LSP server for editor integration

[Nadrieril]

For best editor integration, it'd be awesome to have a LSP (Language Server Protocol) implementation for neuron. It could potentially provide autocompletion of links, automatically showing the link title next to a link, make links clickable, show backlinks, etc., and all this in an editor-agnostic way. That would greatly improve the experience of editing zettels with neuron.

[srid]

In theory one could also do things like new note creation (eg: felko/neuron-mode#28) -- so all editors can just delegate to neuron in that aspect -- however I'm not sure if that's in the scope of LSP.

[sureyeaah]

I think we can have some kind of a hack here. Go-to-definition provides the mechanism for going to a different file and we could provide "note creation" as a codeaction.

[srid]

Some ideas: https://marketplace.visualstudio.com/items?itemName=sheilacat.zknotes

[NullSense]

Renaming of links could also be a functionality for an LSP. Go-to-definition as @sureyeaah mentioned of course. Now sure if views like backlinking would be implemented an an LSP level.

[Nadrieril]

Other suggestion: easy creation of a new note using current selection as the title

[srid]

Another: "refactor" by renaming a tag across the zettelkasten.

cf. felko/neuron-mode@beeb47f

[aca]

@srid Are you working on this? or anyone?
I'm interested.

[srid]

@aca I'm not working on this. But @trevortknguyen was looking into this. Trevor, are you still interested? If so perhaps you two can collaborate?

@aca You might want to join Matrix, the project channel.

[aca]

There will be an "official" neuron lsp implentation in haskell.
Until then, you might want to try https://github.com/aca/neuron-language-server.

[sureyeaah]

@aca I can help with the "official" neuron lsp!

[aca]

@sureyeaah. You should join matrix and discuss about official neuron lsp.
For now, I will just play,experiment with the go lsp..

[trevortknguyen]

I do not think I will be able to work on the "official" neuron lsp.

[hoschi]

the call/caller graph features of LSP can also be utilized for viewing/navigating the z-index.

[srid]

An extension that displays neuron HTML preview to side would be pretty cool: https://twitter.com/owickstrom/status/1301241287727812610

[srid]

Reference implementation in Haskell: https://github.com/dhall-lang/dhall-haskell/tree/master/dhall-lsp-server

[srid]

If anybody feels like taking a stab at this, I added sample LSP server in Haskell here: #539

This provides a neuron lsp command that doesn't do much really, but you can use it with coc.vim in neovim using the below config. There is a hover handler which simply prints the notebook path.

$ cat ~/.config/nvim/coc-settings.json 
{
  "languageserver": {
    "neuron": {
      "command": "neuron",
      "args": ["lsp"],
      "filetypes": ["markdown"]
    }
  }
}

[srid]

One tiny benefit of using LSP would be that renames on large notebooks would be faster (O(1)). cf. svsool/memo#286 (comment)

[srid]

Inspiration: https://github.com/artempyanykh/zeta-note#screenshots

[samwalls]

Is there any other way to use neuron's searching facilities within vscode, currently? Ideally I'd like some quick command, mapped to a keyboard shortcut, where I can search zettel by their title, tag, or contents etc..

Regardless I might like to take a look at this, as I want to get familiar with serving LSP, writing utilities for zettel, and making "production-level" haskell. What's still left to do here? @srid

[srid]

You can try task prompts and then have vscode run the neuron command with prompted values.

For LSP, there is stub code here: https://github.com/srid/neuron/blob/master/src/Neuron/LSP.hs

I'd say the next steps are:

• Configure VSCode (this might require writing an extension in TypeScript?) to use neuron lsp (a command, hidden from --help) as the LSP server. In NeoVim, this is straightforward to do.
• Confirm that the stub LSP.hs hover message works. This ensures that neuron lsp is hooked up to the editor successfully.
• Open neuron source in VSCode and hack away the rest of the LSP features
  • Neuron's LSP.hs uses the same library as haskell-language-server (which provides IDE support for Haskell projects); so you can check the latter's source code for reference. The library's README also has some information: https://github.com/haskell/lsp/blob/master/README.md

As for "writing utilities for zettel" - I imagine that a lot of that can be done by neuron query in the command line, but if you can describe the specific workflows, I can give more info.

For contributing to production Haskell projects more generally, there are some good first issues here in neuron's issue tracker.

Anyway, happy to help with any further specific questions. We also have a Matrix chat room here.

[mickael-menu]

I've integrated an LSP server in zk, which is compatible with neuron. Just run zk init in your neuron notebook, then follow the LSP instructions for your editor.

It supports neuron's tags and wiki-links.

(Kudos to @aca, your neuron LSP Go server got me started)

• Configure VSCode (this might require writing an extension in TypeScript?) to use neuron lsp

That's right, you will need to create a VS Code extension for your LSP server. I've created one for zk that is pretty minimal, which should be easy to modify to start neuron lsp.

One important bit is to configure the activationEvents. I chose to start my LSP client when a .zk folder is detected in the folder, you can probably do the same with neuron.dhall: https://github.com/mickael-menu/zk-vscode/blob/701cfd07db3529b039392067385fa81df594e74d/package.json#L14-L16