ocrd-page2tsv

#!/bin/bash                                                                                                                           
set  -e                                                                                                                               
                                                                                                                                      
ADD_TO_WORKSPACE=true                                                                                                                 
INPUT_FILE_GROUP=OCR-D-OCR-TESS                                                                                                       
OUTPUT_FILE_GROUP=TSV                                                                                                                 
DIRECTORY=$PWD                                                                                                                        
PPN=                                                                                                                                  
IIIF_URL_TEMPLATE="https://content.staatsbibliothek-berlin.de/dc/{{ PPN }}-{{ pageno }}/left,top,width,height/full/0/default.jpg"     
SCALE_FILEGRP=                                                                                                                        
                                                                                                                                      
while [[ $1 = -* ]];do                                                                                                                
        case "$1" in                                                                                                                  
                -I|--input-file-grp) INPUT_FILE_GROUP=$2; shift;;                                                                     
                -O|--output-file-grp) OUTPUT_FILE_GROUP=$2; shift;;                                                                   
                -d|--directory) DIRECTORY=$2; shift;;                                                                                 
                -P|--ppn) PPN=$2; shift;;                                                                                             
                -W|--no-add-to-workspace) ADD_TO_WORKSPACE=false;;                                                                    
                -w|--add-to-workspace) ADD_TO_WORKSPACE=true;;                                                                        
                --scale-by-filegrp) SCALE_FILEGRP=$2; shift;;                                                                         
        esac                                                                                                                          
        shift                                                                                                                         
done                                                                                                                                  
                                                                                                                                      
cd "$DIRECTORY"                                                                                                                       
if [[ -z $PPN ]];then                                                                                                                 
        PPN=$(basename "$DIRECTORY")                                                                                                  
fi                                                                                                                                    
if [[ ! -d "$OUTPUT_FILE_GROUP" ]];then                                                                                               
        mkdir $OUTPUT_FILE_GROUP                                                                                                      
fi                                                                                                                                    
for infile in $INPUT_FILE_GROUP/*;do                                                                                                  
        outfile=$OUTPUT_FILE_GROUP/$(basename $infile)                                                                                
        outfile=${outfile%.xml}.tsv                                                                                                   
        # XXX this does not help, we need IIIF URL                                                                                    
        # img=$(grep -Po 'imageFilename=".*?"' "$infile")                                                                             
        # img=${img:15:-1}                                                                                                            
        scale_factor=1.0                                                                                                              
        if [[ -n "$SCALE_FILEGRP" ]];then                                                                                             
            numeric_part=$(basename $infile|grep -Po '\d+')                                                                           
            compare_file=$(find $SCALE_FILEGRP -name "*$numeric_part*")                                                               
            should_width=$(identify -format '%w' $compare_file)                                                                       
            is_width=$(grep -Po 'imageWidth="[^"]*' $infile|grep -Po '\d+')                                                          
            scale_factor=$(echo "$is_width / $should_width"|bc)                                                                       
        fi                                                                                                                            
        pageno=$(basename $infile|grep -Po '\d\d\d\d'|sed 's,^0*,,')                                                                  
        iiif_url=$(echo $IIIF_URL_TEMPLATE|sed "s/{{ PPN }}/$PPN/"|sed "s/{{ pageno }}/$(printf '%08d' $pageno)/")                    
        page2tsv --purpose OCR --scale-factor $scale_factor --image-url "$iiif_url" $infile $outfile                                  
        if [[ $ADD_TO_WORKSPACE = true ]];then                                                                                        
                ocrd workspace add -i $(basename $outfile) -m text/tsv -G $OUTPUT_FILE_GROUP -g PHYS_$(printf '%04d' $pageno) $outfile
        fi                                                                                                                            
done