diff options
| author | Sergey M․ <dstftw@gmail.com> | 2014-09-05 22:26:31 +0700 | 
|---|---|---|
| committer | Sergey M․ <dstftw@gmail.com> | 2014-09-05 22:26:31 +0700 | 
| commit | 70a1ecd2c1c602164b89512a3646ff4119349018 (patch) | |
| tree | e8f2889ea11531f515d5e262e72655040ca35ecc | |
| parent | 88a23aef5a7d64ab4f4d682d6b336ac9eea2efba (diff) | |
| parent | af8812bb9b50586dd228c624b1bc569639070254 (diff) | |
Merge branch 'unistra_hd' of https://github.com/Rudloff/youtube-dl into Rudloff-unistra_hd
| -rw-r--r-- | youtube_dl/extractor/unistra.py | 28 | 
1 files changed, 20 insertions, 8 deletions
diff --git a/youtube_dl/extractor/unistra.py b/youtube_dl/extractor/unistra.py index 474610eec..bd5a258a5 100644 --- a/youtube_dl/extractor/unistra.py +++ b/youtube_dl/extractor/unistra.py @@ -16,17 +16,29 @@ class UnistraIE(InfoExtractor):      }      def _real_extract(self, url): -        id = re.match(self._VALID_URL, url).group(1) -        webpage = self._download_webpage(url, id) -        file = re.search(r'file: "(.*?)",', webpage).group(1) +        video_id = re.match(self._VALID_URL, url).group(1) +        webpage = self._download_webpage(url, video_id) +        width = re.search(r'width: "(\d*?)",', webpage).group(1) +        height = re.search(r'height: "(\d*?)",', webpage).group(1) +        files = re.findall(r'file: "(.*?)"', webpage) +        video_url = 'http://vod-flash.u-strasbg.fr:8080' +        formats = [{ +            'format_id': 'SD', +            'url': video_url + files[0], +            'ext': 'mp4', +            'resolution': width + 'x' + height +            }] +        if files[1] != files[0]: +            formats.append({ +            'format_id': 'HD', +            'url': video_url + files[1], +            'ext': 'mp4' +            })          title = self._html_search_regex(r'<title>UTV - (.*?)</', webpage, u'title') -        video_url = 'http://vod-flash.u-strasbg.fr:8080/' + file - -        return {'id': id, +        return {'id': video_id,                  'title': title, -                'ext': 'mp4', -                'url': video_url,                  'description': self._html_search_regex(r'<meta name="Description" content="(.*?)"', webpage, u'description', flags=re.DOTALL),                  'thumbnail': self._search_regex(r'image: "(.*?)"', webpage, u'thumbnail'), +                'formats': formats                  }  | 
