]> Raphaƫl G. Git Repositories - youtubedl/blobdiff - youtube_dl/extractor/flickr.py
Update changelog.
[youtubedl] / youtube_dl / extractor / flickr.py
index 791d5b61dd27efcaf0aa94d3bce2e1744aa72678..21ea5ec2bf1c499149809971b780daa9d10d0291 100644 (file)
@@ -1,3 +1,5 @@
+from __future__ import unicode_literals
+
 import re
 
 from .common import InfoExtractor
@@ -9,7 +11,17 @@ from ..utils import (
 
 class FlickrIE(InfoExtractor):
     """Information Extractor for Flickr videos"""
-    _VALID_URL = r'(?:https?://)?(?:www\.)?flickr\.com/photos/(?P<uploader_id>[\w\-_@]+)/(?P<id>\d+).*'
+    _VALID_URL = r'(?:https?://)?(?:www\.|secure\.)?flickr\.com/photos/(?P<uploader_id>[\w\-_@]+)/(?P<id>\d+).*'
+    _TEST = {
+        'url': 'http://www.flickr.com/photos/forestwander-nature-pictures/5645318632/in/photostream/',
+        'file': '5645318632.mp4',
+        'md5': '6fdc01adbc89d72fc9c4f15b4a4ba87b',
+        'info_dict': {
+            "description": "Waterfalls in the Springtime at Dark Hollow Waterfalls. These are located just off of Skyline Drive in Virginia. They are only about 6/10 of a mile hike but it is a pretty steep hill and a good climb back up.", 
+            "uploader_id": "forestwander-nature-pictures", 
+            "title": "Dark Hollow Waterfalls"
+        }
+    }
 
     def _real_extract(self, url):
         mobj = re.match(self._VALID_URL, url)
@@ -19,13 +31,13 @@ class FlickrIE(InfoExtractor):
         webpage_url = 'http://www.flickr.com/photos/' + video_uploader_id + '/' + video_id
         webpage = self._download_webpage(webpage_url, video_id)
 
-        secret = self._search_regex(r"photo_secret: '(\w+)'", webpage, u'secret')
+        secret = self._search_regex(r"photo_secret: '(\w+)'", webpage, 'secret')
 
         first_url = 'https://secure.flickr.com/apps/video/video_mtl_xml.gne?v=x&photo_id=' + video_id + '&secret=' + secret + '&bitrate=700&target=_self'
         first_xml = self._download_webpage(first_url, video_id, 'Downloading first data webpage')
 
         node_id = self._html_search_regex(r'<Item id="id">(\d+-\d+)</Item>',
-            first_xml, u'node_id')
+            first_xml, 'node_id')
 
         second_url = 'https://secure.flickr.com/video_playlist.gne?node_id=' + node_id + '&tech=flash&mode=playlist&bitrate=700&secret=' + secret + '&rd=video.yahoo.com&noad=1'
         second_xml = self._download_webpage(second_url, video_id, 'Downloading second data webpage')
@@ -34,24 +46,15 @@ class FlickrIE(InfoExtractor):
 
         mobj = re.search(r'<STREAM APP="(.+?)" FULLPATH="(.+?)"', second_xml)
         if mobj is None:
-            raise ExtractorError(u'Unable to extract video url')
+            raise ExtractorError('Unable to extract video url')
         video_url = mobj.group(1) + unescapeHTML(mobj.group(2))
 
-        video_title = self._html_search_regex(r'<meta property="og:title" content=(?:"([^"]+)"|\'([^\']+)\')',
-            webpage, u'video title')
-
-        video_description = self._html_search_regex(r'<meta property="og:description" content=(?:"([^"]+)"|\'([^\']+)\')',
-            webpage, u'description', fatal=False)
-
-        thumbnail = self._html_search_regex(r'<meta property="og:image" content=(?:"([^"]+)"|\'([^\']+)\')',
-            webpage, u'thumbnail', fatal=False)
-
         return [{
             'id':          video_id,
             'url':         video_url,
             'ext':         'mp4',
-            'title':       video_title,
-            'description': video_description,
-            'thumbnail':   thumbnail,
+            'title':       self._og_search_title(webpage),
+            'description': self._og_search_description(webpage),
+            'thumbnail':   self._og_search_thumbnail(webpage),
             'uploader_id': video_uploader_id,
         }]