Ability to specify configuration file as command-line argument for unified-language-server

[hjpotter92]

Copying over issue from the project: unifiedjs/unified-language-server#8 (comment)

Subject of the feature

Better documentation around handling of configuration file, or specifying config file as an argument when starting server

Problem

Currently, the docs do not make it clear where and how the configuration file is supposed to be sent to server.

[ChristianMurphy]

https://github.com/unifiedjs/unified-language-server/blob/main/CONFIGURATION.md#for-neovim

Documents where to add the configuration file.
settings.json in the home folder

[ChristianMurphy]

The default settings, and custom settings are merged here: https://github.com/unifiedjs/unified-language-server/blob/ca717b1322a9ff09cbea3442505929a7ae00327c/src/server.js#L153-L157

How would this translate when I'm creating a new client for emacs?

However you want it to, the server just needs a configuration with a settings object, how a client chooses to allow the settings to be set/changed is flexible.

[hjpotter92]

However you want it to, the server just needs a configuration with a settings object, how a client chooses to allow the settings to be set/changed is flexible.

This brings me back to my original question. emacs-lsp uses the stdio pipe processing for providing the LSP support. The command used to start the server is unified-language-server --parser=remark-parse --stdio.

There is no 'default' settings file name, no argument to specify one, and just a rough declaration of a json.

I come from python's environment, where for nearly each server implementation, the config is read from pyproject.toml or pyrightconfig.json etc. (I think something similar happens with ts-ls server etc.)

[ChristianMurphy]

the config is read from

are you asking for what the community convention is?

remark uses .remarkrc (with optional js or json) extensions or a remark filed in package.json, in the project root.
rehype uses .rehyperc (with optional js or json) extensions or a remark filed in package.json, in the project root.

@aecepoglu may be able to comment more on the mix of both config described in the example vim file.

[aecepoglu]

I tried to have a look at emacs-lsp's settings page but I kept getting a 404, so I cannot comment too confidently.
But yes, we rely on the client passing configuration data as per Language Server Protocol specification https://microsoft.github.io/language-server-protocol/specifications/specification-current/#workspace_didChangeConfiguration

Within the data it receives, the server expects to see the shape {settings: {unified-language-server: {...}}, ...}

As for your question: "how can I do this for emacs" I had a look a random item (lsp-rust) from this list https://emacs-lsp.github.io/lsp-mode/page/languages/ . I don't know how emacs works but it seems for each language-server it also requires a middleware like so https://emacs-lsp.github.io/lsp-mode/page/lsp-rust/

the language-server still expects configuration the same way we do: https://github.com/rust-lang/rls
lsp-rust reads config from wherever and passes it to rls as the LSP requires.

This sounds like it requires way more effort than language-server integrations are for other text-editors / IDEs but I suppose that's what you'll have to do.

[aecepoglu]

And I also had a look at python because you mentioned it.

https://github.com/palantir/python-language-server#configuration

It mentions

Overall configuration is computed first from user configuration (in home directory), overridden by configuration passed in by the language client, and then overriden by configuration discovered in the workspace.

For us it is simply

... configuration passed in by the language client ...

[hjpotter92]

Thanks you @aecepoglu and @ChristianMurphy

With your help, I've added a rudimentry client to emacs as well. emacs-lsp/lsp-mode#2800