tools/c10e-html.py

   1 #!/usr/bin/python3
   2 # canonicalize html dirs to ease comaring them
   3 #
   4 # run as:
   5 # ./tools/c10e-html html
   6
   7 import argparse
   8 import glob
   9 import re
  10 import os
  11 import sys
  12
  13 from bs4 import BeautifulSoup
  14
  15
  16 def prettify(filename):
  17     with open(filename, 'r') as doc:
  18         soup = BeautifulSoup(doc.read(), 'lxml')
  19     with open(filename, 'w') as doc:
  20         html = soup.prettify()
  21         # strip things that mkhtml2 is not producing to reduce the diff
  22         html = html.replace('a class="link" href', 'a href')
  23         html = html.replace(' target="_top"', '')
  24         html = html.replace('summary="Navigation header" ', '')
  25         html = html.replace("""   <a name="idx">
  26    </a>
  27 """, '')
  28         html = re.sub("""  <div class="footer">
  29    <hr/>
  30    Generated by GTK-Doc V[.0-9]*
  31   </div>
  32 """, '', html)
  33         html = re.sub(r'\s*<p>\s*</p>', '', html)
  34         html = re.sub(r'\s*<div class="titlepage">\s*</div>', '', html)
  35         html = re.sub(r'\s*<meta content="DocBook[^>]*>', '', html)
  36         html = re.sub(r'\s*<meta content="GTK-Doc[^>]*>', '', html)
  37         doc.write(html)
  38
  39
  40 def main(htmldir):
  41     for filename in glob.glob(os.path.join(htmldir, '*.devhelp2')):
  42         prettify(filename)
  43     for filename in glob.glob(os.path.join(htmldir, '*.html')):
  44         prettify(filename)
  45
  46
  47 if __name__ == '__main__':
  48     parser = argparse.ArgumentParser(
  49         description='c10e-html - canonicalize html files for diffing')
  50     parser.add_argument('args', nargs='*',  help='HTML_DIR')
  51
  52     options = parser.parse_args()
  53     if len(options.args) < 1:
  54         sys.exit('Too few arguments')
  55
  56     main(options.args[0])