'vbr': bitrate,
})
- subtitles={}
- subtitle_matches=re.findall(r'<track kind="captions" src="([^"]+)" srclang="([^"]+)"', webpage)
- for match in subtitle_matches:
- (suburl,sublang)=match
- if not sublang in subtitles.keys():
- subtitles[sublang]=[]
- subext=suburl.split('/')[-1]
- subtitles[sublang].append({'url': 'http://www.funnyordie.com'+suburl,'ext': subext})
+ subtitles = {}
+ for src, src_lang in re.findall(r'<track kind="captions" src="([^"]+)" srclang="([^"]+)"', webpage):
+ subtitles[src_lang] = [{
+ 'ext': src.split('/')[-1],
+ 'url': 'http://www.funnyordie.com%s' % src,
+ }]
post_json = self._search_regex(
r'fb_post\s*=\s*(\{.*?\});', webpage, 'post details')