Consider using Jupyter Cache

[betolink]

I saw a closed issue that describes more or less what I'm running into, I usually need to modify a small part of a notebook and it would be great if mkdocs-jupyter could use a cache to only re-render the parts that changed.

https://jupyter-cache.readthedocs.io/en/latest/#

Would it be possible? Thanks for the library btw!

[danielfrg]

Interesting, i didn't knew about the lib.

Definitely something I want to support, in my mind just checking the modified date and skipping a notebook would be so great to have. It's easy but need to find time to do it.

[fdeheeger]

Same sort of interest here ; Cache would help but there could be something else...

I am building a documentation with a quite large number of notebooks (moving from nbsphinx to mkdocs). I am experiencing a significant time to build the documentation (like way too much). By reducing that number to a manageable one, I have noticed that when the notebook contains a significant number of images, the building process increases a lot (well.. on my machine ;o) ).
I am not yet sure why (conversion time ? process of the converted document by mkdocs ?).
When I am converting notebook to markdown with nbconvert directly, mkdocs is building the site quite easily (but I am losing other options of the mkdocs-jupyter extension) ; so I am not sure where the time goes...

Any hint ?

[danielfrg]

Not sure but agree that cache is something we need to work on.

[logan-markewich]

+1 bumping on the need for a cache. We (LlamaIndex) have like 500 notebooks, and rebuilding them every time even though they haven't changed is a huge pain.

It would be super awesome if there was a way to save the built notebooks, and only rebuild if there was changes

[znichollscr]

The myst-nb project may provide some inspiration for how to implement this: https://myst-nb.readthedocs.io/en/v0.12.2/use/execute.html

If I can find time, I would gladly contribute a merge request