Today I learned about creating and applying patches with git format-patch and git am.

I was working on a remote JupyterLab instance, happily git commit'ing some stuff, only to figure out too late that I lacked git push permissions there. As it was a throwaway remote instance, I didn't want to spend my time on setting up the necessary SSH pubkey configuration. Instead, I took the poor man's route to transport commits:

  • "Export" the commits as a patch file:

    git format-patch origin/main

    This created a patch file for each commit I did on top op origin/main, following an auto-increment prefix filename pattern (e.g. 0001-Fix-wibble.patch).

  • Download the patch files to my local machine (just a right-click away in JupyterLab).

  • Apply the patch files to my local clone of the repo, e.g.

    git am 00*.patch

This locally recreated the git history I created remotely, ready to push.

Note that it is also possible to achieve the same with git diff and git apply. However, with git format-patch and git am more metadata about the original commits seems to be preserved by default.

I just scratch the surface here. Look elsewhere for more in-depth discussion and fine-tuning options of these git subcommands.