Initial impl download; adjust example.sh to show multiple files

Dockerfile

@@ -1,4 +0,0 @@
-FROM busybox
-ADD testfile.txt .
-RUN rm testfile.txt && mkdir app && echo "hello" > /app/othertestfile.txt
-

README.md

@@ -6,7 +6,7 @@
 
 ### Prerequisites
 
-Requires the `jq` and `curl` programs to be installed on your PATH.
+Requires Docker 19.03 or newer and the `jq` and `curl` programs to be installed on your PATH.
 
 ### Installation
 
@@ -25,17 +25,56 @@ mkdir -p ~/.docker/cli-plugins && \
 
 ## Usage
 
-`docker artifact label [image] [file-path-1] [file-path-2] ...`
+`docker artifact` subcommands
 
-Adds labels to an existing image enabling the `download` command below to pull just the layers that contain the file paths specified above.
+* **label**
   
-Note: `download` must download the whole layer; to optimize for artifact download size, add the target files to the image in a separate layer.
+  Usage: `docker artifact label [options] image_name file_paths...`
 
-`docker artifact download [image] [file-path-1] [file-path-2]`
+  Adds a new layer to the end of the existing local image `image_name` with labels indicating in which layer each file from `file_paths` can be found. This enables the `download` subcommand to pull just the layer that contains the desired file without downloading the whole image.
 
-Downloads the image layers associated with the file paths specified and extracts them into the current directory.
+  *Don't forget to push the image after `docker artifact label` completes!*
+
+  Note: The `download` subcommand must download the whole layer; to optimize for artifact download size, add files that will be labeled to the image in a separate layer from other files.
+
+* **download**
+
+  Usage: `docker artifact download [options] image_name file_paths...`
+
+  Queries the remote docker repository api to find labels that indicate which layer to find each file in `file_paths`, then downloads just those layers and extracts `file_paths` from them into the current directory.
+
+Options
+
+* `-v` Prints a verbose description of each operation as the script performs them.
+
+## Example (TODO)
+
+> See the `example/` directory for a complete working example, summarized below:
+
+```
+> printf 'FROM busybox \n RUN mkdir app && echo "Hello World!" > /app/testfile.txt' | docker build -t infogulch/artifact-test -
+ => [internal] load build definition from Dockerfile
+ ...
+ => => writing image sha256:bc8ff9e88a0908f50f3c39b640a2f5a39f6cb5b7bce60510673d5da4f3e6e704
+ => => naming to docker.io/infogulch/artifact-test
+> docker artifact label infogulch/artifact-test /app/testfile.txt
+ ** Rebuilt image 'infogulch/artifact-test' to add 1 labels
+ ** Run 'docker push infogulch/artifact-test' to push it to your container repository
+> docker push infogulch/test-image
+...
+> docker artifact download infogulch/test-image /app/testfile.txt
+Downloaded and extracted '/app/testfile.txt' to the current directory
+> cat testfile.txt
+Hello World!
+```
 
 ## See also
 
-Related to [timwillfixit's original `docker-artifact`](https://github.com/tomwillfixit/docker-artifact) in spirit, though not in history. The primary difference is that this uses a more precice strategy to search for files.
+Related to [timwillfixit's original `docker-artifact`](https://github.com/tomwillfixit/docker-artifact) in spirit, though not in history. Some differences:
+
+* This uses a more precice strategy to find files. Specifically, it searches through layer tars for actual files, where the predecessor only searches through the layer commands for strings that happen to match the specified file names.
+* This requires specifying full file paths both to add labels and download. This prevents indadvertently labeling or downloading the wrong file that happens to have the same name.
+* This doesn't need to rebuild the docker image from its original Dockerfile & directory context. This means you can add file labels to an existing image built on another machine, and you don't need to recreate the exact `docker build` arugments such as --build-arg, --secret, or --target.
+* ???: This doesn't require third-party cli programs such as `ecr` or `az` to connect to cloud-hosted private repositories, it uses `docker login` credentials just like `docker pull` would.
+* ???: This correctly handles internal [`.wh.*` "whiteout files"](http://aufs.sourceforge.net/aufs5/man.html) that indicates when a file is deleted. This helps ensure that the file you download is actually present in the final image and wasn't deleted in some later layer. (Note: this is an anti-footgun, of course a malicious actor can still add any label to their image that they want.)
 

docker-artifact.sh

@@ -1,48 +1,220 @@
+#!/usr/bin/env bash
 set -e
 
-verbose=1
-image="$1"
-export searchpath="$2"
-export tarfile="$(mktemp)-$image.tar"
-
-# Save image tar to temp file
-[ $verbose ] && >&2 echo "Exporting image '$image' to temp file '$tarfile'..."
-docker image save "$image" -o "$tarfile"
-
-# Set up cleanup to not leave image behind
-function onexit() {
-  [ $verbose ] && >&2 echo "Cleaning up exported image '$tarfile'"
-  rm "$tarfile"
+function docker_cli_plugin_metadata {
+	local vendor="infogulch"
+	local version="v0.0.1"
+	local description="Manage Artifacts in Docker Images"
+	local url="https://github.com/infogulch/docker-artifact"
+	printf '{"SchemaVersion":"0.1.0","Vendor":"%s","Version":"%s","ShortDescription":"%s","URL":"%s"}\n' \
+		"$vendor" "$version" "$description" "$url"
 }
-trap onexit EXIT
 
-# Extract manifest and config file contents
-[ $verbose ] && >&2 echo "Collecting metadata from image..."
-manifest=$(tar -xf "$tarfile" -x manifest.json -O | jq)
-config_file=$(echo "$manifest" | jq -r '.[0].Config')
-config=$(tar -f "$tarfile" -x "$config_file" -O | jq)
+__usage="
+Usage: docker artifact [command] [options] [args]
  
-# Combine manifest.json and config json to build a map from tar layer directory to layer id sha
-export idmap=$(echo "$manifest" "$config" | jq -sr '[ [ .[0][0].Layers, .[1].rootfs.diff_ids ] | transpose[] | { (.[0]): .[1] } ] | reduce .[] as $x ({}; . * $x)')
+Commands:
  
-# Search each layer for a file matching $searchpath
-# TODO also search for related whiteout files that start with `.wh.`
-[ $verbose ] && >&2 echo "Searching layers for '$searchpath'..."
-found=$(echo "$manifest" | jq '.[0].Layers[]' | xargs -I {} sh -c 'digest=$(echo "$idmap" | jq -r ".[\"{}\"]"); tar -f "$tarfile" -x {} -O | tar -t | sed s_^_/_ | grep -wx "$searchpath" | xargs -I [] echo "[]=$digest"')
+	ls       - List labeled files available to download directly from an image
+	download - Download labeled files directly from an image
+	label    - Label files in an image 
  
-# If more than one is found, then bail
-if [ $(echo "$found" | wc -l) -gt 1 ]; then
-	>&2 echo "Multiple matches found, aborting:"
-	>&2 echo "$found"
+Command usage:
+ 
+	docker artifact ls [options] image_name
+	docker artifact download [options] image_name file...
+	docker artifact label [options] image_name file...
+ 
+Options:
+	-v       - verbose output
+	-q       - quiet output
+ 
+Examples:
+ 
+	docker artifact ls infogulch/artifact-test
+	docker artifact download infogulch/artifact-test /app/othertestfile.txt
+	docker artifact label infogulch/artifact-test /testfile.txt
+"
+
+function main {
+	case "$1" in
+		docker-cli-plugin-metadata)
+			docker_cli_plugin_metadata
+			;;
+		artifact)
+			case "$2" in
+				ls|list)  list "${@:3}" ;;
+				label)    label "${@:3}" ;;
+				download) download "${@:3}" ;;
+				*)        echo "$__usage" ;;
+			esac
+			;;
+	esac
+}
+
+function download {
+	local image="$1" files=("${@:2}") registry repo tag token labels
+	{ read -r registry; read -r repo; read -r tag; } <<< "$(image_parts "$1")"
+	token="$(get_token "$registry" "$repo")"
+	labels="$(list_json "$registry" "$repo" "$tag" "$token")"
+	# safely convert bash array of args into json array
+	files="$(for i in "${files[@]}"; do jq -n --arg f "$i" '$f'; done | jq -s '.')"
+	# test if all of the requested files are present in manifest labels
+	if ! jq -e --argjson f "$files" 'keys|contains($f)' <<< "$labels" > /dev/null ; then
+		echo " ** These files are not avilable to download from $image:"
+		echo "$(jq -r --argjson f "$files" '$f - keys | .[]' <<< "$labels" | sed 's_^_    _')"
+		exit 1
+	fi
+	# filter labels to just the ones that match files
+	labels="$(jq --argjson f "$files" 'with_entries(select(. as $e | $f | index($e.key)))' <<< "$labels")"
+	local shas="$(jq -r 'to_entries | map({(.value): {(.key): null}}) | reduce .[] as $i {{}; . * $i) | to_entries | map({key:.key, value:(.value|keys)}' <<< "$labels")"
+	echo "$shas"
+	#local a="$(xargs -L1 -I {} bash -c \
+		#'curl -s -L -H "Authorization: Bearer $1" "$2/$3" | tar -xz -O "$3"' \
+		#_ "$token" "https://$registry/v2/$repo/blobs/" {} \
+		#<<< "$shas")"
+	echo "Done!"
+}
+
+function list {
+	local registry repo tag token list
+	{ read -r registry; read -r repo; read -r tag; } <<< "$(image_parts "$1")"
+	token="$(get_token "$registry" "$repo")"
+	list="$(list_json "$registry" "$repo" "$tag" "$token")"
+	echo " ** The following files are available to download from $1: "
+	echo "$(jq -r 'keys | .[]' <<< "$list" | sed 's_^_   _')"
+}
+
+function list_json {
+	local registry="$1" repo="$2" tag="$3" token="$4" manifest labels
+	LOG "Querying manifest to extract labels for '$registry/$repo:$tag"
+	manifest="$(curl --silent \
+		-H "Accept:application/vnd.docker.container.image.v1+json" \
+		-H "Authorization: Bearer $token" \
+		"https://$registry/v2/$repo/manifests/$tag")"
+	# >&2 jq <<< "$manifest"
+	labels="$(jq -r '.history[].v1Compatibility' <<< "$manifest" | jq --slurp '[.[].config | select(.Labels != null) | .Labels] | add')"
+	jq <<< "$labels"
+}
+
+function image_parts {
+	local image="$1" regex='^([-_a-z\.]+)/([-_a-z]+)(:([-_a-z]+))?$'
+	if [[ "$image" =~ $regex ]] ; then
+		local registry="${BASH_REMATCH[1]}"
+		local repo="${BASH_REMATCH[2]}"
+		local tag="${BASH_REMATCH[4]:-latest}"
+		# if no . in registry part, then it must be an image from docker hub; translate appropriately
+		if ! [[ $registry =~ \. ]] ; then
+			repo="$registry/$repo"
+			registry="registry-1.docker.io"
+		fi
+	else
+		echo "Failed to parse image '$image'"
+		exit 1
+	fi
+	# >&2 echo "$image => $registry - $repo - $tag" #debug
+	echo "$registry"
+	echo "$repo"
+	echo "$tag"
+}
+
+function get_token {
+	local registry="$1" image="$2"
+	if [[ ! -z "$REGISTRY_TOKEN" ]] ; then
+		echo "$REGISTRY_TOKEN"
+		return
+	fi
+	LOG "Retrieving registry token for $registry/$image"
+	case "$registry" in
+		registry-1.docker.io)
+			curl --silent "https://auth.docker.io/token?scope=repository:$image:pull&service=registry.docker.io" \
+			| jq -r .token
+			;;
+		*.azurecr.io)
+			echo "$REGISTRY_TOKEN"
+			;;
+		*.dkr.ecr.*.amazonaws.com)
+			aws ecr get-authorization-token | jq -r '.authorizationData[0].authorizationToken'
+			;;
+	esac
+}
+
+function label {
+	local image="$1" searchpath="$2" tarfile="$(mktemp).tar"
+
+	# check to see if image exists locally. If not, Docker already prints an error so just exit
+	if ! docker image inspect "$image" > /dev/null ; then
+		exit 1
+	fi
+
+	# Save image tar to temp file
+	LOG "Exporting image '$image' to temp file '$tarfile'..."
+	docker image save "$image" -o "$tarfile"
+
+	# Set up cleanup to not leave image behind
+	trap "delete_files '$tarfile'; trap - RETURN" RETURN
+
+	# Extract manifest and config file contents
+	LOG "Collecting metadata from image..."
+	local manifest="$(tar -xf "$tarfile" -x manifest.json -O | jq)"
+	local config_file="$(echo "$manifest" | jq -r '.[0].Config')"
+	local config="$(tar -f "$tarfile" -x "$config_file" -O | jq)"
+
+	# Combine manifest.json and config json to build a map from tar layer directory to layer id sha
+	local idmap="$(echo "$manifest" "$config" | jq -s '[ [ .[0][0].Layers, .[1].rootfs.diff_ids ] | transpose[] | { (.[0]): .[1] } ] | reduce .[] as $x ({}; . * $x)')"
+
+	# Search each layer for a file matching $searchpath
+	# TODO also search for related whiteout files that start with `.wh.`
+	LOG "Searching layers for '$searchpath'..."
+	local found="$(echo "$manifest" | jq -r '.[0].Layers[]' | xargs -L1 -I {} bash -c '_search_layer "$@"' _ "$idmap" "$tarfile" {} "$searchpath")"
+	local foundcount="$(echo "$found" | grep -c . || true)"
+
+	# If more than one is found, then bail
+	if [ $foundcount -gt 1 ]; then
+		echo "File was changed in multiple layers, aborting. Found files and layer ids:"
+		echo "$found"
 		exit 2
-fi
+	elif [ $foundcount -eq 0 ]; then
+		echo "No files found matching '$searchpath'"
+		exit 3
+	fi
 
-# print out labels to add during image rebuild
-labels=$(echo "$found" | sed 's_^.*$_--label "\0"_' | paste -d' ')
+	local labels="$(echo "$found" | sed 's_^.*$_--label "\0"_' | paste -d' ')"
 
-echo "$labels"
+	# Add a layer to the existing image to add the labels and tag the new image with the same image name
+	LOG "Rebuilding image and adding labels: $labels"
+	echo "FROM $image" | eval docker build $labels -t "$image" - &> /dev/null
+	LOG "All image labels:"$'\n'"$(docker image inspect "$image" | jq '.[0].Config.Labels' | sed 's_^_     _' )"
 
+	# Remind user to push image
+	echo " ** Rebuilt image '$image' to add $foundcount labels"
+	echo " ** Run 'docker push $image' to push it to your container repository"
+}
 
-[ $verbose ] && >&2 echo "Done!"
-exit
+function _search_layer {
+	local idmap="$1" imagetar="$2" layertar="$3" search="$4"
+	# look up digest associated with layer path
+	local digest="$(jq --arg key "$layertar" -r '.[$key]' <<< "$idmap")"
+	# extract layer from image | list files in layer | add / prefix | search for file | append =$digest to each found file
+	tar -f "$imagetar" -x "$layertar" -O | tar -t | sed s_^_/_ | grep -wxF "$search" | sed 's_.$_\0='"$digest"'_'
+}
+
+function delete_files {
+	LOG "Cleaning up temp files $@"
+	rm "$@"
+}
+
+function LOG {
+	[ $verbose ] && >&2 echo -e "$(tput setaf 4) => $@$(tput sgr0)"
+}
+
+# run in subshell to allow sourcing this file without stomping on parents' namespace
+(
+	# Exports used in subshells
+	export verbose=1
+	export -f LOG
+	export -f _search_layer
+
+	main "$@"
+)
 

example/example.sh

@@ -0,0 +1,26 @@
+#!/usr/bin/env bash
+set -e
+
+pushd "$(dirname "${BASH_SOURCE[0]}")" &> /dev/null
+
+# Build example image
+>&2 printf "Rebuilding docker image infogulch/artifact-test\n"
+printf 'FROM busybox \n RUN mkdir app && echo "Test data" > /app/testfile.txt && echo "Hello world!" > /hello.txt' | docker build -t infogulch/artifact-test -
+
+>&2 printf "\nLabeling /app/testfile.txt and /hello.txt in infogulch/artifact-test\n"
+../docker-artifact.sh artifact label infogulch/artifact-test /app/testfile.txt /hello.txt
+
+>&2 printf "\nPushing image infogulch/artifact-test\n"
+docker push infogulch/artifact-test
+
+>&2 printf "\nListing files in infogulch/artifact-test\n"
+../docker-artifact.sh artifact ls infogulch/artifact-test
+
+>&2 printf "\nAttempt to download missing file /oops.txt\n"
+! ../docker-artifact.sh artifact download infogulch/artifact-test /oops.txt
+
+>&2 printf "\nDownloading /hello.txt and /app/testfile.txt\n"
+../docker-artifact.sh artifact download infogulch/artifact-test /hello.txt /app/testfile.txt
+
+popd &> /dev/null
+

testfile.txt

@@ -1 +0,0 @@
-hi