ocrd-fileformat-transform

#!/usr/bin/env bash
#set -e
#set -u
set -o pipefail
set -m
# set -x

which ocrd >/dev/null 2>/dev/null || { echo "ocrd not in \$PATH. Panicking"; exit 1; }

SHAREDIR="$(cd "$(dirname "$0")" >/dev/null && pwd)"
SCRIPT_NAME="${0##*/}"

MIMETYPE_PAGE=$(ocrd bashlib constants MIMETYPE_PAGE)
NJOBS=2 # maximum number of concurrent processes (to better utilize CPU, since this is mostly I/O bound)

main () {
    # Load ocrd bashlib functions
    # shellcheck source=../core/ocrd/bashlib/lib.bash
    source $(ocrd bashlib filename)

    # Describe calling script to lib.bash
    ocrd__wrap "$SHAREDIR/ocrd-tool.json" "$SCRIPT_NAME" "$@"
    ocrd__minversion 2.67.0

    cd "${ocrd__argv[working_dir]}"
    mets_basename=$(basename ${ocrd__argv[mets_file]})
    page_id=${ocrd__argv[page_id]:-}
    in_file_grp=${ocrd__argv[input_file_grp]}
    out_file_grp=${ocrd__argv[output_file_grp]}
    overwrite=${ocrd__argv[overwrite]}
 
    # read params
    local from_to script_args out_extension out_mimetype
    script_args=(${params['script-args']:-})
    from_to=(${params['from-to']})
    out_extension=${params['ext']}
    if [[ -z "$out_extension" ]];then
        case "${from_to[1]}" in
            page*|alto*) out_extension=".xml" ;;
            hocr*) out_extension=".html" ;;
            text*) out_extension=".txt" ;;
        esac
    fi
    case "${from_to[1]}" in
        alto*) out_mimetype="application/alto+xml" ;;
        page*) out_mimetype="$MIMETYPE_PAGE" ;;
        hocr*) out_mimetype="text/html" ;;
        text) out_mimetype="text/plain" ;;
    esac

    local results=$(mktemp -d)
    for ((n=0; n<${#ocrd__files[*]}; n++)); do
        if (($(jobs -p | wc -l) > $NJOBS + 1)); then
            wait -n # first wait for next subshell (from other iteration) to finish
        fi
        (
        local in_fpath="$(ocrd__input_file $n local_filename)"
        local in_id="$(ocrd__input_file $n ID)"
        local in_mimetype="$(ocrd__input_file $n mimetype)"
        local in_pageId="$(ocrd__input_file $n pageId)"
        local out_id="$(ocrd__input_file $n outputFileId)"
        local out_fpath="$out_file_grp/${out_id}.xml"

        if [[ "$in_mimetype" =~ ^image/.* ]]; then
            ocrd__log debug "ignoring image $in_id ($in_pageId)"
            exit # continue
        elif ! test -f "${in_fpath#file://}"; then
            ocrd__log error "input file \"$in_fpath\" ID=${in_id} (pageId=${in_pageId}) is not on disk"
            exit # continue
        else
            ocrd__log debug "processing input file $in_id ($in_pageId)"
        fi
        mkdir -p $out_file_grp

        # Actual conversion
        ocrd__log info "${from_to[0]} --> ${from_to[1]}: input file $in_id ($in_pageId)"
        if ! ocr-transform "${from_to[@]}" "$in_fpath" "$out_fpath" -- ${script_args[@]} 2>&1; then
            ocrd__log error "exited with status $retval: ocr-transform ${from_to[*]} $in_fpath $out_fpath -- ${script_args[*]}"
            exit # continue
        elif [ ! -e "$out_fpath" ]; then
            ocrd__log error "Transformation exited with return value 0 but no file was written."
            exit # continue
        fi

        # Add the output files
        ocrd__log info "Successfully executed: ocr-transform ${from_to[*]} $in_fpath $out_fpath -- ${script_args[*]}"

        # Aggregate all METS actions
        echo "$in_pageId $out_id $out_fpath" > $results/$BASHPID # no $$ in subshells
        )&
    done
    wait # wait for all remaining jobs to finish

    # Write resulting METS at once
    declare -a bulk_options
    bulk_options=(-G $out_file_grp -m "$out_mimetype")
    bulk_options+=(-r '(?P<page>[^ ]*) (?P<fid>[^ ]*) (?P<fpath>.*)')
    bulk_options+=(-g '{{ page }}' -i '{{ fid }}' -S '{{ fpath }}')
    if [ "$overwrite" = "true" ];then
        bulk_options+=(--force)
    fi
    declare -a workspace_options
    workspace_options=( -m "$mets_basename" )
    if [[ -n "${ocrd__argv[mets_server_url]}" ]];then
        workspace_options+=( -U "${ocrd__argv[mets_server_url]}" )
    fi
    cat $results/* | sort | ocrd workspace "${workspace_options[@]}" bulk-add "${bulk_options[@]}" -
    rm -fr $results
}


main "$@"