[vevo] Some improvements (fixes #1580)

Extract the info from http://videoplayer.vevo.com/VideoService/AuthenticateVideo?isrc={id}
Some videos don't have an smil manifest, extract the video urls directly from the json and use the last version of the video.
Extract all the available formats and set the 'formats' field of the result

youtube_dl/extractor/vevo.py

@@ -1,11 +1,15 @@
 import re
 import json
+import xml.etree.ElementTree
+import datetime
 
 from .common import InfoExtractor
 from ..utils import (
+    determine_ext,
     ExtractorError,
 )
 
+
 class VevoIE(InfoExtractor):
     """
     Accepts urls from vevo.com or in the format 'vevo:{id}'
@@ -15,11 +19,11 @@ class VevoIE(InfoExtractor):
     _TEST = {
         u'url': u'http://www.vevo.com/watch/hurts/somebody-to-die-for/GB1101300280',
         u'file': u'GB1101300280.mp4',
-        u'md5': u'06bea460acb744eab74a9d7dcb4bfd61',
         u'info_dict': {
             u"upload_date": u"20130624",
             u"uploader": u"Hurts",
-            u"title": u"Somebody to Die For"
+            u"title": u"Somebody to Die For",
+            u'duration': 230,
         }
     }
 
@@ -27,27 +31,47 @@ class VevoIE(InfoExtractor):
         mobj = re.match(self._VALID_URL, url)
         video_id = mobj.group('id')
 
-        json_url = 'http://www.vevo.com/data/video/%s' % video_id
-        base_url = 'http://smil.lvl3.vevo.com'
-        videos_url = '%s/Video/V2/VFILE/%s/%sr.smil' % (base_url, video_id, video_id.lower())
+        json_url = 'http://videoplayer.vevo.com/VideoService/AuthenticateVideo?isrc=%s' % video_id
         info_json = self._download_webpage(json_url, video_id, u'Downloading json info')
-        links_webpage = self._download_webpage(videos_url, video_id, u'Downloading videos urls')
 
         self.report_extraction(video_id)
-        video_info = json.loads(info_json)
-        m_urls = list(re.finditer(r'<video src="(?P<ext>.*?):/?(?P<url>.*?)"', links_webpage))
-        if m_urls is None or len(m_urls) == 0:
-            raise ExtractorError(u'Unable to extract video url')
-        # They are sorted from worst to best quality
-        m_url = m_urls[-1]
-        video_url = base_url + '/' + m_url.group('url')
-        ext = m_url.group('ext')
-
-        return {'url': video_url,
-                'ext': ext,
+        video_info = json.loads(info_json)['video']
+        last_version = {'version': -1}
+        for version in video_info['videoVersions']:
+            # These are the HTTP downloads, other types are for different manifests
+            if version['sourceType'] == 2:
+                if version['version'] > last_version['version']:
+                    last_version = version
+        if last_version['version'] == -1:
+            raise ExtractorError(u'Unable to extract last version of the video')
+
+        renditions = xml.etree.ElementTree.fromstring(last_version['data'])
+        formats = []
+        # Already sorted from worst to best quality
+        for rend in renditions.findall('rendition'):
+            attr = rend.attrib
+            f_url = attr['url']
+            formats.append({
+                'url': f_url,
+                'ext': determine_ext(f_url),
+                'height': int(attr['frameheight']),
+                'width': int(attr['frameWidth']),
+            })
+
+        date_epoch = int(self._search_regex(
+            r'/Date\((\d+)\)/', video_info['launchDate'], u'launch date'))/1000
+        upload_date = datetime.datetime.fromtimestamp(date_epoch)
+        info = {
             'id': video_id,
             'title': video_info['title'],
-                'thumbnail': video_info['img'],
-                'upload_date': video_info['launchDate'].replace('/',''),
-                'uploader': video_info['Artists'][0]['title'],
+            'formats': formats,
+            'thumbnail': video_info['imageUrl'],
+            'upload_date': upload_date.strftime('%Y%m%d'),
+            'uploader': video_info['mainArtists'][0]['artistName'],
+            'duration': video_info['duration'],
         }
+
+        # TODO: Remove when #980 has been merged
+        info.update(formats[-1])
+
+        return info