import re from uuid import uuid4 from .common import InfoExtractor from ..compat import compat_HTTPError, compat_str from ..utils import ( ExtractorError, int_or_none, join_nonempty, try_get, url_or_none, urlencode_postdata, ) class ZattooPlatformBaseIE(InfoExtractor): _power_guide_hash = None def _host_url(self): return 'https://%s' % (self._API_HOST if hasattr(self, '_API_HOST') else self._HOST) def _real_initialize(self): if not self._power_guide_hash: self.raise_login_required('An account is needed to access this media', method='password') def _perform_login(self, username, password): try: data = self._download_json( '%s/zapi/v2/account/login' % self._host_url(), None, 'Logging in', data=urlencode_postdata({ 'login': username, 'password': password, 'remember': 'true', }), headers={ 'Referer': '%s/login' % self._host_url(), 'Content-Type': 'application/x-www-form-urlencoded; charset=UTF-8', }) except ExtractorError as e: if isinstance(e.cause, compat_HTTPError) and e.cause.code == 400: raise ExtractorError( 'Unable to login: incorrect username and/or password', expected=True) raise self._power_guide_hash = data['session']['power_guide_hash'] def _initialize_pre_login(self): session_token = self._download_json( f'{self._host_url()}/token.json', None, 'Downloading session token')['session_token'] # Will setup appropriate cookies self._request_webpage( '%s/zapi/v3/session/hello' % self._host_url(), None, 'Opening session', data=urlencode_postdata({ 'uuid': compat_str(uuid4()), 'lang': 'en', 'app_version': '1.8.2', 'format': 'json', 'client_app_token': session_token, })) def _extract_video_id_from_recording(self, recid): playlist = self._download_json( f'{self._host_url()}/zapi/v2/playlist', recid, 'Downloading playlist') try: return next( str(item['program_id']) for item in playlist['recordings'] if item.get('program_id') and str(item.get('id')) == recid) except (StopIteration, KeyError): raise ExtractorError('Could not extract video id from recording') def _extract_cid(self, video_id, channel_name): channel_groups = self._download_json( '%s/zapi/v2/cached/channels/%s' % (self._host_url(), self._power_guide_hash), video_id, 'Downloading channel list', query={'details': False})['channel_groups'] channel_list = [] for chgrp in channel_groups: channel_list.extend(chgrp['channels']) try: return next( chan['cid'] for chan in channel_list if chan.get('cid') and ( chan.get('display_alias') == channel_name or chan.get('cid') == channel_name)) except StopIteration: raise ExtractorError('Could not extract channel id') def _extract_cid_and_video_info(self, video_id): data = self._download_json( '%s/zapi/v2/cached/program/power_details/%s' % ( self._host_url(), self._power_guide_hash), video_id, 'Downloading video information', query={ 'program_ids': video_id, 'complete': True, }) p = data['programs'][0] cid = p['cid'] info_dict = { 'id': video_id, 'title': p.get('t') or p['et'], 'description': p.get('d'), 'thumbnail': p.get('i_url'), 'creator': p.get('channel_name'), 'episode': p.get('et'), 'episode_number': int_or_none(p.get('e_no')), 'season_number': int_or_none(p.get('s_no')), 'release_year': int_or_none(p.get('year')), 'categories': try_get(p, lambda x: x['c'], list), 'tags': try_get(p, lambda x: x['g'], list) } return cid, info_dict def _extract_ondemand_info(self, ondemand_id): """ @returns (ondemand_token, ondemand_type, info_dict) """ data = self._download_json( '%s/zapi/vod/movies/%s' % (self._host_url(), ondemand_id), ondemand_id, 'Downloading ondemand information') info_dict = { 'id': ondemand_id, 'title': data.get('title'), 'description': data.get('description'), 'duration': int_or_none(data.get('duration')), 'release_year': int_or_none(data.get('year')), 'episode_number': int_or_none(data.get('episode_number')), 'season_number': int_or_none(data.get('season_number')), 'categories': try_get(data, lambda x: x['categories'], list), } return data['terms_catalog'][0]['terms'][0]['token'], data['type'], info_dict def _extract_formats(self, cid, video_id, record_id=None, ondemand_id=None, ondemand_termtoken=None, ondemand_type=None, is_live=False): postdata_common = { 'https_watch_urls': True, } if is_live: postdata_common.update({'timeshift': 10800}) url = '%s/zapi/watch/live/%s' % (self._host_url(), cid) elif record_id: url = '%s/zapi/watch/recording/%s' % (self._host_url(), record_id) elif ondemand_id: postdata_common.update({ 'teasable_id': ondemand_id, 'term_token': ondemand_termtoken, 'teasable_type': ondemand_type }) url = '%s/zapi/watch/vod/video' % self._host_url() else: url = '%s/zapi/v3/watch/replay/%s/%s' % (self._host_url(), cid, video_id) formats = [] subtitles = {} for stream_type in ('dash', 'hls7'): postdata = postdata_common.copy() postdata['stream_type'] = stream_type data = self._download_json( url, video_id, 'Downloading %s formats' % stream_type.upper(), data=urlencode_postdata(postdata), fatal=False) if not data: continue watch_urls = try_get( data, lambda x: x['stream']['watch_urls'], list) if not watch_urls: continue for watch in watch_urls: if not isinstance(watch, dict): continue watch_url = url_or_none(watch.get('url')) if not watch_url: continue audio_channel = watch.get('audio_channel') preference = 1 if audio_channel == 'A' else None format_id = join_nonempty(stream_type, watch.get('maxrate'), audio_channel) if stream_type.startswith('dash'): this_formats, subs = self._extract_mpd_formats_and_subtitles( watch_url, video_id, mpd_id=format_id, fatal=False) self._merge_subtitles(subs, target=subtitles) elif stream_type.startswith('hls'): this_formats, subs = self._extract_m3u8_formats_and_subtitles( watch_url, video_id, 'mp4', entry_protocol='m3u8_native', m3u8_id=format_id, fatal=False) self._merge_subtitles(subs, target=subtitles) elif stream_type == 'hds': this_formats = self._extract_f4m_formats( watch_url, video_id, f4m_id=format_id, fatal=False) elif stream_type == 'smooth_playready': this_formats = self._extract_ism_formats( watch_url, video_id, ism_id=format_id, fatal=False) else: assert False for this_format in this_formats: this_format['quality'] = preference formats.extend(this_formats) self._sort_formats(formats) return formats, subtitles def _extract_video(self, video_id, record_id=None): cid, info_dict = self._extract_cid_and_video_info(video_id) info_dict['formats'], info_dict['subtitles'] = self._extract_formats(cid, video_id, record_id=record_id) return info_dict def _extract_live(self, channel_name): cid = self._extract_cid(channel_name, channel_name) formats, subtitles = self._extract_formats(cid, cid, is_live=True) return { 'id': channel_name, 'title': channel_name, 'is_live': True, 'formats': formats, 'subtitles': subtitles } def _extract_record(self, record_id): video_id = self._extract_video_id_from_recording(record_id) cid, info_dict = self._extract_cid_and_video_info(video_id) info_dict['formats'], info_dict['subtitles'] = self._extract_formats(cid, video_id, record_id=record_id) return info_dict def _extract_ondemand(self, ondemand_id): ondemand_termtoken, ondemand_type, info_dict = self._extract_ondemand_info(ondemand_id) info_dict['formats'], info_dict['subtitles'] = self._extract_formats( None, ondemand_id, ondemand_id=ondemand_id, ondemand_termtoken=ondemand_termtoken, ondemand_type=ondemand_type) return info_dict def _real_extract(self, url): video_id, record_id = self._match_valid_url(url).groups() return self._extract_video(video_id, record_id) def _make_valid_url(host): return rf'https?://(?:www\.)?{re.escape(host)}/watch/[^/]+?/(?P[0-9]+)[^/]+(?:/(?P[0-9]+))?' class ZattooBaseIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'zattoo' _HOST = 'zattoo.com' @staticmethod def _create_valid_url(match, qs, base_re=None): match_base = fr'|{base_re}/(?P{match})' if base_re else '(?P)' return rf'''(?x)https?://(?:www\.)?zattoo\.com/(?: [^?#]+\?(?:[^#]+&)?{qs}=(?P{match}) {match_base} )''' def _real_extract(self, url): vid1, vid2 = self._match_valid_url(url).group('vid1', 'vid2') return getattr(self, f'_extract_{self._TYPE}')(vid1 or vid2) class ZattooIE(ZattooBaseIE): _VALID_URL = ZattooBaseIE._create_valid_url(r'\d+', 'program', '(?:program|watch)/[^/]+') _TYPE = 'video' _TESTS = [{ 'url': 'https://zattoo.com/program/zdf/250170418', 'info_dict': { 'id': '250170418', 'ext': 'mp4', 'title': 'Markus Lanz', 'description': 'md5:e41cb1257de008ca62a73bb876ffa7fc', 'thumbnail': 're:http://images.zattic.com/cms/.+/format_480x360.jpg', 'creator': 'ZDF HD', 'release_year': 2022, 'episode': 'Folge 1655', 'categories': 'count:1', 'tags': 'count:2' }, 'params': {'skip_download': 'm3u8'} }, { 'url': 'https://zattoo.com/program/daserste/210177916', 'only_matching': True, }, { 'url': 'https://zattoo.com/guide/german?channel=srf1&program=169860555', 'only_matching': True, }] class ZattooLiveIE(ZattooBaseIE): _VALID_URL = ZattooBaseIE._create_valid_url(r'[^/?&#]+', 'channel', 'live') _TYPE = 'live' _TESTS = [{ 'url': 'https://zattoo.com/channels/german?channel=srf_zwei', 'only_matching': True, }, { 'url': 'https://zattoo.com/live/srf1', 'only_matching': True, }] @classmethod def suitable(cls, url): return False if ZattooIE.suitable(url) else super().suitable(url) class ZattooMoviesIE(ZattooBaseIE): _VALID_URL = ZattooBaseIE._create_valid_url(r'\w+', 'movie_id', 'vod/movies') _TYPE = 'ondemand' _TESTS = [{ 'url': 'https://zattoo.com/vod/movies/7521', 'only_matching': True, }, { 'url': 'https://zattoo.com/ondemand?movie_id=7521&term_token=9f00f43183269484edde', 'only_matching': True, }] class ZattooRecordingsIE(ZattooBaseIE): _VALID_URL = ZattooBaseIE._create_valid_url(r'\d+', 'recording') _TYPE = 'record' _TESTS = [{ 'url': 'https://zattoo.com/recordings?recording=193615508', 'only_matching': True, }, { 'url': 'https://zattoo.com/tc/ptc_recordings_all_recordings?recording=193615420', 'only_matching': True, }] class NetPlusIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'netplus' _HOST = 'netplus.tv' _API_HOST = 'www.%s' % _HOST _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://www.netplus.tv/watch/abc/123-abc', 'only_matching': True, }] class MNetTVIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'mnettv' _HOST = 'tvplus.m-net.de' _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://tvplus.m-net.de/watch/abc/123-abc', 'only_matching': True, }] class WalyTVIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'walytv' _HOST = 'player.waly.tv' _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://player.waly.tv/watch/abc/123-abc', 'only_matching': True, }] class BBVTVIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'bbvtv' _HOST = 'bbv-tv.net' _API_HOST = 'www.%s' % _HOST _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://www.bbv-tv.net/watch/abc/123-abc', 'only_matching': True, }] class VTXTVIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'vtxtv' _HOST = 'vtxtv.ch' _API_HOST = 'www.%s' % _HOST _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://www.vtxtv.ch/watch/abc/123-abc', 'only_matching': True, }] class GlattvisionTVIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'glattvisiontv' _HOST = 'iptv.glattvision.ch' _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://iptv.glattvision.ch/watch/abc/123-abc', 'only_matching': True, }] class SAKTVIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'saktv' _HOST = 'saktv.ch' _API_HOST = 'www.%s' % _HOST _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://www.saktv.ch/watch/abc/123-abc', 'only_matching': True, }] class EWETVIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'ewetv' _HOST = 'tvonline.ewe.de' _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://tvonline.ewe.de/watch/abc/123-abc', 'only_matching': True, }] class QuantumTVIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'quantumtv' _HOST = 'quantum-tv.com' _API_HOST = 'www.%s' % _HOST _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://www.quantum-tv.com/watch/abc/123-abc', 'only_matching': True, }] class OsnatelTVIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'osnateltv' _HOST = 'tvonline.osnatel.de' _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://tvonline.osnatel.de/watch/abc/123-abc', 'only_matching': True, }] class EinsUndEinsTVIE(ZattooPlatformBaseIE): _NETRC_MACHINE = '1und1tv' _HOST = '1und1.tv' _API_HOST = 'www.%s' % _HOST _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://www.1und1.tv/watch/abc/123-abc', 'only_matching': True, }] class SaltTVIE(ZattooPlatformBaseIE): _NETRC_MACHINE = 'salttv' _HOST = 'tv.salt.ch' _VALID_URL = _make_valid_url(_HOST) _TESTS = [{ 'url': 'https://tv.salt.ch/watch/abc/123-abc', 'only_matching': True, }]