mirror of
https://github.com/l1ving/youtube-dl
synced 2024-11-25 19:32:57 +08:00
50 lines
1.7 KiB
Python
50 lines
1.7 KiB
Python
# coding: utf-8
|
|
from __future__ import unicode_literals
|
|
|
|
import re
|
|
|
|
from .common import InfoExtractor
|
|
|
|
|
|
class WeiboIE(InfoExtractor):
|
|
"""
|
|
The videos in Weibo come from different sites, this IE just finds the link
|
|
to the external video and returns it.
|
|
"""
|
|
_VALID_URL = r'https?://video\.weibo\.com/v/weishipin/t_(?P<id>.+?)\.htm'
|
|
|
|
_TEST = {
|
|
'url': 'http://video.weibo.com/v/weishipin/t_zjUw2kZ.htm',
|
|
'info_dict': {
|
|
'id': '98322879',
|
|
'ext': 'flv',
|
|
'title': '魔声耳机最新广告“All Eyes On Us”',
|
|
},
|
|
'params': {
|
|
'skip_download': True,
|
|
},
|
|
'add_ie': ['Sina'],
|
|
}
|
|
|
|
# Additional example videos from different sites
|
|
# Youku: http://video.weibo.com/v/weishipin/t_zQGDWQ8.htm
|
|
# 56.com: http://video.weibo.com/v/weishipin/t_zQ44HxN.htm
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url, flags=re.VERBOSE)
|
|
video_id = mobj.group('id')
|
|
info_url = 'http://video.weibo.com/?s=v&a=play_list&format=json&mix_video_id=t_%s' % video_id
|
|
info = self._download_json(info_url, video_id)
|
|
|
|
videos_urls = map(lambda v: v['play_page_url'], info['result']['data'])
|
|
# Prefer sina video since they have thumbnails
|
|
videos_urls = sorted(videos_urls, key=lambda u: 'video.sina.com' in u)
|
|
player_url = videos_urls[-1]
|
|
m_sina = re.match(r'https?://video\.sina\.com\.cn/v/b/(\d+)-\d+\.html',
|
|
player_url)
|
|
if m_sina is not None:
|
|
self.to_screen('Sina video detected')
|
|
sina_id = m_sina.group(1)
|
|
player_url = 'http://you.video.sina.com.cn/swf/quotePlayer.swf?vid=%s' % sina_id
|
|
return self.url_result(player_url)
|