Closed pierre-haessig closed 6 months ago
@pierre-haessig normally this should have been fixed by the latest release of jupyterlite-pyodide-kernel
and https://github.com/jupyterlite/demo/pull/135.
Just checked from https://jupyterlite.github.io/demo/lab/index.html and it looks good:
Thanks for the fast feedback. Probably an error on my side since now it works also for me.
Normally I always erase the page data/cache before testing, but perhaps it wasn't erased when I made the test this afternoon. Unless something changed in a matter of 5 hours?
Normally nothing changed since #135.
But this might become an issue again in the future when there is a new release of ipywidgets
. Which requires updating the shims in jupyterlite-pyodide-kernel
.
For reference you might also want to try the xeus python kernel (which also supports widgets) with https://jupyterlite-xeus.readthedocs.io/en/latest/ and the template: https://github.com/jupyterlite/xeus-python-demo. This kernel should not have that ipywidgets related issue.
But this might become an issue again in the future when there is a new release of
ipywidgets
. Which requires updating the shims injupyterlite-pyodide-kernel
.For reference you might also want to try the xeus python kernel (which also supports widgets) with https://jupyterlite-xeus.readthedocs.io/en/latest/ and the template: https://github.com/jupyterlite/xeus-python-demo. This kernel should not have that ipywidgets related issue.
Thanks a lot for the feedback.
I think xeus-python kernel already surfaced in a previous discussion and at some point it may become the most reliable option. In the mean time updating pyodide is my shortest path to a fix, since xeus-python can only be bundled with conda-forge packages, not just simple PyPI packages like mine.
since xeus-python can only be bundled with conda-forge packages, not just simple PyPI packages like mine.
jupyterlite-xeus
has some basic support for installing packages from PyPI: https://jupyterlite-xeus.readthedocs.io/en/latest/environment.html#pip-packages
Although with some limitations (as described in the docs). But maybe it can still work with your simple package.
Hello,
This is, as far as I understand, the return of issues #97 and #127 about
ipywidgets
failing to install in the Jupyterlite demo deployment with Pyodide kernel.Reproduce
In the Jupyterlite demo, with the Pyodide kernel based notebook, running the cell magic
(or its function based equivalent
await piplite.install(['ipywidgets'])
) yields an error:Context