODO get example
result = limit(4)(echo)