/tech/ - Technology

>I'll show you my monad if you show me yours
I'm more of a schemer, but I want to learn Haskell. Is "Learn you a Haskell" a good resource? Also, the parallelism seems wonky. It looks straightforward for doing SPMD with Control.Parallel.Strategies, but what about asynchronous IO?

>>24451

My standing rec is for Effective Haskell, by Rebecca Skinner, just as you might recommend Practical Common Lisp.

Parallelism is typically handled via spark-based parallelism (unevaluated lazy thunks are pre-evaluated out of order as per predicted need).

Concurrency usually involves the Async library these days, and some TVars (coordinating mutable variables) thrown in.

>>24454
Thanks, I will check it out and come back here with my progress.

>>24451
Honestly LYAH is not bad but it's very limited. I learned the basics on my own and then the book was kinda moot. It's a fun language for sure, but I'm a Scala evangelist now.

I tried it a bit during uni on the side, I gave up after typeclasses, but I still use Haskell notation to reason about type or processes sometimes. Very interesting language, but I have no idea what to do with it.

>>24466
It’s an algorithms prototyping language limited by performance, otherwise it’d be dominant (cuz 2x Rust isn’t good enough).

Web apps are reasonably mature, and being able to functionally manipulate DOM is quite nice. There’s also adequate GUI apps in Monomer, and fast games terminals up right now.

You don’t need to go past a single monad transformer layer for most uses, and effect systems have replaced deep stacked monad trans.

We also finally have GHCJS and WASM backends, although those are new and not really mature.

We need money to jazz up GHC, unfortunately.

Haskell is bretty cool, despite all the memeing it gets I think monads for structuring programs are an excellent innovation and seem like a great example of the "stop and think about how to really do this correctly" approach winning out versus the "just git 'er dun" style.

Typeclasses are nice in the small (Eq, Ord, etc) but there is an annoying tendency of Haskellers to abuse them to write ostensibly modular code where something like OCaml's modules would be more appropriate. The same Haskellers also want to pack more and more fancy types into Haskell even though fundamentally Haskell is not dependently typed and never will be. Why not leave dependent types to actual dependently-typed languages?

The community also cannot seem to decide if they want industrial buy-in or if they want to stay a "weird" language