"""
HTML Parser using BeautifulSoup
- Written by BroytMann. Copyright (C) 2007, 2008 PhiloSoft Design
+ Written by Broytman. Copyright (C) 2007-2010 PhiloSoft Design
"""
import re
from sgmllib import SGMLParser, SGMLParseError
-from HTMLParser import HTMLParser
from BeautifulSoup import BeautifulSoup, CData
-
-
-class BSoupParser(HTMLParser):
- def __init__(self, charset, meta, title, refresh, icon):
- object.__init__(self)
- self.charset = charset
- self.meta_charset = meta
- self.title = title
- self.refresh = refresh
- self.icon = icon
+from parse_html_util import HTMLParser
# http://groups.google.com/group/beautifulsoup/browse_thread/thread/69093cb0d3a3cf63
title = root.title
if title is None:
- title = ''
- elif title.string:
- title = title.string.encode(_charset)
+ return None
+
+ if title.string:
+ title = title.string
else:
- title = str(title)
+ parts = []
+ for part in title:
+ if not isinstance(part, basestring):
+ part = unicode(part)
+ parts.append(part.strip())
+ title = ''.join(parts)
meta = head.find(_find_contenttype, recursive=False)
if meta:
try:
- __charset = meta.get("content").lower().split('charset=')[1].split(';')[0]
+ meta_content = meta.get("content")
+ if meta_content:
+ __charset = meta_content.lower().split('charset=')[1].split(';')[0]
+ else:
+ __charset = False
except IndexError: # No charset in the META Content-Type
meta_charset = False
else:
else:
meta_charset = False
+ if _charset or meta_charset:
+ title = title.encode(_charset or meta_charset)
+
meta = head.find(_find_refresh, recursive=False)
if meta:
refresh = meta.get("content")
else:
icon = None
- return BSoupParser(_charset, meta_charset, title, refresh, icon)
+ return HTMLParser(_charset, meta_charset, title, refresh, icon)
def _find_contenttype(Tag):
return (Tag.name == "meta") and \