DFG METS: overwrite PAGE-XML @imageFilename from image fileGrp

ocrd_fix_imagefilename.py

import os
import click

from ocrd_utils import MIMETYPE_PAGE
from ocrd_models.ocrd_mets import OcrdMets
from ocrd_models.ocrd_page import to_xml
from ocrd_modelfactory import page_from_file

@click.command()
@click.option('-m', '--mets-file', default="mets.xml", help="path to METS of workspace")
@click.option('-I', '--input-file-grp', required=True, help="fileGrp to pick image files from")
@click.option('-O', '--output-file-grp', required=True, help="fileGrp to modify PAGE files from")
def cli(mets_file, input_file_grp, output_file_grp):
    """
    open METS file, for each physical page in the structmap,
    get the local filename of the respective image file in the input fileGrp,
    get the local filename of the respective PAGE-XML file in the output fileGrp,
    open the PAGE-XML file and overwrite its `@imageFilename` with the local path
    of the image file in the METS. Finally serialise the modified PAGE-XML file.
    """
    os.chdir(os.path.dirname(mets_file))
    mets = OcrdMets(filename=os.path.basename(mets_file))
    page2image = {ocrd_file.pageId: ocrd_file.local_filename
                  for ocrd_file in mets.find_all_files(fileGrp=input_file_grp)
                  if ocrd_file.mimetype.startswith('image/')}
    page2PAGE = {ocrd_file.pageId: ocrd_file.local_filename
                 for ocrd_file in mets.find_all_files(fileGrp=output_file_grp)
                 if ocrd_file.mimetype == MIMETYPE_PAGE}
    for page in page2PAGE:
        PAGE = page2PAGE[page]
        img = page2image[page]
        pcgts = page_from_file(PAGE)
        pcgts.get_Page().set_imageFilename(img)
        with open(PAGE, 'w', encoding='utf-8') as f:
            f.write(to_xml(pcgts))

if __name__ == '__main__':
    cli()