From 712c7530ff4acf3b7ae82cc19b2babb1588a3eab Mon Sep 17 00:00:00 2001 From: Yen Chi Hsuan Date: Tue, 26 Jul 2016 11:03:43 +0800 Subject: [PATCH] [mtv] Extract more metadata and more 1. Remove MTVIggyIE. All www.mtviggy.com URLs now redirects to www.mtv.com 2. Fix MTVDEIE 3. Return multiple URLs from _transform_rtmp_url. This is for tosh.cc.com --- youtube_dl/extractor/mtv.py | 58 +++++++++++++++++++------------------ 1 file changed, 30 insertions(+), 28 deletions(-) diff --git a/youtube_dl/extractor/mtv.py b/youtube_dl/extractor/mtv.py index f3ec2ebbc..ba9c28a56 100644 --- a/youtube_dl/extractor/mtv.py +++ b/youtube_dl/extractor/mtv.py @@ -16,6 +16,7 @@ from ..utils import ( HEADRequest, sanitized_Request, strip_or_none, + timeconvert, unescapeHTML, url_basename, RegexNotFoundError, @@ -36,13 +37,13 @@ class MTVServicesInfoExtractor(InfoExtractor): return uri.split(':')[-1] # This was originally implemented for ComedyCentral, but it also works here - @staticmethod - def _transform_rtmp_url(rtmp_video_url): + @classmethod + def _transform_rtmp_url(cls, rtmp_video_url): m = re.match(r'^rtmpe?://.*?/(?Pgsp\..+?/.*)$', rtmp_video_url) if not m: - return rtmp_video_url + return {'rtmp': rtmp_video_url} base = 'http://viacommtvstrmfs.fplive.net/' - return base + m.group('finalid') + return {'http': base + m.group('finalid')} def _get_feed_url(self, uri): return self._FEED_URL @@ -86,14 +87,14 @@ class MTVServicesInfoExtractor(InfoExtractor): rtmp_video_url = rendition.find('./src').text if rtmp_video_url.endswith('siteunavail.png'): continue - new_url = self._transform_rtmp_url(rtmp_video_url) - formats.append({ + new_urls = self._transform_rtmp_url(rtmp_video_url) + formats.extend([{ 'ext': 'flv' if new_url.startswith('rtmp') else ext, 'url': new_url, - 'format_id': rendition.get('bitrate'), + 'format_id': '-'.join([kind, rendition.get('bitrate')]), 'width': int(rendition.get('width')), 'height': int(rendition.get('height')), - }) + } for kind, new_url in new_urls.items()]) except (KeyError, TypeError): raise ExtractorError('Invalid rendition field.') self._sort_formats(formats) @@ -136,6 +137,8 @@ class MTVServicesInfoExtractor(InfoExtractor): description = strip_or_none(xpath_text(itemdoc, 'description')) + timestamp = timeconvert(xpath_text(itemdoc, 'pubDate')) + title_el = None if title_el is None: title_el = find_xpath_attr( @@ -168,6 +171,7 @@ class MTVServicesInfoExtractor(InfoExtractor): 'thumbnail': self._get_thumbnail_url(uri, itemdoc), 'description': description, 'duration': float_or_none(content_el.attrib.get('duration')), + 'timestamp': timestamp, } def _get_feed_query(self, uri): @@ -186,8 +190,13 @@ class MTVServicesInfoExtractor(InfoExtractor): idoc = self._download_xml( url, video_id, 'Downloading info', transform_source=fix_xml_ampersands) + + title = xpath_text(idoc, './channel/title') + description = xpath_text(idoc, './channel/description') + return self.playlist_result( - [self._get_video_info(item) for item in idoc.findall('.//item')]) + [self._get_video_info(item) for item in idoc.findall('.//item')], + playlist_title=title, playlist_description=description) def _extract_mgid(self, webpage): try: @@ -233,6 +242,8 @@ class MTVServicesEmbeddedIE(MTVServicesInfoExtractor): 'ext': 'mp4', 'title': 'Peter Dinklage Sums Up \'Game Of Thrones\' In 45 Seconds', 'description': '"Sexy sexy sexy, stabby stabby stabby, beautiful language," says Peter Dinklage as he tries summarizing "Game of Thrones" in under a minute.', + 'timestamp': 1400126400, + 'upload_date': '20140515', }, } @@ -275,6 +286,8 @@ class MTVIE(MTVServicesInfoExtractor): 'ext': 'mp4', 'title': 'Taylor Swift - "Ours (VH1 Storytellers)"', 'description': 'Album: Taylor Swift performs "Ours" for VH1 Storytellers at Harvey Mudd College.', + 'timestamp': 1352610000, + 'upload_date': '20121111', }, }, ] @@ -301,20 +314,6 @@ class MTVIE(MTVServicesInfoExtractor): return self._get_videos_info(uri) -class MTVIggyIE(MTVServicesInfoExtractor): - IE_NAME = 'mtviggy.com' - _VALID_URL = r'https?://www\.mtviggy\.com/videos/.+' - _TEST = { - 'url': 'http://www.mtviggy.com/videos/arcade-fire-behind-the-scenes-at-the-biggest-music-experiment-yet/', - 'info_dict': { - 'id': '984696', - 'ext': 'mp4', - 'title': 'Arcade Fire: Behind the Scenes at the Biggest Music Experiment Yet', - } - } - _FEED_URL = 'http://all.mtvworldverticals.com/feed-xml/' - - class MTVDEIE(MTVServicesInfoExtractor): IE_NAME = 'mtv.de' _VALID_URL = r'https?://(?:www\.)?mtv\.de/(?:artists|shows|news)/(?:[^/]+/)*(?P\d+)-[^/#?]+/*(?:[#?].*)?$' @@ -322,7 +321,7 @@ class MTVDEIE(MTVServicesInfoExtractor): 'url': 'http://www.mtv.de/artists/10571-cro/videos/61131-traum', 'info_dict': { 'id': 'music_video-a50bc5f0b3aa4b3190aa', - 'ext': 'mp4', + 'ext': 'flv', 'title': 'MusicVideo_cro-traum', 'description': 'Cro - Traum', }, @@ -335,7 +334,7 @@ class MTVDEIE(MTVServicesInfoExtractor): 'url': 'http://www.mtv.de/shows/933-teen-mom-2/staffeln/5353/folgen/63565-enthullungen', 'info_dict': { 'id': 'local_playlist-f5ae778b9832cc837189', - 'ext': 'mp4', + 'ext': 'flv', 'title': 'Episode_teen-mom-2_shows_season-5_episode-1_full-episode_part1', }, 'params': { @@ -343,7 +342,6 @@ class MTVDEIE(MTVServicesInfoExtractor): 'skip_download': True, }, }, { - # single video in pagePlaylist with different id 'url': 'http://www.mtv.de/news/77491-mtv-movies-spotlight-pixels-teil-3', 'info_dict': { 'id': 'local_playlist-4e760566473c4c8c5344', @@ -355,6 +353,7 @@ class MTVDEIE(MTVServicesInfoExtractor): # rtmp download 'skip_download': True, }, + 'skip': 'Das Video kann zur Zeit nicht abgespielt werden.', }] def _real_extract(self, url): @@ -367,11 +366,14 @@ class MTVDEIE(MTVServicesInfoExtractor): r'window\.pagePlaylist\s*=\s*(\[.+?\]);\n', webpage, 'page playlist'), video_id) + def _mrss_url(item): + return item['mrss'] + item.get('mrssvars', '') + # news pages contain single video in playlist with different id if len(playlist) == 1: - return self._get_videos_info_from_url(playlist[0]['mrss'], video_id) + return self._get_videos_info_from_url(_mrss_url(playlist[0]), video_id) for item in playlist: item_id = item.get('id') if item_id and compat_str(item_id) == video_id: - return self._get_videos_info_from_url(item['mrss'], video_id) + return self._get_videos_info_from_url(_mrss_url(item), video_id)