3.2 Git Branching - Basic Branching and Merging

First, let’s say you’re working on your project and have a couple of commits already on the master branch.

You’ve decided that you’re going to work on issue #53 in whatever issue-tracking system your company uses. To create a new branch and switch to it at the same time, you can run the git checkout command with the -b switch:

This is shorthand for:

You work on your website and do some commits. Doing so moves the iss53 branch forward, because you have it checked out (that is, your HEAD is pointing to it):

Now you get the call that there is an issue with the website, and you need to fix it immediately. With Git, you don’t have to deploy your fix along with the iss53 changes you’ve made, and you don’t have to put a lot of effort into reverting those changes before you can work on applying your fix to what is in production. All you have to do is switch back to your master branch.

However, before you do that, note that if your working directory or staging area has uncommitted changes that conflict with the branch you’re checking out, Git won’t let you switch branches. It’s best to have a clean working state when you switch branches. There are ways to get around this (namely, stashing and commit amending) that we’ll cover later on, in Stashing and Cleaning. For now, let’s assume you’ve committed all your changes, so you can switch back to your master branch:

At this point, your project working directory is exactly the way it was before you started working on issue #53, and you can concentrate on your hotfix. This is an important point to remember: when you switch branches, Git resets your working directory to look like it did the last time you committed on that branch. It adds, removes, and modifies files automatically to make sure your working copy is what the branch looked like on your last commit to it.

Next, you have a hotfix to make. Let’s create a hotfix branch on which to work until it’s completed:

You can run your tests, make sure the hotfix is what you want, and finally merge the hotfix branch back into your master branch to deploy to production. You do this with the git merge command:

You’ll notice the phrase “fast-forward” in that merge. Because the commit C4 pointed to by the branch hotfix you merged in was directly ahead of the commit C2 you’re on, Git simply moves the pointer forward. To phrase that another way, when you try to merge one commit with a commit that can be reached by following the first commit’s history, Git simplifies things by moving the pointer forward because there is no divergent work to merge together — this is called a “fast-forward.”

Your change is now in the snapshot of the commit pointed to by the master branch, and you can deploy the fix.

After your super-important fix is deployed, you’re ready to switch back to the work you were doing before you were interrupted. However, first you’ll delete the hotfix branch, because you no longer need it — the master branch points at the same place. You can delete it with the -d option to git branch:

Now you can switch back to your work-in-progress branch on issue #53 and continue working on it.

It’s worth noting here that the work you did in your hotfix branch is not contained in the files in your iss53 branch. If you need to pull it in, you can merge your master branch into your iss53 branch by running git merge master, or you can wait to integrate those changes until you decide to pull the iss53 branch back into master later.

Suppose you’ve decided that your issue #53 work is complete and ready to be merged into your master branch. In order to do that, you’ll merge your iss53 branch into master, much like you merged your hotfix branch earlier. All you have to do is check out the branch you wish to merge into and then run the git merge command:

This looks a bit different than the hotfix merge you did earlier. In this case, your development history has diverged from some older point. Because the commit on the branch you’re on isn’t a direct ancestor of the branch you’re merging in, Git has to do some work. In this case, Git does a simple three-way merge, using the two snapshots pointed to by the branch tips and the common ancestor of the two.

Instead of just moving the branch pointer forward, Git creates a new snapshot that results from this three-way merge and automatically creates a new commit that points to it. This is referred to as a merge commit, and is special in that it has more than one parent.

It’s worth pointing out that Git determines the best common ancestor to use for its merge base; this is different than older tools like CVS or Subversion (before version 1.5), where the developer doing the merge had to figure out the best merge base for themselves. This makes merging a heck of a lot easier in Git than in these other systems.

Now that your work is merged in, you have no further need for the iss53 branch. You can close the ticket in your ticket-tracking system, and delete the branch:

Occasionally, this process doesn’t go smoothly. If you changed the same part of the same file differently in the two branches you’re merging together, Git won’t be able to merge them cleanly. If your fix for issue #53 modified the same part of a file as the hotfix branch, you’ll get a merge conflict that looks something like this:

Git hasn’t automatically created a new merge commit. It has paused the process while you resolve the conflict. If you want to see which files are unmerged at any point after a merge conflict, you can run git status:

Anything that has merge conflicts and hasn’t been resolved is listed as unmerged. Git adds standard conflict-resolution markers to the files that have conflicts, so you can open them manually and resolve those conflicts. Your file contains a section that looks something like this:

This means the version in HEAD (your master branch, because that was what you had checked out when you ran your merge command) is the top part of that block (everything above the =======), while the version in your iss53 branch looks like everything in the bottom part. In order to resolve the conflict, you have to either choose one side or the other or merge the contents yourself. For instance, you might resolve this conflict by replacing the entire block with this:

This resolution has a little of each section, and the <<<<<<<, =======, and >>>>>>> lines have been completely removed. After you’ve resolved each of these sections in each conflicted file, run git add on each file to mark it as resolved. Staging the file marks it as resolved in Git.

If you want to use a graphical tool to resolve these issues, you can run git mergetool, which fires up an appropriate visual merge tool and walks you through the conflicts:

If you want to use a merge tool other than the default (Git chose opendiff in this case because the command was run on a Mac), you can see all the supported tools listed at the top after “one of the following tools.” Just type the name of the tool you’d rather use.

After you exit the merge tool, Git asks you if the merge was successful. If you tell the script that it was, it stages the file to mark it as resolved for you. You can run git status again to verify that all conflicts have been resolved:

If you’re happy with that, and you verify that everything that had conflicts has been staged, you can type git commit to finalize the merge commit. The commit message by default looks something like this:

If you think it would be helpful to others looking at this merge in the future, you can modify this commit message with details about how you resolved the merge and explain why you did the changes you made if these are not obvious.