toaac

#!/bin/sh

prog=`basename $0`

# check for -f flag to force conversion
force=0

# be verbose in output
verbose=0

# process id of this script
script_pid=$$

# try to make shell's true/false make more sense by using variables
true=0
false=1

outputDir=""

#
# had to read http://wiki.bash-hackers.org/howto/getopts_tutorial to figure out "getopts"!
#
while getopts fhvd: opt; do
	case $opt in
		d)	outputDir="${OPTARG}/"
			if ! mkdir -p "${outputDir}"; then
				echo "Cannot create output dir ${outputDir}"
				exit 1;
			fi ;;

		f) force=1 ;;
		v) verbose=1 ;;

		h)	echo "Usage: $prog # converts all *.mkv files"
			echo "       $prog foo.mkv bar.mkv # only convert specified files"
			echo "       $prog [-f] [-h] [-v]"
			echo "           -f -- force conversion, even if aac already exists (adds another track)"
			echo "           -h -- print this help"
			echo "           -v -- be verbose"
			exit 0
			;;
	esac
done
#
# remove flags from the paramater list
shift $((OPTIND-1))

# for debugging
# set -x

if [ $# = 0 ]; then
	set *.[Mm][Kk][Vv]
	if [ "x$*" = 'x*.mkv' ]; then
		echo "${prog}: No files specified and no files match *.mkv"
		exit 2
	fi
fi

fileCount=0
dirCount=0

## I've been trying various flags to mix aac audio..
ff_aac_opts1="-y -vn -acodec aac -strict experimental -ac 2 -ar 44100 -ab 192k"
ff_aac_opts2="-y -vn -c:a aac -strict experimental -ac 2 -b:a 192k -scodec copy"
ff_aac_opts3='-y -vn -c:a aac -af "pan=stereo|FL=FL+0.5*FC+SL+0.5*LFE|FR=FR+0.5FC+SR+0.5*LFE" -b:a 192k -scodec copy'
## this is the one I'm trying right now
ff_aac_opts="$ff_aac_opts2"

## video options - just copy everything, and mark unknown audio as english
## (should be the aac stream; probably doesn't matter)
ff_mkv_opts="-y -map 0: -map 1: -c copy -flags +global_header -metadata:s:a:0 language=eng"

# check to make sure we have ffmpeg in our path
if which ffmpeg >/dev/null 2>&1; then
	# we can run
	:
else
	echo "$0 requires ffmpeg in its PATH."
	exit 1
fi

# "trash" is a really nice command line tool that moves files to the correct directory
# to have the same effect as dragging files ot the trash, which is nice and undoable!
#
if which trash >/dev/null 2>&1; then
	# yay we can use the trash
	:
else
	echo "Files will not be put in the Trash without the 'trash' command available"
	echo "You will be prompted for every file deletion instead"
fi

# "e" prints out a command and then exexcutes it, filtering out some extra ffmpeg output we don't want
e() {
	if [ $verbose = 1 ]; then
		echo "## `date` -- $@"
	else
		echo "## `date` -- $@" | sed -e "s/$ff_aac_opts//" -e "s/$ff_mkv_opts//"
	fi
	"$@"  </dev/null 2>&1 | egrep -v '(  lib|  config|  built|^ffmpeg|^Press|    encoder|    creation_time|Non-monotonous DTS)'
}

# "del" removes a file using either the nice "trash" command mentioned above, or interactive "rm"
# which at least asks the user if it's ok to delete the file
del() {
	if which trash >/dev/null 2>&1; then
		# use "trash -a" so it works even if you're logged in remotely on the command line
		e trash -a "$@";
	else
		e rm -i "$@";
	fi
}

# check to see if a file already has an aac audio stream.
# Note that shell functions returning "1" for false and "0" for true,
# the opposite of almost all other languages
needs_converting() {
	file="$1"
	has_ac3=0
	has_aac=0
	ffmpeg -i "$file" 2>&1 | grep 'Audio: ac3' > /dev/null && has_ac3=$false
	ffmpeg -i "$file" 2>&1 | grep 'Audio: aac' > /dev/null && has_aac=$false
	needs_it=$false
	[ $has_ac3 = 1 -a $has_aac = 0 ] && needs_it=$true
	[ $force = 1 ] && needs_it=$true
	# echo "# needs_it=$needs_it; has_ac3=$has_ac3, has_aac=$has_aac - $file"
	return $needs_it
}

assert_non_zero_size() {
	file="$1"
	d=($(wc -c "$file"))
	if (( ${d[0]} == 0 )); then
		echo "###!!! file has zero size: $file" >&2
		echo "###!!! aborting.." >&2
		exit 2
	fi
}

larger_than() {
	larger="$1"
	smaller="$2"
	larger_wc=($(wc -c "$larger"))
	larger_size=${larger_wc[0]}
	smaller_wc=($(wc -c "$smaller"))
	smaller_size=${smaller_wc[0]}
	largerMB="$((larger_size/1024/1024)) MB"
	smallerMB="$((smaller_size/1024/1024)) MB"
	if (( larger_size <= smaller_size )); then
		echo "### ❌  file is smaller[$largerMB] than original[$smallerMB]: $larger" >&2
		return $false
	else
		echo "### ✅  file is larger[$largerMB] than original[$smallerMB]: $larger" >&2
		return $true
	fi
}

add_aac_to_file() {
	file="$1"
	[ $verbose = 1 ] && echo "## add_aac_to_file $file"

	# `originalFile` is the new name of the original file,
	# and is the main source for the new version of `file`
	ext="${file##*.}"
	base="${file%.*}"

	# originalFile="${file%.*}-orig.${ext}"
	originalFile="${base}-orig.${ext}"

	# `abortedFile` is used in case of an error
	abortedFile="${base}-aborted.${ext}"

	# rename `file` to `originalFile`
	e mv "$file" "${originalFile}"

	# `aacFile` is the aac-encoded version of the ac3 audio
	# track in `originalFile`
	aacFile="${originalFile%.*}".aac

	#`mkvFile` is the original file with the aac track added to it
	mkvFile="${outputDir}${base}.mkv"

	# create `aacFile` by converting the ac3 track from `originalFile`
	e ffmpeg -i "${originalFile}" $ff_aac_opts "${aacFile}"
	assert_non_zero_size "${aacFile}"

	# Combine the track from `aacFile` and all of `originalFile`,
	# to create `file`
	e ffmpeg -i "${aacFile}" -i "${originalFile}" $ff_mkv_opts "${mkvFile}"
	assert_non_zero_size "${mkvFile}"
	# check that the new file is **larger** than the original, which it **must** be
	if larger_than "${mkvFile}" "${originalFile}"; then
		# all is good, remove now-useless `aacFile` and `originalFile`
		del "${originalFile}" "${aacFile}"
	else
		# restore file to original names, renaming aborted file to `...-aborted...`
		mv "${mkvFile}" "${abortedFile}"
		mv "${originalFile}" "${file}"

		# should it be trashed as well? Hmm ...
		echo "### ❌ aborting.." >&2

		# kill the process instead of just exiting since child processes
		# may be .. still processing
		kill $script_pid
	fi
}

# cpuIndex keeps track of which CPU we are running on; it counts down from
# the # of CPUs and when it hits zero, the script calls "wait" to wait until
# all the CPUs are freed up.
# [In future it would be better to manage this on a CPU-by-CPU basis instead]
#
cpuIndex=1

if which sysctl >/dev/null 2>&1; then
	# Mac-specific way to get the # of physical CPUs
	# (be conservative - "hw.ncpu" rturns max threads which includes non-physical CPUs)
	cpuCount=$(sysctl hw.physicalcpu | awk '{print $2}')
	# less one for zero-based counting
	((cpuCount--))
	cpuIndex=$cpuCount
fi

convert_if_needed() {
	file=$1
	if [ -d "$file" ]; then
		((dirCount++))
		echo "# converting files in $file"
		cd "$file"
		localFiles=(*.[Mm][Kk][Vv])
		if [ "$localFiles" = "*.[Mm][Kk][Vv]" ]; then
			echo "No mkv files in $file"
		else
			for localFile in *.[Mm][Kk][Vv]; do
				convert_if_needed "$localFile" > /dev/null
				echo "## $localFile"
			done
		fi
		echo "# leaving $file"
		echo ""
		cd ..
	else
		((fileCount++))
		[ $verbose = 1 ] && echo "## processing $file"
		if needs_converting "$file"; then
			if ((cpuIndex > 0)); then
				add_aac_to_file "$file" &
				((cpuIndex--))
				# offset the threads a bit to try to make **some** sense of the output
				sleep 2
			else
				add_aac_to_file "$file"
				# wait for any background process to finish
				if [ "x$(jobs)" != x ]; then
					echo "## Waiting for processes: $(jobs -p)"
					wait
				fi
				# start over with all threads
				cpuIndex=$cpuCount
			fi
		else
			echo "does not need converting: $file"
			if [ "x${outputDir}" != "x" ]; then
				mv "$file" "${outputDir}" || echo Cannot move "$file" to "${outputDir}"
			fi
		fi
	fi
}

# iterate over all the files
for file; do
	convert_if_needed "$file"
done

# wait for any background process to finish
if [ "x$(jobs)" != x ]; then
	echo "## Waiting for processes: $(jobs -p)"
	wait
fi

echo "##"
echo "##"
echo "## `date` -- `basename $0`: completed $fileCount files, $dirCount dirs"
echo "##"
echo "##"