[YoutubeDL] Support DASH manifest downloading

[youtube-dl] / youtube_dl / extractor / youtube.py

diff --git a/youtube_dl/extractor/youtube.py b/youtube_dl/extractor/youtube.py
index 9096a29756ca6e1a66ecd442a92977fa1b999b31..5d1297e0d27260dd1e0f389d5add17061fad0644 100644 (file)
--- a/youtube_dl/extractor/youtube.py
+++ b/youtube_dl/extractor/youtube.py
@@ -802,6 +802,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
                     # TODO implement WebVTT downloading
                     pass
                 elif mime_type.startswith('audio/') or mime_type.startswith('video/'):
+                    segment_list = r.find('{urn:mpeg:DASH:schema:MPD:2011}SegmentList')
                     format_id = r.attrib['id']
                     video_url = url_el.text
                     filesize = int_or_none(url_el.attrib.get('{http://youtube.com/yt/2012/10/10}contentLength'))
@@ -815,6 +816,11 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
                         'filesize': filesize,
                         'fps': int_or_none(r.attrib.get('frameRate')),
                     }
+                    if segment_list:
+                        f.update({
+                            'initialization_url': segment_list.find('{urn:mpeg:DASH:schema:MPD:2011}Initialization').attrib['sourceURL'],
+                            'segment_urls': [segment.attrib.get('media') for segment in segment_list.findall('{urn:mpeg:DASH:schema:MPD:2011}SegmentURL')]
+                        })
                     try:
                         existing_format = next(
                             fo for fo in formats
@@ -1126,12 +1132,13 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
                     self.report_warning(
                         'Skipping DASH manifest: %r' % e, video_id)
                 else:
-                    # Hide the formats we found through non-DASH
+                    # Remove the formats we found through non-DASH, they
+                    # contain less info and it can be wrong, because we use
+                    # fixed values (for example the resolution). See
+                    # https://github.com/rg3/youtube-dl/issues/5774 for an
+                    # example.
                     dash_keys = set(df['format_id'] for df in dash_formats)
-                    for f in formats:
-                        if f['format_id'] in dash_keys:
-                            f['format_id'] = 'nondash-%s' % f['format_id']
-                            f['preference'] = f.get('preference', 0) - 10000
+                    formats = [f for f in formats if f['format_id'] not in dash_keys]
                     formats.extend(dash_formats)
 
         # Check for malformed aspect ratio
@@ -1398,6 +1405,22 @@ class YoutubeChannelIE(InfoExtractor):
         channel_id = self._match_id(url)
 
         url = self._TEMPLATE_URL % channel_id
+
+        # Channel by page listing is restricted to 35 pages of 30 items, i.e. 1050 videos total (see #5778)
+        # Workaround by extracting as a playlist if managed to obtain channel playlist URL
+        # otherwise fallback on channel by page extraction
+        channel_page = self._download_webpage(
+            url + '?view=57', channel_id,
+            'Downloading channel page', fatal=False)
+        channel_playlist_id = self._search_regex(
+            [r'<meta itemprop="channelId" content="([^"]+)">',
+             r'data-channel-external-id="([^"]+)"'],
+            channel_page, 'channel id', default=None)
+        if channel_playlist_id and channel_playlist_id.startswith('UC'):
+            playlist_id = 'UU' + channel_playlist_id[2:]
+            return self.url_result(
+                compat_urlparse.urljoin(url, '/playlist?list=%s' % playlist_id), 'YoutubePlaylist')
+
         channel_page = self._download_webpage(url, channel_id, 'Downloading page #1')
         autogenerated = re.search(r'''(?x)
                 class="[^"]*?(?:
@@ -1621,10 +1644,16 @@ class YoutubeFeedsInfoExtractor(YoutubeBaseInfoExtractor):
         # for the video ids doesn't contain an index
         ids = []
         more_widget_html = content_html = page
-
         for page_num in itertools.count(1):
             matches = re.findall(r'href="\s*/watch\?v=([0-9A-Za-z_-]{11})', content_html)
-            new_ids = orderedSet(matches)
+
+            # 'recommended' feed has infinite 'load more' and each new portion spins
+            # the same videos in (sometimes) slightly different order, so we'll check
+            # for unicity and break when portion has no new videos
+            new_ids = filter(lambda video_id: video_id not in ids, orderedSet(matches))
+            if not new_ids:
+                break
+
             ids.extend(new_ids)
 
             mobj = re.search(r'data-uix-load-more-href="/?(?P<more>[^"]+)"', more_widget_html)