Cells not rendering or loading and code seems lost in jupyterlab notebook

[mariajmolina]

Hello! I frequently use JupyterLab and notebooks within the platform.

Certain cells that contain code can stop loading at times, and I am unable to click on them or view the code contained in the cells (see screenshot).

Has anyone else encountered this? Is this related to internet connection or something else? It took about 10 minutes for the cells to show up despite reloading the notebook, stopping and restarting the kernel, etc.

Thanks for your help!

[welcome]

Thank you for opening your first issue in this project! Engagement like this is essential for open source projects! 🤗

If you haven't done so already, check out Jupyter's Code of Conduct. Also, please try to follow the issue template as it helps other other community members to contribute more effectively.

You can meet the other Jovyans by joining our Discourse forum. There is also an intro thread there where you can stop by and say Hi! 👋

Welcome to the Jupyter community! 🎉

[krassowski]

Thanks for opening this issue. It sounds like it could be related to a new performance optimization in JupyterLab 3.1 not working as intended. Could you please confirm which exact version of JupyterLab you are using (you can get it from the top menu bar "Help" → "About JupyterLab")?

[mariajmolina]

Thanks so much, @krassowski ! The version of JupyterLab I am using is 3.1.6. Please let me know if you need any additional information.

[krassowski]

Thank you. A few more questions:

1. When this happens, is there any error message in the browser developer console? That would be very helpful to determine.
2. How many cells is there in your notebook? Is there a notebook in which it happened repeatedly? If so, would you possibly be able to share such a notebook?
3. What browser do you use? What is its version?
4. Could you please go to Advanced Settings Editor -> Notebook and in the right pane (User Preferences) paste the following:

{
    "renderCellOnIdle": false,
    "numberCellsToRenderDirectly": 10000000000000
}

and see if the problem goes away after pressing save and reloading the notebook?

[mariajmolina]

Some answers:

1. Yes! There are quite a few error messages, I am not sure if all are related to this. Screenshots below. Screenshots were taken before the notebook was reloaded and issue (at least temporarily) resolved.

2. There are 87 cells. This issue has happened a few times, but I would not say it is consistent. I raised this github issue because I was scared that I lost my code in the respective cells(!), but then they eventually came back (reloaded). I have not uploaded the notebook to github, but can do so tomorrow and link here if needed.

3. I use google chrome, and upon checking the version, the browser updated. It is now Version 92.0.4515.131 (Official Build) (x86_64). Upon updating and relaunching the browser, the issue appears to have gone away in my JupyterLab notebook, so I am not sure if it was related to the previous google chrome version, or if I just needed to relaunch because my browsers/tabs were open for too many days.

4. I have gone ahead and done this -- thanks for proposing a solution! Upon doing this, all of the cells are present and code is visible (and "run-able").

[krassowski]

Thank you very much! The notebook might help if you can provide it (after removing any content you would prefer not to share - I think that we would be just interested in the layout of cells i.e. how many consecutive code/markdown cells there are and which cells were affected). CC @echarles who may have more ideas.

[krassowski]

I further narrowed it down in #10853 (comment) and I think it is related to naive casting of cell to code cell (but we might be getting a placeholder instead). I can imagine that adding a cell type check could alleviate the issue for now, but I am not sure if that would be a good general solution, or if there is anything else we need to do. Still, a good reproducer would help a lot.

[echarles]

I can imagine that adding a cell type check could alleviate the issue for now, but I am not sure if that would be a good general solution, or if there is anything else we need to do. Still, a good reproducer would help a lot.

Hi, I have never seen the error printed in the console. Looks like it is related to SVG rendering. A Cell type check would avoid the stacktrace, but I don't think it will help to get the cell rendered.

I have not uploaded the notebook to github, but can do so tomorrow and link here if needed.

As @krassowski mentioned, a reproducer will help to understand the issue. @mariajmolina Looking forward your notebook so we can work on it. Thx a lot.

[TomAugspurger]

I believe that https://github.com/microsoft/PlanetaryComputerExamples/blob/main/tutorials/landcover.ipynb reproduces the issue (apologies, it's a bit large and requires a GPU to execute... If I spot another example I'll add it).

The issues start showing up around "Predicting land cover for each pixel"

[zaneselvans]

I think I might be having the same issue. When I do "run all cells" in notebooks that have previously been fine, there are now a bunch of cells that get skipped and don't run. If I step through one cell at a time, the cells will run, but then the cell focus becomes ambiguous and if I hit shift-enter again, the previous output is lost. Sometimes whole cells vanish.

Currently:

• I'm running JupyterLab 3.1.7
• I have disabled all of my extensions
• I'm running the notebook in a private browsing window on Firefox.
• The problem doesn't show up if I run the same notebook in "classic" mode (not jupyterlab)

I've seen the issue in several notebooks, but haven't been able to create a new one that's standalone and replicates the problem. However, it does show up in this notebook starting with code cell 11 which follows the heading "Simple Filtering" and contains:

slow_epacems_dd = dd.read_parquet(epacems_path)
slow_epacems_dd = slow_epacems_dd[
    (slow_epacems_dd.year == 2019)
    & (slow_epacems_dd.state == "ID")
]

The next cell that it attempts to run follows the header "Fuel Consumption" which contains:

unit_ids = comanche_df.unitid.unique()
fig, axs = plt.subplots(nrows=len(unit_ids), ncols=1, sharex=True, figsize=(20,15))
for n, unitid in enumerate(comanche_df.unitid.unique()):
    axs[n].scatter(
        comanche_df.query("unitid==@unitid").operating_datetime_utc,
        comanche_df.query("unitid==@unitid").heat_content_mmbtu,
        s=0.2, alpha=0.1,
    )
    axs[n].set_ylim(-100,10_000)
    axs[n].set_title(f"Comanche Unit {unitid}")
    axs[n].set_ylabel("Fuel consumed [mmbtu / hr]")
plt.show();

However, if after the first run of the notebook in which cells are skipped I then restart-kernel-and-run-all-cells, the notebook executes normally, running every cell.

If I downgrade to JupyterLab v3.1.4 the problem no longer seems to happen, there is no JupyterLab 3.1.5, and the original poster indicated they were on 3.1.6, so I imagine it's only in 3.1.6-7?

I'm running Firefox 91.0 on Ubuntu 21.04.

I got exactly the same behavior running JupyterLab on my laptop, and running it from within a Docker container.

[echarles]

@mariajmolina @TomAugspurger @zaneselvans Thx a lot for reporting. Can you confirm if you face the issue on the initial page load, or when you invoke "Run All Cells".

I have tried to linked notebook, but the need dependencies that don't install on my env. I the simple reproducer is not possible, the error shown in the browser console will help.

[echarles]

Regarding my previous comment, I just wanted to make sure the issue was arising on the initial notebook load (it should be, I have been confused reading comment).

If I downgrade to JupyterLab v3.1.4 the problem no longer seems to happen.

This is interesting. @mariajmolina @TomAugspurger @zaneselvans could I ask you to downgrade step by step from 3.1.7 to 3.1.0 and report when the issue does not arise anymore. Also the full stacktrace in the browser console will help.

[echarles]

I have downloaded the following notebooks and opened them in Chrome, Firefox and Safari (on MacOS). They render correctly, without cells remaining unrendered. I can not run them as I am missing python dependencies on my environment, but the virtual notebook rendering is only used on the page load, not on cell run.

• https://github.com/catalyst-cooperative/pudl-examples/blob/main/notebooks/03-pudl-parquet.ipynb
• https://github.com/microsoft/PlanetaryComputerExamples/blob/main/tutorials/landcover.ipynb

Maybe the indication of a specific browser or operating system can help to identify the cause of the issue?