Test: Run By Line in Jupyter notebooks

[roblourens]

Refs: #126431, #128582

• windows @alexr00
• linux @bamurtaugh
• macOS @misolori

Complexity: 3

Authors: @roblourens, @DavidKutu

Create Issue

---

• Install ipykernel 6 or newer, which has debugging support

pip install ipykernel
python3 -m ipykernel install --user
or
pip install ipykernel --upgrade

• Make sure you have the latest version of the Jupyter extension installed in VS Code Insiders

• Set the experiment setting to enable Run by Line "jupyter.experimental.debugging": true

• Open a .ipynb notebook file

• Pick a kernel, make sure you pick the one where you installed ipykernel 6

• Click the Run by Line button in the cell toolbar

The button takes a bit to appear, as it's checking that you have ipykernel 6

• A debugging session should start, without the status bar changing, without the debug widget and without revealing the debug sidebar automatically
• Press the same button to advance to the next line, or the stop button to end the session
• Once you've stepped to the end of the cell, you have to press the stop button in the cell toolbar

More testing

• It should not step into another cell or file
• Note: the flow above is the only entrypoint. launch.json is not relevant here.
• Test run by line in an untitled notebook
• You should never see an error that looks something like "Unable to open ..." with some tmp file path
• Test run by line in two notebooks simultaneously

Known issues

• The run by line button takes a second to appear on notebook with valid kernels, it will be cached for the final release.
• The jupyter variables view is not enabled yet.
• There are two stop buttons visible (the normal cancel button and the RBL stop button in the cell toolbar) but these will be merged

[sandy081]

@roblourens I do not see Run by Line button in the cell toolbar. I installed ipykernel as you mentioned in the instruction. Is there a way to check if I have done correctly? This is what I see

[bamurtaugh]

I also do not see Run by Line:

I am on Jupyter extension v2021.8.1070546638

[miguelsolorio]

Installing ipykernel only installs v.5.3.4 for me and there doesn't appear an option install v6 via Anaconda Navigator?

[miguelsolorio]

Ok, I was able to upgrade via pip install ipykernel --upgrade, I assumed that installing ipykernel would install the newest version.

[bamurtaugh]

Another question - I needed to update my ipykernel version (as Miguel mentioned), but now that I've done that, should run by line work in a WSL environment?

Before I jump to spinning up a Linux VM, wanted to understand if this was a limitation of WSL, an issue with the actual notebook/run by line, or something else (I usually test Windows or anyOS items so wasn't sure if it was expected WSL may not work).

[roblourens]

Sorry for not being more clear about ipykernel. And I don't know why it wouldn't work in WSL. Did the notebook find your kernel, and were you able to select the right one?

[bamurtaugh]

Yes, I ended up installed Python 3.8 and Python 3.9, and it found both versions. I was able to select each one:

[roblourens]

And you can execute cells with it, and that's the one that you ran when installing ipykernel?

[bamurtaugh]

Yes it's letting me run cells:

and that's the one that you ran when installing ipykernel

I was checking jupyter --version to check on my version of ipykernel, but realized it has the same output regardless of which kernel I have selected.

I'm also unsure what I should be seeing in the status bar. I'm used to seeing the Python version of the interpreter, but now it's gone:

I tried creating a virtual environment. Initially, it defaulted to Python 3.6.9, so I manually changed it to 3.9.6, and now the status bar reads "Python 3.9.6":

However, when I close the virtual env file, the status bar entry disappears:

Apologies if these are obvious setup questions - I haven't experienced these issues when running regular Python projects in WSL before, and for more complex setups, I usually use dev containers. But when I tried the notebook in a dev container in WSL, I also still didn't see the run by line.

[roblourens]

I use a virtual env and I recommend that, but it looks like the venv isn't active in your terminal and isn't selected in the notebook.

I had better instructions on setting up a venv for debugging in last month's test plan item, I should have posted these again

#127362

[bamurtaugh]

Thanks for the links. I created an Azure VM and was just about to give up on WSL and use a VM with a venv for now until I realized the setting was grayed out:

I needed to enabled it explicitly for the WSL or SSH remote (rather than just putting it in my local settings.json). Once I did that in WSL, I see run by line:

(surprised it took me this long to realize this mistake 😅)

Now on to actual testing!

[roblourens]

Oh, that setting should not be machine scoped, I don't know why it is. I'll file an issue. Sorry for the trouble!

[sandy081]

@roblourens Please let me know if I am doing some thing wrong. I have install ipykernel version 6.0.3 and enabled the setting jupyter.experimental.debugging and I still do not see the Run By Line action.

I am blocked on this to test.

[roblourens]

Sorry @sandy081, I think there is a problem with Conda. Thanks for taking it @alexr00

[weinand]

@roblourens just a meta-comment: preparing a Dockerfile for these complex setups is helpful. For my simple-jupyter-notebook I did just that: https://github.com/microsoft/vscode-simple-jupyter-notebook#running-sample-in-docker-container
After installing Conda on my productive OS I saw all kinds of issues...