Simple Git Branch Squashing

The old way

When developing features, I'll often use temporary WIP branches to store WIP commits before squashing them into a single feature commit. To squash the WIP commits, I have always run git rebase -i HEAD~x, where x is the number of commits to squash. To calculate x, I run git log and manually count each commit. This week I found a simpler way write that command.

A simpler way

It turns out that HEAD~x is a commit-ish, which refers to the commit x-1 before HEAD. For trunk based development, HEAD~1 is actually the head of the trunk branch! Conveniently, this commit can also be referred to by the trunk branch's name.

Put more concisely:

When squashing x commits atop a branch named develop, then HEAD~x and develop are equivalent commit-ishes.

That means the command git rebase -i develop can be used instead of git rebase -i HEAD~x. Same outcome, less cognitive overhead!

Example

For this blog, the trunk branch is develop. When implementing the Atom Feed feature, the atom-feed branch was used to track work in progress.

The atom-feed branch contains 4 WIP commits that together comprise the feature. The goal is to squash these 4 commits into a single commit on develop.

First, to perform the squash I ran git rebase -i develop to start an interactive rebase:

Here I reword the first commit (to describe the overall feature) and then fixup to squash the final 3 commits into the first. The resulting commit (a55bf81) is our feature commit. Also, this results in the atom-feed branch being 1 commit ahead of develop:

Finally, after merging the atom-feed branch into develop, the history looks like this: