1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-14 13:25:38 +00:00
youtube-dl/youtube_dl/extractor/tmz.py

33 lines
1.2 KiB
Python
Raw Normal View History

2014-11-25 07:54:13 +00:00
# coding: utf-8
from __future__ import unicode_literals
from .common import InfoExtractor
class TMZIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?tmz\.com/videos/(?P<id>[^/]+)/?'
_TEST = {
'url': 'http://www.tmz.com/videos/0_okj015ty/',
'md5': '791204e3bf790b1426cb2db0706184c0',
'info_dict': {
'id': '0_okj015ty',
'url': 'http://tmz.vo.llnwd.net/o28/2014-03/13/0_okj015ty_0_rt8ro3si_2.mp4',
'ext': 'mp4',
'title': 'Kim Kardashian\'s Boobs Unlock a Mystery!',
'description': 'Did Kim Kardasain try to one-up Khloe by one-upping Kylie??? Or is she just showing off her amazing boobs?',
2014-12-21 14:26:00 +00:00
'thumbnail': r're:http://cdnbakmi\.kaltura\.com/.*thumbnail.*',
2014-11-25 07:54:13 +00:00
}
}
def _real_extract(self, url):
video_id = self._match_id(url)
webpage = self._download_webpage(url, video_id)
return {
'id': video_id,
2014-11-25 08:16:40 +00:00
'url': self._html_search_meta('VideoURL', webpage, fatal=True),
'title': self._og_search_title(webpage),
'description': self._og_search_description(webpage),
'thumbnail': self._html_search_meta('ThumbURL', webpage),
2014-11-25 07:54:13 +00:00
}