1 """HTML Parser using ElementTree+TidyLib.
3 This file is a part of Bookmarks database and Internet robot.
6 __version__ = "$Revision$"[11:-2]
7 __revision__ = "$Id$"[5:-2]
8 __date__ = "$Date$"[7:-2]
9 __author__ = "Oleg Broytman <phd@phdru.name>"
10 __copyright__ = "Copyright (C) 2010, 2011 PhiloSoft Design"
11 __license__ = "GNU GPL"
13 __all__ = ['parse_html']
16 from elementtidy import TidyHTMLTreeBuilder
17 from .bkmk_ph_util import HTMLParser
20 def parse_html(filename, charset=None, log=None):
22 html_tree = TidyHTMLTreeBuilder.parse(filename)
26 XHTML = "{http://www.w3.org/1999/xhtml}"
28 for elem in html_tree.getiterator():
29 if elem.tag.startswith(XHTML):
30 elem.tag = elem.tag[len(XHTML):]
32 title = html_tree.findtext('head/title')
34 title = html_tree.findtext('title')
38 meta = html_tree.findall('head/meta')
40 if m.get('http-equiv', '').lower() == 'content-type':
41 meta_content = m.get("content")
45 meta_content.lower().split('charset=')[1].split(';')[0]
52 if charset or meta_charset:
53 title = title.encode(charset or meta_charset)
56 if m.get('http-equiv', '').lower() == 'refresh':
57 refresh = m.get("content")
62 for link in html_tree.findall('head/link'):
63 if link.get('rel', '').lower() in ('icon', 'shortcut icon'):
64 icon = link.get("href")
69 return HTMLParser(charset, meta_charset, title, refresh, icon)