X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/453698570f26bebd37b39df8537d993b57d77b8b..ff48920ff0d29557ad28d3353d596d7a335eb97b:/youtube_dl/extractor/npo.py diff --git a/youtube_dl/extractor/npo.py b/youtube_dl/extractor/npo.py index 175b145..87f5675 100644 --- a/youtube_dl/extractor/npo.py +++ b/youtube_dl/extractor/npo.py @@ -1,5 +1,7 @@ from __future__ import unicode_literals +import re + from .common import InfoExtractor from ..utils import ( fix_xml_ampersands, @@ -7,7 +9,6 @@ from ..utils import ( qualities, strip_jsonp, unified_strdate, - url_basename, ) @@ -16,13 +17,42 @@ class NPOBaseIE(InfoExtractor): token_page = self._download_webpage( 'http://ida.omroep.nl/npoplayer/i.js', video_id, note='Downloading token') - return self._search_regex( + token = self._search_regex( r'npoplayer\.token = "(.+?)"', token_page, 'token') + # Decryption algorithm extracted from http://npoplayer.omroep.nl/csjs/npoplayer-min.js + token_l = list(token) + first = second = None + for i in range(5, len(token_l) - 4): + if token_l[i].isdigit(): + if first is None: + first = i + elif second is None: + second = i + if first is None or second is None: + first = 12 + second = 13 + + token_l[first], token_l[second] = token_l[second], token_l[first] + + return ''.join(token_l) class NPOIE(NPOBaseIE): - IE_NAME = 'npo.nl' - _VALID_URL = r'https?://www\.npo\.nl/[^/]+/[^/]+/(?P[^/?]+)' + IE_NAME = 'npo' + IE_DESC = 'npo.nl and ntr.nl' + _VALID_URL = r'''(?x) + (?: + npo:| + https?:// + (?:www\.)? + (?: + npo\.nl/(?!live|radio)(?:[^/]+/){2}| + ntr\.nl/(?:[^/]+/){2,}| + omroepwnl\.nl/video/fragment/[^/]+__ + ) + ) + (?P[^/?#]+) + ''' _TESTS = [ { @@ -42,7 +72,7 @@ class NPOIE(NPOBaseIE): 'info_dict': { 'id': 'VARA_101191800', 'ext': 'm4v', - 'title': 'De Mega Mike & Mega Thomas show', + 'title': 'De Mega Mike & Mega Thomas show: The best of.', 'description': 'md5:3b74c97fc9d6901d5a665aac0e5400f4', 'upload_date': '20090227', 'duration': 2400, @@ -54,8 +84,8 @@ class NPOIE(NPOBaseIE): 'info_dict': { 'id': 'VPWON_1169289', 'ext': 'm4v', - 'title': 'Tegenlicht', - 'description': 'md5:d6476bceb17a8c103c76c3b708f05dd1', + 'title': 'Tegenlicht: De toekomst komt uit Afrika', + 'description': 'md5:52cf4eefbc96fffcbdc06d024147abea', 'upload_date': '20130225', 'duration': 3000, }, @@ -84,6 +114,30 @@ class NPOIE(NPOBaseIE): 'title': 'Hoe gaat Europa verder na Parijs?', }, }, + { + 'url': 'http://www.ntr.nl/Aap-Poot-Pies/27/detail/Aap-poot-pies/VPWON_1233944#content', + 'md5': '01c6a2841675995da1f0cf776f03a9c3', + 'info_dict': { + 'id': 'VPWON_1233944', + 'ext': 'm4v', + 'title': 'Aap, poot, pies', + 'description': 'md5:c9c8005d1869ae65b858e82c01a91fde', + 'upload_date': '20150508', + 'duration': 599, + }, + }, + { + 'url': 'http://www.omroepwnl.nl/video/fragment/vandaag-de-dag-verkiezingen__POMS_WNL_853698', + 'md5': 'd30cd8417b8b9bca1fdff27428860d08', + 'info_dict': { + 'id': 'POW_00996502', + 'ext': 'm4v', + 'title': '''"Dit is wel een 'landslide'..."''', + 'description': 'md5:f8d66d537dfb641380226e31ca57b8e8', + 'upload_date': '20150508', + 'duration': 462, + }, + } ] def _real_extract(self, url): @@ -92,12 +146,24 @@ class NPOIE(NPOBaseIE): def _get_info(self, video_id): metadata = self._download_json( - 'http://e.omroep.nl/metadata/aflevering/%s' % video_id, + 'http://e.omroep.nl/metadata/%s' % video_id, video_id, # We have to remove the javascript callback transform_source=strip_jsonp, ) + # For some videos actual video id (prid) is different (e.g. for + # http://www.omroepwnl.nl/video/fragment/vandaag-de-dag-verkiezingen__POMS_WNL_853698 + # video id is POMS_WNL_853698 but prid is POW_00996502) + video_id = metadata.get('prid') or video_id + + # titel is too generic in some cases so utilize aflevering_titel as well + # when available (e.g. http://tegenlicht.vpro.nl/afleveringen/2014-2015/access-to-africa.html) + title = metadata['titel'] + sub_title = metadata.get('aflevering_titel') + if sub_title and sub_title != title: + title += ': %s' % sub_title + token = self._get_token(video_id) formats = [] @@ -123,7 +189,7 @@ class NPOIE(NPOBaseIE): if not video_url: continue if format_id == 'adaptive': - formats.extend(self._extract_m3u8_formats(video_url, video_id)) + formats.extend(self._extract_m3u8_formats(video_url, video_id, 'mp4')) else: formats.append({ 'url': video_url, @@ -161,20 +227,28 @@ class NPOIE(NPOBaseIE): self._sort_formats(formats) + subtitles = {} + if metadata.get('tt888') == 'ja': + subtitles['nl'] = [{ + 'ext': 'vtt', + 'url': 'http://e.omroep.nl/tt888/%s' % video_id, + }] + return { 'id': video_id, - 'title': metadata['titel'], - 'description': metadata['info'], + 'title': title, + 'description': metadata.get('info'), 'thumbnail': metadata.get('images', [{'url': None}])[-1]['url'], 'upload_date': unified_strdate(metadata.get('gidsdatum')), 'duration': parse_duration(metadata.get('tijdsduur')), 'formats': formats, + 'subtitles': subtitles, } class NPOLiveIE(NPOBaseIE): IE_NAME = 'npo.nl:live' - _VALID_URL = r'https?://www\.npo\.nl/live/(?P.+)' + _VALID_URL = r'https?://(?:www\.)?npo\.nl/live/(?P.+)' _TEST = { 'url': 'http://www.npo.nl/live/npo-1', @@ -211,7 +285,8 @@ class NPOLiveIE(NPOBaseIE): if streams: for stream in streams: stream_type = stream.get('type').lower() - if stream_type == 'ss': + # smooth streaming is not supported + if stream_type in ['ss', 'ms']: continue stream_info = self._download_json( 'http://ida.omroep.nl/aapi/?stream=%s&token=%s&type=jsonp' @@ -222,7 +297,10 @@ class NPOLiveIE(NPOBaseIE): stream_url = self._download_json( stream_info['stream'], display_id, 'Downloading %s URL' % stream_type, - transform_source=strip_jsonp) + 'Unable to download %s URL' % stream_type, + transform_source=strip_jsonp, fatal=False) + if not stream_url: + continue if stream_type == 'hds': f4m_formats = self._extract_f4m_formats(stream_url, display_id) # f4m downloader downloads only piece of live stream @@ -234,6 +312,7 @@ class NPOLiveIE(NPOBaseIE): else: formats.append({ 'url': stream_url, + 'preference': -10, }) self._sort_formats(formats) @@ -249,9 +328,119 @@ class NPOLiveIE(NPOBaseIE): } -class TegenlichtVproIE(NPOIE): - IE_NAME = 'tegenlicht.vpro.nl' - _VALID_URL = r'https?://tegenlicht\.vpro\.nl/afleveringen/.*?' +class NPORadioIE(InfoExtractor): + IE_NAME = 'npo.nl:radio' + _VALID_URL = r'https?://(?:www\.)?npo\.nl/radio/(?P[^/]+)/?$' + + _TEST = { + 'url': 'http://www.npo.nl/radio/radio-1', + 'info_dict': { + 'id': 'radio-1', + 'ext': 'mp3', + 'title': 're:^NPO Radio 1 [0-9]{4}-[0-9]{2}-[0-9]{2} [0-9]{2}:[0-9]{2}$', + 'is_live': True, + }, + 'params': { + 'skip_download': True, + } + } + + @staticmethod + def _html_get_attribute_regex(attribute): + return r'{0}\s*=\s*\'([^\']+)\''.format(attribute) + + def _real_extract(self, url): + video_id = self._match_id(url) + + webpage = self._download_webpage(url, video_id) + + title = self._html_search_regex( + self._html_get_attribute_regex('data-channel'), webpage, 'title') + + stream = self._parse_json( + self._html_search_regex(self._html_get_attribute_regex('data-streams'), webpage, 'data-streams'), + video_id) + + codec = stream.get('codec') + + return { + 'id': video_id, + 'url': stream['url'], + 'title': self._live_title(title), + 'acodec': codec, + 'ext': codec, + 'is_live': True, + } + + +class NPORadioFragmentIE(InfoExtractor): + IE_NAME = 'npo.nl:radio:fragment' + _VALID_URL = r'https?://(?:www\.)?npo\.nl/radio/[^/]+/fragment/(?P\d+)' + + _TEST = { + 'url': 'http://www.npo.nl/radio/radio-5/fragment/174356', + 'md5': 'dd8cc470dad764d0fdc70a9a1e2d18c2', + 'info_dict': { + 'id': '174356', + 'ext': 'mp3', + 'title': 'Jubileumconcert Willeke Alberti', + }, + } + + def _real_extract(self, url): + audio_id = self._match_id(url) + + webpage = self._download_webpage(url, audio_id) + + title = self._html_search_regex( + r'href="/radio/[^/]+/fragment/%s" title="([^"]+)"' % audio_id, + webpage, 'title') + + audio_url = self._search_regex( + r"data-streams='([^']+)'", webpage, 'audio url') + + return { + 'id': audio_id, + 'url': audio_url, + 'title': title, + } + + +class SchoolTVIE(InfoExtractor): + IE_NAME = 'schooltv' + _VALID_URL = r'https?://(?:www\.)?schooltv\.nl/video/(?P[^/?#&]+)' + + _TEST = { + 'url': 'http://www.schooltv.nl/video/ademhaling-de-hele-dag-haal-je-adem-maar-wat-gebeurt-er-dan-eigenlijk-in-je-lichaam/', + 'info_dict': { + 'id': 'WO_NTR_429477', + 'display_id': 'ademhaling-de-hele-dag-haal-je-adem-maar-wat-gebeurt-er-dan-eigenlijk-in-je-lichaam', + 'title': 'Ademhaling: De hele dag haal je adem. Maar wat gebeurt er dan eigenlijk in je lichaam?', + 'ext': 'mp4', + 'description': 'md5:abfa0ff690adb73fd0297fd033aaa631' + }, + 'params': { + # Skip because of m3u8 download + 'skip_download': True + } + } + + def _real_extract(self, url): + display_id = self._match_id(url) + webpage = self._download_webpage(url, display_id) + video_id = self._search_regex( + r'data-mid=(["\'])(?P.+?)\1', webpage, 'video_id', group='id') + return { + '_type': 'url_transparent', + 'ie_key': 'NPO', + 'url': 'npo:%s' % video_id, + 'display_id': display_id + } + + +class VPROIE(NPOIE): + IE_NAME = 'vpro' + _VALID_URL = r'https?://(?:www\.)?(?:tegenlicht\.)?vpro\.nl/(?:[^/]+/){2,}(?P[^/]+)\.html' _TESTS = [ { @@ -260,17 +449,72 @@ class TegenlichtVproIE(NPOIE): 'info_dict': { 'id': 'VPWON_1169289', 'ext': 'm4v', - 'title': 'Tegenlicht', - 'description': 'md5:d6476bceb17a8c103c76c3b708f05dd1', + 'title': 'De toekomst komt uit Afrika', + 'description': 'md5:52cf4eefbc96fffcbdc06d024147abea', 'upload_date': '20130225', }, }, + { + 'url': 'http://www.vpro.nl/programmas/2doc/2015/sergio-herman.html', + 'info_dict': { + 'id': 'sergio-herman', + 'title': 'Sergio Herman: Fucking perfect', + }, + 'playlist_count': 2, + }, + { + # playlist with youtube embed + 'url': 'http://www.vpro.nl/programmas/2doc/2015/education-education.html', + 'info_dict': { + 'id': 'education-education', + 'title': '2Doc', + }, + 'playlist_count': 2, + } ] def _real_extract(self, url): - name = url_basename(url) - webpage = self._download_webpage(url, name) - urn = self._html_search_meta('mediaurn', webpage) - info_page = self._download_json( - 'http://rs.vpro.nl/v2/api/media/%s.json' % urn, name) - return self._get_info(info_page['mid']) + playlist_id = self._match_id(url) + + webpage = self._download_webpage(url, playlist_id) + + entries = [ + self.url_result('npo:%s' % video_id if not video_id.startswith('http') else video_id) + for video_id in re.findall(r'data-media-id="([^"]+)"', webpage) + ] + + playlist_title = self._search_regex( + r'\s*([^>]+?)\s*-\s*Teledoc\s*-\s*VPRO\s*', + webpage, 'playlist title', default=None) or self._og_search_title(webpage) + + return self.playlist_result(entries, playlist_id, playlist_title) + + +class WNLIE(InfoExtractor): + _VALID_URL = r'https?://(?:www\.)?omroepwnl\.nl/video/detail/(?P[^/]+)__\d+' + + _TEST = { + 'url': 'http://www.omroepwnl.nl/video/detail/vandaag-de-dag-6-mei__060515', + 'info_dict': { + 'id': 'vandaag-de-dag-6-mei', + 'title': 'Vandaag de Dag 6 mei', + }, + 'playlist_count': 4, + } + + def _real_extract(self, url): + playlist_id = self._match_id(url) + + webpage = self._download_webpage(url, playlist_id) + + entries = [ + self.url_result('npo:%s' % video_id, 'NPO') + for video_id, part in re.findall( + r']+href="([^"]+)"[^>]+class="js-mid"[^>]*>(Deel \d+)', webpage) + ] + + playlist_title = self._html_search_regex( + r'(?s)]+class="subject"[^>]*>(.+?)', + webpage, 'playlist title') + + return self.playlist_result(entries, playlist_id, playlist_title)