#!/bin/bash
##############################################################################
# clean-registry
# -----------
# Deletes old registry blobs that are not used in the latest version of the tag.
# This script assumes your registry is hosted on Google Cloud Storage.
#
# Usage:
#       clean-registry
#
# :authors: Jess Frazelle, @jessfraz
# :date: 20 September 2016
# :version: 0.0.1
##############################################################################
set -e
set -o pipefail

REPO_URL="${REPO_URL:-r.j3ss.co}"
BUCKET_PREFIX="gs://${REPO_URL}/docker/registry/v2"
JOBS=${JOBS:-10}

# get the repos
mapfile -t repos < <(gsutil ls "${BUCKET_PREFIX}/repositories/" | awk -F"/" '{print $(NF-1)}')

for repo in "${repos[@]}"; do
	# get the tags
	mapfile -t tags < <(gsutil ls "${BUCKET_PREFIX}/repositories/${repo}/_manifests/tags/" | awk -F"/" '{print $(NF-1)}')

	for tag in "${tags[@]}"; do
		echo "repo: $repo | tag: $tag"

		# get the latest index for the tag
		current=$(gsutil cat "${BUCKET_PREFIX}/repositories/${repo}/_manifests/tags/${tag}/current/link" || true)
		if [[ "$current" == "true" ]]; then
			continue
		fi
		# split on : since the link comes back as:
		# sha256:97bf033cf680a2fe49a6f0df40ac76dba97bc71745aeb06018dbf9df5de74bd6
		ha=${current%:*}
		id=${current#*:}

		echo "repo: $repo | tag: $tag | current: ${ha}:${id}"

		# get the shas
		mapfile -t shas < <(gsutil ls "${BUCKET_PREFIX}/repositories/${repo}/_manifests/tags/${tag}/index/${ha}/" | awk -F"/" '{print $(NF-1)}')

		# remove the current sha
		shas=( "${shas[@]/$id}" )

		# shellcheck disable=SC2199
		if [[ "${shas[@]}" != "" ]] && [[ ! -z "${shas[@]}" ]]; then
			parallel -j"${JOBS}" reg rm "${REPO_URL}/${repo}@${ha}:{1}" ::: "${shas[@]}" || true
		fi
	done
done