youtube_dl/extractor/nbc.py

   1 from __future__ import unicode_literals
   2
   3 import re
   4
   5 from .common import InfoExtractor
   6 from .theplatform import ThePlatformIE
   7 from .adobepass import AdobePassIE
   8 from ..utils import (
   9     find_xpath_attr,
  10     smuggle_url,
  11     unescapeHTML,
  12     update_url_query,
  13     int_or_none,
  14 )
  15
  16
  17 class NBCIE(AdobePassIE):
  18     _VALID_URL = r'(?P<permalink>https?://(?:www\.)?nbc\.com/[^/]+/video/[^/]+/(?P<id>n?\d+))'
  19
  20     _TESTS = [
  21         {
  22             'url': 'http://www.nbc.com/the-tonight-show/video/jimmy-fallon-surprises-fans-at-ben-jerrys/2848237',
  23             'info_dict': {
  24                 'id': '2848237',
  25                 'ext': 'mp4',
  26                 'title': 'Jimmy Fallon Surprises Fans at Ben & Jerry\'s',
  27                 'description': 'Jimmy gives out free scoops of his new "Tonight Dough" ice cream flavor by surprising customers at the Ben & Jerry\'s scoop shop.',
  28                 'timestamp': 1424246400,
  29                 'upload_date': '20150218',
  30                 'uploader': 'NBCU-COM',
  31             },
  32             'params': {
  33                 # m3u8 download
  34                 'skip_download': True,
  35             },
  36         },
  37         {
  38             'url': 'http://www.nbc.com/saturday-night-live/video/star-wars-teaser/2832821',
  39             'info_dict': {
  40                 'id': '2832821',
  41                 'ext': 'mp4',
  42                 'title': 'Star Wars Teaser',
  43                 'description': 'md5:0b40f9cbde5b671a7ff62fceccc4f442',
  44                 'timestamp': 1417852800,
  45                 'upload_date': '20141206',
  46                 'uploader': 'NBCU-COM',
  47             },
  48             'params': {
  49                 # m3u8 download
  50                 'skip_download': True,
  51             },
  52             'skip': 'Only works from US',
  53         },
  54         {
  55             # HLS streams requires the 'hdnea3' cookie
  56             'url': 'http://www.nbc.com/Kings/video/goliath/n1806',
  57             'info_dict': {
  58                 'id': '101528f5a9e8127b107e98c5e6ce4638',
  59                 'ext': 'mp4',
  60                 'title': 'Goliath',
  61                 'description': 'When an unknown soldier saves the life of the King\'s son in battle, he\'s thrust into the limelight and politics of the kingdom.',
  62                 'timestamp': 1237100400,
  63                 'upload_date': '20090315',
  64                 'uploader': 'NBCU-COM',
  65             },
  66             'params': {
  67                 'skip_download': True,
  68             },
  69             'skip': 'Only works from US',
  70         }
  71     ]
  72
  73     def _real_extract(self, url):
  74         permalink, video_id = re.match(self._VALID_URL, url).groups()
  75         video_data = self._download_json(
  76             'https://api.nbc.com/v3/videos', video_id, query={
  77                 'filter[permalink]': permalink,
  78             })['data'][0]['attributes']
  79         query = {
  80             'mbr': 'true',
  81             'manifest': 'm3u',
  82         }
  83         video_id = video_data['guid']
  84         title = video_data['title']
  85         if video_data.get('entitlement') == 'auth':
  86             resource = self._get_mvpd_resource(
  87                 'nbcentertainment', title, video_id,
  88                 video_data.get('vChipRating'))
  89             query['auth'] = self._extract_mvpd_auth(
  90                 url, video_id, 'nbcentertainment', resource)
  91         theplatform_url = smuggle_url(update_url_query(
  92             'http://link.theplatform.com/s/NnzsPC/media/guid/2410887629/' + video_id,
  93             query), {'force_smil_url': True})
  94         return {
  95             '_type': 'url_transparent',
  96             'id': video_id,
  97             'title': title,
  98             'url': theplatform_url,
  99             'description': video_data.get('description'),
 100             'keywords': video_data.get('keywords'),
 101             'season_number': int_or_none(video_data.get('seasonNumber')),
 102             'episode_number': int_or_none(video_data.get('episodeNumber')),
 103             'series': video_data.get('showName'),
 104             'ie_key': 'ThePlatform',
 105         }
 106
 107
 108 class NBCSportsVPlayerIE(InfoExtractor):
 109     _VALID_URL = r'https?://vplayer\.nbcsports\.com/(?:[^/]+/)+(?P<id>[0-9a-zA-Z_]+)'
 110
 111     _TESTS = [{
 112         'url': 'https://vplayer.nbcsports.com/p/BxmELC/nbcsports_share/select/9CsDKds0kvHI',
 113         'info_dict': {
 114             'id': '9CsDKds0kvHI',
 115             'ext': 'flv',
 116             'description': 'md5:df390f70a9ba7c95ff1daace988f0d8d',
 117             'title': 'Tyler Kalinoski hits buzzer-beater to lift Davidson',
 118             'timestamp': 1426270238,
 119             'upload_date': '20150313',
 120             'uploader': 'NBCU-SPORTS',
 121         }
 122     }, {
 123         'url': 'http://vplayer.nbcsports.com/p/BxmELC/nbc_embedshare/select/_hqLjQ95yx8Z',
 124         'only_matching': True,
 125     }]
 126
 127     @staticmethod
 128     def _extract_url(webpage):
 129         iframe_m = re.search(
 130             r'<iframe[^>]+src="(?P<url>https?://vplayer\.nbcsports\.com/[^"]+)"', webpage)
 131         if iframe_m:
 132             return iframe_m.group('url')
 133
 134     def _real_extract(self, url):
 135         video_id = self._match_id(url)
 136         webpage = self._download_webpage(url, video_id)
 137         theplatform_url = self._og_search_video_url(webpage)
 138         return self.url_result(theplatform_url, 'ThePlatform')
 139
 140
 141 class NBCSportsIE(InfoExtractor):
 142     # Does not include https because its certificate is invalid
 143     _VALID_URL = r'https?://(?:www\.)?nbcsports\.com//?(?:[^/]+/)+(?P<id>[0-9a-z-]+)'
 144
 145     _TEST = {
 146         'url': 'http://www.nbcsports.com//college-basketball/ncaab/tom-izzo-michigan-st-has-so-much-respect-duke',
 147         'info_dict': {
 148             'id': 'PHJSaFWbrTY9',
 149             'ext': 'flv',
 150             'title': 'Tom Izzo, Michigan St. has \'so much respect\' for Duke',
 151             'description': 'md5:ecb459c9d59e0766ac9c7d5d0eda8113',
 152             'uploader': 'NBCU-SPORTS',
 153             'upload_date': '20150330',
 154             'timestamp': 1427726529,
 155         }
 156     }
 157
 158     def _real_extract(self, url):
 159         video_id = self._match_id(url)
 160         webpage = self._download_webpage(url, video_id)
 161         return self.url_result(
 162             NBCSportsVPlayerIE._extract_url(webpage), 'NBCSportsVPlayer')
 163
 164
 165 class CSNNEIE(InfoExtractor):
 166     _VALID_URL = r'https?://(?:www\.)?csnne\.com/video/(?P<id>[0-9a-z-]+)'
 167
 168     _TEST = {
 169         'url': 'http://www.csnne.com/video/snc-evening-update-wright-named-red-sox-no-5-starter',
 170         'info_dict': {
 171             'id': 'yvBLLUgQ8WU0',
 172             'ext': 'mp4',
 173             'title': 'SNC evening update: Wright named Red Sox\' No. 5 starter.',
 174             'description': 'md5:1753cfee40d9352b19b4c9b3e589b9e3',
 175             'timestamp': 1459369979,
 176             'upload_date': '20160330',
 177             'uploader': 'NBCU-SPORTS',
 178         }
 179     }
 180
 181     def _real_extract(self, url):
 182         display_id = self._match_id(url)
 183         webpage = self._download_webpage(url, display_id)
 184         return {
 185             '_type': 'url_transparent',
 186             'ie_key': 'ThePlatform',
 187             'url': self._html_search_meta('twitter:player:stream', webpage),
 188             'display_id': display_id,
 189         }
 190
 191
 192 class NBCNewsIE(ThePlatformIE):
 193     _VALID_URL = r'''(?x)https?://(?:www\.)?(?:nbcnews|today|msnbc)\.com/
 194         (?:video/.+?/(?P<id>\d+)|
 195         ([^/]+/)*(?:.*-)?(?P<mpx_id>[^/?]+))
 196         '''
 197
 198     _TESTS = [
 199         {
 200             'url': 'http://www.nbcnews.com/video/nbc-news/52753292',
 201             'md5': '47abaac93c6eaf9ad37ee6c4463a5179',
 202             'info_dict': {
 203                 'id': '52753292',
 204                 'ext': 'flv',
 205                 'title': 'Crew emerges after four-month Mars food study',
 206                 'description': 'md5:24e632ffac72b35f8b67a12d1b6ddfc1',
 207             },
 208         },
 209         {
 210             'url': 'http://www.nbcnews.com/watch/nbcnews-com/how-twitter-reacted-to-the-snowden-interview-269389891880',
 211             'md5': 'af1adfa51312291a017720403826bb64',
 212             'info_dict': {
 213                 'id': 'p_tweet_snow_140529',
 214                 'ext': 'mp4',
 215                 'title': 'How Twitter Reacted To The Snowden Interview',
 216                 'description': 'md5:65a0bd5d76fe114f3c2727aa3a81fe64',
 217                 'uploader': 'NBCU-NEWS',
 218                 'timestamp': 1401363060,
 219                 'upload_date': '20140529',
 220             },
 221         },
 222         {
 223             'url': 'http://www.nbcnews.com/feature/dateline-full-episodes/full-episode-family-business-n285156',
 224             'md5': 'fdbf39ab73a72df5896b6234ff98518a',
 225             'info_dict': {
 226                 'id': '529953347624',
 227                 'ext': 'mp4',
 228                 'title': 'FULL EPISODE: Family Business',
 229                 'description': 'md5:757988edbaae9d7be1d585eb5d55cc04',
 230             },
 231             'skip': 'This page is unavailable.',
 232         },
 233         {
 234             'url': 'http://www.nbcnews.com/nightly-news/video/nightly-news-with-brian-williams-full-broadcast-february-4-394064451844',
 235             'md5': '73135a2e0ef819107bbb55a5a9b2a802',
 236             'info_dict': {
 237                 'id': 'nn_netcast_150204',
 238                 'ext': 'mp4',
 239                 'title': 'Nightly News with Brian Williams Full Broadcast (February 4)',
 240                 'description': 'md5:1c10c1eccbe84a26e5debb4381e2d3c5',
 241                 'timestamp': 1423104900,
 242                 'uploader': 'NBCU-NEWS',
 243                 'upload_date': '20150205',
 244             },
 245         },
 246         {
 247             'url': 'http://www.nbcnews.com/business/autos/volkswagen-11-million-vehicles-could-have-suspect-software-emissions-scandal-n431456',
 248             'md5': 'a49e173825e5fcd15c13fc297fced39d',
 249             'info_dict': {
 250                 'id': 'x_lon_vwhorn_150922',
 251                 'ext': 'mp4',
 252                 'title': 'Volkswagen U.S. Chief:\xa0 We Have Totally Screwed Up',
 253                 'description': 'md5:c8be487b2d80ff0594c005add88d8351',
 254                 'upload_date': '20150922',
 255                 'timestamp': 1442917800,
 256                 'uploader': 'NBCU-NEWS',
 257             },
 258         },
 259         {
 260             'url': 'http://www.today.com/video/see-the-aurora-borealis-from-space-in-stunning-new-nasa-video-669831235788',
 261             'md5': '118d7ca3f0bea6534f119c68ef539f71',
 262             'info_dict': {
 263                 'id': 'tdy_al_space_160420',
 264                 'ext': 'mp4',
 265                 'title': 'See the aurora borealis from space in stunning new NASA video',
 266                 'description': 'md5:74752b7358afb99939c5f8bb2d1d04b1',
 267                 'upload_date': '20160420',
 268                 'timestamp': 1461152093,
 269                 'uploader': 'NBCU-NEWS',
 270             },
 271         },
 272         {
 273             'url': 'http://www.msnbc.com/all-in-with-chris-hayes/watch/the-chaotic-gop-immigration-vote-314487875924',
 274             'md5': '6d236bf4f3dddc226633ce6e2c3f814d',
 275             'info_dict': {
 276                 'id': 'n_hayes_Aimm_140801_272214',
 277                 'ext': 'mp4',
 278                 'title': 'The chaotic GOP immigration vote',
 279                 'description': 'The Republican House votes on a border bill that has no chance of getting through the Senate or signed by the President and is drawing criticism from all sides.',
 280                 'thumbnail': r're:^https?://.*\.jpg$',
 281                 'timestamp': 1406937606,
 282                 'upload_date': '20140802',
 283                 'uploader': 'NBCU-NEWS',
 284             },
 285         },
 286         {
 287             'url': 'http://www.nbcnews.com/watch/dateline/full-episode--deadly-betrayal-386250819952',
 288             'only_matching': True,
 289         },
 290         {
 291             # From http://www.vulture.com/2016/06/letterman-couldnt-care-less-about-late-night.html
 292             'url': 'http://www.nbcnews.com/widget/video-embed/701714499682',
 293             'only_matching': True,
 294         },
 295     ]
 296
 297     def _real_extract(self, url):
 298         mobj = re.match(self._VALID_URL, url)
 299         video_id = mobj.group('id')
 300         if video_id is not None:
 301             all_info = self._download_xml('http://www.nbcnews.com/id/%s/displaymode/1219' % video_id, video_id)
 302             info = all_info.find('video')
 303
 304             return {
 305                 'id': video_id,
 306                 'title': info.find('headline').text,
 307                 'ext': 'flv',
 308                 'url': find_xpath_attr(info, 'media', 'type', 'flashVideo').text,
 309                 'description': info.find('caption').text,
 310                 'thumbnail': find_xpath_attr(info, 'media', 'type', 'thumbnail').text,
 311             }
 312         else:
 313             # "feature" and "nightly-news" pages use theplatform.com
 314             video_id = mobj.group('mpx_id')
 315             webpage = self._download_webpage(url, video_id)
 316
 317             filter_param = 'byId'
 318             bootstrap_json = self._search_regex(
 319                 [r'(?m)(?:var\s+(?:bootstrapJson|playlistData)|NEWS\.videoObj)\s*=\s*({.+});?\s*$',
 320                  r'videoObj\s*:\s*({.+})', r'data-video="([^"]+)"',
 321                  r'jQuery\.extend\(Drupal\.settings\s*,\s*({.+?})\);'],
 322                 webpage, 'bootstrap json', default=None)
 323             if bootstrap_json:
 324                 bootstrap = self._parse_json(
 325                     bootstrap_json, video_id, transform_source=unescapeHTML)
 326
 327                 info = None
 328                 if 'results' in bootstrap:
 329                     info = bootstrap['results'][0]['video']
 330                 elif 'video' in bootstrap:
 331                     info = bootstrap['video']
 332                 elif 'msnbcVideoInfo' in bootstrap:
 333                     info = bootstrap['msnbcVideoInfo']['meta']
 334                 elif 'msnbcThePlatform' in bootstrap:
 335                     info = bootstrap['msnbcThePlatform']['videoPlayer']['video']
 336                 else:
 337                     info = bootstrap
 338
 339                 if 'guid' in info:
 340                     video_id = info['guid']
 341                     filter_param = 'byGuid'
 342                 elif 'mpxId' in info:
 343                     video_id = info['mpxId']
 344
 345             return {
 346                 '_type': 'url_transparent',
 347                 'id': video_id,
 348                 # http://feed.theplatform.com/f/2E2eJC/nbcnews also works
 349                 'url': update_url_query('http://feed.theplatform.com/f/2E2eJC/nnd_NBCNews', {filter_param: video_id}),
 350                 'ie_key': 'ThePlatformFeed',
 351             }
 352
 353
 354 class NBCOlympicsIE(InfoExtractor):
 355     _VALID_URL = r'https?://www\.nbcolympics\.com/video/(?P<id>[a-z-]+)'
 356
 357     _TEST = {
 358         # Geo-restricted to US
 359         'url': 'http://www.nbcolympics.com/video/justin-roses-son-leo-was-tears-after-his-dad-won-gold',
 360         'md5': '54fecf846d05429fbaa18af557ee523a',
 361         'info_dict': {
 362             'id': 'WjTBzDXx5AUq',
 363             'display_id': 'justin-roses-son-leo-was-tears-after-his-dad-won-gold',
 364             'ext': 'mp4',
 365             'title': 'Rose\'s son Leo was in tears after his dad won gold',
 366             'description': 'Olympic gold medalist Justin Rose gets emotional talking to the impact his win in men\'s golf has already had on his children.',
 367             'timestamp': 1471274964,
 368             'upload_date': '20160815',
 369             'uploader': 'NBCU-SPORTS',
 370         },
 371     }
 372
 373     def _real_extract(self, url):
 374         display_id = self._match_id(url)
 375
 376         webpage = self._download_webpage(url, display_id)
 377
 378         drupal_settings = self._parse_json(self._search_regex(
 379             r'jQuery\.extend\(Drupal\.settings\s*,\s*({.+?})\);',
 380             webpage, 'drupal settings'), display_id)
 381
 382         iframe_url = drupal_settings['vod']['iframe_url']
 383         theplatform_url = iframe_url.replace(
 384             'vplayer.nbcolympics.com', 'player.theplatform.com')
 385
 386         return {
 387             '_type': 'url_transparent',
 388             'url': theplatform_url,
 389             'ie_key': ThePlatformIE.ie_key(),
 390             'display_id': display_id,
 391         }