Prune remote stale and merged branches that are >= 1 year old

prune-stale-branches.sh

#!/usr/bin/env bash

# used to execute branch deletes in parallel, shows progress bar
if ! [ -x "$(command -v parallel)" ]; then
  echo "ERROR: Missing command 'parallel'"
  exit 1
fi

if ! [ -x "$(command -v git)" ]; then
  echo "ERROR: Missing command 'git'"
  exit 1
fi

repo_url="$1"
if [ -z "$repo_url" ] || [ "$repo_url" == "-h" ] || [ "$repo_url" == "--help" ]; then
  cat << __EOF__
Usage: $(basename "$0") repo_url

Clones a bare git repository in order to fetch all remote branches. Then
deletes anything >= 1 year (relative time). Can be modified for absolute dates
swapping out \`\$format_relative\` -> \`\$format_absolute\`, and updating the
\`grep -v\` pattern. 

To finalize, you must run git push --mirror to delete the branches upstream.
__EOF__
  exit 1
fi

repo_name="$(basename "$repo_url")"
target_dir="${repo_name}-mirror"

if [ ! -d "$target_dir" ]; then
  # clone a bare directory with all remote branches, which allows for
  # a "dry run" committing to deletes (e.g. git push --mirror)
  git clone "$repo_url" "$target_dir" --mirror
fi

if [ ! -d "$target_dir" ]; then
  echo "Failed to clone: $repo_url"
  exit 1
fi

pushd "$target_dir" > /dev/null || exit 1

# merged branches
git branch --merged |
  cut -c3- |
  grep --invert-match --extended-regexp 'master|deployable|HEAD' |
  parallel --jobs 16 --bar git branch --delete --force {}

# shellcheck disable=SC2034
format_absolute='%(committerdate:format:%Y-%m-%d)|%(refname:short)'
# stale branches (edge case for 12 months != 1 year)
format_relative='%(committerdate:relative)|%(refname:short)'

git for-each-ref --format="$format_relative" --sort=committerdate refs/heads |
  awk 'BEGIN { FS = "|"  }; { printf "%-25s%s\n", $1, $2 }' |
  grep --extended-regexp "(\d+) year" |
  cut -c26- |
  parallel --jobs 16 --bar git branch --delete --force {}

cat << __EOF__
To complete purging of stale branches, run

  cd $target_dir
  git push --mirror

Note: Errors may occur due if the target repository if Github but these can safely
be ignored

Excerpt from https://github.com/rtyley/bfg-repo-cleaner/issues/36#issuecomment-37877829:
    The refs beginning 'refs/pull' are synthetic read-only refs created by
    GitHub - you can't update (and therefore 'clean') them, because they
    reflect branches that may well actually come from other repositories - ones
    that submitted pull-requests to you.

    So, while you've pushed all your real refs, the pull requests don't get
    updated. There's no real way to fix them with your existing repo with the
    GitHub admin tools. The severity of an issue depends on whether you were
    trying to remove any 'private' data:

    - Private data: the pull requests still have that private data in their
      history - they will remain as a way for other people to retrieve that
      data. The only ways to recover from this are to a) delete the repo using
      the GitHub admin tools, and then recreate it, or b) contact GitHub
      support and ask them to delete the PRs for you (they're pretty good about
      it)
    - Nothing Private: the pull requests still reference your old history - if
      any of them are still open, you don't want to merge them into your new
      fresh history! Get the submitters to rebase them on your new fresh
      history before merging.
__EOF__