# coding: utf-8 from __future__ import unicode_literals import re from .common import InfoExtractor from ..compat import compat_urllib_parse_urlparse from ..utils import ( determine_ext, ExtractorError, int_or_none, merge_dicts, parse_iso8601, qualities, try_get, urljoin, ) class NDRBaseIE(InfoExtractor): def _real_extract(self, url): mobj = re.match(self._VALID_URL, url) display_id = next(group for group in mobj.groups() if group) webpage = self._download_webpage(url, display_id) return self._extract_embed(webpage, display_id, url) class NDRIE(NDRBaseIE): IE_NAME = 'ndr' IE_DESC = 'NDR.de - Norddeutscher Rundfunk' _VALID_URL = r'https?://(?:\w+\.)?ndr\.de/(?:[^/]+/)*(?P[^/?#]+),[\da-z]+\.html' _TESTS = [{ # httpVideo, same content id 'url': 'http://www.ndr.de/fernsehen/Party-Poette-und-Parade,hafengeburtstag988.html', 'md5': '6515bc255dc5c5f8c85bbc38e035a659', 'info_dict': { 'id': 'hafengeburtstag988', 'display_id': 'Party-Poette-und-Parade', 'ext': 'mp4', 'title': 'Party, Pötte und Parade', 'description': 'md5:ad14f9d2f91d3040b6930c697e5f6b4c', 'uploader': 'ndrtv', 'timestamp': 1431108900, 'upload_date': '20150510', 'duration': 3498, }, 'params': { 'skip_download': True, }, }, { # httpVideo, different content id 'url': 'http://www.ndr.de/sport/fussball/40-Osnabrueck-spielt-sich-in-einen-Rausch,osna270.html', 'md5': '1043ff203eab307f0c51702ec49e9a71', 'info_dict': { 'id': 'osna272', 'display_id': '40-Osnabrueck-spielt-sich-in-einen-Rausch', 'ext': 'mp4', 'title': 'Osnabrück - Wehen Wiesbaden: Die Highlights', 'description': 'md5:32e9b800b3d2d4008103752682d5dc01', 'uploader': 'ndrtv', 'timestamp': 1442059200, 'upload_date': '20150912', 'duration': 510, }, 'params': { 'skip_download': True, }, }, { # httpAudio, same content id 'url': 'http://www.ndr.de/info/La-Valette-entgeht-der-Hinrichtung,audio51535.html', 'md5': 'bb3cd38e24fbcc866d13b50ca59307b8', 'info_dict': { 'id': 'audio51535', 'display_id': 'La-Valette-entgeht-der-Hinrichtung', 'ext': 'mp3', 'title': 'La Valette entgeht der Hinrichtung', 'description': 'md5:22f9541913a40fe50091d5cdd7c9f536', 'uploader': 'ndrinfo', 'timestamp': 1290626100, 'upload_date': '20140729', 'duration': 884, }, 'params': { 'skip_download': True, }, }, { # with subtitles 'url': 'https://www.ndr.de/fernsehen/sendungen/extra_3/extra-3-Satiremagazin-mit-Christian-Ehring,sendung1091858.html', 'info_dict': { 'id': 'extra18674', 'display_id': 'extra-3-Satiremagazin-mit-Christian-Ehring', 'ext': 'mp4', 'title': 'Extra 3 vom 11.11.2020 mit Christian Ehring', 'description': 'md5:42ee53990a715eaaf4dc7f13a3bd56c6', 'uploader': 'ndrtv', 'upload_date': '20201113', 'duration': 1749, 'subtitles': { 'de': [{ 'ext': 'ttml', 'url': r're:^https://www\.ndr\.de.+', }], }, }, 'params': { 'skip_download': True, }, 'expected_warnings': ['Unable to download f4m manifest'], }, { 'url': 'https://www.ndr.de/Fettes-Brot-Ferris-MC-und-Thees-Uhlmann-live-on-stage,festivalsommer116.html', 'only_matching': True, }] def _extract_embed(self, webpage, display_id, url): embed_url = ( self._html_search_meta( 'embedURL', webpage, 'embed URL', default=None) or self._search_regex( r'\bembedUrl["\']\s*:\s*(["\'])(?P(?:(?!\1).)+)\1', webpage, 'embed URL', group='url', default=None) or self._search_regex( r'\bvar\s*sophoraID\s*=\s*(["\'])(?P(?:(?!\1).)+)\1', webpage, 'embed URL', group='url', default='')) # some more work needed if we only found sophoraID if re.match(r'^[a-z]+\d+$', embed_url): # get the initial part of the url path,. eg /panorama/archiv/2022/ parsed_url = compat_urllib_parse_urlparse(url) path = self._search_regex(r'(.+/)%s' % display_id, parsed_url.path or '', 'embed URL', default='') # find tell-tale image with the actual ID ndr_id = self._search_regex(r'%s([a-z]+\d+)(?!\.)\b' % (path, ), webpage, 'embed URL', default=None) # or try to use special knowledge! NDR_INFO_URL_TPL = 'https://www.ndr.de/info/%s-player.html' embed_url = 'ndr:%s' % (ndr_id, ) if ndr_id else NDR_INFO_URL_TPL % (embed_url, ) if not embed_url: raise ExtractorError('Unable to extract embedUrl') description = self._search_regex( r']+itemprop="description">([^<]+)

', webpage, 'description', default=None) or self._og_search_description(webpage) timestamp = parse_iso8601( self._search_regex( (r']+itemprop="(?:datePublished|uploadDate)"[^>]+content="(?P[^"]+)"', r'\bvar\s*pdt\s*=\s*(?P["\'])(?P(?:(?!(?P=q)).)+)(?P=q)', ), webpage, 'upload date', group='cont', default=None)) info = self._search_json_ld(webpage, display_id, default={}) return merge_dicts({ '_type': 'url_transparent', 'url': embed_url, 'display_id': display_id, 'description': description, 'timestamp': timestamp, }, info) class NJoyIE(NDRBaseIE): IE_NAME = 'njoy' IE_DESC = 'N-JOY' _VALID_URL = r'https?://(?:www\.)?n-joy\.de/(?:[^/]+/)*(?:(?P[^/?#]+),)?(?P[\da-z]+)\.html' _TESTS = [{ # httpVideo, same content id 'url': 'http://www.n-joy.de/entertainment/comedy/comedy_contest/Benaissa-beim-NDR-Comedy-Contest,comedycontest2480.html', 'md5': 'cb63be60cd6f9dd75218803146d8dc67', 'info_dict': { 'id': 'comedycontest2480', 'display_id': 'Benaissa-beim-NDR-Comedy-Contest', 'ext': 'mp4', 'title': 'Benaissa beim NDR Comedy Contest', 'description': 'md5:f057a6c4e1c728b10d33b5ffd36ddc39', 'uploader': 'ndrtv', 'upload_date': '20141129', 'duration': 654, }, 'params': { 'skip_download': True, }, }, { # httpVideo, different content id 'url': 'http://www.n-joy.de/musik/Das-frueheste-DJ-Set-des-Nordens-live-mit-Felix-Jaehn-,felixjaehn168.html', 'md5': '417660fffa90e6df2fda19f1b40a64d8', 'info_dict': { 'id': 'dockville882', 'display_id': 'Das-frueheste-DJ-Set-des-Nordens-live-mit-Felix-Jaehn-', 'ext': 'mp4', 'title': '"Ich hab noch nie" mit Felix Jaehn', 'description': 'md5:85dd312d53be1b99e1f998a16452a2f3', 'uploader': 'njoy', 'upload_date': '20150822', 'duration': 211, }, 'params': { 'skip_download': True, }, }, { 'url': 'http://www.n-joy.de/radio/webradio/morningshow209.html', 'only_matching': True, }] def _extract_embed(self, webpage, display_id, url=None): # find tell-tale URL with the actual ID, or ... video_id = self._search_regex( (r'''\bsrc\s*=\s*(?:"|')?(?:/\w+)+/([a-z]+\d+)(?!\.)\b''', r']+id="pp_([\da-z]+)"', ), webpage, 'NDR id', default=None) description = ( self._html_search_meta('description', webpage) or self._search_regex( r']+class="subline"[^>]*>[^<]+\s*

([^<]+)

', webpage, 'description', fatal=False)) return { '_type': 'url_transparent', 'ie_key': 'NDREmbedBase', 'url': 'ndr:%s' % video_id, 'display_id': display_id, 'description': description, 'title': display_id.replace('-', ' ').strip(), } class NDREmbedBaseIE(InfoExtractor): IE_NAME = 'ndr:embed:base' _VALID_URL = r'(?:ndr:(?P[\da-z]+)|https?://www\.ndr\.de/(?P[\da-z]+)-ppjson\.json)' _TESTS = [{ 'url': 'ndr:soundcheck3366', 'only_matching': True, }, { 'url': 'http://www.ndr.de/soundcheck3366-ppjson.json', 'only_matching': True, }] def _real_extract(self, url): mobj = re.match(self._VALID_URL, url) video_id = mobj.group('id') or mobj.group('id_s') ppjson = self._download_json( 'http://www.ndr.de/%s-ppjson.json' % video_id, video_id) playlist = ppjson['playlist'] formats = [] quality_key = qualities(('xs', 's', 'm', 'l', 'xl')) for format_id, f in playlist.items(): src = f.get('src') if not src: continue ext = determine_ext(src, None) if ext == 'f4m': formats.extend(self._extract_f4m_formats( src + '?hdcore=3.7.0&plugin=aasp-3.7.0.39.44', video_id, f4m_id='hds', fatal=False)) elif ext == 'm3u8': formats.extend(self._extract_m3u8_formats( src, video_id, 'mp4', m3u8_id='hls', entry_protocol='m3u8_native', fatal=False)) else: quality = f.get('quality') ff = { 'url': src, 'format_id': quality or format_id, 'quality': quality_key(quality), } type_ = f.get('type') if type_ and type_.split('/')[0] == 'audio': ff['vcodec'] = 'none' ff['ext'] = ext or 'mp3' formats.append(ff) self._sort_formats(formats) config = playlist['config'] live = playlist.get('config', {}).get('streamType') in ['httpVideoLive', 'httpAudioLive'] title = config['title'] if live: title = self._live_title(title) uploader = ppjson.get('config', {}).get('branding') upload_date = ppjson.get('config', {}).get('publicationDate') duration = int_or_none(config.get('duration')) thumbnails = [] poster = try_get(config, lambda x: x['poster'], dict) or {} for thumbnail_id, thumbnail in poster.items(): thumbnail_url = urljoin(url, thumbnail.get('src')) if not thumbnail_url: continue thumbnails.append({ 'id': thumbnail.get('quality') or thumbnail_id, 'url': thumbnail_url, 'preference': quality_key(thumbnail.get('quality')), }) subtitles = {} tracks = config.get('tracks') if tracks and isinstance(tracks, list): for track in tracks: if not isinstance(track, dict): continue track_url = urljoin(url, track.get('src')) if not track_url: continue subtitles.setdefault(track.get('srclang') or 'de', []).append({ 'url': track_url, 'ext': 'ttml', }) return { 'id': video_id, 'title': title, 'is_live': live, 'uploader': uploader if uploader != '-' else None, 'upload_date': upload_date[0:8] if upload_date else None, 'duration': duration, 'thumbnails': thumbnails, 'formats': formats, 'subtitles': subtitles, } class NDREmbedIE(NDREmbedBaseIE): IE_NAME = 'ndr:embed' _VALID_URL = r'https?://(?:www\.)?ndr\.de/(?:[^/]+/)*(?P[\da-z]+)-(?:(?:ard)?player|externalPlayer)\.html' _TESTS = [{ 'url': 'http://www.ndr.de/fernsehen/sendungen/ndr_aktuell/ndraktuell28488-player.html', 'md5': '8b9306142fe65bbdefb5ce24edb6b0a9', 'info_dict': { 'id': 'ndraktuell28488', 'ext': 'mp4', 'title': 'Norddeutschland begrüßt Flüchtlinge', 'is_live': False, 'uploader': 'ndrtv', 'upload_date': '20150907', 'duration': 132, }, }, { 'url': 'http://www.ndr.de/ndr2/events/soundcheck/soundcheck3366-player.html', 'md5': '002085c44bae38802d94ae5802a36e78', 'info_dict': { 'id': 'soundcheck3366', 'ext': 'mp4', 'title': 'Ella Henderson braucht Vergleiche nicht zu scheuen', 'is_live': False, 'uploader': 'ndr2', 'upload_date': '20150912', 'duration': 3554, }, 'params': { 'skip_download': True, }, }, { 'url': 'http://www.ndr.de/info/audio51535-player.html', 'md5': 'bb3cd38e24fbcc866d13b50ca59307b8', 'info_dict': { 'id': 'audio51535', 'ext': 'mp3', 'title': 'La Valette entgeht der Hinrichtung', 'is_live': False, 'uploader': 'ndrinfo', 'upload_date': '20140729', 'duration': 884, }, 'params': { 'skip_download': True, }, }, { 'url': 'http://www.ndr.de/fernsehen/sendungen/visite/visite11010-externalPlayer.html', 'md5': 'ae57f80511c1e1f2fd0d0d3d31aeae7c', 'info_dict': { 'id': 'visite11010', 'ext': 'mp4', 'title': 'Visite - die ganze Sendung', 'is_live': False, 'uploader': 'ndrtv', 'upload_date': '20150902', 'duration': 3525, }, 'params': { 'skip_download': True, }, }, { # httpVideoLive 'url': 'http://www.ndr.de/fernsehen/livestream/livestream217-externalPlayer.html', 'info_dict': { 'id': 'livestream217', 'ext': 'flv', 'title': r're:^NDR Fernsehen Niedersachsen \d{4}-\d{2}-\d{2} \d{2}:\d{2}$', 'is_live': True, 'upload_date': '20150910', }, 'params': { 'skip_download': True, }, }, { 'url': 'http://www.ndr.de/ndrkultur/audio255020-player.html', 'only_matching': True, }, { 'url': 'http://www.ndr.de/fernsehen/sendungen/nordtour/nordtour7124-player.html', 'only_matching': True, }, { 'url': 'http://www.ndr.de/kultur/film/videos/videoimport10424-player.html', 'only_matching': True, }, { 'url': 'http://www.ndr.de/fernsehen/sendungen/hamburg_journal/hamj43006-player.html', 'only_matching': True, }, { 'url': 'http://www.ndr.de/fernsehen/sendungen/weltbilder/weltbilder4518-player.html', 'only_matching': True, }, { 'url': 'http://www.ndr.de/fernsehen/doku952-player.html', 'only_matching': True, }] class NJoyEmbedIE(NDREmbedBaseIE): IE_NAME = 'njoy:embed' _VALID_URL = r'https?://(?:www\.)?n-joy\.de/(?:[^/]+/)*(?P[\da-z]+)-(?:player|externalPlayer)_[^/]+\.html' _TESTS = [{ # httpVideo 'url': 'http://www.n-joy.de/events/reeperbahnfestival/doku948-player_image-bc168e87-5263-4d6d-bd27-bb643005a6de_theme-n-joy.html', 'md5': '8483cbfe2320bd4d28a349d62d88bd74', 'info_dict': { 'id': 'doku948', 'ext': 'mp4', 'title': 'Zehn Jahre Reeperbahn Festival - die Doku', 'is_live': False, 'upload_date': '20150807', 'duration': 1011, }, }, { # httpAudio 'url': 'http://www.n-joy.de/news_wissen/stefanrichter100-player_image-d5e938b1-f21a-4b9a-86b8-aaba8bca3a13_theme-n-joy.html', 'md5': 'd989f80f28ac954430f7b8a48197188a', 'info_dict': { 'id': 'stefanrichter100', 'ext': 'mp3', 'title': 'Interview mit einem Augenzeugen', 'is_live': False, 'uploader': 'njoy', 'upload_date': '20150909', 'duration': 140, }, 'params': { 'skip_download': True, }, }, { # httpAudioLive, no explicit ext 'url': 'http://www.n-joy.de/news_wissen/webradioweltweit100-player_image-3fec0484-2244-4565-8fb8-ed25fd28b173_theme-n-joy.html', 'info_dict': { 'id': 'webradioweltweit100', 'ext': 'mp3', 'title': r're:^N-JOY Weltweit \d{4}-\d{2}-\d{2} \d{2}:\d{2}$', 'is_live': True, 'uploader': 'njoy', 'upload_date': '20150810', }, 'params': { 'skip_download': True, }, }, { 'url': 'http://www.n-joy.de/musik/dockville882-player_image-3905259e-0803-4764-ac72-8b7de077d80a_theme-n-joy.html', 'only_matching': True, }, { 'url': 'http://www.n-joy.de/radio/sendungen/morningshow/urlaubsfotos190-player_image-066a5df1-5c95-49ec-a323-941d848718db_theme-n-joy.html', 'only_matching': True, }, { 'url': 'http://www.n-joy.de/entertainment/comedy/krudetv290-player_image-ab261bfe-51bf-4bf3-87ba-c5122ee35b3d_theme-n-joy.html', 'only_matching': True, }]