yt_dlp/extractor/bfmtv.py

   1 import re
   2
   3 from .common import InfoExtractor
   4 from ..utils import ExtractorError, extract_attributes
   5
   6
   7 class BFMTVBaseIE(InfoExtractor):
   8     _VALID_URL_BASE = r'https?://(?:www\.|rmc\.)?bfmtv\.com/'
   9     _VALID_URL_TMPL = _VALID_URL_BASE + r'(?:[^/]+/)*[^/?&#]+_%s[A-Z]-(?P<id>\d{12})\.html'
  10     _VIDEO_BLOCK_REGEX = r'(<div[^>]+class="video_block[^"]*"[^>]*>.*?</div>)'
  11     _VIDEO_ELEMENT_REGEX = r'(<video-js[^>]+>)'
  12     BRIGHTCOVE_URL_TEMPLATE = 'http://players.brightcove.net/%s/%s_default/index.html?videoId=%s'
  13
  14     def _extract_video(self, video_block):
  15         video_element = self._search_regex(
  16             self._VIDEO_ELEMENT_REGEX, video_block, 'video element', default=None)
  17         if video_element:
  18             video_element_attrs = extract_attributes(video_element)
  19             video_id = video_element_attrs.get('data-video-id')
  20             if not video_id:
  21                 return
  22             account_id = video_element_attrs.get('data-account') or '876450610001'
  23             player_id = video_element_attrs.get('adjustplayer') or '19dszYXgm'
  24         else:
  25             video_block_attrs = extract_attributes(video_block)
  26             video_id = video_block_attrs.get('videoid')
  27             if not video_id:
  28                 return
  29             account_id = video_block_attrs.get('accountid') or '876630703001'
  30             player_id = video_block_attrs.get('playerid') or 'KbPwEbuHx'
  31         return self.url_result(
  32             self.BRIGHTCOVE_URL_TEMPLATE % (account_id, player_id, video_id),
  33             'BrightcoveNew', video_id)
  34
  35
  36 class BFMTVIE(BFMTVBaseIE):
  37     IE_NAME = 'bfmtv'
  38     _VALID_URL = BFMTVBaseIE._VALID_URL_TMPL % 'V'
  39     _TESTS = [{
  40         'url': 'https://www.bfmtv.com/politique/emmanuel-macron-l-islam-est-une-religion-qui-vit-une-crise-aujourd-hui-partout-dans-le-monde_VN-202010020146.html',
  41         'info_dict': {
  42             'id': '6196747868001',
  43             'ext': 'mp4',
  44             'title': 'Emmanuel Macron: "L\'Islam est une religion qui vit une crise aujourd’hui, partout dans le monde"',
  45             'description': 'Le Président s\'exprime sur la question du séparatisme depuis les Mureaux, dans les Yvelines.',
  46             'uploader_id': '876450610001',
  47             'upload_date': '20201002',
  48             'timestamp': 1601629620,
  49             'duration': 44.757,
  50             'tags': ['bfmactu', 'politique'],
  51             'thumbnail': 'https://cf-images.eu-west-1.prod.boltdns.net/v1/static/876450610001/5041f4c1-bc48-4af8-a256-1b8300ad8ef0/cf2f9114-e8e2-4494-82b4-ab794ea4bc7d/1920x1080/match/image.jpg',
  52         },
  53     }]
  54
  55     def _real_extract(self, url):
  56         bfmtv_id = self._match_id(url)
  57         webpage = self._download_webpage(url, bfmtv_id)
  58         video = self._extract_video(self._search_regex(
  59             self._VIDEO_BLOCK_REGEX, webpage, 'video block'))
  60         if not video:
  61             raise ExtractorError('Failed to extract video')
  62         return video
  63
  64
  65 class BFMTVLiveIE(BFMTVBaseIE):
  66     IE_NAME = 'bfmtv:live'
  67     _VALID_URL = BFMTVBaseIE._VALID_URL_BASE + '(?P<id>(?:[^/]+/)?en-direct)'
  68     _TESTS = [{
  69         'url': 'https://www.bfmtv.com/en-direct/',
  70         'info_dict': {
  71             'id': '6346069778112',
  72             'ext': 'mp4',
  73             'title': r're:^Le Live BFM TV \d{4}-\d{2}-\d{2} \d{2}:\d{2}$',
  74             'uploader_id': '876450610001',
  75             'upload_date': '20240202',
  76             'timestamp': 1706887572,
  77             'live_status': 'is_live',
  78             'thumbnail': r're:https://.+/image\.jpg',
  79             'tags': [],
  80         },
  81         'params': {
  82             'skip_download': True,
  83         },
  84     }, {
  85         'url': 'https://www.bfmtv.com/economie/en-direct/',
  86         'only_matching': True,
  87     }]
  88
  89     def _real_extract(self, url):
  90         bfmtv_id = self._match_id(url)
  91         webpage = self._download_webpage(url, bfmtv_id)
  92         video = self._extract_video(self._search_regex(
  93             self._VIDEO_BLOCK_REGEX, webpage, 'video block'))
  94         if not video:
  95             raise ExtractorError('Failed to extract video')
  96         return video
  97
  98
  99 class BFMTVArticleIE(BFMTVBaseIE):
 100     IE_NAME = 'bfmtv:article'
 101     _VALID_URL = BFMTVBaseIE._VALID_URL_TMPL % 'A'
 102     _TESTS = [{
 103         'url': 'https://www.bfmtv.com/sante/covid-19-un-responsable-de-l-institut-pasteur-se-demande-quand-la-france-va-se-reconfiner_AV-202101060198.html',
 104         'info_dict': {
 105             'id': '202101060198',
 106             'title': 'Covid-19: un responsable de l\'Institut Pasteur se demande "quand la France va se reconfiner"',
 107             'description': 'md5:947974089c303d3ac6196670ae262843',
 108         },
 109         'playlist_count': 2,
 110     }, {
 111         'url': 'https://www.bfmtv.com/international/pour-bolsonaro-le-bresil-est-en-faillite-mais-il-ne-peut-rien-faire_AD-202101060232.html',
 112         'only_matching': True,
 113     }, {
 114         'url': 'https://www.bfmtv.com/sante/covid-19-oui-le-vaccin-de-pfizer-distribue-en-france-a-bien-ete-teste-sur-des-personnes-agees_AN-202101060275.html',
 115         'only_matching': True,
 116     }, {
 117         'url': 'https://rmc.bfmtv.com/actualites/societe/transports/ce-n-est-plus-tout-rentable-le-bioethanol-e85-depasse-1eu-le-litre-des-automobilistes-regrettent_AV-202301100268.html',
 118         'info_dict': {
 119             'id': '6318445464112',
 120             'ext': 'mp4',
 121             'title': 'Le plein de bioéthanol fait de plus en plus mal à la pompe',
 122             'uploader_id': '876630703001',
 123             'upload_date': '20230110',
 124             'timestamp': 1673341692,
 125             'duration': 109.269,
 126             'tags': ['rmc', 'show', 'apolline de malherbe', 'info', 'talk', 'matinale', 'radio'],
 127             'thumbnail': 'https://cf-images.eu-west-1.prod.boltdns.net/v1/static/876630703001/5bef74b8-9d5e-4480-a21f-60c2e2480c46/96c88b74-f9db-45e1-8040-e199c5da216c/1920x1080/match/image.jpg',
 128         },
 129     }]
 130
 131     def _entries(self, webpage):
 132         for video_block_el in re.findall(self._VIDEO_BLOCK_REGEX, webpage):
 133             video = self._extract_video(video_block_el)
 134             if video:
 135                 yield video
 136
 137     def _real_extract(self, url):
 138         bfmtv_id = self._match_id(url)
 139         webpage = self._download_webpage(url, bfmtv_id)
 140
 141         return self.playlist_result(
 142             self._entries(webpage), bfmtv_id, self._og_search_title(webpage, fatal=False),
 143             self._html_search_meta(['og:description', 'description'], webpage))