workspaces mixed when using in jupyter notebooks

[TTTPOB]

First I thought this may be a jupyter-lsp bug but according to krassowski/jupyterlab-lsp#642, this may be a bug in R languageserver
I installed #452 version

I don't know what information should I provide.

[randy3k]

We are right, we do not support having different workspaces at all. I believe that most clients spawn a new R process for different workspace (for a definition of workspace suitable for that editor, says Windows for VSCode). But it seems that Jupyter-lsp is reusing the same process R process for different notebooks. I am not sure if we want to fix it or how we should address it.

[renkun-ken]

This is exactly what REditorSupport/vscode-r-lsp#45 is trying to solve using multi-server approach.

[TTTPOB]

This is exactly what REditorSupport/vscode-r-lsp#45 is trying to solve using multi-server approach.

We are right, we do not support having different workspaces at all. I believe that most clients spawn a new R process for different workspace (for a definition of workspace suitable for that editor, says Windows for VSCode). But it seems that Jupyter-lsp is reusing the same process R process for different notebooks. I am not sure if we want to fix it or how we should address it.

hmmm, I think jupyter is runnig multiple R but only one language server is launched by jupyter-lsp, so is this expected behavior in current version?

[randy3k]

FYI, Sublime Text LSP also takes a similar one-server-per-workspacefolder approach.

While it would be lovely to have support of workspaces, but it seems that it only benefits a few users. Also, without supports from the clients and editors, it is difficult to test things.

[randy3k]

hmmm, I think jupyter is runnig multiple R but only one language server is launched by jupyter-lsp, so is this expected behavior in current version?

Yes, the main process was launched by jupyter-lsp and some of them are helper R sessions for linting etc (or R sessions which are irrelevant to us at all, eg iRkernel).

[TTTPOB]

thank you for your reply, I am happy with it then :-)

[randy3k]

Ai, a quick check with the codebase of jupyterlab-lsp, they also don't support multiple workspace folders. So enabling multiple workspace folders is not a solution for jupyterlab.

We observe such an issue simply because each notebook is treated as an open R file. And of course, in other editors, we would like to have completions from other R files while editing a particular file.