Branching and merging#
Objectives
Be able to create and merge branches.
Know the difference between a branch and a tag.
Instructor note
35 min teaching/type-along
15 min exercise
Motivation for branches#
In the previous section we tracked a guacamole recipe with Git.
Up until now our repository had only one branch with one commit coming after the other:
Linear Git repository.#
Commits are depicted here as little boxes with abbreviated hashes.
Here the branch
master
points to a commit.“HEAD” is the current position (remember the recording head of tape recorders?).
When we talk about branches, we often mean all parent commits, not only the commit pointed to.
Now we want to do this:#

Source: https://twitter.com/jay_gee/status/703360688618536960#
Software development is often not linear:
We typically need at least one version of the code to “work” (to compile, to give expected results, …).
At the same time we work on new features, often several features concurrently. Often they are unfinished.
We need to be able to separate different lines of work really well.
The strength of version control is that it permits the researcher to isolate different tracks of work, which can later be merged to create a composite version that contains all changes:
Isolated tracks of work.#
We see branching points and merging points.
Main line development is often called
master
ormain
.Other than this convention there is nothing special about
master
ormain
, it is just a branch.Commits form a directed acyclic graph (we have left out the arrows to avoid confusion about the time arrow).
A group of commits that create a single narrative are called a branch. There are different branching strategies, but it is useful to think that a branch tells the story of a feature, e.g. “fast sequence extraction” or “Python interface” or “fixing bug in matrix inversion algorithm”.
An important alias
We will now define an alias in Git, to be able to nicely visualize branch structure in the terminal without having to remember a long Git command (more details about aliases are given in a later section). This is extensively used in the rest of this and other lessons:
$ git config --global alias.graph "log --all --graph --decorate --oneline"
Instructor note
Instructors, please demonstrate how to set this alias and ensure that all create it. This is very important for this lesson and git-collaborative.
Let us inspect the project history using the git graph
alias:
$ git graph
* dd4472c (HEAD -> master) we should not forget to enjoy
* 2bb9bb4 add half an onion
* 2d79e7e adding ingredients and instructions
We have three commits and only one development line (branch) and this branch is called
master
.Commits are states characterized by a 40-character hash (checksum).
git graph
print abbreviations of these checksums.Branches are pointers that point to a commit.
Branch
master
points to commitdd4472c8093b7bbcdaa15e3066da6ca77fcabadd
.HEAD
is another pointer, it points to where we are right now (currentlymaster
)
On which branch are we?#
To see where we are (where HEAD points to) use git branch
:
$ git branch
* master
This command shows where we are, it does not create a branch.
There is only
master
and we are onmaster
(star represents theHEAD
).
In the following we will learn how to create branches, how to switch between them, how to merge branches, and how to remove them afterwards.
Creating and working with branches#
Instructor note
We do the following part together. Encourage participants to type along.
Let’s create a branch called experiment
where we add cilantro to ingredients.txt
.
$ git branch experiment master # creates branch "experiment" from master
$ git checkout experiment # switch to branch "experiment"
$ git branch # list all local branches and show on which branch we are
Verify that you are on the
experiment
branch (note thatgit graph
also makes it clear what branch you are on:HEAD -> branchname
):$ git branch * experiment master
Then add 2 tbsp cilantro on top of the
ingredients.txt
:* 2 tbsp cilantro * 2 avocados * 1 lime * 2 tsp salt * 1/2 onion
Stage this and commit it with the message “let us try with some cilantro”.
Then reduce the amount of cilantro to 1 tbsp, stage and commit again with “maybe little bit less cilantro”.
We have created two new commits:
$ git graph
* 6feb49d (HEAD -> experiment) maybe little bit less cilantro
* 7cf6d8c let us try with some cilantro
* dd4472c (master) we should not forget to enjoy
* 2bb9bb4 add half an onion
* 2d79e7e adding ingredients and instructions
The branch
experiment
is two commits ahead ofmaster
.We commit our changes to this branch.
Exercise: Create and commit to branches#
Branch-1: Create and commit to branches
In this exercise, you will create another new branch and few more commits. We will use this in the next section, to practice merging. The goal of the exercise is to end up with 3 branches.
Change to the branch
master
.Create another branch called
less-salt
Note! makes sure you are on master branch when you create the less-salt branch
A safer way would be to explicitly mention to create from the master branch as shown below:
$ git branch less-salt master
Where you reduce the amount of salt.
Commit your changes to the
less-salt
branch.
Use the same commands as we used above.
We now have three branches (in this case HEAD
points to less-salt
):
$ git branch
experiment
* less-salt
master
$ git graph
* bf59be6 (HEAD -> less-salt) reduce amount of salt
| * 6feb49d (experiment) maybe little bit less cilantro
| * 7cf6d8c let us try with some cilantro
|/
* dd4472c (master) we should not forget to enjoy
* 2bb9bb4 add half an onion
* 2d79e7e adding ingredients and instructions
Here is a graphical representation of what we have created:
Now switch to
master
.Add and commit the following
README.md
tomaster
:# Guacamole recipe Used in teaching Git.
Now you should have this situation:
$ git graph
* 40fbb90 (HEAD -> master) draft a readme
| * bf59be6 (less-salt) reduce amount of salt
|/
| * 6feb49d (experiment) maybe little bit less cilantro
| * 7cf6d8c let us try with some cilantro
|/
* dd4472c we should not forget to enjoy
* 2bb9bb4 add half an onion
* 2d79e7e adding ingredients and instructions
And for comparison this is how it looks on GitHub.
Merging branches#
Instructor note
We do the rest as type-along. Instructors, encourage learners to type-along.
It turned out that our experiment with cilantro was a good idea.
Our goal now is to merge experiment
into master
.
If you got stuck in the above exercises or joined later
If you got stuck in the above exercises or joined later, you can apply the commands below. But skip this box if you managed to create branches.
$ cd .. # step out of the current directory
$ git clone https://github.com/coderefinery/recipe-before-merge.git
$ cd recipe-before-merge
$ git checkout experiment
$ git checkout less-salt
$ git checkout master
$ git remote remove origin
$ git graph
Or call a helper to un-stuck it for you.
First we make sure we are on the branch we wish to merge into:
$ git branch
experiment
less-salt
* master
Then we merge experiment
into master
:
$ git merge experiment
We can verify the result in the terminal:
$ git graph
* c43b24c (HEAD -> master) Merge branch 'experiment'
|\
| * 6feb49d (experiment) maybe little bit less cilantro
| * 7cf6d8c let us try with some cilantro
* | 40fbb90 draft a readme
|/
| * bf59be6 (less-salt) reduce amount of salt
|/
* dd4472c we should not forget to enjoy
* 2bb9bb4 add half an onion
* 2d79e7e adding ingredients and instructions
What happens internally when you merge two branches is that Git creates a new commit, attempts to incorporate changes from both branches and records the state of all files in the new commit. While a regular commit has one parent, a merge commit has two (or more) parents.
To view the branches that are merged into the current branch we can use the command:
$ git branch --merged
experiment
* master
We are also happy with the work on the less-salt
branch. Let us merge that
one, too, into master
:
$ git branch # make sure you are on master
$ git merge less-salt
Commit graph after merge.#
We can verify the result in the terminal:
$ git graph
* 4f00317 (HEAD -> master) Merge branch 'less-salt'
|\
| * bf59be6 (less-salt) reduce amount of salt
* | c43b24c Merge branch 'experiment'
|\ \
| * | 6feb49d (experiment) maybe little bit less cilantro
| * | 7cf6d8c let us try with some cilantro
| |/
* | 40fbb90 draft a readme
|/
* dd4472c we should not forget to enjoy
* 2bb9bb4 add half an onion
* 2d79e7e adding ingredients and instructions
Observe how Git nicely merged the changed amount of salt and the new ingredient in the same file without us merging it manually:
$ cat ingredients.txt
* 1 tbsp cilantro
* 2 avocados
* 1 lime
* 1 tsp salt
* 1/2 onion
If the same file is changed in both branches, Git attempts to incorporate both changes into the merged file. If the changes overlap then the user has to manually settle merge conflicts (we will do that later).
Deleting branches safely#
Both feature branches are merged:
$ git branch --merged
experiment
less-salt
* master
This means we can delete the branches:
$ git branch -d experiment less-salt
Deleted branch experiment (was 6feb49d).
Deleted branch less-salt (was bf59be6).
This is the result:
Commit graph after merged branches were deleted.#
Compare in the terminal:
$ git graph
* 4f00317 (HEAD -> master) Merge branch 'less-salt'
|\
| * bf59be6 reduce amount of salt
* | c43b24c Merge branch 'experiment'
|\ \
| * | 6feb49d maybe little bit less cilantro
| * | 7cf6d8c let us try with some cilantro
| |/
* | 40fbb90 draft a readme
|/
* dd4472c we should not forget to enjoy
* 2bb9bb4 add half an onion
* 2d79e7e adding ingredients and instructions
Comparing figures “Commit graph after merge” and “Commit graph after merged branches were deleted”, we observe that only the pointers (“sticky notes”) disappeared, not the commits.
Git will not let you delete a branch which has not been reintegrated unless you
insist using git branch -D
. Even then your commits will not be lost but you
may have a hard time finding them as there is no branch pointing to them.
Optional exercises with branches#
The following exercises are advanced, absolutely no problem to postpone them to a few months later. If you give them a go, keep in mind that you might run into conflicts, which we will learn to resolve in the next section.
(optional) Branch-2: Perform a fast-forward merge
Create a new branch from
master
and switch to it.Create a couple of commits on the new branch (for instance edit
README.md
):Now switch to
master
.Merge the new branch to
master
.Examine the result with
git graph
.Have you expected the result? Discuss what you see.
Solution
You will see that in this case no merge commit was created and Git merged the two branches by moving (fast-forwarding) the “master” branch (label) three commits forward.
This was possible since one branch is the ancestor of the other and their developments did not diverge.
A merge that does not require any merge commit is a fast-forward merge.
(optional) Branch-3: Rebase a branch (instead of merge)
As an alternative to merging branches, one can also rebase branches. Rebasing means that the new commits are replayed on top of another branch (instead of creating an explicit merge commit). Note that rebasing changes history and should not be done on public commits!
Create a new branch, and make a couple of commits on it.
Switch back to
master
, and make a couple of commits on it.Inspect the situation with
git graph
.Now rebase the new branch on top of
master
by first switching to the new branch, and thengit rebase master
.Inspect again the situation with
git graph
. Notice that the commit hashes have changed - think about why!
Solution
You will notice two things:
History is now linear and does not contain merge commits.
All the commit hashes that were on the branch that got rebased, have changed. This also demonstrates that
git rebase
is a command that alters history. The commit history looks as if the rebased commits were all done after themaster
commits.
Summary#
Let us pause for a moment and recapitulate what we have just learned:
$ git branch # see where we are
$ git branch <name> # create branch <name>
$ git checkout <name> # switch to branch <name>
$ git merge <name> # merge branch <name> (to current branch)
$ git branch -d <name> # delete branch <name>
$ git branch -D <name> # delete unmerged branch
Since the following command combo is so frequent:
$ git branch <name> # create branch <name>
$ git checkout <name> # switch to branch <name>
There is a shortcut for it:
$ git checkout -b <name> # create branch <name> and switch to it
Typical workflows#
With this there are two typical workflows:
$ git checkout -b new-feature # create branch, switch to it
$ git commit # work, work, work, ..., and test
$ git checkout master # once feature is ready, switch to master
$ git merge new-feature # merge work to master
$ git branch -d new-feature # remove branch
Sometimes you have a wild idea which does not work. Or you want some throw-away branch for debugging:
$ git checkout -b wild-idea # create branch, switch to it, work, work, work ...
$ git checkout master # realize it was a bad idea, back to master
$ git branch -D wild-idea # it is gone, off to a new idea
No problem: we worked on a branch, branch is deleted, master
is clean.
Branch-4: Test your understanding
Which of the following combos (one or more) creates a new branch and makes a commit to it?
$ git branch new-branch $ git add file.txt $ git commit
$ git add file.txt $ git branch new-branch $ git checkout new-branch $ git commit
$ git checkout -b new-branch $ git add file.txt $ git commit
$ git checkout new-branch $ git add file.txt $ git commit
Solution
Both 2 and 3 would do the job. Note that in 2 we first stage the file, and then create the
branch and commit to it. In 1 we create the branch but do not switch to it, while in 4 we
don’t give the -b
flag to git checkout
to create the new branch.
Different meanings of “checkout”
Depending on the context git checkout
can do very different actions:
Switch to a branch:
$ git checkout <branchname>
Bring the working tree to a specific state (commit):
$ git checkout <hash>
Set a file/path to a specific state (throws away all unstaged/uncommitted changes):
$ git checkout <path/file>
This is unfortunate from the user’s point of view but the way Git is implemented it makes sense.
Picture git checkout
as an operation that brings the working tree to a specific state.
The state can be a commit or a branch (pointing to a commit).
In Git 2.23 (2019-08-16) and later this is much nicer:
$ git switch <branchname> # switch to a different branch
$ git restore <path/file> # discard changes in working directory
Keypoints
A branch is a division unit of work, to be merged with other units of work.
A tag is a pointer to a moment in the history of a project.