May 04, 2015
It is very important to keep a project’s Git history clean. There are several huge advantages to this like making it easier to track down bugs and understand why technical decisions were made. To this end, it is often times needed to refine your working history before pushing it to remote.
Let’s take the following example: You are working on some functionality and you make some changes to one area of the application. While in there, you notice something else and make a quick change. You commit all of this in one commit. But when going back and look at the commit, you realize they touch two different areas of the application and would make more sense as two distinct commits. Git provides us with an easy way to go back and edit this history.
To demo how this is done, we’ll take a look at a simple history where we have two files (file1.txt and file2.txt) and three total commits. We will split the second commit into separate commits: one that touches file1.txt and one that touches file2.txt.
To start with our Git history looks like:
ebe053a Added line to file2 f2504cd Added line to file1 and added file2 3064113 Add file1.txt
We want to edit the second commit,
f2504cd and split it into two commits. First we need to rebase.
git rebase -i HEAD~2
Then in the editor, we need to select
edit for the first commit in the list.
edit f2504cd Added line to file1 and added file2 pick ebe053a Added line to file2
At this point, the rebase will be stopped with the last commit being the one we want to edit. If you want the existing commit message, it is a good idea here to use
git log and copy the last message. We then need to reset our head to before the commit.
git reset HEAD~
This will keep the changes from the commit we want to modify in the working directory. Now we can commit each part individually.
git add file1.txt git commit -m 'Added line to file1' git add file2.txt git commit -m 'Added file2.txt'
This will create two new commits rather than the original one. To finish off, we just need to complete the rebase.
git rebase --continue
Now if we look at our history, we have four commits instead of three:
a5b0a6b Added line to file2 3011d73 Added file2.txt 44e23d7 Added line to file1 3064113 Add file1.txt
This highlights how simple it is to split commits in Git using rebase and edit.
Written by Greg Babiars who builds things for the web. You can follow me on Twitter.