deploy.sh

#!/usr/bin/env bash
set -o errexit #abort if any command fails

deploy_directory=public
deploy_branch=gh-pages

#if no user identity is already set in the current git environment, use this:
default_username=deploy.sh
default_email=

#repository to deploy to. must be readable and writable.
repo=origin

# Parse arg flags
while : ; do
  if [[ $1 = "-v" || $1 = "--verbose" ]]; then
    verbose=true
    shift
  elif [[ $1 = "-s" || $1 = "--setup" ]]; then
    setup=true
    shift
  elif [[ $1 = "-e" || $1 = "--allow-empty" ]]; then
    allow_empty=true
    shift
  else
    break
  fi
done

#echo expanded commands as they are executed (for debugging)
function enable_expanded_output {
  if [ $verbose ]; then
    set -o xtrace
    set +o verbose
  fi
}

#this is used to avoid outputting the repo URL, which may contain a secret token
function disable_expanded_output {
  if [ $verbose ]; then
    set +o xtrace
    set -o verbose
  fi
}

enable_expanded_output

function set_user_id {
  if [[ -z `git config user.name` ]]; then
    git config user.name "$default_username"
  fi
  if [[ -z `git config user.email` ]]; then
    git config user.email "$default_email"
  fi
}

function restore_head {
  if [[ $previous_branch = "HEAD" ]]; then
    #we weren't on any branch before, so just set HEAD back to the commit it was on
    git update-ref --no-deref HEAD $commit_hash $deploy_branch
  else
    git symbolic-ref HEAD refs/heads/$previous_branch
  fi

  git reset --mixed
}

if ! git diff --exit-code --quiet --cached; then
  echo Aborting due to uncommitted changes in the index >&2
  exit 1
fi

commit_title=`git log -n 1 --format="%s" HEAD`
commit_hash=`git log -n 1 --format="%H" HEAD`
previous_branch=`git rev-parse --abbrev-ref HEAD`

if [ $setup ]; then
  mkdir -p "$deploy_directory"
  git --work-tree "$deploy_directory" checkout --orphan $deploy_branch
  git --work-tree "$deploy_directory" rm -r "*"
  git --work-tree "$deploy_directory" add --all
  git --work-tree "$deploy_directory" commit -m "initial publish"$'\n\n'"generated from commit $commit_hash"
  git push $repo $deploy_branch
  restore_head
  exit
fi

if [ ! -d "$deploy_directory" ]; then
  echo "Deploy directory '$deploy_directory' does not exist. Aborting." >&2
  exit 1
fi

if [[ -z `ls -A "$deploy_directory" 2> /dev/null` && -z $allow_empty ]]; then
  echo "Deploy directory '$deploy_directory' is empty. Aborting. If you're sure you want to deploy an empty tree, use the -e flag." >&2
  exit 1
fi

disable_expanded_output
git fetch --force $repo $deploy_branch:$deploy_branch
enable_expanded_output

#make deploy_branch the current branch
git symbolic-ref HEAD refs/heads/$deploy_branch

#put the previously committed contents of deploy_branch branch into the index
git --work-tree "$deploy_directory" reset --mixed --quiet

git --work-tree "$deploy_directory" add --all

set +o errexit
diff=$(git --work-tree "$deploy_directory" diff --exit-code --quiet HEAD)$?
set -o errexit
case $diff in
  0) echo No changes to files in $deploy_directory. Skipping commit.;;
  1)
    set_user_id
    git --work-tree "$deploy_directory" commit -m \
      "publish: $commit_title"$'\n\n'"generated from commit $commit_hash"

    disable_expanded_output
    #--quiet is important here to avoid outputting the repo URL, which may contain a secret token
    git push --quiet $repo $deploy_branch
    enable_expanded_output
    ;;
  *)
    echo git diff exited with code $diff. Aborting. Staying on branch $deploy_branch so you can debug. To switch back to master, use: git symbolic-ref HEAD refs/heads/master && git reset --mixed >&2
    exit $diff
    ;;
esac

restore_head