diff options
author | Sergey M․ <dstftw@gmail.com> | 2014-12-30 22:28:07 +0600 |
---|---|---|
committer | Sergey M․ <dstftw@gmail.com> | 2014-12-30 22:28:07 +0600 |
commit | a349873226c873a2f3bea9f5dffe167053cb1666 (patch) | |
tree | 0bf9507e1733a2135bc7936223708e3f917efb0c | |
parent | fccae2b911970d0ffa97800b27e70b1937cd3058 (diff) |
[atresplayer] Add extractor (Closes #2341)
-rw-r--r-- | youtube_dl/extractor/__init__.py | 1 | ||||
-rw-r--r-- | youtube_dl/extractor/atresplayer.py | 111 |
2 files changed, 112 insertions, 0 deletions
diff --git a/youtube_dl/extractor/__init__.py b/youtube_dl/extractor/__init__.py index 4f5a1ce18..c15786ad7 100644 --- a/youtube_dl/extractor/__init__.py +++ b/youtube_dl/extractor/__init__.py @@ -25,6 +25,7 @@ from .arte import ( ArteTVDDCIE, ArteTVEmbedIE, ) +from .atresplayer import AtresPlayerIE from .audiomack import AudiomackIE from .auengine import AUEngineIE from .azubu import AzubuIE diff --git a/youtube_dl/extractor/atresplayer.py b/youtube_dl/extractor/atresplayer.py new file mode 100644 index 000000000..7e987b2a0 --- /dev/null +++ b/youtube_dl/extractor/atresplayer.py @@ -0,0 +1,111 @@ +from __future__ import unicode_literals + +import time +import hmac + +from .common import InfoExtractor +from ..utils import ( + compat_str, + compat_urllib_request, + int_or_none, + float_or_none, + xpath_text, + ExtractorError, +) + + +class AtresPlayerIE(InfoExtractor): + _VALID_URL = r'https?://(?:www\.)?atresplayer\.com/television/[^/]+/[^/]+/[^/]+/(?P<id>.+?)_\d+\.html' + _TESTS = [ + { + 'url': 'http://www.atresplayer.com/television/programas/el-club-de-la-comedia/temporada-4/capitulo-10-especial-solidario-nochebuena_2014122100174.html', + 'md5': 'efd56753cda1bb64df52a3074f62e38a', + 'info_dict': { + 'id': 'capitulo-10-especial-solidario-nochebuena', + 'ext': 'mp4', + 'title': 'Especial Solidario de Nochebuena', + 'description': 'md5:e2d52ff12214fa937107d21064075bf1', + 'duration': 5527.6, + 'thumbnail': 're:^https?://.*\.jpg$', + }, + }, + { + 'url': 'http://www.atresplayer.com/television/series/el-secreto-de-puente-viejo/el-chico-de-los-tres-lunares/capitulo-977-29-12-14_2014122400174.html', + 'only_matching': True, + }, + ] + + _USER_AGENT = 'Dalvik/1.6.0 (Linux; U; Android 4.3; GT-I9300 Build/JSS15J' + _MAGIC = 'QWtMLXs414Yo+c#_+Q#K@NN)' + _TIMESTAMP_SHIFT = 30000 + + _TIME_API_URL = 'http://servicios.atresplayer.com/api/admin/time.json' + _URL_VIDEO_TEMPLATE = 'https://servicios.atresplayer.com/api/urlVideo/{1}/{0}/{1}|{2}|{3}.json' + _PLAYER_URL_TEMPLATE = 'https://servicios.atresplayer.com/episode/getplayer.json?episodePk=%s' + _EPISODE_URL_TEMPLATE = 'http://www.atresplayer.com/episodexml/%s' + + def _real_extract(self, url): + video_id = self._match_id(url) + + webpage = self._download_webpage(url, video_id) + + episode_id = self._search_regex( + r'episode="([^"]+)"', webpage, 'episode id') + + timestamp = int_or_none(self._download_webpage( + self._TIME_API_URL, + video_id, 'Downloading timestamp', fatal=False), 1000, time.time()) + timestamp_shifted = compat_str(timestamp + self._TIMESTAMP_SHIFT) + token = hmac.new(self._MAGIC.encode('utf-8'), episode_id + timestamp_shifted).hexdigest() + + formats = [] + for fmt in ['windows', 'android_tablet']: + request = compat_urllib_request.Request( + self._URL_VIDEO_TEMPLATE.format(fmt, episode_id, timestamp_shifted, token)) + request.add_header('Youtubedl-user-agent', self._USER_AGENT) + + fmt_json = self._download_json( + request, video_id, 'Downloading %s video JSON' % fmt) + + result = fmt_json.get('resultDes') + if result.lower() != 'ok': + raise ExtractorError( + '%s returned error: %s' % (self.IE_NAME, result), expected=True) + + for _, video_url in fmt_json['resultObject'].items(): + if video_url.endswith('/Manifest'): + formats.extend(self._extract_f4m_formats(video_url[:-9] + '/manifest.f4m', video_id)) + else: + formats.append({ + 'url': video_url, + 'format_id': 'android', + 'preference': 1, + }) + self._sort_formats(formats) + + player = self._download_json( + self._PLAYER_URL_TEMPLATE % episode_id, + episode_id) + + path_data = player.get('pathData') + + episode = self._download_xml( + self._EPISODE_URL_TEMPLATE % path_data, + video_id, 'Downloading episode XML') + + duration = float_or_none(xpath_text( + episode, './media/asset/info/technical/contentDuration', 'duration')) + + art = episode.find('./media/asset/info/art') + title = xpath_text(art, './name', 'title') + description = xpath_text(art, './description', 'description') + thumbnail = xpath_text(episode, './media/asset/files/background', 'thumbnail') + + return { + 'id': video_id, + 'title': title, + 'description': description, + 'thumbnail': thumbnail, + 'duration': duration, + 'formats': formats, + } |