Julia + Jupyter notebooks in VSCodium?

Hi. I’m not much for telemetry and such, so I’m trying to use VSCodium to work with Jupyter Notebooks running the Julia kernel. The VSCodium I’m using is the current up to date 1.59.0 AppImage download for Linux x64

VSCodium has the built-in .ipynb editor which opens the notebook just fine, but it doesn’t offer anything when I try to select a kernel:

Any suggestions?
Also when I manually download the jupyter notebook vsx file from microsoft and install it from vsx VSCodium says it can’t because it’s not compatible with 1.59.0

If I install the Python extension and then try to open the ipynb with that it just spins and never opens

Did you enable the julia.notebookController setting?

1 Like

I didn’t. How is that done?

Just like any other setting. Hit Ctrl-, to open the settings and paste that id into the search bar.

awesome, found it, clicked it, quit, restarted vscode… now it shows julia kernel in the selection menu! am fooling with it as we speak but it seems to be working!

What documentation or other resource was I supposed to have read to know about that setting? And why is it not enabled by default? Too new/experimental?

Yes. I wouldn’t recommend the current integration for serious usage :wink:

Yeah, but jupyter lab itself literally took me multiple Days to install successfully on several computers using Conda, and starts up and runs slow as a dog… whereas download VSCodium, click the right thing, stuff runs in seconds in VSCodium, so in some ways Jupyterlab is kinda a joke (on my machines). I do think it’s something specific to some aspect of my setup, but right now VSCodium is dramatically more usable.

(from the time I run “jupyter lab” in my bash shell, to the time I have a working window in Firefox is 55 seconds? And then it’s like another minute to just Pkg.activate(".") and using several things)

Compare to 18 seconds to start VSCodium and open the same notebook… and 86 seconds to execute that first using cell… Definitely more usable.

The reason we are hiding this behind a flag at the moment is that on save we are not properly saving kernel information, and we consider that serious enough that we don’t want to make this available out of the box. We are waiting for an upstream API to fix that, as soon as that is sorted out, we’ll remove the flag and just make the kernels available by default.

Great to hear that this all works on VSCodium!

1 Like