Merge 0500dbd905 into 12d8ea8246

Closes #9554
Authored by: coletdjnz, bashonly

Co-authored-by: bashonly <88596187+bashonly@users.noreply.github.com>

README.md

@@ -666,7 +666,7 @@ ## Filesystem Options:
                                     The name of the browser to load cookies
                                     from. Currently supported browsers are:
                                     brave, chrome, chromium, edge, firefox,
-                                    opera, safari, vivaldi. Optionally, the
+                                    opera, safari, vivaldi, whale. Optionally, the
                                     KEYRING used for decrypting Chromium cookies
                                     on Linux, the name/path of the PROFILE to
                                     load cookies from, and the CONTAINER name
@@ -1760,7 +1760,7 @@ # EXTRACTOR ARGUMENTS
 #### youtube
 * `lang`: Prefer translated metadata (`title`, `description` etc) of this language code (case-sensitive). By default, the video primary language metadata is preferred, with a fallback to `en` translated. See [youtube.py](https://github.com/yt-dlp/yt-dlp/blob/c26f9b991a0681fd3ea548d535919cec1fbbd430/yt_dlp/extractor/youtube.py#L381-L390) for list of supported content language codes
 * `skip`: One or more of `hls`, `dash` or `translated_subs` to skip extraction of the m3u8 manifests, dash manifests and [auto-translated subtitles](https://github.com/yt-dlp/yt-dlp/issues/4090#issuecomment-1158102032) respectively
-* `player_client`: Clients to extract video data from. The main clients are `web`, `android` and `ios` with variants `_music`, `_embedded`, `_embedscreen`, `_creator` (e.g. `web_embedded`); and `mweb`, `mweb_embedscreen`, `mediaconnect` and `tv_embedded` (agegate bypass) with no variants. By default, `ios,android,web` is used, but `tv_embedded` and `creator` variants are added as required for age-gated videos. Similarly, the music variants are added for `music.youtube.com` urls. You can use `all` to use all the clients, and `default` for the default clients.
+* `player_client`: Clients to extract video data from. The main clients are `web`, `ios` and `android`, with variants `_music`, `_embedded`, `_embedscreen`, `_creator` (e.g. `web_embedded`); and `mweb`, `mweb_embedscreen` and `tv_embedded` (agegate bypass) with no variants. By default, `ios,web` is used, but `tv_embedded` and `creator` variants are added as required for age-gated videos. Similarly, the music variants are added for `music.youtube.com` urls. The `android` clients will always be given lowest priority since their formats are broken. You can use `all` to use all the clients, and `default` for the default clients.
 * `player_skip`: Skip some network requests that are generally needed for robust extraction. One or more of `configs` (skip client configs), `webpage` (skip initial webpage), `js` (skip js player). While these options can help reduce the number of requests needed or avoid some rate-limiting, they could cause some issues. See [#860](https://github.com/yt-dlp/yt-dlp/pull/860) for more details
 * `player_params`: YouTube player parameters to use for player requests. Will overwrite any default ones set by yt-dlp.
 * `comment_sort`: `top` or `new` (default) - choose comment sorting mode (on YouTube's side)

yt_dlp/cookies.py

@@ -46,7 +46,7 @@
 from .utils._utils import _YDLLogger
 from .utils.networking import normalize_url
 
-CHROMIUM_BASED_BROWSERS = {'brave', 'chrome', 'chromium', 'edge', 'opera', 'vivaldi'}
+CHROMIUM_BASED_BROWSERS = {'brave', 'chrome', 'chromium', 'edge', 'opera', 'vivaldi', 'whale'}
 SUPPORTED_BROWSERS = CHROMIUM_BASED_BROWSERS | {'firefox', 'safari'}
 
 
@@ -219,6 +219,7 @@ def _get_chromium_based_browser_settings(browser_name):
             'edge': os.path.join(appdata_local, R'Microsoft\Edge\User Data'),
             'opera': os.path.join(appdata_roaming, R'Opera Software\Opera Stable'),
             'vivaldi': os.path.join(appdata_local, R'Vivaldi\User Data'),
+            'whale': os.path.join(appdata_local, R'Naver\Naver Whale\User Data'),
         }[browser_name]
 
     elif sys.platform == 'darwin':
@@ -230,6 +231,7 @@ def _get_chromium_based_browser_settings(browser_name):
             'edge': os.path.join(appdata, 'Microsoft Edge'),
             'opera': os.path.join(appdata, 'com.operasoftware.Opera'),
             'vivaldi': os.path.join(appdata, 'Vivaldi'),
+            'whale': os.path.join(appdata, 'Naver/Whale'),
         }[browser_name]
 
     else:
@@ -241,6 +243,7 @@ def _get_chromium_based_browser_settings(browser_name):
             'edge': os.path.join(config, 'microsoft-edge'),
             'opera': os.path.join(config, 'opera'),
             'vivaldi': os.path.join(config, 'vivaldi'),
+            'whale': os.path.join(config, 'naver-whale'),
         }[browser_name]
 
     # Linux keyring names can be determined by snooping on dbus while opening the browser in KDE:
@@ -252,6 +255,7 @@ def _get_chromium_based_browser_settings(browser_name):
         'edge': 'Microsoft Edge' if sys.platform == 'darwin' else 'Chromium',
         'opera': 'Opera' if sys.platform == 'darwin' else 'Chromium',
         'vivaldi': 'Vivaldi' if sys.platform == 'darwin' else 'Chrome',
+        'whale': 'Whale',
     }[browser_name]
 
     browsers_without_profiles = {'opera'}

yt_dlp/extractor/common.py

@@ -957,7 +957,8 @@ def _download_webpage_handle(self, url_or_request, video_id, note=None, errnote=
         if urlh is False:
             assert not fatal
             return False
-        content = self._webpage_read_content(urlh, url_or_request, video_id, note, errnote, fatal, encoding=encoding)
+        content = self._webpage_read_content(urlh, url_or_request, video_id, note, errnote, fatal,
+                                             encoding=encoding, data=data)
         return (content, urlh)
 
     @staticmethod
@@ -1005,8 +1006,10 @@ def __check_blocked(self, content):
                 'Visit http://blocklist.rkn.gov.ru/ for a block reason.',
                 expected=True)
 
-    def _request_dump_filename(self, url, video_id):
-        basen = f'{video_id}_{url}'
+    def _request_dump_filename(self, url, video_id, data=None):
+        if data is not None:
+            data = hashlib.md5(data).hexdigest()
+        basen = join_nonempty(video_id, data, url, delim='_')
         trim_length = self.get_param('trim_file_name') or 240
         if len(basen) > trim_length:
             h = '___' + hashlib.md5(basen.encode('utf-8')).hexdigest()
@@ -1028,16 +1031,18 @@ def __decode_webpage(self, webpage_bytes, encoding, headers):
         except LookupError:
             return webpage_bytes.decode('utf-8', 'replace')
 
-    def _webpage_read_content(self, urlh, url_or_request, video_id, note=None, errnote=None, fatal=True, prefix=None, encoding=None):
+    def _webpage_read_content(self, urlh, url_or_request, video_id, note=None, errnote=None, fatal=True,
+                              prefix=None, encoding=None, data=None):
         webpage_bytes = urlh.read()
         if prefix is not None:
             webpage_bytes = prefix + webpage_bytes
+        url_or_request = self._create_request(url_or_request, data)
         if self.get_param('dump_intermediate_pages', False):
             self.to_screen('Dumping request to ' + urlh.url)
             dump = base64.b64encode(webpage_bytes).decode('ascii')
             self._downloader.to_screen(dump)
         if self.get_param('write_pages'):
-            filename = self._request_dump_filename(urlh.url, video_id)
+            filename = self._request_dump_filename(urlh.url, video_id, url_or_request.data)
             self.to_screen(f'Saving request to {filename}')
             with open(filename, 'wb') as outf:
                 outf.write(webpage_bytes)
@@ -1098,7 +1103,7 @@ def download_content(self, url_or_request, video_id, note=note, errnote=errnote,
                              impersonate=None, require_impersonation=False):
             if self.get_param('load_pages'):
                 url_or_request = self._create_request(url_or_request, data, headers, query)
-                filename = self._request_dump_filename(url_or_request.url, video_id)
+                filename = self._request_dump_filename(url_or_request.url, video_id, url_or_request.data)
                 self.to_screen(f'Loading request from {filename}')
                 try:
                     with open(filename, 'rb') as dumpf:

yt_dlp/extractor/twitch.py

@@ -54,6 +54,7 @@ class TwitchBaseIE(InfoExtractor):
         'VideoMetadata': '49b5b8f268cdeb259d75b58dcb0c1a748e3b575003448a2333dc5cdafd49adad',
         'VideoPlayer_ChapterSelectButtonVideo': '8d2793384aac3773beab5e59bd5d6f585aedb923d292800119e03d40cd0f9b41',
         'VideoPlayer_VODSeekbarPreviewVideo': '07e99e4d56c5a7c67117a154777b0baf85a5ffefa393b213f4bc712ccaf85dd6',
+        'VideoCommentsByOffsetOrCursor': 'b70a3591ff0f4e0313d126c6a1502d79a1c02baebb288227c582044aa76adf6a',
     }
 
     @property
@@ -61,6 +62,14 @@ def _CLIENT_ID(self):
         return self._configuration_arg(
             'client_id', ['ue6666qo983tsx6so1t0vnawi233wa'], ie_key='Twitch', casesense=True)[0]
 
+    @property
+    def _DEVICE_ID(self):
+        return self._configuration_arg('device_id', [None], ie_key='Twitch', casesense=True)[0]
+
+    @property
+    def _CLIENT_INTEGRITY(self):
+        return self._configuration_arg('client_integrity', [None], ie_key='Twitch', casesense=True)[0]
+
     def _perform_login(self, username, password):
         def fail(message):
             raise ExtractorError(
@@ -145,6 +154,14 @@ def _download_base_gql(self, video_id, ops, note, fatal=True):
         gql_auth = self._get_cookies('https://gql.twitch.tv').get('auth-token')
         if gql_auth:
             headers['Authorization'] = 'OAuth ' + gql_auth.value
+
+        # TODO: remove existence checks when the values will be generated
+        if self._DEVICE_ID:
+            headers["X-Device-Id"] = self._DEVICE_ID
+
+        if self._CLIENT_INTEGRITY:
+            headers["Client-Integrity"] = self._CLIENT_INTEGRITY
+
         return self._download_json(
             'https://gql.twitch.tv/gql', video_id, note,
             data=json.dumps(ops).encode(),
@@ -319,6 +336,7 @@ class TwitchVodIE(TwitchBaseIE):
             'view_count': int,
         },
         'params': {
+            'subtitleslangs': ['live_chat'],
             'skip_download': True
         },
     }, {
@@ -545,6 +563,64 @@ def _extract_storyboard(self, item_id, storyboard_json_url, duration):
                 } for path in images],
             }
 
+    def _get_subtitles(self, vod_id):
+        chat_history = []
+        has_more_pages = True
+        pagenum = 1
+        gql_ops = [{
+            'operationName': 'VideoCommentsByOffsetOrCursor',
+            'variables': {'videoID': vod_id}
+            # 'variables.cursor': <filled in in subsequent requests>
+        }]
+
+        while has_more_pages:
+            response = None
+
+            for retry in self.RetryManager():
+                response = self._download_gql(vod_id, gql_ops, 'Downloading chat fragment page %d' % pagenum, fatal=False)
+
+                if response is False:
+                    retry.error = ExtractorError("f'Unable to fetch next chat history fragment.'", video_id=vod_id, ie=self)
+
+                    # TODO: when this happens, should I forget a partial chat history, or is it better to keep it?
+                    #       I think if I keep it, it might be better to persist a warning that it is incomplete
+
+            response_errors = traverse_obj(response, (..., 'errors'))
+            if response_errors:
+                self.report_warning(f"Error response recevied for fetching next chat history fragment: {response_errors}")
+
+            comments_obj = traverse_obj(response, (0, 'data', 'video', 'comments'))
+            chat_history.extend(traverse_obj(comments_obj, ('edges', ..., 'node')))
+
+            has_more_pages = traverse_obj(comments_obj, ('pageInfo', 'hasNextPage'))
+
+            if has_more_pages:
+                cursor = traverse_obj(comments_obj, ('edges', 0, 'cursor'))
+                if cursor is None:
+                    self.report_warning("Cannot continue downloading chat history: cursor is missing. There are additional chat pages to download.")
+                    break
+
+                pagenum += 1
+                gql_ops[0]['variables']['cursor'] = cursor
+
+            if has_more_pages is None:
+                cursor = traverse_obj(comments_obj, ('edges', 0, 'cursor'))
+
+                if cursor is not None:
+                    self.report_warning("Next page indication is missing, but found cursor. Continuing chat history download.")
+                else:  # In this case maintenance might be needed. Purpose is to prevent silent errors.
+                    self.report_warning("Next page indication is missing, and cursor not found.")
+
+        if not chat_history:
+            return
+        else:
+            self.write_debug(f'Extracted {len(chat_history)} chat messages')
+
+        return {'rechat': [{
+            'data': json.dumps(chat_history),
+            'ext': 'twitch-gql-20221228.json'
+        }]}
+
     def _real_extract(self, url):
         vod_id = self._match_id(url)
 
@@ -564,16 +640,8 @@ def _real_extract(self, url):
         if 't' in query:
             info['start_time'] = parse_duration(query['t'][0])
 
-        if info.get('timestamp') is not None:
-            info['subtitles'] = {
-                'rechat': [{
-                    'url': update_url_query(
-                        'https://api.twitch.tv/v5/videos/%s/comments' % vod_id, {
-                            'client_id': self._CLIENT_ID,
-                        }),
-                    'ext': 'json',
-                }],
-            }
+        if info.get('timestamp'):
+            info['__post_extractor'] = lambda: {'requested_subtitles': {'rechat': traverse_obj(self.extract_subtitles(vod_id), ['rechat', 0])}}
 
         return info
 

yt_dlp/extractor/youtube.py

@@ -2353,6 +2353,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
                 'format': '17',  # 3gp format available on android
                 'extractor_args': {'youtube': {'player_client': ['android']}},
             },
+            'skip': 'android client broken',
         },
         {
             # Skip download of additional client configs (remix client config in this case)
@@ -2730,7 +2731,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
                 'heatmap': 'count:100',
             },
             'params': {
-                'extractor_args': {'youtube': {'player_client': ['android'], 'player_skip': ['webpage']}},
+                'extractor_args': {'youtube': {'player_client': ['ios'], 'player_skip': ['webpage']}},
             },
         },
     ]
@@ -3317,7 +3318,36 @@ def _extract_heatmap(self, data):
                 'value': ('intensityScoreNormalized', {float_or_none}),
             })) or None
 
-    def _extract_comment(self, comment_renderer, parent=None):
+    def _extract_comment(self, entities, parent=None):
+        comment_entity_payload = get_first(entities, ('payload', 'commentEntityPayload', {dict}))
+        if not (comment_id := traverse_obj(comment_entity_payload, ('properties', 'commentId', {str}))):
+            return
+
+        toolbar_entity_payload = get_first(entities, ('payload', 'engagementToolbarStateEntityPayload', {dict}))
+        time_text = traverse_obj(comment_entity_payload, ('properties', 'publishedTime', {str})) or ''
+
+        return {
+            'id': comment_id,
+            'parent': parent or 'root',
+            **traverse_obj(comment_entity_payload, {
+                'text': ('properties', 'content', 'content', {str}),
+                'like_count': ('toolbar', 'likeCountA11y', {parse_count}),
+                'author_id': ('author', 'channelId', {self.ucid_or_none}),
+                'author': ('author', 'displayName', {str}),
+                'author_thumbnail': ('author', 'avatarThumbnailUrl', {url_or_none}),
+                'author_is_uploader': ('author', 'isCreator', {bool}),
+                'author_is_verified': ('author', 'isVerified', {bool}),
+                'author_url': ('author', 'channelCommand', 'innertubeCommand', (
+                    ('browseEndpoint', 'canonicalBaseUrl'), ('commandMetadata', 'webCommandMetadata', 'url')
+                ), {lambda x: urljoin('https://www.youtube.com', x)}),
+            }, get_all=False),
+            'is_favorited': (None if toolbar_entity_payload is None else
+                             toolbar_entity_payload.get('heartState') == 'TOOLBAR_HEART_STATE_HEARTED'),
+            '_time_text': time_text,  # FIXME: non-standard, but we need a way of showing that it is an estimate.
+            'timestamp': self._parse_time_text(time_text),
+        }
+
+    def _extract_comment_old(self, comment_renderer, parent=None):
         comment_id = comment_renderer.get('commentId')
         if not comment_id:
             return
@@ -3398,21 +3428,39 @@ def extract_header(contents):
                 break
             return _continuation
 
-        def extract_thread(contents):
+        def extract_thread(contents, entity_payloads):
             if not parent:
                 tracker['current_page_thread'] = 0
             for content in contents:
                 if not parent and tracker['total_parent_comments'] >= max_parents:
                     yield
                 comment_thread_renderer = try_get(content, lambda x: x['commentThreadRenderer'])
-                comment_renderer = get_first(
-                    (comment_thread_renderer, content), [['commentRenderer', ('comment', 'commentRenderer')]],
-                    expected_type=dict, default={})
 
-                comment = self._extract_comment(comment_renderer, parent)
+                # old comment format
+                if not entity_payloads:
+                    comment_renderer = get_first(
+                        (comment_thread_renderer, content), [['commentRenderer', ('comment', 'commentRenderer')]],
+                        expected_type=dict, default={})
+
+                    comment = self._extract_comment_old(comment_renderer, parent)
+
+                # new comment format
+                else:
+                    view_model = (
+                        traverse_obj(comment_thread_renderer, ('commentViewModel', 'commentViewModel', {dict}))
+                        or traverse_obj(content, ('commentViewModel', {dict})))
+                    comment_keys = traverse_obj(view_model, (('commentKey', 'toolbarStateKey'), {str}))
+                    if not comment_keys:
+                        continue
+                    entities = traverse_obj(entity_payloads, lambda _, v: v['entityKey'] in comment_keys)
+                    comment = self._extract_comment(entities, parent)
+                    if comment:
+                        comment['is_pinned'] = traverse_obj(view_model, ('pinnedText', {str})) is not None
+
                 if not comment:
                     continue
                 comment_id = comment['id']
+
                 if comment.get('is_pinned'):
                     tracker['pinned_comment_ids'].add(comment_id)
                 # Sometimes YouTube may break and give us infinite looping comments.
@@ -3505,7 +3553,7 @@ def extract_thread(contents):
             check_get_keys = None
             if not is_forced_continuation and not (tracker['est_total'] == 0 and tracker['running_total'] == 0):
                 check_get_keys = [[*continuation_items_path, ..., (
-                    'commentsHeaderRenderer' if is_first_continuation else ('commentThreadRenderer', 'commentRenderer'))]]
+                    'commentsHeaderRenderer' if is_first_continuation else ('commentThreadRenderer', 'commentViewModel', 'commentRenderer'))]]
             try:
                 response = self._extract_response(
                     item_id=None, query=continuation,
@@ -3529,6 +3577,7 @@ def extract_thread(contents):
                 raise
             is_forced_continuation = False
             continuation = None
+            mutations = traverse_obj(response, ('frameworkUpdates', 'entityBatchUpdate', 'mutations', ..., {dict}))
             for continuation_items in traverse_obj(response, continuation_items_path, expected_type=list, default=[]):
                 if is_first_continuation:
                     continuation = extract_header(continuation_items)
@@ -3537,7 +3586,7 @@ def extract_thread(contents):
                         break
                     continue
 
-                for entry in extract_thread(continuation_items):
+                for entry in extract_thread(continuation_items, mutations):
                     if not entry:
                         return
                     yield entry
@@ -3614,8 +3663,6 @@ def _extract_player_response(self, client, video_id, master_ytcfg, player_ytcfg,
         yt_query = {
             'videoId': video_id,
         }
-        if _split_innertube_client(client)[0] in ('android', 'android_embedscreen'):
-            yt_query['params'] = 'CgIIAQ=='
 
         pp_arg = self._configuration_arg('player_params', [None], casesense=True)[0]
         if pp_arg:
@@ -3631,19 +3678,24 @@ def _extract_player_response(self, client, video_id, master_ytcfg, player_ytcfg,
 
     def _get_requested_clients(self, url, smuggled_data):
         requested_clients = []
-        default = ['ios', 'android', 'web']
+        android_clients = []
+        default = ['ios', 'web']
         allowed_clients = sorted(
             (client for client in INNERTUBE_CLIENTS.keys() if client[:1] != '_'),
             key=lambda client: INNERTUBE_CLIENTS[client]['priority'], reverse=True)
         for client in self._configuration_arg('player_client'):
-            if client in allowed_clients:
-                requested_clients.append(client)
-            elif client == 'default':
+            if client == 'default':
                 requested_clients.extend(default)
             elif client == 'all':
                 requested_clients.extend(allowed_clients)
-            else:
+            elif client not in allowed_clients:
                 self.report_warning(f'Skipping unsupported client {client}')
+            elif client.startswith('android'):
+                android_clients.append(client)
+            else:
+                requested_clients.append(client)
+        # Force deprioritization of broken Android clients for format de-duplication
+        requested_clients.extend(android_clients)
         if not requested_clients:
             requested_clients = default
 
@@ -3862,6 +3914,14 @@ def build_fragments(f):
                     f'{video_id}: Some formats are possibly damaged. They will be deprioritized', only_once=True)
 
             client_name = fmt.get(STREAMING_DATA_CLIENT_NAME)
+            # Android client formats are broken due to integrity check enforcement
+            # Ref: https://github.com/yt-dlp/yt-dlp/issues/9554
+            is_broken = client_name and client_name.startswith(short_client_name('android'))
+            if is_broken:
+                self.report_warning(
+                    f'{video_id}: Android client formats are broken and may yield HTTP Error 403. '
+                    'They will be deprioritized', only_once=True)
+
             name = fmt.get('qualityLabel') or quality.replace('audio_quality_', '') or ''
             fps = int_or_none(fmt.get('fps')) or 0
             dct = {
@@ -3874,7 +3934,7 @@ def build_fragments(f):
                     name, fmt.get('isDrc') and 'DRC',
                     try_get(fmt, lambda x: x['projectionType'].replace('RECTANGULAR', '').lower()),
                     try_get(fmt, lambda x: x['spatialAudioType'].replace('SPATIAL_AUDIO_TYPE_', '').lower()),
-                    throttled and 'THROTTLED', is_damaged and 'DAMAGED',
+                    throttled and 'THROTTLED', is_damaged and 'DAMAGED', is_broken and 'BROKEN',
                     (self.get_param('verbose') or all_formats) and client_name,
                     delim=', '),
                 # Format 22 is likely to be damaged. See https://github.com/yt-dlp/yt-dlp/issues/3372
@@ -3892,8 +3952,8 @@ def build_fragments(f):
                 'language': join_nonempty(audio_track.get('id', '').split('.')[0],
                                           'desc' if language_preference < -1 else '') or None,
                 'language_preference': language_preference,
-                # Strictly de-prioritize damaged and 3gp formats
-                'preference': -10 if is_damaged else -2 if itag == '17' else None,
+                # Strictly de-prioritize broken, damaged and 3gp formats
+                'preference': -20 if is_broken else -10 if is_damaged else -2 if itag == '17' else None,
             }
             mime_mobj = re.match(
                 r'((?:[^/]+)/(?:[^;]+))(?:;\s*codecs="([^"]+)")?', fmt.get('mimeType') or '')