TL;DR

Sometimes you need to find out the Greatest Common Divisor of two integers.. Euclid’s algorithm is your friend. Sometimes you need slightly more… the extended Euclid’s algorithm can help too!

A few days ago Julia Evans posted a tweet about Euclid’s algorithm, linking to a video about it. While I highly appreciate her work, and wholeheartedly encourage everyone to look at it, in this specific case I didn’t go through the whole 15 minutes explanation in the video because… well, I already know Euclid’s algorithm for finding the greatest common divisor of two integers. OK I’m cheating, I know what it is for and where to find about it.

So that should be the end of the story, right? Wrong.

## Six Lines of Code

The first thing that popped to my eyes is that the implementation was going to be six lines of code long. It seemed a bit too much (didn’t know the language to be honest), considering that it can be as low as five when done in Perl without signatures and with the sub closing brace on its own line. See the implementation at the end if you don’t believe me.

Then I looked at the code:

1  def gcf(a, b):
2      print(f"gcf({a},{b})")
3      if a < b:
4          return gcf(b, a)
5      if a % b == 0:
6          return b
7      return gcf(b, a % b)

Well, first observation is that it’s seven lines 😁. To be fair, the print at line 2 is actually only there to show what’s going on and can be removed in “the real world”.

Jokes apart, lines 3 and 4 are not really necessary. Sure, the algorithm “works” by taking the remainder of the greater number in a division by the smaller one, but if a is smaller than b:

• a % b is just a;
• if a is zero then the greater common divisor is indeed b and lines 5-6 work fine;
• otherwise, line 7 becomes the same as return gcf(b, a) i.e. line 4.

I’m not particularly fond of recursion-based solutions in general, but it’s totally my bias because I usually program in Perl and there’s no tail-recursion optimization there (that I know of, at least). No clue about Python.

In this case, though, I think that the recursive solution is perfect to explain the algorithm.

## The extended Euclid’s algorithm

Julia’s video touches upon a generalization of the algorithm that is applied over polynomials, so you should definitely watch the video to get the gist of it. Here, I’d like to elaborate a bit more on an evolution towards a different direction.

The extended Euclid’s algoritm computes the greater common divisor and something more within the same complexity bounds. Yes, it has a bigger multiplicative constant 😅. In particular, it is capable of calculating the coefficients $x$ and $y$ of the Bézout’s identity:

$a \cdot x + b \cdot y = gcd(a, b)$

Why should we care?

Sometimes we might want to do operations modulo some prime $a$ and find the inverse of $b < a$ in the field $\mathbb{Z}_a$. The greatest commond divisor between them is 1 by definition ($a$ is prime!) so the identity becomes:

$a \cdot x + b \cdot y = 1$

Going modulo $a$ simply yields:

$b \cdot y = 1\ (mod\ a)$

i.e. $y$ from the algorithm is the inverse of $b$ in $\mathbb{Z}_a$. Yay!

## The two algorithms, in Perl

The two algorithms above, written in Perl:

In case GitLab does not work, you can find also a local version.

As promised, the basic algorithm is 5 lines long, including the input parameters unwrapping line and the closing brace - otherwise it would be 2 😁.

The extended algorithm is more or less a direct implementation of what explained in the Wikipedia page, apart from taking the stylistic choice to change some variables’ names. Translating the code to lesser programming languages should be easy 😁.

Both implementations are iterative, which helps keeping the number of function calls to a minimum and avoid bloating perl’s stack.

These functions are part of cglib-perl, my library of copy-and-paste functions for CodinGame.

## So long!

Nothing much to wrap up this time, only a suggestion: if you don’t know Julia Evans and her programming zines you SHOULD definitely check them out because they are amazing!