--- /dev/null
+"""HTML Parser using lxml.html
+
+This file is a part of Bookmarks database and Internet robot.
+"""
+
+__version__ = "$Revision$"[11:-2]
+__revision__ = "$Id$"[5:-2]
+__date__ = "$Date$"[7:-2]
+__author__ = "Oleg Broytman <phd@phdru.name>"
+__copyright__ = "Copyright (C) 2010, 2011 PhiloSoft Design"
+__license__ = "GNU GPL"
+
+__all__ = ['parse_html']
+
+
+from lxml.html import parse
+from .bkmk_ph_util import HTMLParser
+
+
+def parse_html(filename, charset=None, log=None):
+ html_tree = parse(filename)
+
+ if html_tree.getroot() is None:
+ return None
+
+ title = html_tree.findtext('head/title')
+ if title is None:
+ title = html_tree.findtext('title')
+ if title is None:
+ return None
+
+ meta = html_tree.findall('head/meta')
+ for m in meta:
+ if m.get('http-equiv', '').lower() == 'content-type':
+ meta_content = m.get("content")
+ if meta_content:
+ try:
+ meta_charset = \
+ meta_content.lower().split('charset=')[1].split(';')[0]
+ break
+ except IndexError:
+ meta_charset = False
+ else:
+ meta_charset = False
+
+ if charset or meta_charset:
+ title = title.encode(charset or meta_charset)
+
+ for m in meta:
+ if m.get('http-equiv', '').lower() == 'refresh':
+ refresh = m.get("content")
+ break
+ else:
+ refresh = None
+
+ for link in html_tree.findall('head/link'):
+ if link.get('rel', '').lower() in ('icon', 'shortcut icon'):
+ icon = link.get("href")
+ break
+ else:
+ icon = None
+
+ return HTMLParser(charset, meta_charset, title, refresh, icon)