X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/cb32b5fcf6f680ea9caf47919b33779491b79038..0e3c5183a16fb2d7b1ffb838220cc72b7578234e:/youtube_dl/extractor/nrk.py?ds=sidebyside diff --git a/youtube_dl/extractor/nrk.py b/youtube_dl/extractor/nrk.py index 7157e23..9411553 100644 --- a/youtube_dl/extractor/nrk.py +++ b/youtube_dl/extractor/nrk.py @@ -4,12 +4,19 @@ from __future__ import unicode_literals import re from .common import InfoExtractor -from ..compat import compat_urllib_parse_unquote +from ..compat import ( + compat_str, + compat_urllib_parse_unquote, +) from ..utils import ( ExtractorError, int_or_none, + JSON_LD_RE, + js_to_json, + NO_DEFAULT, parse_age_limit, parse_duration, + try_get, ) @@ -39,8 +46,8 @@ class NRKBaseIE(InfoExtractor): entries = [] conviva = data.get('convivaStatistics') or {} - live = (data.get('mediaElementType') == 'Live' or - data.get('isLive') is True or conviva.get('isLive')) + live = (data.get('mediaElementType') == 'Live' + or data.get('isLive') is True or conviva.get('isLive')) def make_title(t): return self._live_title(t) if live else t @@ -99,6 +106,7 @@ class NRKBaseIE(InfoExtractor): MESSAGES = { 'ProgramRightsAreNotReady': 'Du kan dessverre ikke se eller høre programmet', 'ProgramRightsHasExpired': 'Programmet har gått ut', + 'NoProgramRights': 'Ikke tilgjengelig', 'ProgramIsGeoBlocked': 'NRK har ikke rettigheter til å vise dette programmet utenfor Norge', } message_type = data.get('messageType', '') @@ -205,13 +213,13 @@ class NRKIE(NRKBaseIE): _TESTS = [{ # video 'url': 'http://www.nrk.no/video/PS*150533', - 'md5': '2f7f6eeb2aacdd99885f355428715cfa', + 'md5': '706f34cdf1322577589e369e522b50ef', 'info_dict': { 'id': '150533', 'ext': 'mp4', 'title': 'Dompap og andre fugler i Piip-Show', 'description': 'md5:d9261ba34c43b61c812cb6b0269a5c8f', - 'duration': 263, + 'duration': 262, } }, { # audio @@ -242,24 +250,36 @@ class NRKTVIE(NRKBaseIE): _VALID_URL = r'''(?x) https?:// (?:tv|radio)\.nrk(?:super)?\.no/ - (?:serie/[^/]+|program)/ + (?:serie(?:/[^/]+){1,2}|program)/ (?![Ee]pisodes)%s (?:/\d{2}-\d{2}-\d{4})? (?:\#del=(?P\d+))? ''' % _EPISODE_RE _API_HOSTS = ('psapi-ne.nrk.no', 'psapi-we.nrk.no') _TESTS = [{ + 'url': 'https://tv.nrk.no/program/MDDP12000117', + 'md5': '8270824df46ec629b66aeaa5796b36fb', + 'info_dict': { + 'id': 'MDDP12000117AA', + 'ext': 'mp4', + 'title': 'Alarm Trolltunga', + 'description': 'md5:46923a6e6510eefcce23d5ef2a58f2ce', + 'duration': 2223, + 'age_limit': 6, + }, + }, { 'url': 'https://tv.nrk.no/serie/20-spoersmaal-tv/MUHH48000314/23-05-2014', - 'md5': '4e9ca6629f09e588ed240fb11619922a', + 'md5': '9a167e54d04671eb6317a37b7bc8a280', 'info_dict': { 'id': 'MUHH48000314AA', 'ext': 'mp4', 'title': '20 spørsmål 23.05.2014', 'description': 'md5:bdea103bc35494c143c6a9acdd84887a', 'duration': 1741, - 'series': '20 spørsmål - TV', + 'series': '20 spørsmål', 'episode': '23.05.2014', }, + 'skip': 'NoProgramRights', }, { 'url': 'https://tv.nrk.no/program/mdfp15000514', 'info_dict': { @@ -295,7 +315,7 @@ class NRKTVIE(NRKBaseIE): 'id': 'MSPO40010515AH', 'ext': 'mp4', 'title': 'Sprint fri teknikk, kvinner og menn 06.01.2015 (Part 1)', - 'description': 'md5:c03aba1e917561eface5214020551b7a', + 'description': 'md5:1f97a41f05a9486ee00c56f35f82993d', 'duration': 772, 'series': 'Tour de Ski', 'episode': '06.01.2015', @@ -308,7 +328,7 @@ class NRKTVIE(NRKBaseIE): 'id': 'MSPO40010515BH', 'ext': 'mp4', 'title': 'Sprint fri teknikk, kvinner og menn 06.01.2015 (Part 2)', - 'description': 'md5:c03aba1e917561eface5214020551b7a', + 'description': 'md5:1f97a41f05a9486ee00c56f35f82993d', 'duration': 6175, 'series': 'Tour de Ski', 'episode': '06.01.2015', @@ -320,7 +340,7 @@ class NRKTVIE(NRKBaseIE): 'info_dict': { 'id': 'MSPO40010515', 'title': 'Sprint fri teknikk, kvinner og menn 06.01.2015', - 'description': 'md5:c03aba1e917561eface5214020551b7a', + 'description': 'md5:1f97a41f05a9486ee00c56f35f82993d', }, 'expected_warnings': ['Video is geo restricted'], }, { @@ -356,9 +376,233 @@ class NRKTVIE(NRKBaseIE): }, { 'url': 'https://radio.nrk.no/serie/dagsnytt/NPUB21019315/12-07-2015#', 'only_matching': True, + }, { + 'url': 'https://tv.nrk.no/serie/lindmo/2018/MUHU11006318/avspiller', + 'only_matching': True, }] +class NRKTVEpisodeIE(InfoExtractor): + _VALID_URL = r'https?://tv\.nrk\.no/serie/(?P[^/]+/sesong/\d+/episode/\d+)' + _TESTS = [{ + 'url': 'https://tv.nrk.no/serie/hellums-kro/sesong/1/episode/2', + 'info_dict': { + 'id': 'MUHH36005220BA', + 'ext': 'mp4', + 'title': 'Kro, krig og kjærlighet 2:6', + 'description': 'md5:b32a7dc0b1ed27c8064f58b97bda4350', + 'duration': 1563, + 'series': 'Hellums kro', + 'season_number': 1, + 'episode_number': 2, + 'episode': '2:6', + 'age_limit': 6, + }, + 'params': { + 'skip_download': True, + }, + }, { + 'url': 'https://tv.nrk.no/serie/backstage/sesong/1/episode/8', + 'info_dict': { + 'id': 'MSUI14000816AA', + 'ext': 'mp4', + 'title': 'Backstage 8:30', + 'description': 'md5:de6ca5d5a2d56849e4021f2bf2850df4', + 'duration': 1320, + 'series': 'Backstage', + 'season_number': 1, + 'episode_number': 8, + 'episode': '8:30', + }, + 'params': { + 'skip_download': True, + }, + 'skip': 'ProgramRightsHasExpired', + }] + + def _real_extract(self, url): + display_id = self._match_id(url) + + webpage = self._download_webpage(url, display_id) + + nrk_id = self._parse_json( + self._search_regex(JSON_LD_RE, webpage, 'JSON-LD', group='json_ld'), + display_id)['@id'] + + assert re.match(NRKTVIE._EPISODE_RE, nrk_id) + return self.url_result( + 'nrk:%s' % nrk_id, ie=NRKIE.ie_key(), video_id=nrk_id) + + +class NRKTVSerieBaseIE(InfoExtractor): + def _extract_series(self, webpage, display_id, fatal=True): + config = self._parse_json( + self._search_regex( + (r'INITIAL_DATA(?:_V\d)?_*\s*=\s*({.+?})\s*;', + r'({.+?})\s*,\s*"[^"]+"\s*\)\s*'), + webpage, 'config', default='{}' if not fatal else NO_DEFAULT), + display_id, fatal=False, transform_source=js_to_json) + if not config: + return + return try_get( + config, + (lambda x: x['initialState']['series'], lambda x: x['series']), + dict) + + def _extract_seasons(self, seasons): + if not isinstance(seasons, list): + return [] + entries = [] + for season in seasons: + entries.extend(self._extract_episodes(season)) + return entries + + def _extract_episodes(self, season): + if not isinstance(season, dict): + return [] + return self._extract_entries(season.get('episodes')) + + def _extract_entries(self, entry_list): + if not isinstance(entry_list, list): + return [] + entries = [] + for episode in entry_list: + nrk_id = episode.get('prfId') + if not nrk_id or not isinstance(nrk_id, compat_str): + continue + entries.append(self.url_result( + 'nrk:%s' % nrk_id, ie=NRKIE.ie_key(), video_id=nrk_id)) + return entries + + +class NRKTVSeasonIE(NRKTVSerieBaseIE): + _VALID_URL = r'https?://tv\.nrk\.no/serie/[^/]+/sesong/(?P\d+)' + _TEST = { + 'url': 'https://tv.nrk.no/serie/backstage/sesong/1', + 'info_dict': { + 'id': '1', + 'title': 'Sesong 1', + }, + 'playlist_mincount': 30, + } + + @classmethod + def suitable(cls, url): + return (False if NRKTVIE.suitable(url) or NRKTVEpisodeIE.suitable(url) + else super(NRKTVSeasonIE, cls).suitable(url)) + + def _real_extract(self, url): + display_id = self._match_id(url) + + webpage = self._download_webpage(url, display_id) + + series = self._extract_series(webpage, display_id) + + season = next( + s for s in series['seasons'] + if int(display_id) == s.get('seasonNumber')) + + title = try_get(season, lambda x: x['titles']['title'], compat_str) + return self.playlist_result( + self._extract_episodes(season), display_id, title) + + +class NRKTVSeriesIE(NRKTVSerieBaseIE): + _VALID_URL = r'https?://(?:tv|radio)\.nrk(?:super)?\.no/serie/(?P[^/]+)' + _ITEM_RE = r'(?:data-season=["\']|id=["\']season-)(?P\d+)' + _TESTS = [{ + 'url': 'https://tv.nrk.no/serie/blank', + 'info_dict': { + 'id': 'blank', + 'title': 'Blank', + 'description': 'md5:7664b4e7e77dc6810cd3bca367c25b6e', + }, + 'playlist_mincount': 30, + }, { + # new layout, seasons + 'url': 'https://tv.nrk.no/serie/backstage', + 'info_dict': { + 'id': 'backstage', + 'title': 'Backstage', + 'description': 'md5:c3ec3a35736fca0f9e1207b5511143d3', + }, + 'playlist_mincount': 60, + }, { + # new layout, instalments + 'url': 'https://tv.nrk.no/serie/groenn-glede', + 'info_dict': { + 'id': 'groenn-glede', + 'title': 'Grønn glede', + 'description': 'md5:7576e92ae7f65da6993cf90ee29e4608', + }, + 'playlist_mincount': 10, + }, { + # old layout + 'url': 'https://tv.nrksuper.no/serie/labyrint', + 'info_dict': { + 'id': 'labyrint', + 'title': 'Labyrint', + 'description': 'md5:318b597330fdac5959247c9b69fdb1ec', + }, + 'playlist_mincount': 3, + }, { + 'url': 'https://tv.nrk.no/serie/broedrene-dal-og-spektralsteinene', + 'only_matching': True, + }, { + 'url': 'https://tv.nrk.no/serie/saving-the-human-race', + 'only_matching': True, + }, { + 'url': 'https://tv.nrk.no/serie/postmann-pat', + 'only_matching': True, + }] + + @classmethod + def suitable(cls, url): + return ( + False if any(ie.suitable(url) + for ie in (NRKTVIE, NRKTVEpisodeIE, NRKTVSeasonIE)) + else super(NRKTVSeriesIE, cls).suitable(url)) + + def _real_extract(self, url): + series_id = self._match_id(url) + + webpage = self._download_webpage(url, series_id) + + # New layout (e.g. https://tv.nrk.no/serie/backstage) + series = self._extract_series(webpage, series_id, fatal=False) + if series: + title = try_get(series, lambda x: x['titles']['title'], compat_str) + description = try_get( + series, lambda x: x['titles']['subtitle'], compat_str) + entries = [] + entries.extend(self._extract_seasons(series.get('seasons'))) + entries.extend(self._extract_entries(series.get('instalments'))) + entries.extend(self._extract_episodes(series.get('extraMaterial'))) + return self.playlist_result(entries, series_id, title, description) + + # Old layout (e.g. https://tv.nrksuper.no/serie/labyrint) + entries = [ + self.url_result( + 'https://tv.nrk.no/program/Episodes/{series}/{season}'.format( + series=series_id, season=season_id)) + for season_id in re.findall(self._ITEM_RE, webpage) + ] + + title = self._html_search_meta( + 'seriestitle', webpage, + 'title', default=None) or self._og_search_title( + webpage, fatal=False) + if title: + title = self._search_regex( + r'NRK (?:Super )?TV\s*[-–]\s*(.+)', title, 'title', default=title) + + description = self._html_search_meta( + 'series_description', webpage, + 'description', default=None) or self._og_search_description(webpage) + + return self.playlist_result(entries, series_id, title, description) + + class NRKTVDirekteIE(NRKTVIE): IE_DESC = 'NRK TV Direkte and NRK Radio Direkte' _VALID_URL = r'https?://(?:tv|radio)\.nrk\.no/direkte/(?P[^/?#&]+)' @@ -411,7 +655,7 @@ class NRKPlaylistIE(NRKPlaylistBaseIE): 'title': 'Rivertonprisen til Karin Fossum', 'description': 'Første kvinne på 15 år til å vinne krimlitteraturprisen.', }, - 'playlist_count': 5, + 'playlist_count': 2, }] def _extract_title(self, webpage): @@ -438,71 +682,13 @@ class NRKTVEpisodesIE(NRKPlaylistBaseIE): r'

([^<]+)

', webpage, 'title', fatal=False) -class NRKTVSeriesIE(InfoExtractor): - _VALID_URL = r'https?://(?:tv|radio)\.nrk(?:super)?\.no/serie/(?P[^/]+)' - _ITEM_RE = r'(?:data-season=["\']|id=["\']season-)(?P\d+)' - _TESTS = [{ - 'url': 'https://tv.nrk.no/serie/groenn-glede', - 'info_dict': { - 'id': 'groenn-glede', - 'title': 'Grønn glede', - 'description': 'md5:7576e92ae7f65da6993cf90ee29e4608', - }, - 'playlist_mincount': 9, - }, { - 'url': 'http://tv.nrksuper.no/serie/labyrint', - 'info_dict': { - 'id': 'labyrint', - 'title': 'Labyrint', - 'description': 'md5:58afd450974c89e27d5a19212eee7115', - }, - 'playlist_mincount': 3, - }, { - 'url': 'https://tv.nrk.no/serie/broedrene-dal-og-spektralsteinene', - 'only_matching': True, - }, { - 'url': 'https://tv.nrk.no/serie/saving-the-human-race', - 'only_matching': True, - }, { - 'url': 'https://tv.nrk.no/serie/postmann-pat', - 'only_matching': True, - }] - - @classmethod - def suitable(cls, url): - return False if NRKTVIE.suitable(url) else super(NRKTVSeriesIE, cls).suitable(url) - - def _real_extract(self, url): - series_id = self._match_id(url) - - webpage = self._download_webpage(url, series_id) - - entries = [ - self.url_result( - 'https://tv.nrk.no/program/Episodes/{series}/{season}'.format( - series=series_id, season=season_id)) - for season_id in re.findall(self._ITEM_RE, webpage) - ] - - title = self._html_search_meta( - 'seriestitle', webpage, - 'title', default=None) or self._og_search_title( - webpage, fatal=False) - - description = self._html_search_meta( - 'series_description', webpage, - 'description', default=None) or self._og_search_description(webpage) - - return self.playlist_result(entries, series_id, title, description) - - class NRKSkoleIE(InfoExtractor): IE_DESC = 'NRK Skole' _VALID_URL = r'https?://(?:www\.)?nrk\.no/skole/?\?.*\bmediaId=(?P\d+)' _TESTS = [{ 'url': 'https://www.nrk.no/skole/?page=search&q=&mediaId=14099', - 'md5': '6bc936b01f9dd8ed45bc58b252b2d9b6', + 'md5': '18c12c3d071953c3bf8d54ef6b2587b7', 'info_dict': { 'id': '6021', 'ext': 'mp4',