Closed ebiiii closed 5 years ago
This is exactly what is planned with our CrowdAI integration by building an extension that makes this easy: https://jupyterlab.readthedocs.io/en/stable/developer/extension_dev.html#developer-extensions
Note that i fear this can end up as implementation nightmare :tm:
why?
Because you need to wrap renku run
around and start the kernel inside.
oh I had in mind something much simpler... basically do the cleanup, then do renku run
with either papermill
or nbconvert
+ python
in a (sub)shell. You could do it on a branch and create a merge request or just do it on the same branch and commit the result.
@ebiiii have a look at https://github.com/SwissDataScienceCenter/renku-jupyterlab-ts -- for the time being, this only adds a command to the command palette.
Looks like it is doing the job. Thanks!
Maybe for the future, a reminder for the user that there are un-pushed changes before he closes the notebook or the lab. ;-)
closed by #8
Currently every execution inside the notebook is not tracking the output (this is fine for experimenting and exploration). But if someone want to do so, she would be required to open a terminal and type the renku run command. The idea here would be some shortcut to clean previous executions (untracked ones), and run the whole notebook in a deterministic manner (all cells in the right order) and track all outpus.