mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-27 02:42:30 +00:00
[Youtube] Extract alerts from continuation (#144)
Related: #143 Authored by: colethedj
This commit is contained in:
parent
9e631877f8
commit
f3eaa8dd1c
@ -2796,7 +2796,11 @@ def extract_entries(parent_renderer): # this needs to called again for continua
|
||||
# See: https://github.com/ytdl-org/youtube-dl/issues/28194
|
||||
if response.get('continuationContents') or response.get('onResponseReceivedActions'):
|
||||
break
|
||||
last_error = 'Incomplete data recieved'
|
||||
|
||||
# Youtube may send alerts if there was an issue with the continuation page
|
||||
self._extract_alerts(response, expected=False)
|
||||
|
||||
last_error = 'Incomplete data received'
|
||||
if count >= retries:
|
||||
self._downloader.report_error(last_error)
|
||||
|
||||
@ -2982,23 +2986,35 @@ def _extract_from_playlist(self, item_id, url, data, playlist):
|
||||
self._extract_mix_playlist(playlist, playlist_id),
|
||||
playlist_id=playlist_id, playlist_title=title)
|
||||
|
||||
@staticmethod
|
||||
def _extract_alerts(data):
|
||||
for alert_dict in try_get(data, lambda x: x['alerts'], list) or []:
|
||||
if not isinstance(alert_dict, dict):
|
||||
continue
|
||||
for renderer in alert_dict:
|
||||
alert = alert_dict[renderer]
|
||||
alert_type = alert.get('type')
|
||||
if not alert_type:
|
||||
def _extract_alerts(self, data, expected=False):
|
||||
|
||||
def _real_extract_alerts():
|
||||
for alert_dict in try_get(data, lambda x: x['alerts'], list) or []:
|
||||
if not isinstance(alert_dict, dict):
|
||||
continue
|
||||
message = try_get(alert, lambda x: x['text']['simpleText'], compat_str)
|
||||
if message:
|
||||
yield alert_type, message
|
||||
for run in try_get(alert, lambda x: x['text']['runs'], list) or []:
|
||||
message = try_get(run, lambda x: x['text'], compat_str)
|
||||
for alert in alert_dict.values():
|
||||
alert_type = alert.get('type')
|
||||
if not alert_type:
|
||||
continue
|
||||
message = try_get(alert, lambda x: x['text']['simpleText'], compat_str)
|
||||
if message:
|
||||
yield alert_type, message
|
||||
for run in try_get(alert, lambda x: x['text']['runs'], list) or []:
|
||||
message = try_get(run, lambda x: x['text'], compat_str)
|
||||
if message:
|
||||
yield alert_type, message
|
||||
|
||||
err_msg = None
|
||||
for alert_type, alert_message in _real_extract_alerts():
|
||||
if alert_type.lower() == 'error':
|
||||
if err_msg:
|
||||
self._downloader.report_warning('YouTube said: %s - %s' % ('ERROR', err_msg))
|
||||
err_msg = alert_message
|
||||
else:
|
||||
self._downloader.report_warning('YouTube said: %s - %s' % (alert_type, alert_message))
|
||||
|
||||
if err_msg:
|
||||
raise ExtractorError('YouTube said: %s' % err_msg, expected=expected)
|
||||
|
||||
def _extract_identity_token(self, webpage, item_id):
|
||||
ytcfg = self._extract_ytcfg(item_id, webpage)
|
||||
@ -3024,16 +3040,7 @@ def _extract_webpage(self, url, item_id):
|
||||
url, item_id,
|
||||
'Downloading webpage%s' % (' (retry #%d)' % count if count else ''))
|
||||
data = self._extract_yt_initial_data(item_id, webpage)
|
||||
err_msg = None
|
||||
for alert_type, alert_message in self._extract_alerts(data):
|
||||
if alert_type.lower() == 'error':
|
||||
if err_msg:
|
||||
self._downloader.report_warning('YouTube said: %s - %s' % ('ERROR', err_msg))
|
||||
err_msg = alert_message
|
||||
else:
|
||||
self._downloader.report_warning('YouTube said: %s - %s' % (alert_type, alert_message))
|
||||
if err_msg:
|
||||
raise ExtractorError('YouTube said: %s' % err_msg, expected=True)
|
||||
self._extract_alerts(data, expected=True)
|
||||
if data.get('contents') or data.get('currentVideoEndpoint'):
|
||||
break
|
||||
if count >= retries:
|
||||
|
Loading…
Reference in New Issue
Block a user