From 4ce57d3b873c2887814cbec03d029533e82f7db5 Mon Sep 17 00:00:00 2001 From: Alard Date: Mon, 27 Mar 2023 19:04:23 +0200 Subject: [PATCH] [ie] Support multi-period MPD streams (#6654) --- yt_dlp/YoutubeDL.py | 3 +- yt_dlp/extractor/common.py | 65 ++++++++++++++++++++++++++++++-------- 2 files changed, 54 insertions(+), 14 deletions(-) diff --git a/yt_dlp/YoutubeDL.py b/yt_dlp/YoutubeDL.py index e7d654d0f..bd20d0896 100644 --- a/yt_dlp/YoutubeDL.py +++ b/yt_dlp/YoutubeDL.py @@ -3483,7 +3483,8 @@ def ffmpeg_fixup(cndn, msg, cls): or info_dict.get('is_live') and self.params.get('hls_use_mpegts') is None, 'Possible MPEG-TS in MP4 container or malformed AAC timestamps', FFmpegFixupM3u8PP) - ffmpeg_fixup(info_dict.get('is_live') and downloader == 'dashsegments', + ffmpeg_fixup(downloader == 'dashsegments' + and (info_dict.get('is_live') or info_dict.get('is_dash_periods')), 'Possible duplicate MOOV atoms', FFmpegFixupDuplicateMoovPP) ffmpeg_fixup(downloader == 'web_socket_fragment', 'Malformed timestamps detected', FFmpegFixupTimestampPP) diff --git a/yt_dlp/extractor/common.py b/yt_dlp/extractor/common.py index af534775f..f56ccaf7e 100644 --- a/yt_dlp/extractor/common.py +++ b/yt_dlp/extractor/common.py @@ -247,6 +247,8 @@ class InfoExtractor: (For internal use only) * http_chunk_size Chunk size for HTTP downloads * ffmpeg_args Extra arguments for ffmpeg downloader + * is_dash_periods Whether the format is a result of merging + multiple DASH periods. RTMP formats can also have the additional fields: page_url, app, play_path, tc_url, flash_version, rtmp_live, rtmp_conn, rtmp_protocol, rtmp_real_time @@ -2530,7 +2532,11 @@ def _extract_mpd_formats(self, *args, **kwargs): self._report_ignoring_subs('DASH') return fmts - def _extract_mpd_formats_and_subtitles( + def _extract_mpd_formats_and_subtitles(self, *args, **kwargs): + periods = self._extract_mpd_periods(*args, **kwargs) + return self._merge_mpd_periods(periods) + + def _extract_mpd_periods( self, mpd_url, video_id, mpd_id=None, note=None, errnote=None, fatal=True, data=None, headers={}, query={}): @@ -2543,17 +2549,16 @@ def _extract_mpd_formats_and_subtitles( errnote='Failed to download MPD manifest' if errnote is None else errnote, fatal=fatal, data=data, headers=headers, query=query) if res is False: - return [], {} + return [] mpd_doc, urlh = res if mpd_doc is None: - return [], {} + return [] # We could have been redirected to a new url when we retrieved our mpd file. mpd_url = urlh.url mpd_base_url = base_url(mpd_url) - return self._parse_mpd_formats_and_subtitles( - mpd_doc, mpd_id, mpd_base_url, mpd_url) + return self._parse_mpd_periods(mpd_doc, mpd_id, mpd_base_url, mpd_url) def _parse_mpd_formats(self, *args, **kwargs): fmts, subs = self._parse_mpd_formats_and_subtitles(*args, **kwargs) @@ -2561,8 +2566,39 @@ def _parse_mpd_formats(self, *args, **kwargs): self._report_ignoring_subs('DASH') return fmts - def _parse_mpd_formats_and_subtitles( - self, mpd_doc, mpd_id=None, mpd_base_url='', mpd_url=None): + def _parse_mpd_formats_and_subtitles(self, *args, **kwargs): + periods = self._parse_mpd_periods(*args, **kwargs) + return self._merge_mpd_periods(periods) + + def _merge_mpd_periods(self, periods): + """ + Combine all formats and subtitles from an MPD manifest into a single list, + by concatenate streams with similar formats. + """ + formats, subtitles = {}, {} + for period in periods: + for f in period['formats']: + assert 'is_dash_periods' not in f, 'format already processed' + f['is_dash_periods'] = True + format_key = tuple(v for k, v in f.items() if k not in ( + ('format_id', 'fragments', 'manifest_stream_number'))) + if format_key not in formats: + formats[format_key] = f + elif 'fragments' in f: + formats[format_key].setdefault('fragments', []).extend(f['fragments']) + + if subtitles and period['subtitles']: + self.report_warning(bug_reports_message( + 'Found subtitles in multiple periods in the DASH manifest; ' + 'if part of the subtitles are missing,' + ), only_once=True) + + for sub_lang, sub_info in period['subtitles'].items(): + subtitles.setdefault(sub_lang, []).extend(sub_info) + + return list(formats.values()), subtitles + + def _parse_mpd_periods(self, mpd_doc, mpd_id=None, mpd_base_url='', mpd_url=None): """ Parse formats from MPD manifest. References: @@ -2641,9 +2677,13 @@ def extract_Initialization(source): return ms_info mpd_duration = parse_duration(mpd_doc.get('mediaPresentationDuration')) - formats, subtitles = [], {} stream_numbers = collections.defaultdict(int) - for period in mpd_doc.findall(_add_ns('Period')): + for period_idx, period in enumerate(mpd_doc.findall(_add_ns('Period'))): + period_entry = { + 'id': period.get('id', f'period-{period_idx}'), + 'formats': [], + 'subtitles': collections.defaultdict(list), + } period_duration = parse_duration(period.get('duration')) or mpd_duration period_ms_info = extract_multisegment_info(period, { 'start_number': 1, @@ -2893,11 +2933,10 @@ def add_segment_url(): if content_type in ('video', 'audio', 'image/jpeg'): f['manifest_stream_number'] = stream_numbers[f['url']] stream_numbers[f['url']] += 1 - formats.append(f) + period_entry['formats'].append(f) elif content_type == 'text': - subtitles.setdefault(lang or 'und', []).append(f) - - return formats, subtitles + period_entry['subtitles'][lang or 'und'].append(f) + yield period_entry def _extract_ism_formats(self, *args, **kwargs): fmts, subs = self._extract_ism_formats_and_subtitles(*args, **kwargs)