tower-lsp VS rust-analyzer

I am using nom / nom_locate to build the parser side because I've done a handful of other projects with it, and I plan to use tower-lsp to hook up the language server side.

I realize this might not be for everyone, but I'm writing it in Rust using Lib-ruby-parser and tower-lsp: two existing libraries that handle a bunch of the heavy lifting for me. I'm more productive in Rust than with Ruby at this point, despite doing Ruby full time for 15 years, plus I really really don't want to have to deal with a slow LSP--that was the whome impetus for this project. I started in the spring, made a bunch of headway, then backtracked to redo the internals to make it easier to handle monkeypatching, overriding/redefining of methods, etc. across your project.

https://github.com/ebkalderon/tower-lsp is a generalized LSP implementation in a lower-level language (Rust) so you may get a better idea by reading through that repo. It seems that the server opens a TCP socket that the client later connects to, but I'm not really sure.

A LSP is just an api implemented to Microsoft's LSP spec. You can implement that API however you wish but something like tower-lsp can do a lot of the heavy lifting for you.

Mine all use [tower-lsp](https://github.com/ebkalderon/tower-lsp/) for the LSP protocol stuff, and then either [Tree-sitter](https://github.com/tree-sitter/tree-sitter) or [Nom](https://github.com/Geal/nom). If I do another I'll probably try [Chumsky](https://github.com/zesterer/chumsky) which combines some of the advantages of both.

I've contributed to rust-analyzer and nushell and had a great experience in both! Tons of open issues with a huge range of difficulties, and the maintainers are really helpful in providing hints to get started.

For something more concrete you can try and ask around on their zulip or browse their issues.

More info here: https://rust-analyzer.github.io/ and here: https://rust-analyzer.github.io/manual.html#installation

You can submit feature requests in the repository: https://github.com/rust-lang/rust-analyzer/

sh git clone https://github.com/rust-lang/rust-analyzer.git && cd rust-analyzer cargo xtask install --server

Thanks for the ideas! Separating out presentation crate into it's own crate sounds like a great idea. Re "temporarily stopping rust-analyzer", I came across this trick[0] to make RA use a different build path (trading off disk-space for speed, but I mean, disk space is cheap).

Re: cranelift, I gotta check that out! Thanks for the reminder!

[0] https://github.com/rust-lang/rust-analyzer/issues/6007#issue...

I also filled a new rust-analyzer issue for the cases that remain broken here.

Arc (or Rc if single threaded) sounds like a good fit since you mention immutability and no parent-pointers. If the lack of parent pointers or contextual information gets frustrating, I just want to mention the "Red/Green trees" approach used by Roslyn and rust-analyzer (link) where the "green tree" stores the data using basic Arcs and child pointers only, but later on you can create a "red tree" rooted at any green node which provides parent pointers as you navigate through the tree.

Neovim has a LSP-Client built in, for Vim there are several plugins. For Rust, the LSP-Server is Rust Analyzer.

The vscode status bar button that shows rust-analyzer can be clicked now and it stops the server. I also found this commit that adds a cancel button to the notifications.