Use unescapeHTML for OpenGraph properties

These are attribute values, so we don't need the more complex and whitespace-destroying cleanHTML - we just need to unescape quotes, that's it.
This commit is contained in:
Philipp Hagemeister 2013-07-17 10:38:23 +02:00
parent 58e7d46d1b
commit f38de77f6e

View file

@ -14,6 +14,7 @@
clean_html, clean_html,
compiled_regex_type, compiled_regex_type,
ExtractorError, ExtractorError,
unescapeHTML,
) )
class InfoExtractor(object): class InfoExtractor(object):
@ -270,7 +271,8 @@ def _og_regex(prop):
def _og_search_property(self, prop, html, name=None, **kargs): def _og_search_property(self, prop, html, name=None, **kargs):
if name is None: if name is None:
name = 'OpenGraph %s' % prop name = 'OpenGraph %s' % prop
return self._html_search_regex(self._og_regex(prop), html, name, flags=re.DOTALL, **kargs) escaped = self._search_regex(self._og_regex(prop), html, name, flags=re.DOTALL, **kargs)
return unescapeHTML(escaped)
def _og_search_thumbnail(self, html, **kargs): def _og_search_thumbnail(self, html, **kargs):
return self._og_search_property('image', html, u'thumbnail url', fatal=False, **kargs) return self._og_search_property('image', html, u'thumbnail url', fatal=False, **kargs)