yt_dlp/extractor/harpodeon.py

   1 from .common import InfoExtractor
   2 from ..utils import int_or_none
   3
   4
   5 class HarpodeonIE(InfoExtractor):
   6     _VALID_URL = r'https?://(?:www\.)?harpodeon\.com/(?:video|preview)/\w+/(?P<id>\d+)'
   7     _TESTS = [{
   8         'url': 'https://www.harpodeon.com/video/The_Smoking_Out_of_Bella_Butts/268068288',
   9         'md5': '727371564a6a9ebccef2073535b5b6bd',
  10         'skip': 'Free video could become unavailable',
  11         'info_dict': {
  12             'id': '268068288',
  13             'ext': 'mp4',
  14             'title': 'The Smoking Out of Bella Butts',
  15             'description': 'md5:47e16bdb41fc8a79c83ab83af11c8b77',
  16             'creator': 'Vitagraph Company of America',
  17             'release_year': 1915,
  18         },
  19     }, {
  20         'url': 'https://www.harpodeon.com/preview/The_Smoking_Out_of_Bella_Butts/268068288',
  21         'md5': '6dfea5412845f690c7331be703f884db',
  22         'info_dict': {
  23             'id': '268068288',
  24             'ext': 'mp4',
  25             'title': 'The Smoking Out of Bella Butts',
  26             'description': 'md5:47e16bdb41fc8a79c83ab83af11c8b77',
  27             'creator': 'Vitagraph Company of America',
  28             'release_year': 1915,
  29         },
  30     }, {
  31         'url': 'https://www.harpodeon.com/preview/Behind_the_Screen/421838710',
  32         'md5': '7979df9ca04637282cb7d172ab3a9c3b',
  33         'info_dict': {
  34             'id': '421838710',
  35             'ext': 'mp4',
  36             'title': 'Behind the Screen',
  37             'description': 'md5:008972a3dc51fba3965ee517d2ba9155',
  38             'creator': 'Lone Star Corporation',
  39             'release_year': 1916,
  40         },
  41     }]
  42
  43     def _real_extract(self, url):
  44         video_id = self._match_id(url)
  45         webpage = self._download_webpage(url, video_id)
  46
  47         title, creator, release_year = self._search_regex(
  48             r'''(?x)
  49                 <div[^>]+videoInfo[^<]*<h2[^>]*>(?P<title>[^>]+)</h2>
  50                 (?:\s*<p[^>]*>\((?P<creator>.+),\s*)?(?P<release_year>\d{4})?''',
  51             webpage, 'title', group=('title', 'creator', 'release_year'),
  52             fatal=False) or (None, None, None)
  53
  54         hp_base = self._html_search_regex(r'hpBase\(\s*["\']([^"\']+)', webpage, 'hp_base')
  55
  56         hp_inject_video, hp_resolution = self._search_regex(
  57             r'''(?x)
  58                 hpInjectVideo\([\'\"](?P<hp_inject_video>\w+)[\'\"],
  59                 [\'\"](?P<hp_resolution>\d+)[\'\"]''',
  60             webpage, 'hp_inject_video', group=['hp_inject_video', 'hp_resolution'])
  61
  62         return {
  63             'id': video_id,
  64             'title': title,
  65             'url': f'{hp_base}{hp_inject_video}_{hp_resolution}.mp4',
  66             'http_headers': {'Referer': url},
  67             'description': self._html_search_meta('description', webpage, fatal=False),
  68             'creator': creator,
  69             'release_year': int_or_none(release_year),
  70         }