diff options
author | Philipp Hagemeister <phihag@phihag.de> | 2013-06-23 20:29:46 +0200 |
---|---|---|
committer | Philipp Hagemeister <phihag@phihag.de> | 2013-06-23 20:29:56 +0200 |
commit | 9f4e6bbaeb50fd27f90c799ed8d2531532cfdad7 (patch) | |
tree | e28562cc02028c29eb2182a703b0a511019c1365 /youtube_dl/InfoExtractors.py | |
parent | b05654f0e36278c8e109e3ab591609275093d04e (diff) |
Move gametrailers IE into its own file
Diffstat (limited to 'youtube_dl/InfoExtractors.py')
-rwxr-xr-x | youtube_dl/InfoExtractors.py | 51 |
1 files changed, 1 insertions, 50 deletions
diff --git a/youtube_dl/InfoExtractors.py b/youtube_dl/InfoExtractors.py index acf11a960..999521feb 100755 --- a/youtube_dl/InfoExtractors.py +++ b/youtube_dl/InfoExtractors.py @@ -26,6 +26,7 @@ from .extractor.common import InfoExtractor, SearchInfoExtractor from .extractor.ard import ARDIE from .extractor.arte import ArteTvIE from .extractor.dailymotion import DailymotionIE +from .extractor.gametrailers import GametrailersIE from .extractor.metacafe import MetacafeIE from .extractor.statigram import StatigramIE from .extractor.photobucket import PhotobucketIE @@ -2910,56 +2911,6 @@ class Vbox7IE(InfoExtractor): 'thumbnail': thumbnail_url, }] -class GametrailersIE(InfoExtractor): - _VALID_URL = r'http://www.gametrailers.com/(?P<type>videos|reviews|full-episodes)/(?P<id>.*?)/(?P<title>.*)' - - def _real_extract(self, url): - mobj = re.match(self._VALID_URL, url) - if mobj is None: - raise ExtractorError(u'Invalid URL: %s' % url) - video_id = mobj.group('id') - video_type = mobj.group('type') - webpage = self._download_webpage(url, video_id) - if video_type == 'full-episodes': - mgid_re = r'data-video="(?P<mgid>mgid:.*?)"' - else: - mgid_re = r'data-contentId=\'(?P<mgid>mgid:.*?)\'' - mgid = self._search_regex(mgid_re, webpage, u'mgid') - data = compat_urllib_parse.urlencode({'uri': mgid, 'acceptMethods': 'fms'}) - - info_page = self._download_webpage('http://www.gametrailers.com/feeds/mrss?' + data, - video_id, u'Downloading video info') - links_webpage = self._download_webpage('http://www.gametrailers.com/feeds/mediagen/?' + data, - video_id, u'Downloading video urls info') - - self.report_extraction(video_id) - info_re = r'''<title><!\[CDATA\[(?P<title>.*?)\]\]></title>.* - <description><!\[CDATA\[(?P<description>.*?)\]\]></description>.* - <image>.* - <url>(?P<thumb>.*?)</url>.* - </image>''' - - m_info = re.search(info_re, info_page, re.VERBOSE|re.DOTALL) - if m_info is None: - raise ExtractorError(u'Unable to extract video info') - video_title = m_info.group('title') - video_description = m_info.group('description') - video_thumb = m_info.group('thumb') - - m_urls = list(re.finditer(r'<src>(?P<url>.*)</src>', links_webpage)) - if m_urls is None or len(m_urls) == 0: - raise ExtractError(u'Unable to extrat video url') - # They are sorted from worst to best quality - video_url = m_urls[-1].group('url') - - return {'url': video_url, - 'id': video_id, - 'title': video_title, - # Videos are actually flv not mp4 - 'ext': 'flv', - 'thumbnail': video_thumb, - 'description': video_description, - } def gen_extractors(): """ Return a list of an instance of every supported extractor. |