yt_dlp/extractor/vrt.py

   1 from .common import InfoExtractor
   2 from ..utils import (
   3     extract_attributes,
   4     float_or_none,
   5     get_element_by_class,
   6     strip_or_none,
   7     unified_timestamp,
   8 )
   9
  10
  11 class VRTIE(InfoExtractor):
  12     IE_DESC = 'VRT NWS, Flanders News, Flandern Info and Sporza'
  13     _VALID_URL = r'https?://(?:www\.)?(?P<site>vrt\.be/vrtnws|sporza\.be)/[a-z]{2}/\d{4}/\d{2}/\d{2}/(?P<id>[^/?&#]+)'
  14     _TESTS = [{
  15         'url': 'https://www.vrt.be/vrtnws/nl/2019/05/15/beelden-van-binnenkant-notre-dame-een-maand-na-de-brand/',
  16         'md5': 'e1663accf5cf13f375f3cd0d10476669',
  17         'info_dict': {
  18             'id': 'pbs-pub-7855fc7b-1448-49bc-b073-316cb60caa71$vid-2ca50305-c38a-4762-9890-65cbd098b7bd',
  19             'ext': 'mp4',
  20             'title': 'Beelden van binnenkant Notre-Dame, één maand na de brand',
  21             'description': 'Op maandagavond 15 april ging een deel van het dakgebinte van de Parijse kathedraal in vlammen op.',
  22             'timestamp': 1557924660,
  23             'upload_date': '20190515',
  24             'duration': 31.2,
  25         },
  26     }, {
  27         'url': 'https://sporza.be/nl/2019/05/15/de-belgian-cats-zijn-klaar-voor-het-ek/',
  28         'md5': '910bba927566e9ab992278f647eb4b75',
  29         'info_dict': {
  30             'id': 'pbs-pub-f2c86a46-8138-413a-a4b9-a0015a16ce2c$vid-1f112b31-e58e-4379-908d-aca6d80f8818',
  31             'ext': 'mp4',
  32             'title': 'De Belgian Cats zijn klaar voor het EK mét Ann Wauters',
  33             'timestamp': 1557923760,
  34             'upload_date': '20190515',
  35             'duration': 115.17,
  36         },
  37     }, {
  38         'url': 'https://www.vrt.be/vrtnws/en/2019/05/15/belgium_s-eurovision-entry-falls-at-the-first-hurdle/',
  39         'only_matching': True,
  40     }, {
  41         'url': 'https://www.vrt.be/vrtnws/de/2019/05/15/aus-fuer-eliott-im-halbfinale-des-eurosongfestivals/',
  42         'only_matching': True,
  43     }]
  44     _CLIENT_MAP = {
  45         'vrt.be/vrtnws': 'vrtnieuws',
  46         'sporza.be': 'sporza',
  47     }
  48
  49     def _real_extract(self, url):
  50         site, display_id = self._match_valid_url(url).groups()
  51         webpage = self._download_webpage(url, display_id)
  52         attrs = extract_attributes(self._search_regex(
  53             r'(<[^>]+class="vrtvideo( [^"]*)?"[^>]*>)', webpage, 'vrt video'))
  54
  55         asset_id = attrs['data-video-id']
  56         publication_id = attrs.get('data-publication-id')
  57         if publication_id:
  58             asset_id = publication_id + '$' + asset_id
  59         client = attrs.get('data-client-code') or self._CLIENT_MAP[site]
  60
  61         title = strip_or_none(get_element_by_class(
  62             'vrt-title', webpage) or self._html_search_meta(
  63             ['og:title', 'twitter:title', 'name'], webpage))
  64         description = self._html_search_meta(
  65             ['og:description', 'twitter:description', 'description'], webpage)
  66         if description == '…':
  67             description = None
  68         timestamp = unified_timestamp(self._html_search_meta(
  69             'article:published_time', webpage))
  70
  71         return {
  72             '_type': 'url_transparent',
  73             'id': asset_id,
  74             'display_id': display_id,
  75             'title': title,
  76             'description': description,
  77             'thumbnail': attrs.get('data-posterimage'),
  78             'timestamp': timestamp,
  79             'duration': float_or_none(attrs.get('data-duration'), 1000),
  80             'url': 'https://mediazone.vrt.be/api/v1/%s/assets/%s' % (client, asset_id),
  81             'ie_key': 'Canvas',
  82         }