mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-04 20:17:22 +01:00
46 lines
1.4 KiB
Python
46 lines
1.4 KiB
Python
from __future__ import unicode_literals
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import ExtractorError
|
|
|
|
|
|
class RedTubeIE(InfoExtractor):
|
|
_VALID_URL = r'http://(?:www\.)?redtube\.com/(?P<id>[0-9]+)'
|
|
_TEST = {
|
|
'url': 'http://www.redtube.com/66418',
|
|
'md5': '7b8c22b5e7098a3e1c09709df1126d2d',
|
|
'info_dict': {
|
|
'id': '66418',
|
|
'ext': 'mp4',
|
|
'title': 'Sucked on a toilet',
|
|
'age_limit': 18,
|
|
}
|
|
}
|
|
|
|
def _real_extract(self, url):
|
|
video_id = self._match_id(url)
|
|
webpage = self._download_webpage(url, video_id)
|
|
|
|
if any(s in webpage for s in ['video-deleted-info', '>This video has been removed']):
|
|
raise ExtractorError('Video %s has been removed' % video_id, expected=True)
|
|
|
|
video_url = self._html_search_regex(
|
|
r'<source src="(.+?)" type="video/mp4">', webpage, 'video URL')
|
|
video_title = self._html_search_regex(
|
|
r'<h1 class="videoTitle[^"]*">(.+?)</h1>',
|
|
webpage, 'title')
|
|
video_thumbnail = self._og_search_thumbnail(webpage)
|
|
|
|
# No self-labeling, but they describe themselves as
|
|
# "Home of Videos Porno"
|
|
age_limit = 18
|
|
|
|
return {
|
|
'id': video_id,
|
|
'url': video_url,
|
|
'ext': 'mp4',
|
|
'title': video_title,
|
|
'thumbnail': video_thumbnail,
|
|
'age_limit': age_limit,
|
|
}
|