2 HTML Parser using ElementTree+TidyLib.
4 Written by Broytman. Copyright (C) 2010 PhiloSoft Design
7 from elementtidy import TidyHTMLTreeBuilder
8 from parse_html_util import HTMLParser
11 def parse_html(filename, charset=None, log=None):
13 html_tree = TidyHTMLTreeBuilder.parse(filename)
17 XHTML = "{http://www.w3.org/1999/xhtml}"
19 for elem in html_tree.getiterator():
20 if elem.tag.startswith(XHTML):
21 elem.tag = elem.tag[len(XHTML):]
23 title = html_tree.findtext('head/title')
25 title = html_tree.findtext('title')
29 meta = html_tree.findall('head/meta')
31 if m.get('http-equiv', '').lower() == 'content-type':
32 meta_content = m.get("content")
36 meta_content.lower().split('charset=')[1].split(';')[0]
43 if charset or meta_charset:
44 title = title.encode(charset or meta_charset)
47 if m.get('http-equiv', '').lower() == 'refresh':
48 refresh = m.get("content")
53 for link in html_tree.findall('head/link'):
54 if link.get('rel', '').lower() in ('icon', 'shortcut icon'):
55 icon = link.get("href")
60 return HTMLParser(charset, meta_charset, title, refresh, icon)