Hacker News Reader

Comment by open-source-ux | Hacker News Reader

Crystal is a lovely language. The language community is small though, and learning resources are also small. The story for Crystal is the same story for many programming languages: grow the interest of users, and find generous funding.

My impression is that not many Ruby programmers have switched to Crystal. The slow(ish) Crystal compilation will not please Ruby users. And Crystal for Windows is still in beta.

Despite these factors, Crystal is a pleasure to use - fast, readable, and a well-featured standard libary.

View on HN

> The slow(ish) Crystal compilation will not please Ruby users

I don't know what's the general opinion, but distribution for dynamic languages is a nightmare (and I'm including intra-machine, ie. switching between environments).

Personally, I'd be thrilled to trade off speed of compilation for distribution simplicity!

> My impression is that not many Ruby programmers have switched to Crystal.

This depends on the context of moving - professional projects, or hobby programming. Every once in a while somebody pops up saying that they've switched for hobby projects (mostly, scripts).

I'd love to switch, also professionally (that is, for certain parts of my professional project), however, lack of (release quality) parallelism is a dealbraker for me. It's mostly a matter of long-term trust - I personally don't trust a programming language that in 2022 doesn't support parallelism. Ironically, Ruby now has it (even if in limited form).

Development of 3d party libraries are also in vicious circle (few libraries -> few devs -> few libraries). Lack of (release quality) AWS SDK, for starters, is a dealbreaker.

> Personally, I'd be thrilled to trade off speed of compilation for distribution simplicity!

Technologies like Docker have made distribution a lot easier. I haven't used tools like chruby or rbenv since 2014 when I started using Docker because Docker manages your app's runtime environment. You pull down your built image on your server and run it. There's no complication, sprawling of version managers or surprises.

Personally I think having a fast dev feedback loop is critically important. It's something you experience hundreds of times a day as you're developing your project.

With static typing and analysis in the editor there’s no reason to run the code for every second line you write.

I usually only need to run it a few times towards the end of writing a feature to see if the compiler or specs catches any errors.

If you're building web apps chances are you want to see your changes frequently since modifying a model, controller action or view may yield a different UI state.

I'm constantly editing code then directing my attention to my browser to see the results. This is independent of CSS or JS changes that might be picked up by asset watching tools like esbuild or tailwind's CLI tool.

Fair enough, I build json rest apis in Crystal, then someone else does the js/css front end.

> distribution for dynamic languages is a nightmare

Weird statement. Software distribution generally is just a terrible mess, regardless of what sort of language it's written in. The one exception is JS, which happens to be dynamic and also has what is probably the most reliable and ubiquitous platform available—so much so that its distribution story is often credited, rightly or wrongly, as the only reason anyone actually uses it.

There are very significant differences between distributing programs in a static vs. dynamic language.

Github's Hub tool is a good example.

It was originally written in Ruby. This requires that the user:

- may need to handle a certain environment in terms of Ruby interpreter (potentially install a new version)

- may need to handle a certain libraries group (gemsets in RVM)

- needs to install dependencies

- needs to handle switching to the appropriate Ruby interpreter version and libraries group

Especially point 3 is a pain.

Even just installing everything as system Ruby won't work, because often, different Ruby interpreter versions are required.

Once they moved to Golang:

- the distributor compiles for different platforms

- the user downloads the given version for their O/S

That's all.

Okay. Sounds like you're on the road to a convincing argument against Ruby. (I myself have avoided Ruby in the past for similar reasons, despite other things about it that are attractive.) And gesturing towards Golang is really just highlighting a conscious design decision by the Golang team—who didn't have to design it as a static language in order to achieve the same effect; they're orthogonal issues.

Where does the generalization to dynamic vs. static languages come from?

> Github's Hub tool is a good example.

I gave an example already (a counterexample, that is—JS). Was there something wrong with that?

Yes, there something wrong with it, because the distribution of the runtime is done via binaries i.e. browsers. Using Go or Crystal is mimicking that level of ease. Ruby, or JS (which is more of a mess than Ruby even though the runtime is already distributed) are still a pain to share. No credit goes to the JS distribution system(s) nor the form that the code itself is distributed in for the most difficult bit being done by browser vendors bundling the runtime.

This another weird comment—for different reasons compared to the original. There are several statements where the truth status of the claims are hard to ascertain, because the statements they're part of are incomprehensible. The claims in the statements that are comprehensible are untrue.

I wouldn't shrug my shoulders at

>> the distributor compiles for different platforms

Most people aren't doing that for you. There are lots of tools out there where you do need to still manually compile it yourself and make sure you have the appropriate dependencies which will still be painful. That being said, I agree that shipping a script can be a problem, but lots of people use Brew and that seems to work ok.

One minor note:

Crystal achieved basic Windows support as of 2021/11/18. https://github.com/crystal-lang/crystal/issues/5430

As of Crystal 1.5.0, Windows support is basically there. https://crystal-lang.org/2022/07/06/1.5.0-released.html

However, Crystal aims to support the lowest common denominator between POSIX and Windows, so not all of the Windows standard APIs are supported out-of-the-box. Other Windows specific libraries will need to be created specifically for Windows users/developers.

> The story for Crystal is the same story for many programming languages: grow the interest of users, and find generous funding.

it's the same thing for Elixir as well. they have successfully wooed a few Ruby programmers but not many have completely switched to using Elixir. Again some companies using Elixir are those that want to migrate away from Ruby.

it takes quite a while for a new hot language to become extremely popular and then go on to replace other popular mainstream languages.

I love Elixir, but it is an extremely niche language. It's great for network application, but just doesn't seem to be suitable for a daily driver language. Meanwhile crystal can fit all the bill: you can use it to write small scripts, you can write it for write long running service, you can write it to crunch data (the multi-threading feature of crystal is still an experiment though, but it works well for basic stuff)...

As I mainly use ruby for scripting, Crystal has been replacing ruby's role for me for 3+ years and I don't even find anything that really missing, except for maybe a decent ORM like Sequel.

To be fair, the slow compilation time is annoying even compared to other compiled languages, particularly Go and Nim, which are similarly "high level".

I also found that the Crystal LSP server was painfully slow, and there seems to be a lack of robust developer tooling in general. Has that improved in the last year or so?

Reply on news.ycombinator.com