hocr-merge-dc

#!/usr/bin/env python

import argparse
import re

from lxml import etree, html

dcknown = [
    "dc:title", "dc:creator", "dc:subject", "dc:description", "dc:publisher",
    "dc:contributor", "dc:date", "dc:type", "dc:format", "dc:identifier",
    "dc:source", "dc:language", "dc:relation", "dc:coverage", "dc:rights"
]


def get_text(node):
    textnodes = node.xpath(".//text()")
    s = "".join([text for text in textnodes])
    return re.sub(r'\s+', ' ', s)


parser = argparse.ArgumentParser(
    description="merge Dublin Core metadata into hOCR header files")
parser.add_argument(
    "dc",
    help="XML file with Dublin Core metadata",
    type=argparse.FileType('r'))
parser.add_argument("hocr", help="hOCR file", type=argparse.FileType('r'))
args = parser.parse_args()

dc_doc = etree.parse(args.dc, html.XHTMLParser())
hocr_doc = html.parse(args.hocr)

# remove all existing META tags representing Dublin Core metadata

hocr_meta = hocr_doc.xpath("//HEAD|//head")
assert hocr_meta != []
hocr_meta = hocr_meta[0]

hocr_nodes = hocr_doc.xpath("//head//meta[starts-with(@name,'DC.')]")
for node in hocr_nodes:
    node.getparent().remove(node)

# find all the Dublin Core tags in the Dublin Core metadata

dc_nodes = dc_doc.xpath(
    "//dc:*", namespaces={"dc": "http://purl.org/dc/elements/1.1/"})
for node in dc_nodes:
    nt = re.sub(r'^{http://purl.org/dc/elements/1.1/}', 'dc:', node.tag)
    if nt in dcknown:
        name = re.sub(r'^dc:', 'DC.', nt)
        value = get_text(node)
        value = re.sub("[\t\r\n'\"]", " ", value).strip()
        value = value[:500]
        hnode = etree.Element(
            "meta", nsmap={'DC': 'http://purl.org/dc/elements/1.1'})
        hnode.attrib['name'] = name
        hnode.attrib['content'] = value
        hocr_meta.append(hnode)

print(etree.tostring(hocr_doc, pretty_print=True))