Emacs mode for Rhombus

[kiranandcode]

Not sure if there's already an effort in this direction, but I've been playing around with Rhombus recently, and have gotten tired of having to rely on Python's highlighting as a workaround for making rhombus code pretty.

I've written a basic Emacs mode for Rhombus (very incomplete, but good enough to get the job done for now): https://github.com/Gopiandcode/rhombus-mode

Here are some screenshots on my local machine:

[jackfirth]

You might be interested in some of the work on Racket Mode to make it derive more information from the #lang infrastructure. As I understand it, that would in theory make it work and highlight non-racket #langs correctly.

[samth]

I've been writing more rhombus recently, and I definitely want racket-hash-lang-mode. So I'm happy to dogfood it.

[greghendershott]

Great! I just need to figure out a practical concurrency strategy for me to juggle working on both this and pdb.

Until you said this: I assumed I'd wrap up pdb fully(ish) before really looking at racket-hash-lang-mode again.

Instead: I guess I could get pdb to some clean(ish) pause point soon, and while waiting to sync up with Robby, reload my brain with racket-hash-lang-mode?

[rfindler]

One option (dare I suggest...) would be to use racket-hash-lang-mode even for #lang racket and then you'd be in the happy dogfooding case again ... :)

[greghendershott]

Well.

IME eating my own dog food works well when, sure I need to detour from my intended task to fix a bug in my tool, but at least I'm still eating from the same bowl (same git branch). Shorter, simpler detour. Juggling even just two branches (say main and pdb, or main and hash-lang) can entail possibly needing to restart Emacs, and using possibly buggy variants of Racket Mode to edit Racket Mode's own Racket source code. And periodically syncing the branches. At three branches... it can start to feel like stepping in dog $NOUN I forgot to pick up earlier *cough*.

Also: So far, I just don't like the hash-lang approach as much for editing s-expression langs in Emacs. The syntax highlighting is more vanilla, indentation isn't as customizable, other Emacs modes like paredit know nothing about hash-langs, and you need the whole back end running just to edit a plain text file. It's nice that various lexer edge cases are handled correctly compared to the "pile o' regexps etc."... but the overall net benefit for a sexp lang? idk yet.

Of course, some of those deficiencies could be improved through more dog-food detours, and someday probably will, but I do have some other work I'd like to get one, too. 😄

So TL;DR if someone like Sam is willing/able to alpha test, especially on something like #lang rhombus, I think that's the ROI sweet spot in the near future.

To facilitate that: Maybe I should slap some "EXPERIMENTAL!" caveats in the docs and just go ahead and merge the hash-lang branch to the main branch. If people don't enable racket-hash-lang-mode, it shouldn't break anything (in theory).

[greghendershott]

p.s. @samth Meanwhile I did go ahead and rebase hash-lang as well as push a few more commits. And somewhat reload my brain. So if you wanted to try that with #lang rhombus, I think I'd be ready to field questions or problem reports via Slack or the racket-mode repo issues.

[Lazerbeak12345]

While I think this is very cool, I think this shows a greater need for proper LSP in the racket ecosystem. With LSP you'd do this exactly once, then use it across all compatible IDEs, such as VScode, (n)vim (with appropriate plugin), or even emacs (with appropriate plugin).

Further, that "doing once" could be done in the repo where the language is defined - we already have some of this, but it's very DrRacket specific, and doesn't seem general enough to me.

Great work! Looks very readable.

[kiranandcode]

When I tried it I wasn't able to get syntax highlighting through lsp-mode, so in the end, I wrote my own syntax highlighting rules directly in Emacs --- the repository includes a lsp-rhombus file to allow running lsp-mode on rhombus buffers, because lsp-mode does show warnings etc. correctly.

[Lazerbeak12345]

LSP

https://github.com/jeapostrophe/racket-langserver wraps functionality provided by things like drracket/check-syntax (the library, as opposed to Dr Racket the app) and syntax-color.

It looks like rhombus-mode uses this via Emacs lsp-mode.

sorry, been busy.

I actually have been using that LSP for a very long time (years? has time gone by that fast?). My issue with it is it doesn't work on headless servers - while everything else in my setup does (neovim with lspconfig). Cario is still needed at runtime - even though nothing is shown on screen. (jeapostrophe/racket-langserver#45)

the racket server relies (heavily) on the editor class text% provided by racket/gui/base, and even just requiring this package causes it to try and initialize a display, even if one won't actually be used. I'm doubtful that there's anything we can do here without basically re-implementing all of the navigation and paragraph features from text%.

This is why I think the problem lies more that racket needs first party support for LSP.

More here: racket/drracket#512 (comment)

[samth]

I don't think there are any fundamental issues here; racket-mode is built on the drracket infrastructure as well but doesn't need a GUI. I believe the Racket LSP was written that way because it was the easiest approach and it hasn't been changed.

[greghendershott]

Partly to support expeditor (an expression editor for the REPL), Matthew added syntax-color/color-textoid which defines a color-textoid<%> interface with a subset of the methods from text% that are needed for these purposes -- and which can be implemented without racket/gui.

That defines an interface. Each tool -- like expeditor, racket-lang-server, Racket Mode back end -- needs to supply a suitable implementation.

So I think that is probably the direction racket-lang-server would need to take. (Meanwhile an option to use xvfb-run would help; maybe that already exists.)

[rfindler]

(I'd replied by email with a much less helpful message than @greghendershott 's message above; it didn't actually go into the right place so I just deleted it now. See Greg's!)

[greghendershott]

Oh, sorry; I overlooked that. I only glanced in lsp-rhombus.el. And even there, if I'd looked more carefully, I would have noticed mentions of both https://github.com/jeapostrophe/racket-langserver and https://github.com/theia-ide/racket-language-server.

Glancing at the latter, the good news is it provides a custom (not standard LSP) notification for fancier coloring from check-syntax information, as well as a custom method for indent. The bad news is the project is in Archived status.

It sounds like you wanted to use both LSP servers, if possible, but that didn't work out?

I guess ideally the extra things from the archived LSP server project could be added to active one, unless there are practical or licensing obstacles.

[Disclaimer: I don't know LSP in any detail, or its use in Emacs (Racket Mode predates LSP).]

[kiranandcode]

@greghendershott , ah, no, lsp-rhombus.el is just lsp-racket.el from lsp-mode, with the substitution 's/racket/rhombus/' (see https://github.com/emacs-lsp/lsp-mode/blob/master/clients/lsp-racket.el)

I included both because lsp-racket.el includes both, and I didn't know which one I needed.

(Looking at my Emacs, I guess I was using racket-langserver (the non-archived one))