Why I don’t like the lens library

As some of you know I’m the author of fclabels, that other lens library that isn’t lens. I recently released a new version and I’m very happy with the result. But I knew beforehand that usage would probably be extremely low. Mainly because of the popularity of the lens library.

And I’m fine with that.

I never built fclabels because I wanted people to use my software (maybe just a bit), but I wanted a nice solution for Haskell’s non-composable record labels. I think both fclabels and lens nicely solve this problem, both with slightly different technology and philosophy.

Van Laarhoven lenses are a very clever trick that gives rise to great extensibility, a great basis for a lens library. One of the reasons I actually considered giving up on fclabels in favor of lens. It will save me time as a maintainer and community consensus is mostly in favor of lens.

But after closer inspection I didn’t. Partially because I still think the design choices in fclabels add something to the ecosystem, but mostly because I think that lens is a vastly over-engineered, intimidating, complex beast. I tried really hard, but I can’t get myself to like the package.

Maybe an extreme example, but opening the package and just randomly clicking some modules could end up with things like this:

class (Choice p, Corepresentable p,
       Comonad (Corep p), Traversable (Corep p),
       Strong p, Representable p, Monad (Rep p),
       MonadFix (Rep p), Distributive (Rep p),
       ArrowLoop p, ArrowApply p, ArrowChoice p
       ) => Conjoined p where
  distrib :: Functor f => p a b -> p (f a) (f b)
  conjoined :: (p ~ (->) => q (a -> b) r) -> q (p a b) r -> q (p a b) r

I cannot be the only that thinks this might be just a little bit too much.

In my opinion lens doesn’t have enough focus, has an extremely intimidating API and exposes way to many utilities that are only useful in handful of use-cases. It wants to solve every problem of data access at once, at the cost of simplicity and elegance. Simply put, I think lens is very cool technology, everything in there has some nice theoretical underpinning, but from an end-user perspective it’s very badly designed.

I know that not everyone agrees with me, but I also know a lot of people do think the package is intimidating.

Haskell always attracted me because of small and simple libraries, that focus on solving small problems. Bigger problems are mostly solved by composition of smaller solutions instead of adding more primitives. Van Laarhoven lenses allow for exactly that, lens as a library feels like the exact opposite.

If we all agree that lenses are an essential addition to the haskell ecosystem, can’t we work towards a small and simple solution? One that we can safely put in the platform and show to first time Haskellers?

No, I do not propose fclabels. Can’t we just get the essential core ideas out of lens and put that in a library? A small library, with a very carefully crafted interface? No corepresentable distributive indexed coreps, no category theory, not tons of ‘convenient’ operators. Just lenses. Maybe traversables and isomorphisms. Everything else can be built on top.