]> Raphaƫl G. Git Repositories - youtubedl/blobdiff - youtube_dl/extractor/niconico.py
New upstream version 2020.06.16.1
[youtubedl] / youtube_dl / extractor / niconico.py
index df7f528be2d4c8da7cfe826609608754f2f8e088..eb07ca7765e6ccfe08f856f44647f1ea7d7f706f 100644 (file)
@@ -163,7 +163,7 @@ class NiconicoIE(InfoExtractor):
         self._login()
 
     def _login(self):
         self._login()
 
     def _login(self):
-        (username, password) = self._get_login_info()
+        username, password = self._get_login_info()
         # No authentication to be performed
         if not username:
             return True
         # No authentication to be performed
         if not username:
             return True
@@ -252,7 +252,7 @@ class NiconicoIE(InfoExtractor):
                     },
                     'timing_constraint': 'unlimited'
                 }
                     },
                     'timing_constraint': 'unlimited'
                 }
-            }))
+            }).encode())
 
         resolution = video_quality.get('resolution', {})
 
 
         resolution = video_quality.get('resolution', {})
 
@@ -369,14 +369,14 @@ class NiconicoIE(InfoExtractor):
         video_detail = watch_api_data.get('videoDetail', {})
 
         thumbnail = (
         video_detail = watch_api_data.get('videoDetail', {})
 
         thumbnail = (
-            get_video_info(['thumbnail_url', 'thumbnailURL']) or
-            self._html_search_meta('image', webpage, 'thumbnail', default=None) or
-            video_detail.get('thumbnail'))
+            get_video_info(['thumbnail_url', 'thumbnailURL'])
+            or self._html_search_meta('image', webpage, 'thumbnail', default=None)
+            or video_detail.get('thumbnail'))
 
         description = get_video_info('description')
 
 
         description = get_video_info('description')
 
-        timestamp = (parse_iso8601(get_video_info('first_retrieve')) or
-                     unified_timestamp(get_video_info('postedDateTime')))
+        timestamp = (parse_iso8601(get_video_info('first_retrieve'))
+                     or unified_timestamp(get_video_info('postedDateTime')))
         if not timestamp:
             match = self._html_search_meta('datePublished', webpage, 'date published', default=None)
             if match:
         if not timestamp:
             match = self._html_search_meta('datePublished', webpage, 'date published', default=None)
             if match:
@@ -395,9 +395,9 @@ class NiconicoIE(InfoExtractor):
                 view_count = int_or_none(match.replace(',', ''))
         view_count = view_count or video_detail.get('viewCount')
 
                 view_count = int_or_none(match.replace(',', ''))
         view_count = view_count or video_detail.get('viewCount')
 
-        comment_count = (int_or_none(get_video_info('comment_num')) or
-                         video_detail.get('commentCount') or
-                         try_get(api_data, lambda x: x['thread']['commentCount']))
+        comment_count = (int_or_none(get_video_info('comment_num'))
+                         or video_detail.get('commentCount')
+                         or try_get(api_data, lambda x: x['thread']['commentCount']))
         if not comment_count:
             match = self._html_search_regex(
                 r'>Comments: <strong[^>]*>([^<]+)</strong>',
         if not comment_count:
             match = self._html_search_regex(
                 r'>Comments: <strong[^>]*>([^<]+)</strong>',
@@ -406,11 +406,11 @@ class NiconicoIE(InfoExtractor):
                 comment_count = int_or_none(match.replace(',', ''))
 
         duration = (parse_duration(
                 comment_count = int_or_none(match.replace(',', ''))
 
         duration = (parse_duration(
-            get_video_info('length') or
-            self._html_search_meta(
-                'video:duration', webpage, 'video duration', default=None)) or
-            video_detail.get('length') or
-            get_video_info('duration'))
+            get_video_info('length')
+            or self._html_search_meta(
+                'video:duration', webpage, 'video duration', default=None))
+            or video_detail.get('length')
+            or get_video_info('duration'))
 
         webpage_url = get_video_info('watch_url') or url
 
 
         webpage_url = get_video_info('watch_url') or url