Hi all,
I have been using Julia for most of my course over the last 5 years. The students usually install VScode + Julia on their own machines and then the course mostly relies on notebooks. Habitually, things go smooth, despite the diversity of student computers (new-old machines, mac-linux-windows).
Today, we just had some new problems, I can not (right now) give enough detail for reproducing the error but maybe the symptoms could give an idea of what the causes could be?
Julia is installed with Juliaup on a mac. 1.10 set as default. VScodeserver picks it up correctly. A script works fine. The execution of notebook cells hang forever (although the kernel is selected).
Julia is installed with Juliaup on a mac. 1.10 set as default. Upon launching VSCode an error message pops up: You must have the "release" channel in Juliaup installed for the best Julia experience in VS Code. Nevertheless, VScodeserver seems to pick up 1.10, script execution works, but notebook hangs (but show release channel as kernel). This seems to relate that closed issue. I’m not sure why another version (besides 1.10 and 1.11) needs to be installed for this to work.
If anyone has hint on what could cause, it would be very welcome!
Yes, exactly. We run the jupyter notebooks within VScode.
Yesterday, we managed to fix one Windows machine (Julia working in VScode using a simple script but not using Jupyter notebooks within VScode) by making a “clean restart”. I will write the steps explicitly, in case it can be helpful to anyone: (1) delete Julia, (2) delete vscode, (3) delete folder .vscode, (4) delete folder .julia, (5) restart the computer, (6) install VScode with microsoft store, (7) install juliaup, (8) install 1.10 via juliaup, (9) install julia server extension in VScode. Then, the Julia server would complain (You must have the "release" channel in Juliaup installed for the best Julia experience in VS Code) so one more step needed (10) install an additional release channel in juliaup (which we don’t use in the end), then restart VScode… After this both scripts and notebooks started to work, however I sensed that the student was feeling like “things were so much easier in the previous lectures using matlab”…
Today, I will try fixing two other mac computers (symptoms described in the first post), I will try a similar “clean restart”.
Unfortunately no, we did a “clean restart” this morning on both macs. This did not help. This time only 1.10 and release channel were installed and this did not help. Next, I will try using JupiterLab instead of VScode notebooks, which will require explicitly the installation of IJulia. I hope this will do the job. A last option could be to use CoLab but the Julia support seems weak (one needs to install Julia in each notebook somehow).
It was also failing before I installed the release channel. Since there was the warning, I thought it would anyway be good to get rid of it, and that it may help things to run.
I think you should install IJulia explicitly, then the jupyter notebook within vscode should be able to communicate with julia. (even if the julia extension of vscode is not installed)
We’ve tried it but it did not help either. Eventually one student installed JupyterLab and can now use Julia’s kernel in there, so it’s one success. Now there is still one issue with a mac, which appears too old (2017) to reliably support installation of JupyterLab (warning during homebrew install steps).
Thanks! This explain why removing .vscode was not sufficient to erase all previous settings (e.g. last files opened).
So far, it seems that students manage to install JupyterLab with Julia kernel - so at least they can follow the lectures! I will try again later to get everything running in VScode together with them (erasing ~/.config/VSCode also, then reinstalling).