delete all remote branches that have already been merged into master

gistfile1.txt

$ git branch -r --merged | 
  grep origin | 
  grep -v '>' | 
  grep -v master | 
  xargs -L1 | 
  awk '{split($0,a,"/"); print a[2]}' | 
  xargs git push origin --delete

Hi, thanks all for the tip to drop a huge list of merged branches but I believe that we can save time pushing all branches together to have one unique transaction:

git push origin --delete $(git branch -r --merged origin/master |  grep origin | egrep -v '>|master|develop' | cut -d/ -f2-)

I like your solution @voiski. Can this somehow also be improved to avoid this message when no refs exist:

fatal: --delete doesn't make sense without any refs

Best answer in my opinion @voiski 👏 👏 👏

I created a script named git-merged on my PATH. Git recognizes this syntax and allows you to execute these scripts as if they are aliases. In this case, this script allows you to run git merged. The contents of the script are as follows:

#!/usr/bin/env bash
remote="${1:-origin}"
branch="$2"
git branch -r --list "$remote/*" --merged $branch \
    | sed "s/\s*$remote\///" \
    | egrep -v "^(HEAD|release|hotfix|master|develop)"

With this script, you can list merged branches on the remote (default behavior is to check remote origin for branches merged to HEAD):

$ git merged

You can fully specify the remote and target branch to check for merges:

$ git merged origin my-topic

Example above checks all remote tracking branches on remote origin that are merged to local branch my-topic. Using xargs, you can use this to effectively delete all merged branches on the remote:

$ git merged | xargs git push origin --delete

You can add the -n option to do a dry push to verify what will happen before you actually delete anything:

$ git merged | xargs git push origin --delete -n

The following branch patterns are ignored (supports git-flow branch naming):

• release/1.2.3
• hotfix/1.2.3
• develop
• master
• origin/HEAD

The intention is to explicitly clean up stale, merged branches for feature development such as:

• feature/my-thing
• bugfix/crash-issue
• my-topic-branch

Could you replace the current awk statement that resides in the gist file with the suggestion by @catsby, i.e. awk '{sub(/origin\//,"");print}'?

Here's a bash version, that doesn't rely on awk or sed or xargs.

for branch in $(git branch -r --merged master | grep origin | grep -v develop | grep -v master);
do
  git push origin --delete "${branch##*/}";
done

Here's a bash version, that doesn't rely on awk or sed or xargs.

for branch in $(git branch -r --merged master | grep origin | grep -v develop | grep -v master);
do
  git push origin --delete "${branch##*/}";
done

Clearest solution I've seen, it makes the intent much more explicit than a long chain of awk/sed/xarg calls.

Only thing I'd add is that it doesn't quite work for branches that have a '/' in the name - if branch="origin/foo/bar" then ${branch##*/} will be "bar" and not "foo/bar"! You can fix by using the non-greedy single # to match the substring:

for branch in $(git branch -r --merged master | grep origin | grep -v develop | grep -v master)
do
  git push origin --delete "${branch#*/}"
done

My solution to prune merged branches from local + multiple remotes, based on snippets above:

https://gist.github.com/ryanc414/f7686d2c97808b41ed8518a5840e2d78