lsp-docker VS robe

I believe this has cost me a ton of time but I did learn a lot along the way. lsp-docker highest level function is lsp-docker-start. It calls lsp-workspace-root which calls buffer-file-name which, if starting from a dired buffer, returns nil so lsp-workspace-root returns nil and the LSP server is not started.

I'm trying to figure that out right now. There is lsp-docker which is designed (I'm still half guessing) to have a separate container hold the language server. What I want is a pre-existing container to hold the server. There is partial code in that repo to do that but it appears not fully baked.

If you're on WSL, you may want to take a look at something like lsp-docker which bypass all quirk of Tramp and works directly with your files in WSL. I've never been able to make Tramp work reliably (too much hanging with or without lsp-mode) from Windows -> WSL so bypassing it provides a much better experience.

This is a reasonable point, and something that is being worked on. I definitely think Docker can provide LSP servers as necessary, we're just not quite there. This is something that I'm interested in working on, though I'm not the only one: https://github.com/emacs-lsp/lsp-docker

Yeah, I'm a NixOS user, and nix helps a lot with managing LSP servers on my system at the point that this is not a issue for me, but I agree a solution runing then in a docker or something would fit nice as well, we already have this conecpt: https://github.com/emacs-lsp/lsp-docker/

[0] https://github.com/emacs-lsp/lsp-docker#docker-over-tramp-tbd

I also did this for over 20 years (well, with screen(1) back in the day first) and I managed a lot of scripts for rsync(1)ing configs and source code. Emacs was designed to run as a GUI and makes heavy use of control sequences that affect terminal emulators (such as C-c, C-s, C-q, C-h, C-z, etc.); I spent considerable time forcing Emacs to work nicely in my terminal under tmux (where I've always used M-t as my prefix key, but that's a story for another time!), and I had a pretty nice terminal workflow that way, but it was laughably and ironically a PITA to get Control-H and backspace working in a way that was satisfiable to me in all cases of all layers (terminal emulator -> tmux -> zsh -> emacs). I eventually gave up on that for reasons related to an insane thing I discovered and make heavy use of, called Org Mode, which I primarily use to organize and maintain and evaluate code blocks of scripting languages with a REPL integration (the whole reason I switched to Emacs one day in a fury of ragequitting Vim was to try Robe Mode). So I now run Emacs in the GUI as demigods intended, but this will likely not be an option for you given your SSH workflow.

This means I can be editing a shell script and easily inject arbitrary regions into a shell buffer for immediate testing (point never leaves the window where I am editing, and I can view the shell output in an adjacent window). This is similar to what Robe does with Pry within an inferior Ruby process using comint.

I work on a large rails project with Emacs and I had a lot of trouble with Eglot and LSP mode with Solargraph, and decided to try Robe and it has been working well for me the past few days