Hacker News Reader

Comment by Taikonerd | Hacker News Reader

OCaml makes so much sense to me -- it's just a shame that the syntax has some weird decisions. For example, an array is:

    let arr = [| 1; 2; 3 |];;

instead of something like

    let arr = [1, 2, 3];

I wish ReasonML (https://reasonml.github.io/) would come back -- it's a new syntax for the same language, kind of the same relationship Elixir has to Erlang.

View on HN

Reason never went anywhere. And, like Ocaml, its array syntax has pipes in it (and like Ocaml, the pipeless version makes a list).

I think you’re doing Elixir a bit of a disservice by that comparison, though - Elixir is very much its own language on the Erlang platform, whereas Reason really is pretty much just an alternate syntax.

That said, one of the big problems with Ocaml is that the ecosystem is so fragmented; I don’t think Reason really adds much (at least if you’re not using with JSX) to make it worthwhile and I’m kind of mad that it exists.

> Elixir is very much its own language on the Erlang platform

Huh -- so, Elixir has some new semantics, rather than just being "Erlang with Ruby syntax?" I hadn't known that -- but admittedly I'm not an Erlang/Elixir dev.

The semantics are very close if not exactly identical. But it definitely feels and works as a separate language, mostly by having its own standard lib and toolchain.

It's not ruby syntax either. There are some very obvious inspirations, but it also explicitly rejects some standard ruby moves in favor of FP conventions. The whole "ruby for beam" angle was dropped years ago, and I'm not sure how much if any support that marketing tactic ever had from the official elixir devs.

reason isn't taking anything away from ocaml or fragmenting the ecosystem - just think of it as a ppx on steroids. at worst it needs its own source tooling to work with the different syntax.

This one makes sense to me, since lists are way more canonically used than arrays in functional languages. I rarely find myself constructing array literals -- I'm far more likely to simply convert from a list.

Though, I agree the semicolon delimiting is regrettable.

dune supports Reason syntax out of the box, you just have to change the file extension to .re.

https://dune.readthedocs.io/en/stable/reference/dune-project...

Yeah, Reason is supported in that sense. But when the last blog post on their site is from Aug. 2018 [0], that gives me an uneasy feeling.

[0] https://reasonml.github.io/blog/

Some implementations of Standard ML (SML/NJ and MLton being the two I know of) support an extension to the language for vector (immutable array) literals. That would look like:

#[1, 2, 3]

It's still not a feature you can expect all implementations to have, so OCaml is not in a bad position there I think. (Although I prefer SML most of the time.)

> It's still not a feature you can expect all implementations to have, so OCaml is not in a bad position there I think.

The number of SML implementations supporting this feature is greater than the number of OCaml implementations. ;)

ReScript eventually emerged out of the whole reason deal and is worth looking into. It targets JS runtimes so uses arrays instead of lists as the main sequence data structure. Uses the "normal" JS syntax for array, has no linked list literal though.

There's also F# if you're looking for an OCaml with better syntax.

F#, for better or for worse, is missing some of the things that make OCaml really interesting - the powerful module system and the structurally typed object system.

It does have its own neat features, though. Not to mention access to the .NET ecosystem, of course.

I agree that F# is worth checking out, but it actually has the same array syntax that the OP was complaining about.

ReasonML is never left, the syntax has been stable but the compilation to JS now happens through melange instead of bucklescript

I guess it’s due to tuples?

Reply on news.ycombinator.com