Clean Git house

This guide is about deleting branches in Git. There’s a few ways I break down how to do it from manual effort to maximum automation and every step explained in between. To skip that and get to the good stuff, click here .

Deleting branches in Git can be confusing and at times harrowing for folks getting started. The distributed nature of Git allows for remotes to have copies of branches and your local machine to have copies of branches. Over time, you can create a large number of branches that exist in both places. If you delete your remote branches outside of your Git repository, such as on GitHub, then you still need to delete them from your local repository. And, you can delete local branches, but that doesn’t actually delete them from the remote repository.

Git has some features that easily allow you to do either of these activities easily. To really clean house though, you need to combine these commands to delete your local and remote stale branches from your local Git repository.

Deleting local branches

There are a lot of ways to delete local branches for Git. One of the more popular ways is to delete them each on by one running the delete flag on the branch command.


git branch --delete ${local_branch_name}

When you do this with the -d, --delete flag, it’ll delete the branch you choose. The command will give you an error if you attempt to delete a branch that hasn’t been merged upstream or if its commits aren’t currently in your HEAD. If you want to go ahead and delete the branch regardless, you run the -D flag which is a shortcut for --delete --force.


git branch --delete --force ${local_branch_name}

Deleting remote branches

There’s a couple of ways to delete remote branches. If you’re using GitHub, you can do it from the /branches endpoint on your GitHub repository and delete any branches from there. Also, you can set your branches to automatically delete branches from the Options section of the /settings endpoint on your GitHub repository.

On the command-line, you can delete branches by pushing nothing to the branch on your repository.


git push origin :${branch_name}
                ^ # the empty space in front of the colon is the nothing you
                  # are pushing to the ${branch_name}.

This will cause the remote branch to be deleted. Combining this with the local way of deleting branches allows you to keep your output from git branch --all neat and tidy.

Deleting by pruning stale remote branches

For the sake of this post, a stale remote branch is a branch that has been deleted from your remote repository and your local repository still thinks it exists when you run git branch --remote. Git has a pretty straight forward way of deleting these stale branches by pruning them.


git remote prune ${remote_name}
                 ^ # this is usually `origin` but you can name remotes whatever
                   # you'd like to name them.

Helpfully, there isn’t an equivalent command that takes into account local and remote branch relationships and deletes the local branches that your local repository doesn’t know about.

The main reason for this is that a command like that would be dangerous to any local work that you’ve done on a local branch that you haven’t pushed up yet. After all, local branches are usually the beginning of new work on a repository and remote branches are considered already distributed across clones of the repository. It should be easy to delete these remote branches locally, but deleting local branches based on the non-existence of a remote branch is a tricky thing to do that could potentially cause you to lose your local work.

So let’s build something

So even though it doesn’t ship with Git, you can pipe together various commands to get Git to do this local branch pruning based on what remote branches don’t exist on the repository but do exist in your local repository.


git branch -r | \
  awk '{print $1}' | \
  egrep -v -f /dev/fd/0 <(git branch -vv | grep origin) | \
  awk '{print $1}' | \
  xargs git branch --delete

Breaking down the Bash

Now there’s quite a lot to this command. I’ve broken it down into multiple lines for legibility, but it is still running 7 separate commands in two separate shells. It’s a lot, so I’m going to break it down further.

Fetch remote branches

This command gets all the remote branches from all your remotes. For the sake of these examples, let’s assume the remote is name origin as that’s the default.


git branch --remote

Get the first column of output

This command leverages awk to print out the first column of output from the previous command. In this case, it outputs the names of all the remote branches with origin/ in front of them. e.g. origin/main.


awk '{print $1}'

Grep the remote branch list for local branch names

This command leverages egrep with flags of -v and -f. The -v, --invert-match flag is used to exclude any remote branches from the list of local branches. The list of local branches is being redirected into a file called /dev/fd/0. The explanation of this is a bit out-of-scope for this post , but essentially it’s allowing egrep to read from STDIN.

This command outputs an inverted match of remote branches comparing them to a list of your local branches. Because the match is inverted, it means that the names of the branches that are matched locally don’t exist in the list of remote branches.


egrep -v -f /dev/fd/0 <(git branch -vv | grep origin)

Get the first column of output again

This command leverages awk to print out the first column of output from the previous command. In this case, it outputs the full-name of the local branch.


awk '{print $1}'

Leverage xargs to run git commands

This command leverages xargs which takes a string that’s broken up by delimiters such as spaces or newlines and runs the command passed into it. This command runs the git branch --delete on the output of the previous command which is a newline delimited string of all the local branches that don’t exist on the remote repository.


xargs git branch --delete

Automating and aliasing

So I don’t like to have to remember all these steps when running this locally. There’s quite a few things you’d need to remember such as pruning the remote branches first, then running the 7 commands in the right order and piping the output of one into the next.

To make this easier, let’s create a named function which takes an argument of the remote name for your Git repository and iterates through all the steps I previously mentioned. It looks like this.


limpia-git() {
  remote_name="${1:-origin}"
  if [ -d $(pwd)/.git ]
  then
    git remote prune $remote_name

    git branch -r | \
    awk '{print $1}' | \
    egrep -v -f /dev/fd/0 <(git branch -vv | grep ${remote_name}) | \
    awk '{print $1}' | \
    xargs git branch --delete

  else
    echo "There's not \".git/\" directory in \"$(pwd)\"."
  fi
}

Finally with all this setup in a file that you source into your shell session, you can simply run limpia-git or limpia-git origin and have it clean up all your stale local and remote branches. And because of the egrep matching and the git remote prune ${remote_name} commands, this is a no-op when there aren’t branches to delete or stale local branches to delete.

Also, to be extra careful, the git branch --delete command will check if your commits have been merged into the HEAD of any of your remote branches and won’t delete local branches that aren’t merged yet. You can remove this safety net by changing the command to git branch --delete --force.