Making notebooks from rsts without clearing outputs

[jstac]

One of the things I loved about Jupytext is that, when you edit the Markdown file and then reload the corresponding notebook, you don't lose the outputs or what's currently stored in kernel memory.

So if I make a small edit at the bottom of a markdown file and then go back to the notebook, I don't need to run all the cells again. I just refresh and off I go.

In contrast, we lose everything when we make ipynbs from rsts, and hence need to run again from the top. That's almost prohibitive in some notebooks, making our editing system close to non-functional in such settings.

@AakashGfude @mmcky Could we please get together soon and discuss this?

[mmcky]

@jstac the outputs (or notebooks) using sphinx-build states are not checked (only the source RST files) so we would need to engineer this into the extension to get this kind of behaviour. It would be nice to have improved pipeline caching. The build process only checks RST's in the process and then rebuilds notebooks fresh each time. I think this task would be better suited to the RST to IPYNB converter that we have as an additional project (in similar fashion to JupyText, or perhaps teach JupytText RST)

[mmcky]

@AakashGfude let's review this after lunch today.

[mmcky]

@AakashGfude if we have ipynb notebooks available and a user doesn't use make clean do you think we could just replace all the incoming blocks and keep the outputs in place. The issue here would be we would need to know which code-blocks or code-cells changed (if any) and then re-run from that point forward.

[mmcky]

@jstac I have tagged this for v0.6 milestone which follows pdf support v0.5. Just flagging this as this is tagged with high-priority? Is that OK?

[jstac]

Sure, thanks @mmcky . I flagged it as high priority because it's something that would be super useful, but it's a side track from what you're trying to do now so please put it aside, in the way that you suggest.

I've removed the high priority flag.