cognitive psychology. PhD 
The Future of Notebooks: Lessons from #JupyterCon http://willcrichton.net/notes/lessons-from-jupytercon/ …
-
Best we can do today is strip outputs before committing (see: http://timstaley.co.uk/posts/making-git-and-jupyter-notebooks-play-nice/ …). The challenge is that Jupyter's cell abstraction doesn't cleanly match text file abstraction expected by git et al.
You could always flatten a notebook into a text file and version control that, but there's logistical issues (mapping from derived file to source) and conceptual issues (do we want diffs to span cells?).
-
-
Arguably this is also an artifact of VCS design. You can't version notebooks for the same reason you can't version Photoshop files, Unity scenes, PDFs--no way to extend VCS with domain-specific structures. Can't capture "these cells swapped" as a 1-byte diff, only "JSON changed."
-
Slightly tangential: you could if you had a more contextual VCS - e.g. check out Memo https://github.com/atom/xray/tree/master/memo …
End of conversation
New conversation -
-
-
Right. The thing that's missing is the domain specific diff representation and diff visualization. I'd argue that this is not the responsibility of the vcs: the vcs should merely support pluggable diffs.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.