Tail Recursion In Python

Some programming languages are tail-recursive, essentially this means is that they're able to make optimizations to functions that return the result of calling themselves. That is, the function returns only a call to itself.

Confusing, I know, but stick with me. It turns out that most recursive functions can be reworked into the tail-call form. Here's an example of the factorial function in it's original form, then reworked into the tail-call form.

def factorial(n):
  if n == 0: return 1
  else: return factorial(n-1) * n

def tail_factorial(n, accumulator=1):
  if n == 0: return accumulator
  else: return tail_factorial(n-1, accumulator * n)

They both look similar, and in fact the original even looks like it's in the tail call form, but since there's that pesky multiplication which is outside of the recursive call it can't be optimized away. In the non-tail version the computer needs to keep track of the number you're going to multiply it with, whereas in the tail-call version the computer can realize that the only work left to do is another function call and it can forget about all of the variables and state used in the current function (or if it's really smart, it can re-use the memory of the last function call for the new one)

This is all great, but there's a problem with that example, namely that python doesn't support tail-call optimization. There's a few reasons for this, the simplest of which is just that python is built more around the idea of iteration than recursion.

But hey, I don't really care if this is something we should or shouldn't be doing, I'm just curious if we can! Let's see if we can make it happen.

# factorial.py
from tail_recursion import tail_recursive, recurse

# Normal recursion depth maxes out at 980, this one works indefinitely
@tail_recursive
def factorial(n, accumulator=1):
    if n == 0:
        return accumulator
    recurse(n-1, accumulator=accumulator*n)

# tail_recursion.py
class Recurse(Exception):
    def __init__(self, *args, **kwargs):
        self.args = args
        self.kwargs = kwargs

def recurse(*args, **kwargs):
    raise Recurse(*args, **kwargs)
        
def tail_recursive(f):
    def decorated(*args, **kwargs):
        while True:
            try:
                return f(*args, **kwargs)
            except Recurse as r:
                args = r.args
                kwargs = r.kwargs
                continue
    return decorated

Now, don't get scared by decorators if you haven't seen them before, in fact go read about them now, basically they're functions which are called on other functions and change the behaviour in some way.

This decorator will call the function it's given and will check to see if it wants to 'recurse'. We signal a 'recursion' by simply raising an exception with the arguments we'd like to recurse with. Then our decorator simply unpacks the variables from the exception and tries calling the function again.

Eventually we'll reach our exit condition (we hope) and the function will return instead of raising an exception. At this point the decorator just passes along that return value to whoever was asking for it.

This particular method helps out with doing recursive calls in python because python has a rather small limit to how many recursive calls can be made (typically ~1000). The reason for this limit is (among other things) doing recursive calls takes a lot of memory and resources because each frame in the call stack must be persisted until the call is complete. Our decorator gets around that problem by continually entering and exiting a single call, so technically our function isn't actually recursive anymore and we avoid the limits.

I tested out both versions, the normal version hits the tail-recursion limit at factorial(980) whereas the tail-recursive version will happily compute numbers as large as your computer can handle.

There's an alternative approach that actually uses stack introspection to do it, but it's a bit more complex than the one we built here.

Hope you learned something, cheers!

Hopefully you learned something 🤞! If you did, please consider checking out my book: It teaches the principles of using optics in Haskell and other functional programming languages and takes you all the way from an beginner to wizard in all types of optics! You can get it here. Every sale helps me justify more time writing blog posts like this one and helps me to continue writing educational functional programming content. Cheers!