image-best-format.sh

image-best-format.sh

#!/bin/sh
# shellcheck disable=SC3043
set -e
set -u
#(set -o | grep -q pipefail) && set -o pipefail
#(set -o | grep -q posix) && set -o posix
#shopt -s failglob
#set -x

# Compute the most optimal (file size) format for an input image

# input:
# * a set of images to study
# * a threshold from wich a format change is recommanded (net gain of x% in file size)

# output:
# * an Excel spreadsheet detailing the most optimal format for each image, along with the transcoded image sizes for all supported output formats
# * the transcoded images where changing format makes sense

# If running this script publicly, don't forget to setup an ImageMagick policy:
# https://imagemagick.org/script/security-policy.php

# Copyright 2020-2024 Cédrik LIME


CONST_OUTPUT_FORMATS="${OUTPUT_FORMATS:-GIF JPEG PNG WebP AVIF}"

##############################################################################
# Shell utilities
##############################################################################

# Usage:
# local reset_e=0 ; internal_is_shell_attribute_set "e" && set +e && reset_e=1
# ...
# [ "$reset_e" -eq 1 ] && set -e
internal_is_shell_attribute_set() { # attribute, like "e"
	# Alternative implementation (e.g. for set -x): [ ${-/x} != ${-} ] && tracing=1 || tracing=0
	#local search_attribute=$1
	case "$-" in
		*"$1"*) return 0 ;;
		*)      return 1 ;;
	esac
}

internal_is_shell_option_set() { # option, like "pipefail"
	# Note: bash-specific alternative: `test -o`
	local search_option="$1"
	case $(set -o | grep "$search_option" | cut -f2) in
		on) return 0 ;;
		off) return 1 ;;
		*) echo "Error: unknown shell option value \"$search_option\"!" >&2; return 1 ;;
	esac
}

# shellcheck disable=SC2034
LOG_LEVEL_TRACE=0
# shellcheck disable=SC2034
LOG_LEVEL_DEBUG=1
# shellcheck disable=SC2034
LOG_LEVEL_INFO=2
# shellcheck disable=SC2034
LOG_LEVEL_WARN=3
# shellcheck disable=SC2034
LOG_LEVEL_ERROR=4
LOG_LEVEL=$LOG_LEVEL_INFO

log() {
	local REQUESTED_LOG_LEVEL="${1}"
	shift
	# shellcheck disable=SC2086
	if [ $REQUESTED_LOG_LEVEL -ge $LOG_LEVEL ]; then
		echo "$@"
	fi
}

##############################################################################

print_usage() {
	cat << EOT
Optimise images by computing the best (file size) format.
Supported input formats: all of underlying ImageMagick (see https://imagemagick.org/script/formats.php).
Supported output formats: JPEG, PNG, WebP, GIF.
${0##*/} will output (on stdout) the results in tabular format (tsv).
Usage
    ${0##*/} [-q|-v] [-t threshold] [-o directory] file [file...]
    -t threshold in percent for which we report a better image format, from 0 to 100; default: 10
    -o directory to which write images in new format
       no transcoding will be done if not specified
    -q quiet
    -v verbose
    -h this help
    E.g.: ${0##*/} -t 5 myImage1.jpg myImage2.png
    E.g.: ${0##*/} -q -t 5 -d target_directory myImage1.jpg myImage2.png
EOT
}


##############################################################################

# See https://imagemagick.org/script/defines.php and https://imagemagick.org/script/webp.php
IMAGICK_STD_DEFINES="-define preserve-timestamp=true -define heic:depth-image=true -define jpeg:arithmetic-coding=on -define jpeg:optimize-coding=on -define webp:alpha-compression=1 -define webp:auto-filter=true -define webp:thread-level=1"

compute_image_size_for_format() {
	local FORMAT="$1"
	local FILE="$2"
	local reset_e=0
	internal_is_shell_attribute_set "e" && set +e && reset_e=1
	local optimised_size=$(magick -define stream:buffer-size=0 $IMAGICK_STD_DEFINES "$FILE" "$FORMAT":- | wc -c | sed -e 's/^[[:space:]]*//')
	[ "$reset_e" -eq 1 ] && set -e
	echo "$optimised_size"
}

convert_image_to_format() {
	local FORMAT="$1"
	local SOURCE_FILE="$2"
	local TARGET_FILE="$3"
	local reset_e=0
	internal_is_shell_attribute_set "e" && set +e && reset_e=1
	magick -define preserve-timestamp=true $IMAGICK_STD_DEFINES "$SOURCE_FILE" "${FORMAT}:${TARGET_FILE}"
	touch -r "$SOURCE_FILE" "$TARGET_FILE"
	[ "$reset_e" -eq 1 ] && set -e
}

##############################################################################

main() {
	local THRESHOLD_PERCENT="${THRESHOLD_PERCENT:-10}"
	local OUTPUT_DIR="${OUTPUT_DIR:-}"
	
	# Options
	while getopts "t:o:qvh" option; do
		case "$option" in
			t) THRESHOLD_PERCENT="$OPTARG" ;;
			o) OUTPUT_DIR="$OPTARG" ;;
			q) LOG_LEVEL=$LOG_LEVEL_ERROR ;;
			v) LOG_LEVEL=$LOG_LEVEL_DEBUG ;;
			h) print_usage; exit 1 ;;
			*) print_usage; exit 1 ;;
		esac
	done
	shift $((OPTIND - 1))  # Shift off the options and optional --
	
	if [ -n "$THRESHOLD_PERCENT" ]; then
		case "$THRESHOLD_PERCENT" in
			*[![:digit:]]*)
				print_usage
				exit 1
				;;
		esac
		if [ "$THRESHOLD_PERCENT" -lt 0 ] || [ "$THRESHOLD_PERCENT" -gt 100 ]; then
			print_usage
			exit 1
		fi
	fi
	
	# $# should be at least 1 (the file to optimize), however it may be strictly
	# greater than 1 if multiple files are specified.
	if [ $# -eq 0 ]; then
		print_usage
		exit 1
	fi
	
	if [ -n "$OUTPUT_DIR" ]; then
		mkdir -p "${OUTPUT_DIR}"
	fi
	
	log $LOG_LEVEL_INFO "file_name	size\c"
	for format in $CONST_OUTPUT_FORMATS; do
		log $LOG_LEVEL_INFO "	${format}\c"
	done
	log $LOG_LEVEL_INFO "	best_image_format"
	for file in "$@"; do
		[ -s "$file" ] || continue
		local original_size=$(wc -c < "$file" | sed -e 's/^[[:space:]]*//')
		local best_size=$original_size
		local best_format=
		log $LOG_LEVEL_INFO "${file}	${original_size}\c"
		for format in $CONST_OUTPUT_FORMATS; do
			local optimised_size=$(compute_image_size_for_format "${format}" "${file}")
			log $LOG_LEVEL_INFO "	${optimised_size}\c"
			if [ "${optimised_size}" -lt "${best_size}" ]; then
				best_size=$optimised_size
				if [ "$(expr \( "${original_size}" - "${optimised_size}" \) '*' 100)" -gt "$(expr "${original_size}" '*' "${THRESHOLD_PERCENT}")" ]; then
					best_format=$format
				fi
			fi
		done
		log $LOG_LEVEL_INFO "	${best_format}"
		if [ -n "$best_format" ] && [ -n "$OUTPUT_DIR" ]; then
			local target_file="${OUTPUT_DIR}/$(basename "${file%.*}").$(echo ${best_format} | tr '[:upper:]' '[:lower:]')"
			convert_image_to_format "${best_format}" "${file}" "${target_file}"
		fi
	done
}
main "$@"