tei2textpages.py

# tei2textpages.py
import lxml.etree as ET
import sys
import re

tei_file = sys.argv[1]

with open(tei_file, 'r') as f:
    tei_xml = f.read()
    tei_xml = re.sub(r'<\?xml version="1.0" encoding="UTF-8"\?>', '', tei_xml)
    tei_xml = re.sub(r'<teiHeader>.*</teiHeader>', '', tei_xml)
    # tei_xml = re.sub(r'<pb ?facs="#([^"]*)"? ?[^>]*>', r'!!FORMFEED!!\1', tei_xml)
    tei_xml = re.sub(r'<pb[^>]*>', '!!FORMFEED!!', tei_xml)
    tei_str = ''.join(ET.fromstring(tei_xml).itertext())
    for n, chunk in enumerate(tei_str.split('!!FORMFEED!!')):
        with open('%s_%04d.txt' % (tei_file, n), 'w') as outf:
            outf.write(chunk)