yt_dlp/extractor/mlb.py

   1 import json
   2 import re
   3 import time
   4 import uuid
   5
   6 from .common import InfoExtractor
   7 from ..networking.exceptions import HTTPError
   8 from ..utils import (
   9     ExtractorError,
  10     determine_ext,
  11     int_or_none,
  12     join_nonempty,
  13     jwt_decode_hs256,
  14     parse_duration,
  15     parse_iso8601,
  16     try_get,
  17     url_or_none,
  18     urlencode_postdata,
  19 )
  20 from ..utils.traversal import traverse_obj
  21
  22
  23 class MLBBaseIE(InfoExtractor):
  24     def _real_extract(self, url):
  25         display_id = self._match_id(url)
  26         video = self._download_video_data(display_id)
  27         video_id = video['id']
  28         title = video['title']
  29         feed = self._get_feed(video)
  30
  31         formats = []
  32         for playback in (feed.get('playbacks') or []):
  33             playback_url = playback.get('url')
  34             if not playback_url:
  35                 continue
  36             name = playback.get('name')
  37             ext = determine_ext(playback_url)
  38             if ext == 'm3u8':
  39                 formats.extend(self._extract_m3u8_formats(
  40                     playback_url, video_id, 'mp4',
  41                     'm3u8_native', m3u8_id=name, fatal=False))
  42             else:
  43                 f = {
  44                     'format_id': name,
  45                     'url': playback_url,
  46                 }
  47                 mobj = re.search(r'_(\d+)K_(\d+)X(\d+)', name)
  48                 if mobj:
  49                     f.update({
  50                         'height': int(mobj.group(3)),
  51                         'tbr': int(mobj.group(1)),
  52                         'width': int(mobj.group(2)),
  53                     })
  54                 mobj = re.search(r'_(\d+)x(\d+)_(\d+)_(\d+)K\.mp4', playback_url)
  55                 if mobj:
  56                     f.update({
  57                         'fps': int(mobj.group(3)),
  58                         'height': int(mobj.group(2)),
  59                         'tbr': int(mobj.group(4)),
  60                         'width': int(mobj.group(1)),
  61                     })
  62                 formats.append(f)
  63
  64         thumbnails = []
  65         for cut in (try_get(feed, lambda x: x['image']['cuts'], list) or []):
  66             src = cut.get('src')
  67             if not src:
  68                 continue
  69             thumbnails.append({
  70                 'height': int_or_none(cut.get('height')),
  71                 'url': src,
  72                 'width': int_or_none(cut.get('width')),
  73             })
  74
  75         language = (video.get('language') or 'EN').lower()
  76
  77         return {
  78             'id': video_id,
  79             'title': title,
  80             'formats': formats,
  81             'description': video.get('description'),
  82             'duration': parse_duration(feed.get('duration')),
  83             'thumbnails': thumbnails,
  84             'timestamp': parse_iso8601(video.get(self._TIMESTAMP_KEY)),
  85             'subtitles': self._extract_mlb_subtitles(feed, language),
  86         }
  87
  88
  89 class MLBIE(MLBBaseIE):
  90     _VALID_URL = r'''(?x)
  91                     https?://
  92                         (?:[\da-z_-]+\.)*mlb\.com/
  93                         (?:
  94                             (?:
  95                                 (?:[^/]+/)*video/[^/]+/c-|
  96                                 (?:
  97                                     shared/video/embed/(?:embed|m-internal-embed)\.html|
  98                                     (?:[^/]+/)+(?:play|index)\.jsp|
  99                                 )\?.*?\bcontent_id=
 100                             )
 101                             (?P<id>\d+)
 102                         )
 103                     '''
 104     _EMBED_REGEX = [
 105         r'<iframe[^>]+?src=(["\'])(?P<url>https?://m(?:lb)?\.mlb\.com/shared/video/embed/embed\.html\?.+?)\1',
 106         r'data-video-link=["\'](?P<url>http://m\.mlb\.com/video/[^"\']+)',
 107     ]
 108     _TESTS = [
 109         {
 110             'url': 'https://www.mlb.com/mariners/video/ackleys-spectacular-catch/c-34698933',
 111             'md5': '632358dacfceec06bad823b83d21df2d',
 112             'info_dict': {
 113                 'id': '34698933',
 114                 'ext': 'mp4',
 115                 'title': "Ackley's spectacular catch",
 116                 'description': 'md5:7f5a981eb4f3cbc8daf2aeffa2215bf0',
 117                 'duration': 66,
 118                 'timestamp': 1405995000,
 119                 'upload_date': '20140722',
 120                 'thumbnail': r're:^https?://.*\.jpg$',
 121             },
 122         },
 123         {
 124             'url': 'https://www.mlb.com/video/stanton-prepares-for-derby/c-34496663',
 125             'md5': 'bf2619bf9cacc0a564fc35e6aeb9219f',
 126             'info_dict': {
 127                 'id': '34496663',
 128                 'ext': 'mp4',
 129                 'title': 'Stanton prepares for Derby',
 130                 'description': 'md5:d00ce1e5fd9c9069e9c13ab4faedfa57',
 131                 'duration': 46,
 132                 'timestamp': 1405120200,
 133                 'upload_date': '20140711',
 134                 'thumbnail': r're:^https?://.*\.jpg$',
 135             },
 136         },
 137         {
 138             'url': 'https://www.mlb.com/video/cespedes-repeats-as-derby-champ/c-34578115',
 139             'md5': '99bb9176531adc600b90880fb8be9328',
 140             'info_dict': {
 141                 'id': '34578115',
 142                 'ext': 'mp4',
 143                 'title': 'Cespedes repeats as Derby champ',
 144                 'description': 'md5:08df253ce265d4cf6fb09f581fafad07',
 145                 'duration': 488,
 146                 'timestamp': 1405414336,
 147                 'upload_date': '20140715',
 148                 'thumbnail': r're:^https?://.*\.jpg$',
 149             },
 150         },
 151         {
 152             'url': 'https://www.mlb.com/video/bautista-on-home-run-derby/c-34577915',
 153             'md5': 'da8b57a12b060e7663ee1eebd6f330ec',
 154             'info_dict': {
 155                 'id': '34577915',
 156                 'ext': 'mp4',
 157                 'title': 'Bautista on Home Run Derby',
 158                 'description': 'md5:b80b34031143d0986dddc64a8839f0fb',
 159                 'duration': 52,
 160                 'timestamp': 1405405122,
 161                 'upload_date': '20140715',
 162                 'thumbnail': r're:^https?://.*\.jpg$',
 163             },
 164         },
 165         {
 166             'url': 'https://www.mlb.com/video/hargrove-homers-off-caldwell/c-1352023483?tid=67793694',
 167             'only_matching': True,
 168         },
 169         {
 170             'url': 'http://m.mlb.com/shared/video/embed/embed.html?content_id=35692085&topic_id=6479266&width=400&height=224&property=mlb',
 171             'only_matching': True,
 172         },
 173         {
 174             'url': 'http://mlb.mlb.com/shared/video/embed/embed.html?content_id=36599553',
 175             'only_matching': True,
 176         },
 177         {
 178             'url': 'http://mlb.mlb.com/es/video/play.jsp?content_id=36599553',
 179             'only_matching': True,
 180         },
 181         {
 182             'url': 'https://www.mlb.com/cardinals/video/piscottys-great-sliding-catch/c-51175783',
 183             'only_matching': True,
 184         },
 185         {
 186             # From http://m.mlb.com/news/article/118550098/blue-jays-kevin-pillar-goes-spidey-up-the-wall-to-rob-tim-beckham-of-a-homer
 187             'url': 'http://mlb.mlb.com/shared/video/embed/m-internal-embed.html?content_id=75609783&property=mlb&autoplay=true&hashmode=false&siteSection=mlb/multimedia/article_118550098/article_embed&club=mlb',
 188             'only_matching': True,
 189         },
 190     ]
 191     _TIMESTAMP_KEY = 'date'
 192
 193     @staticmethod
 194     def _get_feed(video):
 195         return video
 196
 197     @staticmethod
 198     def _extract_mlb_subtitles(feed, language):
 199         subtitles = {}
 200         for keyword in (feed.get('keywordsAll') or []):
 201             keyword_type = keyword.get('type')
 202             if keyword_type and keyword_type.startswith('closed_captions_location_'):
 203                 cc_location = keyword.get('value')
 204                 if cc_location:
 205                     subtitles.setdefault(language, []).append({
 206                         'url': cc_location,
 207                     })
 208         return subtitles
 209
 210     def _download_video_data(self, display_id):
 211         return self._download_json(
 212             f'http://content.mlb.com/mlb/item/id/v1/{display_id}/details/web-v1.json',
 213             display_id)
 214
 215
 216 class MLBVideoIE(MLBBaseIE):
 217     _VALID_URL = r'https?://(?:www\.)?mlb\.com/(?:[^/]+/)*video/(?P<id>[^/?&#]+)'
 218     _TEST = {
 219         'url': 'https://www.mlb.com/mariners/video/ackley-s-spectacular-catch-c34698933',
 220         'md5': '632358dacfceec06bad823b83d21df2d',
 221         'info_dict': {
 222             'id': 'c04a8863-f569-42e6-9f87-992393657614',
 223             'ext': 'mp4',
 224             'title': "Ackley's spectacular catch",
 225             'description': 'md5:7f5a981eb4f3cbc8daf2aeffa2215bf0',
 226             'duration': 66,
 227             'timestamp': 1405995000,
 228             'upload_date': '20140722',
 229             'thumbnail': r're:^https?://.+',
 230         },
 231     }
 232     _TIMESTAMP_KEY = 'timestamp'
 233
 234     @classmethod
 235     def suitable(cls, url):
 236         return False if MLBIE.suitable(url) else super().suitable(url)
 237
 238     @staticmethod
 239     def _get_feed(video):
 240         return video['feeds'][0]
 241
 242     @staticmethod
 243     def _extract_mlb_subtitles(feed, language):
 244         subtitles = {}
 245         for cc_location in (feed.get('closedCaptions') or []):
 246             subtitles.setdefault(language, []).append({
 247                 'url': cc_location,
 248             })
 249
 250     def _download_video_data(self, display_id):
 251         # https://www.mlb.com/data-service/en/videos/[SLUG]
 252         return self._download_json(
 253             'https://fastball-gateway.mlb.com/graphql',
 254             display_id, query={
 255                 'query': '''{
 256   mediaPlayback(ids: "%s") {
 257     description
 258     feeds(types: CMS) {
 259       closedCaptions
 260       duration
 261       image {
 262         cuts {
 263           width
 264           height
 265           src
 266         }
 267       }
 268       playbacks {
 269         name
 270         url
 271       }
 272     }
 273     id
 274     timestamp
 275     title
 276   }
 277 }''' % display_id,  # noqa: UP031
 278             })['data']['mediaPlayback'][0]
 279
 280
 281 class MLBTVIE(InfoExtractor):
 282     _VALID_URL = r'https?://(?:www\.)?mlb\.com/tv/g(?P<id>\d{6})'
 283     _NETRC_MACHINE = 'mlb'
 284     _TESTS = [{
 285         'url': 'https://www.mlb.com/tv/g661581/vee2eff5f-a7df-4c20-bdb4-7b926fa12638',
 286         'info_dict': {
 287             'id': '661581',
 288             'ext': 'mp4',
 289             'title': '2022-07-02 - St. Louis Cardinals @ Philadelphia Phillies',
 290             'release_date': '20220702',
 291             'release_timestamp': 1656792300,
 292         },
 293         'params': {'skip_download': 'm3u8'},
 294     }, {
 295         # makeup game: has multiple dates, need to avoid games with 'rescheduleDate'
 296         'url': 'https://www.mlb.com/tv/g747039/vd22541c4-5a29-45f7-822b-635ec041cf5e',
 297         'info_dict': {
 298             'id': '747039',
 299             'ext': 'mp4',
 300             'title': '2024-07-29 - Toronto Blue Jays @ Baltimore Orioles',
 301             'release_date': '20240729',
 302             'release_timestamp': 1722280200,
 303         },
 304         'params': {'skip_download': 'm3u8'},
 305     }]
 306     _GRAPHQL_INIT_QUERY = '''\
 307 mutation initSession($device: InitSessionInput!, $clientType: ClientType!, $experience: ExperienceTypeInput) {
 308     initSession(device: $device, clientType: $clientType, experience: $experience) {
 309         deviceId
 310         sessionId
 311         entitlements {
 312             code
 313         }
 314         location {
 315             countryCode
 316             regionName
 317             zipCode
 318             latitude
 319             longitude
 320         }
 321         clientExperience
 322         features
 323     }
 324   }'''
 325     _GRAPHQL_PLAYBACK_QUERY = '''\
 326 mutation initPlaybackSession(
 327         $adCapabilities: [AdExperienceType]
 328         $mediaId: String!
 329         $deviceId: String!
 330         $sessionId: String!
 331         $quality: PlaybackQuality
 332     ) {
 333         initPlaybackSession(
 334             adCapabilities: $adCapabilities
 335             mediaId: $mediaId
 336             deviceId: $deviceId
 337             sessionId: $sessionId
 338             quality: $quality
 339         ) {
 340             playbackSessionId
 341             playback {
 342                 url
 343                 token
 344                 expiration
 345                 cdn
 346             }
 347         }
 348     }'''
 349     _APP_VERSION = '7.8.2'
 350     _device_id = None
 351     _session_id = None
 352     _access_token = None
 353     _token_expiry = 0
 354
 355     @property
 356     def _api_headers(self):
 357         if (self._token_expiry - 120) <= time.time():
 358             self.write_debug('Access token has expired; re-logging in')
 359             self._perform_login(*self._get_login_info())
 360         return {'Authorization': f'Bearer {self._access_token}'}
 361
 362     def _real_initialize(self):
 363         if not self._access_token:
 364             self.raise_login_required(
 365                 'All videos are only available to registered users', method='password')
 366
 367     def _set_device_id(self, username):
 368         if not self._device_id:
 369             self._device_id = self.cache.load(
 370                 self._NETRC_MACHINE, 'device_ids', default={}).get(username)
 371         if self._device_id:
 372             return
 373         self._device_id = str(uuid.uuid4())
 374         self.cache.store(self._NETRC_MACHINE, 'device_ids', {username: self._device_id})
 375
 376     def _perform_login(self, username, password):
 377         try:
 378             self._access_token = self._download_json(
 379                 'https://ids.mlb.com/oauth2/aus1m088yK07noBfh356/v1/token', None,
 380                 'Logging in', 'Unable to log in', headers={
 381                     'User-Agent': 'okhttp/3.12.1',
 382                     'Content-Type': 'application/x-www-form-urlencoded',
 383                 }, data=urlencode_postdata({
 384                     'grant_type': 'password',
 385                     'username': username,
 386                     'password': password,
 387                     'scope': 'openid offline_access',
 388                     'client_id': '0oa3e1nutA1HLzAKG356',
 389                 }))['access_token']
 390         except ExtractorError as error:
 391             if isinstance(error.cause, HTTPError) and error.cause.status == 400:
 392                 raise ExtractorError('Invalid username or password', expected=True)
 393             raise
 394
 395         self._token_expiry = traverse_obj(self._access_token, ({jwt_decode_hs256}, 'exp', {int})) or 0
 396         self._set_device_id(username)
 397
 398         self._session_id = self._call_api({
 399             'operationName': 'initSession',
 400             'query': self._GRAPHQL_INIT_QUERY,
 401             'variables': {
 402                 'device': {
 403                     'appVersion': self._APP_VERSION,
 404                     'deviceFamily': 'desktop',
 405                     'knownDeviceId': self._device_id,
 406                     'languagePreference': 'ENGLISH',
 407                     'manufacturer': '',
 408                     'model': '',
 409                     'os': '',
 410                     'osVersion': '',
 411                 },
 412                 'clientType': 'WEB',
 413             },
 414         }, None, 'session ID')['data']['initSession']['sessionId']
 415
 416     def _call_api(self, data, video_id, description='GraphQL JSON', fatal=True):
 417         return self._download_json(
 418             'https://media-gateway.mlb.com/graphql', video_id,
 419             f'Downloading {description}', f'Unable to download {description}', fatal=fatal,
 420             headers={
 421                 **self._api_headers,
 422                 'Accept': 'application/json',
 423                 'Content-Type': 'application/json',
 424                 'x-client-name': 'WEB',
 425                 'x-client-version': self._APP_VERSION,
 426             }, data=json.dumps(data, separators=(',', ':')).encode())
 427
 428     def _extract_formats_and_subtitles(self, broadcast, video_id):
 429         feed = traverse_obj(broadcast, ('homeAway', {str.title}))
 430         medium = traverse_obj(broadcast, ('type', {str}))
 431         language = traverse_obj(broadcast, ('language', {str.lower}))
 432         format_id = join_nonempty(feed, medium, language)
 433
 434         response = self._call_api({
 435             'operationName': 'initPlaybackSession',
 436             'query': self._GRAPHQL_PLAYBACK_QUERY,
 437             'variables': {
 438                 'adCapabilities': ['GOOGLE_STANDALONE_AD_PODS'],
 439                 'deviceId': self._device_id,
 440                 'mediaId': broadcast['mediaId'],
 441                 'quality': 'PLACEHOLDER',
 442                 'sessionId': self._session_id,
 443             },
 444         }, video_id, f'{format_id} broadcast JSON', fatal=False)
 445
 446         playback = traverse_obj(response, ('data', 'initPlaybackSession', 'playback', {dict}))
 447         m3u8_url = traverse_obj(playback, ('url', {url_or_none}))
 448         token = traverse_obj(playback, ('token', {str}))
 449
 450         if not (m3u8_url and token):
 451             errors = '; '.join(traverse_obj(response, ('errors', ..., 'message', {str})))
 452             if 'not entitled' in errors:
 453                 raise ExtractorError(errors, expected=True)
 454             elif errors:  # Only warn when 'blacked out' since radio formats are available
 455                 self.report_warning(f'API returned errors for {format_id}: {errors}')
 456             else:
 457                 self.report_warning(f'No formats available for {format_id} broadcast; skipping')
 458             return [], {}
 459
 460         cdn_headers = {'x-cdn-token': token}
 461         fmts, subs = self._extract_m3u8_formats_and_subtitles(
 462             m3u8_url.replace(f'/{token}/', '/'), video_id, 'mp4',
 463             m3u8_id=format_id, fatal=False, headers=cdn_headers)
 464         for fmt in fmts:
 465             fmt['http_headers'] = cdn_headers
 466             fmt.setdefault('format_note', join_nonempty(feed, medium, delim=' '))
 467             fmt.setdefault('language', language)
 468             if fmt.get('vcodec') == 'none' and fmt['language'] == 'en':
 469                 fmt['source_preference'] = 10
 470
 471         return fmts, subs
 472
 473     def _real_extract(self, url):
 474         video_id = self._match_id(url)
 475         data = self._download_json(
 476             'https://statsapi.mlb.com/api/v1/schedule', video_id, query={
 477                 'gamePk': video_id,
 478                 'hydrate': 'broadcasts(all),statusFlags',
 479             })
 480         metadata = traverse_obj(data, (
 481             'dates', ..., 'games',
 482             lambda _, v: str(v['gamePk']) == video_id and not v.get('rescheduleDate'), any))
 483
 484         broadcasts = traverse_obj(metadata, (
 485             'broadcasts', lambda _, v: v['mediaId'] and v['mediaState']['mediaStateCode'] != 'MEDIA_OFF'))
 486
 487         formats, subtitles = [], {}
 488         for broadcast in broadcasts:
 489             fmts, subs = self._extract_formats_and_subtitles(broadcast, video_id)
 490             formats.extend(fmts)
 491             self._merge_subtitles(subs, target=subtitles)
 492
 493         return {
 494             'id': video_id,
 495             'title': join_nonempty(
 496                 traverse_obj(metadata, ('officialDate', {str})),
 497                 traverse_obj(metadata, ('teams', ('away', 'home'), 'team', 'name', {str}, all, {' @ '.join})),
 498                 delim=' - '),
 499             'is_live': traverse_obj(broadcasts, (..., 'mediaState', 'mediaStateCode', {str}, any)) == 'MEDIA_ON',
 500             'release_timestamp': traverse_obj(metadata, ('gameDate', {parse_iso8601})),
 501             'formats': formats,
 502             'subtitles': subtitles,
 503         }
 504
 505
 506 class MLBArticleIE(InfoExtractor):
 507     _VALID_URL = r'https?://www\.mlb\.com/news/(?P<id>[\w-]+)'
 508     _TESTS = [{
 509         'url': 'https://www.mlb.com/news/manny-machado-robs-guillermo-heredia-reacts',
 510         'info_dict': {
 511             'id': '36db7394-343c-4ea3-b8ca-ead2e61bca9a',
 512             'title': 'Machado\'s grab draws hilarious irate reaction',
 513             'modified_timestamp': 1675888370,
 514             'description': 'md5:a19d4eb0487b2cb304e9a176f6b67676',
 515             'modified_date': '20230208',
 516         },
 517         'playlist_mincount': 2,
 518     }]
 519
 520     def _real_extract(self, url):
 521         display_id = self._match_id(url)
 522         webpage = self._download_webpage(url, display_id)
 523         apollo_cache_json = self._search_json(r'window\.initState\s*=', webpage, 'window.initState', display_id)['apolloCache']
 524
 525         content_real_info = traverse_obj(
 526             apollo_cache_json, ('ROOT_QUERY', lambda k, _: k.startswith('getArticle')), get_all=False)
 527
 528         return self.playlist_from_matches(
 529             traverse_obj(content_real_info, ('parts', lambda _, v: v['__typename'] == 'Video' or v['type'] == 'video')),
 530             getter=lambda x: f'https://www.mlb.com/video/{x["slug"]}',
 531             ie=MLBVideoIE, playlist_id=content_real_info.get('translationId'),
 532             title=self._html_search_meta('og:title', webpage),
 533             description=content_real_info.get('summary'),
 534             modified_timestamp=parse_iso8601(content_real_info.get('lastUpdatedDate')))