When working on a new feature, I try to group my code under logical commit boundaries. That is to say, I like the code in my commits to be related. Sometimes though, an unrelated change will sneak past my guard. For this reason, I find it helpful to know how to break one large commit into several smaller ones.
A word of caution, however. This technique can be dangerous when applied to branches shared by multiple contributors. To break one commit into several is to rewrite history. Rewriting history can cause headaches for other contributors on a branch, leading to merge conflicts or even loss of work. Because of this, it is only ever safe to rewrite history on branches with no other collaborators. Even though I usually work alone on my projects, I tend to work in feature branches for this reason. Just to be safe.
To break a commit apart, we will use the
edit command in an interactive
rebase. To start an interactive rebase, use
git rebase -i with some kind of
reference. This reference is any pointer to a commit which you would like to
rebase against. It could be the root commit, a particular hash, or even a branch
$ git rebase -i --root $ git rebase -i 133b47e $ git rebase -i origin/master
Running one of the variants above will open something like this in a text editor:
pick 133b47e The first commit message pick 6468919 Another commit message pick e1d4848 The last commit message # Rebase f73e4e2..e1d4848 onto f73e4e2 (3 commands) # # Commands: # p, pick = use commit # r, reword = use commit, but edit the commit message # e, edit = use commit, but stop for amending # s, squash = use commit, but meld into previous commit # f, fixup = like "squash", but discard this commit's log message # x, exec = run command (the rest of the line) using shell # d, drop = remove commit # # These lines can be re-ordered; they are executed from top to bottom. # # If you remove a line here THAT COMMIT WILL BE LOST. # # However, if you remove everything, the rebase will be aborted. # # Note that empty commits are commented out
Breaking the commit apart
Find the commit that you would like to break apart and change
Save and quit to continue. You should see output like this:
Stopped at 6468919... Add "breaking-a-commit-apart" and "other" files You can amend the commit now, with git commit --amend Once you are satisfied with your changes, run git rebase --continue
At this point, we’ve traveled through Git’s commit history and arrived at a target commit. To unstage the relevant changes, run:
git reset HEAD^
From here, you are able to
add changes and
commit them using your normal
process. When you’re satisfied with the more granular commits that you’ve made,
you can run:
git rebase --continue
This will replay the commits we traversed on top of your modified history. The end result is a commit history where your large commit has been broken apart.
Pushing Your Changes
If this is the first time you’re pushing a set of commits upstream, your
git push should work as expected. If, however, your changes modify a commit
that the upstream branch is aware of, you’re going to need to force push.
I recommend using this variant of the
git push --force-with-lease
--force-with-lease flag protects you from overwriting upstream changes.
If, for example, another contributor commits to your feature branch, your push
will fail rather than overwrite their work.