Merge 8647d6e311 into f2a4983df7

Closes #11526
Authored by: jshumphrey

yt_dlp/extractor/archiveorg.py

@@ -205,6 +205,26 @@ class ArchiveOrgIE(InfoExtractor):
                 },
             },
         ],
+    }, {
+        # The reviewbody is None for one of the reviews; just need to extract data without crashing
+        'url': 'https://archive.org/details/gd95-04-02.sbd.11622.sbeok.shnf/gd95-04-02d1t04.shn',
+        'info_dict': {
+            'id': 'gd95-04-02.sbd.11622.sbeok.shnf/gd95-04-02d1t04.shn',
+            'ext': 'mp3',
+            'title': 'Stuck Inside of Mobile with the Memphis Blues Again',
+            'creators': ['Grateful Dead'],
+            'duration': 338.31,
+            'track': 'Stuck Inside of Mobile with the Memphis Blues Again',
+            'description': 'md5:764348a470b986f1217ffd38d6ac7b72',
+            'display_id': 'gd95-04-02d1t04.shn',
+            'location': 'Pyramid Arena',
+            'uploader': 'jon@archive.org',
+            'album': '1995-04-02 - Pyramid Arena',
+            'upload_date': '20040519',
+            'track_number': 4,
+            'release_date': '19950402',
+            'timestamp': 1084927901,
+        },
     }]
 
     @staticmethod
@@ -335,7 +355,7 @@ class ArchiveOrgIE(InfoExtractor):
                 info['comments'].append({
                     'id': review.get('review_id'),
                     'author': review.get('reviewer'),
-                    'text': str_or_none(review.get('reviewtitle'), '') + '\n\n' + review.get('reviewbody'),
+                    'text': join_nonempty('reviewtitle', 'reviewbody', from_dict=review, delim='\n\n'),
                     'timestamp': unified_timestamp(review.get('createdate')),
                     'parent': 'root'})
 

yt_dlp/extractor/facebook.py

@@ -563,13 +563,13 @@ class FacebookIE(InfoExtractor):
                 return extract_video_data(try_get(
                     js_data, lambda x: x['jsmods']['instances'], list) or [])
 
-        def extract_dash_manifest(video, formats):
+        def extract_dash_manifest(vid_data, formats, mpd_url=None):
             dash_manifest = traverse_obj(
-                video, 'dash_manifest', 'playlist', 'dash_manifest_xml_string', expected_type=str)
+                vid_data, 'dash_manifest', 'playlist', 'dash_manifest_xml_string', 'manifest_xml', expected_type=str)
             if dash_manifest:
                 formats.extend(self._parse_mpd_formats(
                     compat_etree_fromstring(urllib.parse.unquote_plus(dash_manifest)),
-                    mpd_url=url_or_none(video.get('dash_manifest_url'))))
+                    mpd_url=url_or_none(video.get('dash_manifest_url')) or mpd_url))
 
         def process_formats(info):
             # Downloads with browser's User-Agent are rate limited. Working around
@@ -619,9 +619,12 @@ class FacebookIE(InfoExtractor):
                         video = video['creation_story']
                         video['owner'] = traverse_obj(video, ('short_form_video_context', 'video_owner'))
                         video.update(reel_info)
-                    fmt_data = traverse_obj(video, ('videoDeliveryLegacyFields', {dict})) or video
+
                     formats = []
                     q = qualities(['sd', 'hd'])
+
+                    # Legacy formats extraction
+                    fmt_data = traverse_obj(video, ('videoDeliveryLegacyFields', {dict})) or video
                     for key, format_id in (('playable_url', 'sd'), ('playable_url_quality_hd', 'hd'),
                                            ('playable_url_dash', ''), ('browser_native_hd_url', 'hd'),
                                            ('browser_native_sd_url', 'sd')):
@@ -629,7 +632,7 @@ class FacebookIE(InfoExtractor):
                         if not playable_url:
                             continue
                         if determine_ext(playable_url) == 'mpd':
-                            formats.extend(self._extract_mpd_formats(playable_url, video_id))
+                            formats.extend(self._extract_mpd_formats(playable_url, video_id, fatal=False))
                         else:
                             formats.append({
                                 'format_id': format_id,
@@ -638,6 +641,28 @@ class FacebookIE(InfoExtractor):
                                 'url': playable_url,
                             })
                     extract_dash_manifest(fmt_data, formats)
+
+                    # New videoDeliveryResponse formats extraction
+                    fmt_data = traverse_obj(video, ('videoDeliveryResponseFragment', 'videoDeliveryResponseResult'))
+                    mpd_urls = traverse_obj(fmt_data, ('dash_manifest_urls', ..., 'manifest_url', {url_or_none}))
+                    dash_manifests = traverse_obj(fmt_data, ('dash_manifests', lambda _, v: v['manifest_xml']))
+                    for idx, dash_manifest in enumerate(dash_manifests):
+                        extract_dash_manifest(dash_manifest, formats, mpd_url=traverse_obj(mpd_urls, idx))
+                    if not dash_manifests:
+                        # Only extract from MPD URLs if the manifests are not already provided
+                        for mpd_url in mpd_urls:
+                            formats.extend(self._extract_mpd_formats(mpd_url, video_id, fatal=False))
+                    for prog_fmt in traverse_obj(fmt_data, ('progressive_urls', lambda _, v: v['progressive_url'])):
+                        format_id = traverse_obj(prog_fmt, ('metadata', 'quality', {str.lower}))
+                        formats.append({
+                            'format_id': format_id,
+                            # sd, hd formats w/o resolution info should be deprioritized below DASH
+                            'quality': q(format_id) - 3,
+                            'url': prog_fmt['progressive_url'],
+                        })
+                    for m3u8_url in traverse_obj(fmt_data, ('hls_playlist_urls', ..., 'hls_playlist_url', {url_or_none})):
+                        formats.extend(self._extract_m3u8_formats(m3u8_url, video_id, 'mp4', fatal=False, m3u8_id='hls'))
+
                     if not formats:
                         # Do not append false positive entry w/o any formats
                         return

yt_dlp/extractor/mediasite.py

@@ -31,6 +31,9 @@ class MediasiteIE(InfoExtractor):
                 'ext': 'mp4',
                 'title': 'Lecture: Tuesday, September 20, 2016 - Sir Andrew Wiles',
                 'description': 'Sir Andrew Wiles: “Equations in arithmetic”\\n\\nI will describe some of the interactions between modern number theory and the problem of solving equations in rational numbers or integers\\u0027.',
+                'thumbnail': 'https://hitsmediaweb.h-its.org/Mediasite/FileServer/Presentation/2db6c271681e4f199af3c60d1f82869b1d/423e2016-e080-4ca8-a9bc-b50c39b87caa.jpg?playbackTicket=1b6a49d3e6074616af4dae9cf6952967',
+                'cast': ['Sir Andrew J. Miles, Silver Plaque of the IMU, 1998 | Abel Prize, 2016'],
+                'duration': 2978.0,
                 'timestamp': 1474268400.0,
                 'upload_date': '20160919',
             },
@@ -44,6 +47,7 @@ class MediasiteIE(InfoExtractor):
                 'title': '5) IT-forum 2015-Dag 1  - Dungbeetle -  How and why Rain created a tiny bug tracker for Unity',
                 'timestamp': 1430311380.0,
             },
+            'skip': 'no longer exist',
         },
         {
             'url': 'https://collegerama.tudelft.nl/Mediasite/Play/585a43626e544bdd97aeb71a0ec907a01d',
@@ -54,10 +58,15 @@ class MediasiteIE(InfoExtractor):
                 'title': 'Een nieuwe wereld: waarden, bewustzijn en techniek van de mensheid 2.0.',
                 'description': '',
                 'thumbnail': r're:^https?://.*\.jpg(?:\?.*)?$',
+                'cast': ['H. Wijffels'],
                 'duration': 7713.088,
                 'timestamp': 1413309600,
                 'upload_date': '20141014',
             },
+            'params': {
+                # format 'video1-1.1' HTTP Error 400: Bad Request
+                'skip_download': True,
+            },
         },
         {
             'url': 'https://collegerama.tudelft.nl/Mediasite/Play/86a9ea9f53e149079fbdb4202b521ed21d?catalog=fd32fd35-6c99-466c-89d4-cd3c431bc8a4',
@@ -68,10 +77,14 @@ class MediasiteIE(InfoExtractor):
                 'title': '64ste Vakantiecursus: Afvalwater',
                 'description': 'md5:7fd774865cc69d972f542b157c328305',
                 'thumbnail': r're:^https?://.*\.jpg(?:\?.*?)?$',
+                'cast': ['D.J. van den Berg', 'L.C. Rietveld', 'D. van Halem', 'N.C. van de Giesen', 'J.Q.J.C. Verberk'],
                 'duration': 10853,
                 'timestamp': 1326446400,
                 'upload_date': '20120113',
             },
+            'params': {
+                'skip_download': True,
+            },
         },
         {
             'url': 'http://digitalops.sandia.gov/Mediasite/Play/24aace4429fc450fb5b38cdbf424a66e1d',
@@ -79,11 +92,17 @@ class MediasiteIE(InfoExtractor):
             'info_dict': {
                 'id': '24aace4429fc450fb5b38cdbf424a66e1d',
                 'ext': 'mp4',
-                'title': 'Xyce Software Training - Section 1',
+                'title': 'Xyce Software Training - Section 1 - Apr. 2012',
                 'description': r're:(?s)SAND Number: SAND 2013-7800.{200,}',
-                'upload_date': '20120409',
+                'thumbnail': 'https://digitalops.sandia.gov/Mediasite/FileServer/Presentation/24aace4429fc450fb5b38cdbf424a66e1d/8590199d-2a48-492c-8614-05c502dea8c0.jpg?playbackTicket=aecb83175ff740daace12a13101c2a7f',
-                'timestamp': 1333983600,
+                'cast': ['01400 Computation, Computers \\u0026 Math'],
                 'duration': 7794,
+                'timestamp': 1333983600,
+                'upload_date': '20120409',
+            },
+            'params': {
+                # format 'video1-1.0' HTTP Error 400: Bad Request
+                'skip_download': True,
             },
         },
         {
@@ -267,7 +286,6 @@ class MediasiteIE(InfoExtractor):
                 })
             formats.extend(stream_formats)
 
-        # XXX: Presentation['Presenters']
         # XXX: Presentation['Transcript']
 
         return {
@@ -278,6 +296,7 @@ class MediasiteIE(InfoExtractor):
             'timestamp': float_or_none(presentation.get('UnixTime'), 1000),
             'formats': formats,
             'thumbnails': thumbnails,
+            'cast': traverse_obj(presentation, ('Presenters', ..., 'Name', {str})),
         }