You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
yt-dlp/yt_dlp/extractor/lenta.py

52 lines
1.6 KiB
Python

from .common import InfoExtractor
class LentaIE(InfoExtractor):
_WORKING = False
_VALID_URL = r'https?://(?:www\.)?lenta\.ru/[^/]+/\d+/\d+/\d+/(?P<id>[^/?#&]+)'
_TESTS = [{
'url': 'https://lenta.ru/news/2018/03/22/savshenko_go/',
'info_dict': {
'id': '964400',
'ext': 'mp4',
'title': 'Надежду Савченко задержали',
'thumbnail': r're:^https?://.*\.jpg$',
'duration': 61,
'view_count': int,
},
'params': {
'skip_download': True,
},
}, {
# EaglePlatform iframe embed
'url': 'http://lenta.ru/news/2015/03/06/navalny/',
'info_dict': {
'id': '227304',
'ext': 'mp4',
'title': 'Навальный вышел на свободу',
'description': 'md5:d97861ac9ae77377f3f20eaf9d04b4f5',
'thumbnail': r're:^https?://.*\.jpg$',
'duration': 87,
'view_count': int,
'age_limit': 0,
},
'params': {
'skip_download': True,
},
}]
def _real_extract(self, url):
display_id = self._match_id(url)
webpage = self._download_webpage(url, display_id)
video_id = self._search_regex(
r'vid\s*:\s*["\']?(\d+)', webpage, 'eagleplatform id',
default=None)
if video_id:
return self.url_result(
'eagleplatform:lentaru.media.eagleplatform.com:%s' % video_id,
ie='EaglePlatform', video_id=video_id)
return self.url_result(url, ie='Generic')