What monads are

Fri 06 May 2016

It’s almost like a rite of passage for a developer to write a monad tutorial.

Well, this isn’t a monad tutorial. This is simply an effort to put an end to some of the nonsense that leads to monad tutorials being written in the first place and all the magical woo surrounding what is actually a very mundane topic every developer in the world understands, but simply doesn’t realise they do.

And let’s not use silly metaphors for any of this. Analogies, yes, but no metaphors.

Say I have the following expression:

1 + 2

Because the type of the values on each side of the + operators is an integer, we know that the + refers to integer addition, and thus the answer will be ‘3’.

Now, let’s say that our language allows operators to be overloaded, and one of the meanings of + is concatenation. Thus in the following expression, + means list concatenation:

[1, 2, 3] + [4, 5, 6]

With me so far? Now’s the bit where I blow your minds. Look at this:

;

That’s a semicolon, which in many languages is used as a statement separator. If you tilt your head a bit, you’ll see that it can also be thought of as an operator, thus in the following:

3 + 4; 5 + 6

The semicolon is acting as an operator that joins two expressions together and evaluates them in a particular order. It’s really no different from + in that regard.

So, what determines the behaviour of ; when though of as an operator? Nothing really, except that the languages we use assume that it behaves a particular way. However, if we can vary the behaviour of + depending on the type of the expressions on either side, why can’t we do the same with ;? This is what a monad does: it’s a type that lets you determine the behaviour of the semicolons—or your language’s equivalent—as you need to.

If you’ve got this far and understand everything, then you understand what’s most essential about monads. Congratulations!

Now, stop reading and writing monad tutorials.

There’s a bit more to it, and I may cover those details in a future post, but suffice it to say that if you’re looking for the closest useful equivalent of the semicolon as an operator in a monad type/typeclass, you’re looking for what’s referred to as bind, or >>=. The related join or >> operator/method is a more direct equivalent, but join can be implemented in terms of bind: join is just bind, but you’re throwing away the result of the expression.

Category: Coding