#!/bin/sh
#
# Run pdftotext on a hierarchy and write to OCR directories
#
# Usage: ocr-all.sh [-no-overwrite] <src-root> 

OVER=1
if [ "$1" = "-no-overwrite" ]; then
	OVER=0
	echo "Not overwriting files"
	shift
fi

SRCDIR=$1
#DSTDIR=$2

if [ ! -d "$SRCDIR" ]; then
	echo "Usage: ocr-all.sh [-no-overwrite] srcdir"
	exit 1
fi

echo "Finding directories to process from: ${SRCDIR}"
#find "$SRCDIR" -type d -name 'images' | while read dir; do
find  "$SRCDIR" -name 'isDataDirectory' -maxdepth 2 | while read dir; do
	#dst=${SRCDIR}`echo $dir | sed "s#$SRCDIR##" | sed "s#FRCapture/images##"`OCR
	#dst=${SRCDIR}`echo $dir | sed "s#$SRCDIR##" | sed "s#/isDataDirectory##"`/output/OCR
	echo $dir
	dir=`echo ${dir} | sed "s#isDataDirectory#output/FRCapture/images#"`
	dst=`echo ${dir} | sed "s#FRCapture/images#OCR#"`
	echo "*** Processing directory: $dir to $dst " `date` " ***"
	(
		echo "*** Processing directory: $dir to $dst " `date` " ***"
		mkdir -p "${dst}"
		find "${dir}" -name '*.pdf' -maxdepth 1 | while read pdf; do
			ptx=`basename $pdf .pdf`.ptx
			if [ ! -f "${dst}/${ptx}" -o $OVER = 1 ]; then
				echo "*** OCR $pdf to $dst/$ptx"
				rm "${dst}/${ptx}*"
				pdftotext -bbox "${pdf}" "${dst}/${ptx}"
			fi
		done 
	) > "${dst}/../_ocr_.log"
done

exit 0