RBS, Ruby’s new type signature language

At some point it makes more sense to switch to Crystal.

Crystal is amazing (we sponsor it at Fly.io), but the sheer amount of stuff that's already built for Ruby makes it hard to switch to another language.

Node has had 10 years and its still not there.

Crystal is a non-starter for me since the entire appeal of Ruby is the runtime metaprogramming.

You don't want to switch to a half baked language for anything serious.

Crystal has a nice AST based Macro system. IMHO it makes up for it. The problem with Crystal is no direct support for Windows OS compilation which is a non starter for many developers.

> still not there

And never will be.

Crystal is pretty unstable at the moment (in that it changes often). I have a very small project in crystal (discord bot with only a couple of commands that CRUDs a database). Every time I deploy to heroku, if heroku updated crystal, the bot breaks. I have to spend an hour determining what broke. The language changes so often that either I update with heroku or I can't use any newer libraries for development.

And then you get stuck in very opaque error loops with the typing where it's expecting eg. a Number. no, not that number, a different type of number. no, not that type either. no, not that type. Most the code i've written has been typecasting.

Crystal does have some significant benefits over Ruby (which is why i'm using it - better memory use, better for scaling for my purpose) but I just spent time rewriting the non-user facing, non-scaling part of the bot in Ruby so I could actually get stuff done instead of fighting the language.

Of course a lot of this might be my inexperience with Crystal, but as a Ruby dev for 13 years, it's not as easy as just switching over from Ruby to Crystal. I've had this bot running 4 years and it hasn't got any easier for me.

> Every time I deploy to heroku, if heroku updated crystal, the bot breaks

Have you tried adding a `.crystal-version` file as described in the buildpack's README? https://github.com/crystal-lang/heroku-buildpack-crystal#cry...

> Crystal is pretty unstable at the moment (in that it changes often)

This had never been my experience, I have a server written fully in crystal running in production serving millions upon millions of requests on heroku and crystal doesn't break a sweat.

Quite happy with it so far.

Not sure what you're saying. What are places where Node isn't caught up to Ruby and can never catch up to Ruby?

Ruby on Rails seems like a kneejerk response, but then again it doesn't exist because nobody really wants it, not for technical reasons.

For example, Python has all the ML/math stuff. Nothing comes to mind for Ruby.

How is Crystal a "half baked language"?

I think the parent comment was referring to the overall market position of RoR, and I'm saying that Node isn't likely to achieve an equivalently dominant share going forward, due at least in part to the wide variety of options available.

You wouldn't know what edge cases it might have unless the language becomes a "major language" to have more eyes.

I don't think average programmers even know about it.

The current state of Crystal is basically a language for PoC during the week-end, it's far from being ready.

I could go on about what's wrong with the language but:

It's not stable, API change all the time, breaking change all the time, cryptic errors, lot of missing basic features, IDE integration etc ...

I touched on performance later on - and it's the primary reason I use crystal. I mean more than the language's API changes often.

Yeah, the issue is more that I can either keep Crystal updated and use all libraries, or keep it frozen and be restricted in what I use. It's not mature enough yet to be able to lock it down and have more or less Crystal's entire shard collection available to me, like I can do in Ruby. Reminds me more of Ruby a decade ago with 1.8.7 and 1.9.2 and 2.0 (which is fine for it to be at, esp for an evolving language, but it's not "there yet" to be a replacement for Ruby)

One incident that stands out is that certain Postgres support was only available in the latest version of a shard, which required the latest version of Crystal, which wasn't compatible with 3 of the other shards I was using.

No one really wants it other than the guys who use Sails I guess...or is the name just a coincidence?

> The current state of Crystal is basically a language for PoC during the week-end, it's far from being ready.

That is changing, earlier this year, the Crystal team are working on getting the language stable for 1.0 [0]

Companies are already now using Crystal in production, most recently Nikola Motor Company [1].

Surely they wouldn't choose crystal if it was a 'half baked' language.

[0] https://crystal-lang.org/2020/03/03/towards-crystal-1.0.html

[1] https://manas.tech/blog/2020/02/11/nikola-motor-company/

stability of the language is their next goal.

sure, and I hope it gets there soon, but it's not there yet, which was the real point of my answer to "why not just use Crystal"

> it doesn't exist because nobody really wants it.

Clearly not true, as there's two separate fullstack frameworks being actively developed[0][1], not mentioning the ones in the past (sails.js, meteor.js).

[0] https://github.com/blitz-js/blitz [1] https://github.com/redwoodjs/redwood

The Crystal macro system is pretty advanced though, I think it works very well to replace metaprogramming.

You will always find someone somewhere picking up a language / framework and thinking it's production ready, from my experience Crystal is not ready and in the example you gave there is absolutly no reason to no use Rust. They needed C binding, and the fact that they started using Crystal years ago when the state was even worse is very worrisom.

It's at version 0.3.5 currently and their latest blog post is about how their main goal is to reach 1.0 which will be stable.