Free Monads

Similarity in free monads and tagless final

Free monads are, in some ways, very similar to the tagless final. In essence, they allow us to turn any functor (the part of a monad that has a map) into a monad, thereby gaining of and chain functionality without any additional work.

This is in itself already quite nifty, but free monads can also be used to build our application as an Abstract Syntax Tree (AST). First, create the functor, which consists of the commands that are important in our application (think of this as a DSL). Next, lift it so that it becomes a free monad. Now we can combine the commands we just created in our program. This will, yet again, look nicer in a language like PureScript or Haskell. But it will work just the same in TypeScript, and it might make our application code more readable and usable.

We’re not finished, though. Our code won’t actually do anything yet! For that, we need to again write an interpreter that will give meaning to each of our commands. This gives us an advantage when it comes to mocking. There are drawbacks as well, besides the added complexity. For one, the recursive nature of this solution means that we might trigger a stack overflow. Even if we avoid this by using trampolining, our solution won’t win any performance prizes. So, don’t use this solution unless you think your domain or application (and its complexity) warrants it.

Some applications are based on a previous version of fp-ts, but most changes required to make these examples work are simple enough. For example, instead of calling chain on our monad, we should now use a pipe to combine chains (as we’ve seen countless times in the previous pages), and Free is now located in the fp-ts-contrib library.