Support the "reload the open livebook from disk" workflow

[nietaki]

I just want to preface this by saying I'm really impressed by Livebook and how easy it was for me to get started with it for the project that needs some experiments/research. The featureset and the quality is great for how young the project is and I'm a fan of your work 💜

One thing that's making it harder for me to adopt it is the clunky workflow I need to use to edit the livebook file with an external editor - even if the project isn't shared and hosted exclusively on my machine.

The underlying flat .livemd files seem great for portability and are perfectly compatible with any and all text editors. And I can edit them with an external editor, but in order for the open session to pick the changes up, I need to close and reopen it every time, like in the repro:

Screen.Recording.2023-11-27.at.14.53.57.mov

TL;DR: This workflow is definitely possible (as long as autosave is off), but it's just enough clicks to be a bit of a pain to execute frequently - having an entry for it in the ... menu and/or a keyboard shortcut for it would make a huge difference for me.

I know that Livebook now supports vim/emacs bindings, but I have just enough personal configuration in my editor for the muscle memory to be messing me up every time I try to use the web editor. Also, using a fully-featured editor allows me to have multiple files from different projects open side by side for reference, easy project navigation, git integration, and other things that would be helpful when working on a set of livebooks over a longer period of time.

I also understand that reloading from disk probably has some implications when it comes to the evaluation of individual cells - I'd be perfectly happy to have to evaluate the whole livebook from scratch after it's been reloaded from the disk. This way one could do the bulk of the work in their editor of choice and use the livebook UI for finishing touches, interactive experiments, presentations and so on.

[josevalim]

There has been previous discussions/issues around this, so I recommend looking into them. But TL;DR is that many of our features, such as smart cells, notebook signing, etc go against editing Livebook in a separate editor, so it is not something we plan to explore. There are also technical complexities, as you mentioned, so I recommend finding past issues. :)

[nietaki]

Thank you for the very quick reply and providing more context. I looked for the previous issues/ideas again and only found this one, which did give some ideas.

I understand what you're saying and respect it's not something you plan to explore.

FWIW as long as the livebooks can be read from a local disk I think I personally will try to fit it into my current workflow, even if I run the risk of corrupting my files or not being able to use some advanced features.

Based on the comments in the other thread I figured out I can run livebook using escript and use a simple script a bit like this one to restart it and read the livebook from file whenever I want to:

pkill -9 -f livebook; livebook server /path/to/notebook.livemd --port 8769 >/dev/null 2>/dev/null &

[jonatanklosko]

If you want strictly the "edit in editor, test in Livebook" workflow, you can also always fork the notebook file, instead opening them. This means that any changes you do in Livebook won't be saved, but you don't have to worry about overriding or closing the session (unless you do something memory intensive, in which case you want to close sessions more frequently). From CLI you can do livebook server file:///path/to/notebook.livemd, or you can star the notebook for easy access from the home page.