Interactive graph in jupyterlab-rise presentation

For a lecture, I produced a slide deck in jupyter lab with interactive graphs and presented it as a slide show with jupyterlab-rise. With the latest version of ipympl, the slide show only gives a static graph. Sadly I cannot reproduce the working setup I used last year, which makes it hard to investigate.

Minimal example notebook:

%matplotlib widget
import matplotlib.pyplot as plt
plt.plot([1,2,3], [4,5,6])

with ipympl 0.9.8

Opening the notebook with uv run --with ipympl==0.9.8 --with matplotlib==3.10.8 --with matplotlib-inline==0.2.1 --with jupyterlab_rise jupyter lab (the latest versions as of now) and running the cell gives a working interactive plot, but when opening the presentation, it shows a static image instead:

(You can tell that the right plot in the screenshot misses the resize handler, I can also tell that it is a plain <img> tag with a png data URL instead of a <canvas>, which seems to be the fallback option.)

ipympl 0.9.3 already shows the same behavior of falling back to a static image, but 0.9.0 is different:

with ipympl 0.9.0

Opening the notebook with uv run --with ipympl==0.9.0 --with matplotlib'<'3.7 --with matplotlib-inline==0.1.7 --with jupyterlab_rise jupyter lab and running the cell gives a working interactive plot, but when opening the presentation, CPU usage goes up to 100%, the two plots constantly redraw and become unusuable:

After countless tries, I once saw the redrawing resolve, with both plots becoming partly usable: I could hover over either of the plots and have the coordinates be displayed beneath, but as soon as I tried drawing a Zoom rectangle, it went into a busy loop again. It seems there is some race condition that can stop the redrawing if lucky.

with ipywidgets 7.8.5

Downgrading ipywidgets instead of ipympl also works, with the same busy loop problem as with ipympl 0.9.0. Downgrading ipywidgets has the advantage that ipympl, matplotlib and matplotlib-inline can be kept at the most recent version:

uv run --with ipympl==0.9.8 --with matplotlib==3.10.8 --with matplotlib-inline==0.2.1 --with jupyterlab_rise --with ipywidgets==7.8.5 jupyter lab

Versions

For the first example, these are the versions:

python -c "import sys; print('\n',sys.version); import ipympl; print('ipympl version:', ipympl.__version__)" && jupyter --version && jupyter nbextension list && jupyter labextension list

-->

 3.12.3 (main, Nov  6 2025, 13:44:16) [GCC 13.3.0]
ipympl version: 0.9.0

Selected Jupyter core packages...
IPython          : 8.38.0
ipykernel        : 7.1.0
ipywidgets       : 7.8.5
jupyter_client   : 8.8.0
jupyter_core     : 5.9.1
jupyter_server   : 2.17.0
jupyterlab       : 4.5.2
nbclient         : 0.10.4
nbconvert        : 7.16.6
nbformat         : 5.10.4
notebook         : 7.5.2
qtconsole        : not installed
traitlets        : 5.14.3

JupyterLab v4.5.2
/home/.../jupyter/labextensions
        jupyterlab_pygments v0.3.0 enabled OK (python, jupyterlab_pygments)
        jupyter-matplotlib v0.11.0 enabled OK
        jupyterlab-rise v0.43.1 enabled OK (python, jupyterlab_rise)
        @jupyterlab/mathjax3-extension v4.3.0 enabled  X (python, jupyterlab-mathjax3)
        @jupyter-notebook/lab-extension v7.5.2 enabled OK
        @jupyter-widgets/jupyterlab-manager v3.1.11 enabled OK (python, jupyterlab_widgets)


   The following extensions may be outdated or specify dependencies that are incompatible with the current version of jupyterlab:
        @jupyterlab/mathjax3-extension
        
   If you are a user, check if an update is available for these packages.
   If you are a developer, re-run with `--verbose` flag for more details.

(I assume mathjax3 is not the issue here since the plot works fine outside of RISE.)

I'm not sure whether this behavior is due to ipympl or jupyterlab-rise. I assume that RISE is to blame for showing the static fallback PNG, but I'm not so sure about the 100% CPU usage / busy loop. I'd be happy about any hints on what could be a working combination of versions, or any hints on how to work around the endless redrawing, if this rings a bell with anybody. I'm also happy to do further investigations if helpful.

首先，在 JupyterLab 中使用 jupyterlab-rise 运行最小 notebook，并使用此处报告的命令和版本比较 ipympl 0.9.8、ipympl 0.9.0 和 ipywidgets 7.8.5。追踪是 RISE 还是 widget 集成导致了静态 PNG 回退和持续重绘循环；完成的标准是演示文稿能够保留交互式绘图，同时不会占用过多 CPU。