diff options
Diffstat (limited to 'youtube_dl/extractor')
-rw-r--r-- | youtube_dl/extractor/ustream.py | 13 |
1 files changed, 8 insertions, 5 deletions
diff --git a/youtube_dl/extractor/ustream.py b/youtube_dl/extractor/ustream.py index f326163a3..488b10df9 100644 --- a/youtube_dl/extractor/ustream.py +++ b/youtube_dl/extractor/ustream.py @@ -15,11 +15,12 @@ class UstreamIE(InfoExtractor): IE_NAME = 'ustream' _TEST = { 'url': 'http://www.ustream.tv/recorded/20274954', - 'file': '20274954.flv', 'md5': '088f151799e8f572f84eb62f17d73e5c', 'info_dict': { - "uploader": "Young Americans for Liberty", - "title": "Young Americans for Liberty February 7, 2012 2:28 AM", + 'id': '20274954', + 'ext': 'flv', + 'uploader': 'Young Americans for Liberty', + 'title': 'Young Americans for Liberty February 7, 2012 2:28 AM', }, } @@ -27,14 +28,16 @@ class UstreamIE(InfoExtractor): m = re.match(self._VALID_URL, url) video_id = m.group('videoID') - if m.group('type') == 'embed/recorded': # some sites use this embed format (see: http://github.com/rg3/youtube-dl/issues/2990) + # some sites use this embed format (see: http://github.com/rg3/youtube-dl/issues/2990) + if m.group('type') == 'embed/recorded': video_id = m.group('videoID') desktop_url = 'http://www.ustream.tv/recorded/' + video_id return self.url_result(desktop_url, 'Ustream') if m.group('type') == 'embed': video_id = m.group('videoID') webpage = self._download_webpage(url, video_id) - desktop_video_id = self._html_search_regex(r'ContentVideoIds=\["([^"]*?)"\]', webpage, 'desktop_video_id') + desktop_video_id = self._html_search_regex( + r'ContentVideoIds=\["([^"]*?)"\]', webpage, 'desktop_video_id') desktop_url = 'http://www.ustream.tv/recorded/' + desktop_video_id return self.url_result(desktop_url, 'Ustream') |