Monads are too powerful: The expressiveness spectrum

Summary

Okay, so you and I both know monads are great, they allow us to sequence effects in a structured way and are in many ways a super-power in the functional-programming toolkit. It's likely none of us would have even heard of Haskell without them. It's my opinion, though, that monads are actually too powerful for their own good. Or to be more clear, monads are more expressive than they need to be, and that we're paying hidden costs to gain expressive power that we rarely, if ever, actually use. In this post we'll take a look at how different approaches to effects lie on the spectrum between expressiveness and strong static analysis, and how, just like Dynamic vs Statically typed programming languages, there's a benefit to limiting the number of programs you can write by adding more structure and constraints to your effects system. The Status Quo A defining feature of the Monadic interface is that it allows the dynamic selection of effects based on the results of previous effects. This is a huge boon, and is what allowed the construction of real programs in Haskell without compromising on its goals of purity and laziness. This ability is what allows us to express normal programming workflows like fetching input from a user before deciding which command to run next, or fetching IDs from the database and then resolving those IDs with subsequent database calls. This form of choice is necessary for writing most moderately complex programs. Alas, as it turns out, this expressiveness isn't free! It exists on a spectrum. As anyone who's maintained any relatively complex JavaScript or Python codebase can tell you, the ability to do anything at any time comes at a cost of readability, perhaps more relevant to the current discussion, at the cost of static analysis. Allow me to present, in all its glory, the Expressiveness Spectrum: Strong Static Analysis <+---------+---------+> Embarrassingly Expressive Code As you can clearly see, as you gain more expressive power you begin to l...