One-step solution to moving change from commit to a new branch off main / other arbitrary point?

[thomasheartman]

Hey!

Is it possible to move parts of a commit onto a different branch in a single command or without having to look up revisions? Or asked differently: can I squash into a new commit at an arbitrary destination?

Lemme explain what I'm looking for with an example. For context: I'm using jj in a colocated git repo.

Say I'm working on a branch A, which comes off of branch B (typically main, but can be any arbitrary git branch / jj bookmark), I guess.

I'm using the "squash workflow", so I've got a commit with a message in @-, and I'm currently making changes in @. That gives us something like:

@
|
A (or @-)
| 
B (main)

But while working on @, I find something that doesn't really belong to the task I'm working on, but that I would like to fix. Could be a typo, unused imports, unused code, etc.

Because it doesn't belong to my current change, I'd like to make the changes but instead of squashing them into @- (which would mix them up with my current changes), I'd like to make a new commit on main with these changes.

Is there an easy way to achieve this?

I've looked up the docs for jj squash, and they give you a destination, but not, it seems a way to create a new commit.

Workarounds

Of course, there are workarounds, but I haven't found any that are as convenient.

The most obvious to me, and the one that achieves what I want (if I understand correctly) is:

1. Create a new commit off of your target, but don't edit it: jj new B --no-edit
2. Read the unique prefix off of the command output
3. jj squash <files> -t <prefix>

But I would like to not have to read off the prefix, and ideally do it in one command.

In some cases, you could probably also rebase @ off the target (B), but that assumes that I want to squash all the changes in the current commit, which I often do not.

I'm sure there's other ways to do it, but I'm not sure how. I could probably make an alias that does this somehow, but I'm not sure exactly how (still pretty new to jj), so any tips would also be appreciated.

Thanks!

[joyously]

Have you looked at split? Along with the --interactive option, it has --insert-after, --insert-before, and --parallel. You still have to be able to tell it where.

[thomasheartman]

Oh, dang. I've seen it around, but haven't quite understood exactly what to use it for. Until now 🥳 Thank you 🙏🏼

Yeah, that seems to work perfectly, and it has the added benefits of

• letting me add a description for the new commit as I make it (and updating the existing one if there is one)
• it's only one operation, so jj undo works smoothly
• it auto-opens the diff editor if you don't specify any files
• it shows you the rev id of the newly created commit, so it's easy to follow up with making a remote branch and pr

But in short, the command it seems I was looking for was:

jj split -A main <files>

Thanks again!

[jennings]

You probably want -d main instead of -A main.

The former makes main the new commit’s parent, but the latter will insert the new commit between main and all of its descendants, including unrelated changes and remote branches.

Only use -A and -B if you want “between”. If you just want “on top of”, use -d.

[thomasheartman]

Good clarification. So if I understand correctly:

• If there's nothing else branched from the tip of main, -A and -d do the same thing
• If I'm on a branch based on the tip of main, -A will insert the new commit after main and rebase my current branch to instead be based off that commit?

The latter might be useful, actually, but it's a good distinction to bear in mind, anyway. Thanks!