+ def _prepare_live_from_start_formats(self, formats, video_id, live_start_time, url, webpage_url, smuggled_data):
+ EXPIRATION_DURATION = 18_000
+ lock = threading.Lock()
+
+ is_live = True
+ expiration_time = time.time() + EXPIRATION_DURATION
+ formats = [f for f in formats if f.get('is_from_start')]
+
+ def refetch_manifest(format_id):
+ nonlocal formats, expiration_time, is_live
+ if time.time() <= expiration_time:
+ return
+
+ _, _, prs, player_url = self._download_player_responses(url, smuggled_data, video_id, webpage_url)
+ video_details = traverse_obj(
+ prs, (..., 'videoDetails'), expected_type=dict, default=[])
+ microformats = traverse_obj(
+ prs, (..., 'microformat', 'playerMicroformatRenderer'),
+ expected_type=dict, default=[])
+ _, is_live, _, formats = self._list_formats(video_id, microformats, video_details, prs, player_url)
+ expiration_time = time.time() + EXPIRATION_DURATION
+
+ def mpd_feed(format_id):
+ """
+ @returns (manifest_url, manifest_stream_number, is_live) or None
+ """
+ with lock:
+ refetch_manifest(format_id)
+
+ f = next((f for f in formats if f['format_id'] == format_id), None)
+ if not f:
+ self.report_warning(
+ f'Cannot find refreshed manifest for format {format_id}{bug_reports_message()}')
+ return None
+ return f['manifest_url'], f['manifest_stream_number'], is_live
+
+ for f in formats:
+ f['protocol'] = 'http_dash_segments_generator'
+ f['fragments'] = functools.partial(
+ self._live_dash_fragments, f['format_id'], live_start_time, mpd_feed)
+
+ def _live_dash_fragments(self, format_id, live_start_time, mpd_feed, ctx):
+ FETCH_SPAN, MAX_DURATION = 5, 432000
+
+ mpd_url, stream_number, is_live = None, None, True
+
+ begin_index = 0
+ download_start_time = ctx.get('start') or time.time()
+
+ lack_early_segments = download_start_time - (live_start_time or download_start_time) > MAX_DURATION
+ if lack_early_segments:
+ self.report_warning(bug_reports_message(
+ 'Starting download from the last 120 hours of the live stream since '
+ 'YouTube does not have data before that. If you think this is wrong,'), only_once=True)
+ lack_early_segments = True
+
+ known_idx, no_fragment_score, last_segment_url = begin_index, 0, None
+ fragments, fragment_base_url = None, None
+
+ def _extract_sequence_from_mpd(refresh_sequence):
+ nonlocal mpd_url, stream_number, is_live, no_fragment_score, fragments, fragment_base_url
+ # Obtain from MPD's maximum seq value
+ old_mpd_url = mpd_url
+ mpd_url, stream_number, is_live = mpd_feed(format_id) or (mpd_url, stream_number, False)
+ if old_mpd_url == mpd_url and not refresh_sequence:
+ return True, last_seq
+ try:
+ fmts, _ = self._extract_mpd_formats_and_subtitles(
+ mpd_url, None, note=False, errnote=False, fatal=False)
+ except ExtractorError:
+ fmts = None
+ if not fmts:
+ no_fragment_score += 1
+ return False, last_seq
+ fmt_info = next(x for x in fmts if x['manifest_stream_number'] == stream_number)
+ fragments = fmt_info['fragments']
+ fragment_base_url = fmt_info['fragment_base_url']
+ assert fragment_base_url
+
+ _last_seq = int(re.search(r'(?:/|^)sq/(\d+)', fragments[-1]['path']).group(1))
+ return True, _last_seq
+
+ while is_live:
+ fetch_time = time.time()
+ if no_fragment_score > 30:
+ return
+ if last_segment_url:
+ # Obtain from "X-Head-Seqnum" header value from each segment
+ try:
+ urlh = self._request_webpage(
+ last_segment_url, None, note=False, errnote=False, fatal=False)
+ except ExtractorError:
+ urlh = None
+ last_seq = try_get(urlh, lambda x: int_or_none(x.headers['X-Head-Seqnum']))
+ if last_seq is None:
+ no_fragment_score += 1
+ last_segment_url = None
+ continue
+ else:
+ should_retry, last_seq = _extract_sequence_from_mpd(True)
+ if not should_retry:
+ continue
+
+ if known_idx > last_seq:
+ last_segment_url = None
+ continue
+
+ last_seq += 1
+
+ if begin_index < 0 and known_idx < 0:
+ # skip from the start when it's negative value
+ known_idx = last_seq + begin_index
+ if lack_early_segments:
+ known_idx = max(known_idx, last_seq - int(MAX_DURATION // fragments[-1]['duration']))
+ try:
+ for idx in range(known_idx, last_seq):
+ # do not update sequence here or you'll get skipped some part of it
+ should_retry, _ = _extract_sequence_from_mpd(False)
+ if not should_retry:
+ # retry when it gets weird state
+ known_idx = idx - 1
+ raise ExtractorError('breaking out of outer loop')
+ last_segment_url = urljoin(fragment_base_url, 'sq/%d' % idx)
+ yield {
+ 'url': last_segment_url,
+ }
+ if known_idx == last_seq:
+ no_fragment_score += 5
+ else:
+ no_fragment_score = 0
+ known_idx = last_seq
+ except ExtractorError:
+ continue
+
+ time.sleep(max(0, FETCH_SPAN + fetch_time - time.time()))
+