From aa4b0545120becc11a5992384ce52c943da8ead5 Mon Sep 17 00:00:00 2001 From: coletdjnz Date: Fri, 10 Dec 2021 12:43:15 +1300 Subject: [PATCH] [web.archive:youtube] Improve metadata extraction (#1785) Authored-by: coletdjnz --- README.md | 3 + yt_dlp/extractor/archiveorg.py | 467 +++++++++++++++++++++++++-------- 2 files changed, 363 insertions(+), 107 deletions(-) diff --git a/README.md b/README.md index 0eecb5536..4fec2c8b4 100644 --- a/README.md +++ b/README.md @@ -1580,6 +1580,9 @@ The following extractors use this feature: #### vikichannel * `video_types`: Types of videos to download - one or more of `episodes`, `movies`, `clips`, `trailers` +#### youtubewebarchive +* `check_all`: Try to check more at the cost of more requests. One or more of `thumbnails`, `captures` + NOTE: These options may be changed/removed in the future without concern for backward compatibility diff --git a/yt_dlp/extractor/archiveorg.py b/yt_dlp/extractor/archiveorg.py index d90fcb13a..467fe4875 100644 --- a/yt_dlp/extractor/archiveorg.py +++ b/yt_dlp/extractor/archiveorg.py @@ -3,33 +3,36 @@ from __future__ import unicode_literals import re import json - from .common import InfoExtractor -from .youtube import YoutubeIE +from .youtube import YoutubeIE, YoutubeBaseInfoExtractor from ..compat import ( compat_urllib_parse_unquote, compat_urllib_parse_unquote_plus, compat_HTTPError ) from ..utils import ( + bug_reports_message, clean_html, - determine_ext, dict_get, extract_attributes, ExtractorError, + get_element_by_id, HEADRequest, int_or_none, KNOWN_EXTENSIONS, merge_dicts, mimetype2ext, + orderedSet, parse_duration, parse_qs, - RegexNotFoundError, str_to_int, str_or_none, + traverse_obj, try_get, unified_strdate, unified_timestamp, + urlhandle_detect_ext, + url_or_none ) @@ -262,12 +265,12 @@ class YoutubeWebArchiveIE(InfoExtractor): _VALID_URL = r"""(?x)^ (?:https?://)?web\.archive\.org/ (?:web/)? - (?:[0-9A-Za-z_*]+/)? # /web and the version index is optional + (?:(?P[0-9]{14})?[0-9A-Za-z_*]*/)? # /web and the version index is optional (?:https?(?::|%3[Aa])//)? (?: - (?:\w+\.)?youtube\.com/watch(?:\?|%3[fF])(?:[^\#]+(?:&|%26))?v(?:=|%3[dD]) # Youtube URL - |(wayback-fakeurl\.archive\.org/yt/) # Or the internal fake url + (?:\w+\.)?youtube\.com(?::(?:80|443))?/watch(?:\.php)?(?:\?|%3[fF])(?:[^\#]+(?:&|%26))?v(?:=|%3[dD]) # Youtube URL + |(?:wayback-fakeurl\.archive\.org/yt/) # Or the internal fake url ) (?P[0-9A-Za-z_-]{11})(?:%26|\#|&|$) """ @@ -278,141 +281,391 @@ class YoutubeWebArchiveIE(InfoExtractor): 'info_dict': { 'id': 'aYAGB11YrSs', 'ext': 'webm', - 'title': 'Team Fortress 2 - Sandviches!' + 'title': 'Team Fortress 2 - Sandviches!', + 'description': 'md5:4984c0f9a07f349fc5d8e82ab7af4eaf', + 'upload_date': '20110926', + 'uploader': 'Zeurel', + 'channel_id': 'UCukCyHaD-bK3in_pKpfH9Eg', + 'duration': 32, + 'uploader_id': 'Zeurel', + 'uploader_url': 'http://www.youtube.com/user/Zeurel' } - }, - { + }, { # Internal link 'url': 'https://web.archive.org/web/2oe/http://wayback-fakeurl.archive.org/yt/97t7Xj_iBv0', 'info_dict': { 'id': '97t7Xj_iBv0', 'ext': 'mp4', - 'title': 'How Flexible Machines Could Save The World' + 'title': 'Why Machines That Bend Are Better', + 'description': 'md5:00404df2c632d16a674ff8df1ecfbb6c', + 'upload_date': '20190312', + 'uploader': 'Veritasium', + 'channel_id': 'UCHnyfMqiRRG1u-2MsSQLbXA', + 'duration': 771, + 'uploader_id': '1veritasium', + 'uploader_url': 'http://www.youtube.com/user/1veritasium' } - }, - { - # Video from 2012, webm format itag 45. + }, { + # Video from 2012, webm format itag 45. Newest capture is deleted video, with an invalid description. + # Should use the date in the link. Title ends with '- Youtube'. Capture has description in eow-description 'url': 'https://web.archive.org/web/20120712231619/http://www.youtube.com/watch?v=AkhihxRKcrs&gl=US&hl=en', 'info_dict': { 'id': 'AkhihxRKcrs', 'ext': 'webm', - 'title': 'Limited Run: Mondo\'s Modern Classic 1 of 3 (SDCC 2012)' + 'title': 'Limited Run: Mondo\'s Modern Classic 1 of 3 (SDCC 2012)', + 'upload_date': '20120712', + 'duration': 398, + 'description': 'md5:ff4de6a7980cb65d951c2f6966a4f2f3', + 'uploader_id': 'machinima', + 'uploader_url': 'http://www.youtube.com/user/machinima' } - }, - { - # Old flash-only video. Webpage title starts with "YouTube - ". + }, { + # FLV video. Video file URL does not provide itag information 'url': 'https://web.archive.org/web/20081211103536/http://www.youtube.com/watch?v=jNQXAC9IVRw', 'info_dict': { 'id': 'jNQXAC9IVRw', - 'ext': 'unknown_video', - 'title': 'Me at the zoo' + 'ext': 'flv', + 'title': 'Me at the zoo', + 'upload_date': '20050423', + 'channel_id': 'UC4QobU6STFB0P71PMvOGN5A', + 'duration': 19, + 'description': 'md5:10436b12e07ac43ff8df65287a56efb4', + 'uploader_id': 'jawed', + 'uploader_url': 'http://www.youtube.com/user/jawed' } - }, - { - # Flash video with .flv extension (itag 34). Title has prefix "YouTube -" - # Title has some weird unicode characters too. + }, { 'url': 'https://web.archive.org/web/20110712231407/http://www.youtube.com/watch?v=lTx3G6h2xyA', 'info_dict': { 'id': 'lTx3G6h2xyA', 'ext': 'flv', - 'title': '‪Madeon - Pop Culture (live mashup)‬‏' + 'title': 'Madeon - Pop Culture (live mashup)', + 'upload_date': '20110711', + 'uploader': 'Madeon', + 'channel_id': 'UCqMDNf3Pn5L7pcNkuSEeO3w', + 'duration': 204, + 'description': 'md5:f7535343b6eda34a314eff8b85444680', + 'uploader_id': 'itsmadeon', + 'uploader_url': 'http://www.youtube.com/user/itsmadeon' } - }, - { # Some versions of Youtube have have "YouTube" as page title in html (and later rewritten by js). + }, { + # First capture is of dead video, second is the oldest from CDX response. + 'url': 'https://web.archive.org/https://www.youtube.com/watch?v=1JYutPM8O6E', + 'info_dict': { + 'id': '1JYutPM8O6E', + 'ext': 'mp4', + 'title': 'Fake Teen Doctor Strikes AGAIN! - Weekly Weird News', + 'upload_date': '20160218', + 'channel_id': 'UCdIaNUarhzLSXGoItz7BHVA', + 'duration': 1236, + 'description': 'md5:21032bae736421e89c2edf36d1936947', + 'uploader_id': 'MachinimaETC', + 'uploader_url': 'http://www.youtube.com/user/MachinimaETC' + } + }, { + # First capture of dead video, capture date in link links to dead capture. + 'url': 'https://web.archive.org/web/20180803221945/https://www.youtube.com/watch?v=6FPhZJGvf4E', + 'info_dict': { + 'id': '6FPhZJGvf4E', + 'ext': 'mp4', + 'title': 'WTF: Video Games Still Launch BROKEN?! - T.U.G.S.', + 'upload_date': '20160219', + 'channel_id': 'UCdIaNUarhzLSXGoItz7BHVA', + 'duration': 798, + 'description': 'md5:a1dbf12d9a3bd7cb4c5e33b27d77ffe7', + 'uploader_id': 'MachinimaETC', + 'uploader_url': 'http://www.youtube.com/user/MachinimaETC' + }, + 'expected_warnings': [ + r'unable to download capture webpage \(it may not be archived\)' + ] + }, { # Very old YouTube page, has - YouTube in title. + 'url': 'http://web.archive.org/web/20070302011044/http://youtube.com/watch?v=-06-KB9XTzg', + 'info_dict': { + 'id': '-06-KB9XTzg', + 'ext': 'flv', + 'title': 'New Coin Hack!! 100% Safe!!' + } + }, { + 'url': 'web.archive.org/https://www.youtube.com/watch?v=dWW7qP423y8', + 'info_dict': { + 'id': 'dWW7qP423y8', + 'ext': 'mp4', + 'title': 'It\'s Bootleg AirPods Time.', + 'upload_date': '20211021', + 'channel_id': 'UC7Jwj9fkrf1adN4fMmTkpug', + 'channel_url': 'http://www.youtube.com/channel/UC7Jwj9fkrf1adN4fMmTkpug', + 'duration': 810, + 'description': 'md5:7b567f898d8237b256f36c1a07d6d7bc', + 'uploader': 'DankPods', + 'uploader_id': 'UC7Jwj9fkrf1adN4fMmTkpug', + 'uploader_url': 'http://www.youtube.com/channel/UC7Jwj9fkrf1adN4fMmTkpug' + } + }, { + # player response contains '};' See: https://github.com/ytdl-org/youtube-dl/issues/27093 + 'url': 'https://web.archive.org/web/20200827003909if_/http://www.youtube.com/watch?v=6Dh-RL__uN4', + 'info_dict': { + 'id': '6Dh-RL__uN4', + 'ext': 'mp4', + 'title': 'bitch lasagna', + 'upload_date': '20181005', + 'channel_id': 'UC-lHJZR3Gqxm24_Vd_AJ5Yw', + 'channel_url': 'http://www.youtube.com/channel/UC-lHJZR3Gqxm24_Vd_AJ5Yw', + 'duration': 135, + 'description': 'md5:2dbe4051feeff2dab5f41f82bb6d11d0', + 'uploader': 'PewDiePie', + 'uploader_id': 'PewDiePie', + 'uploader_url': 'http://www.youtube.com/user/PewDiePie' + } + }, { 'url': 'https://web.archive.org/web/http://www.youtube.com/watch?v=kH-G_aIBlFw', - 'info_dict': { - 'id': 'kH-G_aIBlFw', - 'ext': 'mp4', - 'title': 'kH-G_aIBlFw' - }, - 'expected_warnings': [ - 'unable to extract title', - ] - }, - { - # First capture is a 302 redirect intermediary page. - 'url': 'https://web.archive.org/web/20050214000000/http://www.youtube.com/watch?v=0altSZ96U4M', - 'info_dict': { - 'id': '0altSZ96U4M', - 'ext': 'mp4', - 'title': '0altSZ96U4M' - }, - 'expected_warnings': [ - 'unable to extract title', - ] - }, - { + 'only_matching': True + }, { + 'url': 'https://web.archive.org/web/20050214000000_if/http://www.youtube.com/watch?v=0altSZ96U4M', + 'only_matching': True + }, { # Video not archived, only capture is unavailable video page 'url': 'https://web.archive.org/web/20210530071008/https://www.youtube.com/watch?v=lHJTf93HL1s&spfreload=10', - 'only_matching': True, - }, - { # Encoded url + 'only_matching': True + }, { # Encoded url 'url': 'https://web.archive.org/web/20120712231619/http%3A//www.youtube.com/watch%3Fgl%3DUS%26v%3DAkhihxRKcrs%26hl%3Den', - 'only_matching': True, - }, - { + 'only_matching': True + }, { 'url': 'https://web.archive.org/web/20120712231619/http%3A//www.youtube.com/watch%3Fv%3DAkhihxRKcrs%26gl%3DUS%26hl%3Den', - 'only_matching': True, + 'only_matching': True + }, { + 'url': 'https://web.archive.org/web/20060527081937/http://www.youtube.com:80/watch.php?v=ELTFsLT73fA&search=soccer', + 'only_matching': True + }, { + 'url': 'https://web.archive.org/http://www.youtube.com:80/watch?v=-05VVye-ffg', + 'only_matching': True } ] + _YT_INITIAL_DATA_RE = r'(?:(?:(?:window\s*\[\s*["\']ytInitialData["\']\s*\]|ytInitialData)\s*=\s*({.+?})\s*;)|%s)' % YoutubeBaseInfoExtractor._YT_INITIAL_DATA_RE + _YT_INITIAL_PLAYER_RESPONSE_RE = r'(?:(?:(?:window\s*\[\s*["\']ytInitialPlayerResponse["\']\s*\]|ytInitialPlayerResponse)\s*=[(\s]*({.+?})[)\s]*;)|%s)' % YoutubeBaseInfoExtractor._YT_INITIAL_PLAYER_RESPONSE_RE + _YT_INITIAL_BOUNDARY_RE = r'(?:(?:var\s+meta|= 2: + # format response to make it easier to use + return list(dict(zip(res[0], v)) for v in res[1:]) + elif not isinstance(res, list) or len(res) != 0: + self.report_warning('Error while parsing CDX API response' + bug_reports_message()) + + def _extract_yt_initial_variable(self, webpage, regex, video_id, name): + return self._parse_json(self._search_regex( + (r'%s\s*%s' % (regex, self._YT_INITIAL_BOUNDARY_RE), + regex), webpage, name, default='{}'), video_id, fatal=False) + + def _extract_webpage_title(self, webpage): + page_title = self._html_search_regex( + r'([^<]*)', webpage, 'title', default='') + # YouTube video pages appear to always have either 'YouTube -' as prefix or '- YouTube' as suffix. + return self._html_search_regex( + r'(?:YouTube\s*-\s*(.*)$)|(?:(.*)\s*-\s*YouTube$)', + page_title, 'title', default='') + + def _extract_metadata(self, video_id, webpage): + + search_meta = ((lambda x: self._html_search_meta(x, webpage, default=None)) if webpage else (lambda x: None)) + player_response = self._extract_yt_initial_variable( + webpage, self._YT_INITIAL_PLAYER_RESPONSE_RE, video_id, 'initial player response') or {} + initial_data = self._extract_yt_initial_variable( + webpage, self._YT_INITIAL_DATA_RE, video_id, 'initial player response') or {} + + initial_data_video = traverse_obj( + initial_data, ('contents', 'twoColumnWatchNextResults', 'results', 'results', 'contents', ..., 'videoPrimaryInfoRenderer'), + expected_type=dict, get_all=False, default={}) + + video_details = traverse_obj( + player_response, 'videoDetails', expected_type=dict, get_all=False, default={}) + + microformats = traverse_obj( + player_response, ('microformat', 'playerMicroformatRenderer'), expected_type=dict, get_all=False, default={}) + + video_title = ( + video_details.get('title') + or YoutubeBaseInfoExtractor._get_text(microformats, 'title') + or YoutubeBaseInfoExtractor._get_text(initial_data_video, 'title') + or self._extract_webpage_title(webpage) + or search_meta(['og:title', 'twitter:title', 'title'])) + + channel_id = str_or_none( + video_details.get('channelId') + or microformats.get('externalChannelId') + or search_meta('channelId') + or self._search_regex( + r'data-channel-external-id=(["\'])(?P(?:(?!\1).)+)\1', # @b45a9e6 + webpage, 'channel id', default=None, group='id')) + channel_url = f'http://www.youtube.com/channel/{channel_id}' if channel_id else None + + duration = int_or_none( + video_details.get('lengthSeconds') + or microformats.get('lengthSeconds') + or parse_duration(search_meta('duration'))) + description = ( + video_details.get('shortDescription') + or YoutubeBaseInfoExtractor._get_text(microformats, 'description') + or clean_html(get_element_by_id('eow-description', webpage)) # @9e6dd23 + or search_meta(['description', 'og:description', 'twitter:description'])) + + uploader = video_details.get('author') + + # Uploader ID and URL + uploader_mobj = re.search( + r'', # @fd05024 + webpage) + if uploader_mobj is not None: + uploader_id, uploader_url = uploader_mobj.group('uploader_id'), uploader_mobj.group('uploader_url') + else: + # @a6211d2 + uploader_url = url_or_none(microformats.get('ownerProfileUrl')) + uploader_id = self._search_regex( + r'(?:user|channel)/([^/]+)', uploader_url or '', 'uploader id', default=None) + + upload_date = unified_strdate( + dict_get(microformats, ('uploadDate', 'publishDate')) + or search_meta(['uploadDate', 'datePublished']) + or self._search_regex( + [r'(?s)id="eow-date.*?>(.*?)', + r'(?:id="watch-uploader-info".*?>.*?|["\']simpleText["\']\s*:\s*["\'])(?:Published|Uploaded|Streamed live|Started) on (.+?)[<"\']'], # @7998520 + webpage, 'upload date', default=None)) + + return { + 'title': video_title, + 'description': description, + 'upload_date': upload_date, + 'uploader': uploader, + 'channel_id': channel_id, + 'channel_url': channel_url, + 'duration': duration, + 'uploader_url': uploader_url, + 'uploader_id': uploader_id, + } + + def _extract_thumbnails(self, video_id): + try_all = 'thumbnails' in self._configuration_arg('check_all') + thumbnail_base_urls = ['http://{server}/vi{webp}/{video_id}'.format( + webp='_webp' if ext == 'webp' else '', video_id=video_id, server=server) + for server in (self._YT_ALL_THUMB_SERVERS if try_all else self._YT_DEFAULT_THUMB_SERVERS) for ext in (('jpg', 'webp') if try_all else ('jpg',))] + + thumbnails = [] + for url in thumbnail_base_urls: + response = self._call_cdx_api( + video_id, url, filters=['mimetype:image/(?:webp|jpeg)'], + collapse=['urlkey'], query={'matchType': 'prefix'}) + if not response: + continue + thumbnails.extend( + { + 'url': (self._WAYBACK_BASE_URL % (int_or_none(thumbnail_dict.get('timestamp')) or self._OLDEST_CAPTURE_DATE)) + thumbnail_dict.get('original'), + 'filesize': int_or_none(thumbnail_dict.get('length')), + 'preference': int_or_none(thumbnail_dict.get('length')) + } for thumbnail_dict in response) + if not try_all: + break + + self._remove_duplicate_formats(thumbnails) + return thumbnails + + def _get_capture_dates(self, video_id, url_date): + capture_dates = [] + # Note: CDX API will not find watch pages with extra params in the url. + response = self._call_cdx_api( + video_id, f'https://www.youtube.com/watch?v={video_id}', + filters=['mimetype:text/html'], collapse=['timestamp:6', 'digest'], query={'matchType': 'prefix'}) or [] + all_captures = sorted([int_or_none(r['timestamp']) for r in response if int_or_none(r['timestamp']) is not None]) + + # Prefer the new polymer UI captures as we support extracting more metadata from them + # WBM captures seem to all switch to this layout ~July 2020 + modern_captures = list(filter(lambda x: x >= 20200701000000, all_captures)) + if modern_captures: + capture_dates.append(modern_captures[0]) + capture_dates.append(url_date) + if all_captures: + capture_dates.append(all_captures[0]) + + if 'captures' in self._configuration_arg('check_all'): + capture_dates.extend(modern_captures + all_captures) + + # Fallbacks if any of the above fail + capture_dates.extend([self._OLDEST_CAPTURE_DATE, self._NEWEST_CAPTURE_DATE]) + return orderedSet(capture_dates) def _real_extract(self, url): - video_id = self._match_id(url) - title = video_id # if we are not able get a title - def _extract_title(webpage): - page_title = self._html_search_regex( - r'([^<]*)', webpage, 'title', fatal=False) or '' - # YouTube video pages appear to always have either 'YouTube -' as suffix or '- YouTube' as prefix. - try: - page_title = self._html_search_regex( - r'(?:YouTube\s*-\s*(.*)$)|(?:(.*)\s*-\s*YouTube$)', - page_title, 'title', default='') - except RegexNotFoundError: - page_title = None + url_date, video_id = self._match_valid_url(url).groups() - if not page_title: - self.report_warning('unable to extract title', video_id=video_id) - return - return page_title - - # If the video is no longer available, the oldest capture may be one before it was removed. - # Setting the capture date in url to early date seems to redirect to earliest capture. - webpage = self._download_webpage( - 'https://web.archive.org/web/20050214000000/http://www.youtube.com/watch?v=%s' % video_id, - video_id=video_id, fatal=False, errnote='unable to download video webpage (probably not archived).') - if webpage: - title = _extract_title(webpage) or title - - # Use link translator mentioned in https://github.com/ytdl-org/youtube-dl/issues/13655 - internal_fake_url = 'https://web.archive.org/web/2oe_/http://wayback-fakeurl.archive.org/yt/%s' % video_id + urlh = None try: - video_file_webpage = self._request_webpage( - HEADRequest(internal_fake_url), video_id, - note='Fetching video file url', expected_status=True) + urlh = self._request_webpage( + HEADRequest('https://web.archive.org/web/2oe_/http://wayback-fakeurl.archive.org/yt/%s' % video_id), + video_id, note='Fetching archived video file url', expected_status=True) except ExtractorError as e: # HTTP Error 404 is expected if the video is not saved. if isinstance(e.cause, compat_HTTPError) and e.cause.code == 404: - raise ExtractorError( - 'HTTP Error %s. Most likely the video is not archived or issue with web.archive.org.' % e.cause.code, + self.raise_no_formats( + 'The requested video is not archived, indexed, or there is an issue with web.archive.org', expected=True) - raise - video_file_url = compat_urllib_parse_unquote(video_file_webpage.url) - video_file_url_qs = parse_qs(video_file_url) + else: + raise - # Attempt to recover any ext & format info from playback url - format = {'url': video_file_url} - itag = try_get(video_file_url_qs, lambda x: x['itag'][0]) - if itag and itag in YoutubeIE._formats: # Naughty access but it works - format.update(YoutubeIE._formats[itag]) - format.update({'format_id': itag}) - else: - mime = try_get(video_file_url_qs, lambda x: x['mime'][0]) - ext = mimetype2ext(mime) or determine_ext(video_file_url) - format.update({'ext': ext}) - return { - 'id': video_id, - 'title': title, - 'formats': [format], - 'duration': str_to_int(try_get(video_file_url_qs, lambda x: x['dur'][0])) - } + capture_dates = self._get_capture_dates(video_id, int_or_none(url_date)) + self.write_debug('Captures to try: ' + ', '.join(str(i) for i in capture_dates if i is not None)) + info = {'id': video_id} + for capture in capture_dates: + if not capture: + continue + webpage = self._download_webpage( + (self._WAYBACK_BASE_URL + 'http://www.youtube.com/watch?v=%s') % (capture, video_id), + video_id=video_id, fatal=False, errnote='unable to download capture webpage (it may not be archived)', + note='Downloading capture webpage') + current_info = self._extract_metadata(video_id, webpage or '') + # Try avoid getting deleted video metadata + if current_info.get('title'): + info = merge_dicts(info, current_info) + if 'captures' not in self._configuration_arg('check_all'): + break + + info['thumbnails'] = self._extract_thumbnails(video_id) + + if urlh: + url = compat_urllib_parse_unquote(urlh.url) + video_file_url_qs = parse_qs(url) + # Attempt to recover any ext & format info from playback url & response headers + format = {'url': url, 'filesize': int_or_none(urlh.headers.get('x-archive-orig-content-length'))} + itag = try_get(video_file_url_qs, lambda x: x['itag'][0]) + if itag and itag in YoutubeIE._formats: + format.update(YoutubeIE._formats[itag]) + format.update({'format_id': itag}) + else: + mime = try_get(video_file_url_qs, lambda x: x['mime'][0]) + ext = (mimetype2ext(mime) + or urlhandle_detect_ext(urlh) + or mimetype2ext(urlh.headers.get('x-archive-guessed-content-type'))) + format.update({'ext': ext}) + info['formats'] = [format] + if not info.get('duration'): + info['duration'] = str_to_int(try_get(video_file_url_qs, lambda x: x['dur'][0])) + + if not info.get('title'): + info['title'] = video_id + return info