X-Git-Url: http://git.bitcoin.ninja/index.cgi?a=blobdiff_plain;f=youtube_dl%2Fextractor%2Ftvp.py;h=3954f0b93812e5bd58d950e8401c954d0ea0a7c6;hb=4c76aa06665621c7689938afd7bbdbc797b5c7ea;hp=5070082da7ba3b34078a01bd214d02a9e8dcac33;hpb=5b6a74856babce30e9e72701259f790322281d3a;p=youtube-dl diff --git a/youtube_dl/extractor/tvp.py b/youtube_dl/extractor/tvp.py index 5070082da..3954f0b93 100644 --- a/youtube_dl/extractor/tvp.py +++ b/youtube_dl/extractor/tvp.py @@ -15,15 +15,16 @@ from ..utils import ( class TVPIE(InfoExtractor): IE_NAME = 'tvp' IE_DESC = 'Telewizja Polska' - _VALID_URL = r'https?://[^/]+\.tvp\.(?:pl|info)/(?:(?!\d+/)[^/]+/)*(?P\d+)' + _VALID_URL = r'https?://[^/]+\.tvp\.(?:pl|info)/(?:video/(?:[^,\s]*,)*|(?:(?!\d+/)[^/]+/)*)(?P\d+)' _TESTS = [{ - 'url': 'http://vod.tvp.pl/194536/i-seria-odc-13', + 'url': 'https://vod.tvp.pl/video/czas-honoru,i-seria-odc-13,194536', 'md5': '8aa518c15e5cc32dfe8db400dc921fbb', 'info_dict': { 'id': '194536', 'ext': 'mp4', 'title': 'Czas honoru, I seria – odc. 13', + 'description': 'md5:381afa5bca72655fe94b05cfe82bf53d', }, }, { 'url': 'http://www.tvp.pl/there-can-be-anything-so-i-shortened-it/17916176', @@ -32,6 +33,17 @@ class TVPIE(InfoExtractor): 'id': '17916176', 'ext': 'mp4', 'title': 'TVP Gorzów pokaże filmy studentów z podroży dookoła świata', + 'description': 'TVP Gorzów pokaże filmy studentów z podroży dookoła świata', + }, + }, { + # page id is not the same as video id(#7799) + 'url': 'https://wiadomosci.tvp.pl/33908820/28092017-1930', + 'md5': '84cd3c8aec4840046e5ab712416b73d0', + 'info_dict': { + 'id': '33908820', + 'ext': 'mp4', + 'title': 'Wiadomości, 28.09.2017, 19:30', + 'description': 'Wydanie główne codziennego serwisu informacyjnego.' }, }, { 'url': 'http://vod.tvp.pl/seriale/obyczajowe/na-sygnale/sezon-2-27-/odc-39/17834272', @@ -53,6 +65,40 @@ class TVPIE(InfoExtractor): 'only_matching': True, }] + def _real_extract(self, url): + page_id = self._match_id(url) + webpage = self._download_webpage(url, page_id) + video_id = self._search_regex([ + r']+src="[^"]*?object_id=(\d+)', + r"object_id\s*:\s*'(\d+)'", + r'data-video-id="(\d+)"'], webpage, 'video id', default=page_id) + return { + '_type': 'url_transparent', + 'url': 'tvp:' + video_id, + 'description': self._og_search_description(webpage, default=None), + 'thumbnail': self._og_search_thumbnail(webpage), + 'ie_key': 'TVPEmbed', + } + + +class TVPEmbedIE(InfoExtractor): + IE_NAME = 'tvp:embed' + IE_DESC = 'Telewizja Polska' + _VALID_URL = r'(?:tvp:|https?://[^/]+\.tvp\.(?:pl|info)/sess/tvplayer\.php\?.*?object_id=)(?P\d+)' + + _TESTS = [{ + 'url': 'http://www.tvp.pl/sess/tvplayer.php?object_id=22670268', + 'md5': '8c9cd59d16edabf39331f93bf8a766c7', + 'info_dict': { + 'id': '22670268', + 'ext': 'mp4', + 'title': 'Panorama, 07.12.2015, 15:40', + }, + }, { + 'url': 'tvp:22670268', + 'only_matching': True, + }] + def _real_extract(self, url): video_id = self._match_id(url) @@ -89,11 +135,14 @@ class TVPIE(InfoExtractor): r'(https?://.+?/video)(?:\.(?:ism|f4m|m3u8)|-\d+\.mp4)', video_url, 'video base url', default=None) if video_url_base: - # TODO: Current DASH formats are broken - $Time$ pattern in - # not implemented yet + # TODO: found instead of in MPD manifest. + # It's not mentioned in MPEG-DASH standard. Figure that out. # formats.extend(self._extract_mpd_formats( # video_url_base + '.ism/video.mpd', # video_id, mpd_id='dash', fatal=False)) + formats.extend(self._extract_ism_formats( + video_url_base + '.ism/Manifest', + video_id, 'mss', fatal=False)) formats.extend(self._extract_f4m_formats( video_url_base + '.ism/video.f4m', video_id, f4m_id='hds', fatal=False)) @@ -102,8 +151,7 @@ class TVPIE(InfoExtractor): 'mp4', 'm3u8_native', m3u8_id='hls', fatal=False) self._sort_formats(m3u8_formats) m3u8_formats = list(filter( - lambda f: f.get('vcodec') != 'none' and f.get('resolution') != 'multiple', - m3u8_formats)) + lambda f: f.get('vcodec') != 'none', m3u8_formats)) formats.extend(m3u8_formats) for i, m3u8_format in enumerate(m3u8_formats, 2): http_url = '%s-%d.mp4' % (video_url_base, i)