EBNF grammar for Rhombus

[asumbek]

Is there an EBNF grammar for Rhombus? I perhaps need it to integrate syntax highlighting to editors that uses tree-sitter.

[spdegabrielle]

I don't know but the latest release of Magic Racket now includes specialised highlighting for Rhombus files.

0.7.0 (2025-05-17) added

• Rhombus syntax highlighting and comment toggling (@TimWhiting)
• Rhombus line length (@jryans)

https://github.com/Eugleo/magic-racket

[usaoc]

We had some discussion on Discord, and the short answer is that Tree-sitter integration isn’t as useful for Rhombus (and Racket alike). The reason is that Rhombus grammar is fully extensible by macros, so its grammar cannot be specified without looking for all imports and local macro bindings. The seemingly very fundamental core forms, like fun and match, are all implemented as normal macros. Some degree of Tree-sitter integration can be built for shrubbery notation alone, but we still need DrRacket-style integration (that is, semantic highlighting) for a better experience. Racket mode for Emacs is an example of that.

[Lazerbeak12345]

Lots of treesitter languages have this problem. (Vim script and markdown being just two of them)

The solution is that embedded grammars get detected, and have their own TS, and whatever TS absolutely can't do is done with LSP.

[Bogdanp]

So I suggest combining a grammar for shrubberies with a list of known keywords and known functions.
...
The list of known keywords and known functions can be generated automatically.

This is what racket-hash-lang-mode for Emacs does (#666).

[asumbek]

I have this in my mind for a while, where the parser is made to be less strict. It will parses it as a macro rule if no other rules are matched, it'll still parses it as Shrubbery notation, but the user can inject a grammar on the node and take control of what to do with it.

[asumbek]

For example, the fun syntax. It's considered as a macro in the Shrubbery notation grammar, and then it uses queries to inject Rhombus grammar into that macro node. It partially solves with problem of parsing user-defined macros.

[benknoble]

Lots of treesitter languages have this problem. (Vim script and markdown being just two of them)

The solution is that embedded grammars get detected, and have their own TS, and whatever TS absolutely can't do is done with LSP.

@Lazerbeak12345 really? Vimscript not having macros (of the kind being discussed here) makes me wonder: what about parsing it for Treesitter is so hard?

The grammar is almost certainly not context-free, so it's probably not easy to write a declarative TS spec: is that the issue you're talking about?