Forty minutes to the punch line or "we'll never look at functions the same way again""

How many times do we teach something and leave the kids thinking:

"what's the point of this?" "When will I use this?" or even just the plain old fashioned "that's weird."

It's pretty cool when a lesson starts out that way but you get to the payoff by the end of the class.

Today we started exploring some advanced python.

We started by showing that you can assign functions to variables or pass them as parameters:

def inc(x):
    return x+1
def dec(x):
    return x-1

f = inc
print f(5)

flist = [inc,dec]

print flist[1](5)

We then looked at closures in python:

def makeAdder(n):
    def inner(x):
        return x+n
    return inner

add3 = makeAdder(3)
add5 = makeAdder(5)

print add3(10)
print add5(10)

The idea that we can make a function that builds and returns a function. When we call makeAdder(3), the 3 is passed as parameter n so the inner function reduces to return x+3 and then we return that inner function. When we later run it: add3(10) it adds 10+3. makeAdder(5) works similarly but passing a 5 in for n instead of a 3.

Even a somewhat clunky way of doing class type abstractions:

def make_counter():
    # private "instance" data
    # has to be a list due to weird python scoping rules
    count = [0]

    # public methods
    def inc():
        count[0]=count[0]+1
    def dec():
        count[0]=count[0]-1
    def reset():
        count[0]=0
    def get():
        return count[0]

    # send back a dictionary so we have access to all the methods
    return {'inc': inc, 'dec': dec, 'reset': reset, 'get': get}

c1 = make_counter()
c2 = make_counter()

# we've got to use the clunky list notation 
c1['inc']()
c1['inc']()
c1['inc']()
print c1['get']()
c2['inc']()
print c2['get']()
c1['reset']()
print c1['get']()

Up to now the students are able to see how this works but the why is unclear.

So now we'll look at where this is useful.

Let's suppose we have routines like the following. It returns a string:

import random
def get_name():
    names = ['tom','sue','harry','lisa','sarah','horatio']
    return random.choice(names)

get_name and routines like it might be scattered throughout our code. Let's suppose, for some strange reason, we want to double the name every time we use it. A "traditional" way of doing this might be:

def dble(f):
    name = f()
    return name+name

print dble(get_name) # returns something like tomtom

Here, we pass a function which returns a string and then dble returns that string repeated twice. The problem here is that if we've got get_name all over our code base, we have to find it and change each instance to dble(get_name)

What if we write it as a function that returns a function:

def doubler(f):
    def inner():
        name=f()
        return name+name
    return inner

Now, in this case, we can do something like:

# make a new function that wraps get_name in "inner"
# when we call inner, it returns the string twice
f = doubler(get_name)
# and then later
print f() # will print something like tomtom

The cool part is that we can also do this:

get_name = double(get_name)

And then whenever we call our new get_name we end up calling the wrapped function.

This was the first A-ha. We don't have to change get_name all over our code base - we only have to change it once.

Once we write a function like doubler, instead of re-assigning get_name as we did above, we can do the following:

@doubler
def get_name():
    # rest of code as defined above

@doubler():
def demo():
    return "hello"

The second example will have demo return "hellohello" whenever we invoke it.

A Python decorator is merely shorthand for calling a wrapper type function like doubler.

This was the second A-ha moment – we can write functions that transform functions.

We'll cover more decorator and closure plumbing tomorrow and then start doing some fun stuff with these concepts.