import sys if sys.version_info < (3, 9): raise ImportError( f'You are using an unsupported version of Python. Only Python versions 3.9 and above are supported by yt-dlp') # noqa: F541 __license__ = 'The Unlicense' import collections import getpass import itertools import optparse import os import re import traceback import json from .cookies import SUPPORTED_BROWSERS, SUPPORTED_KEYRINGS, CookieLoadError from .downloader.external import get_external_downloader from .extractor import list_extractor_classes from .extractor.adobepass import MSO_INFO from .networking.impersonate import ImpersonateTarget from .options import parseOpts from .postprocessor import ( FFmpegExtractAudioPP, FFmpegMergerPP, FFmpegPostProcessor, FFmpegSubtitlesConvertorPP, FFmpegThumbnailsConvertorPP, FFmpegVideoConvertorPP, FFmpegVideoRemuxerPP, MetadataFromFieldPP, MetadataParserPP, ) from .update import Updater from .utils import ( Config, NO_DEFAULT, POSTPROCESS_WHEN, DateRange, DownloadCancelled, DownloadError, FormatSorter, GeoUtils, PlaylistEntries, SameFileError, download_range_func, expand_path, float_or_none, format_field, int_or_none, join_nonempty, match_filter_func, parse_bytes, parse_duration, preferredencoding, read_batch_urls, read_stdin, render_table, setproctitle, shell_quote, traverse_obj, variadic, write_string, ) from .utils.networking import std_headers from .utils._utils import _UnsafeExtensionError from .YoutubeDL import YoutubeDL _IN_CLI = False def _exit(status=0, *args): for msg in args: sys.stderr.write(msg) raise SystemExit(status) def get_urls(urls, batchfile, verbose): """ @param verbose -1: quiet, 0: normal, 1: verbose """ batch_urls = [] if batchfile is not None: try: batch_urls = read_batch_urls( read_stdin(None if verbose == -1 else 'URLs') if batchfile == '-' else open(expand_path(batchfile), encoding='utf-8', errors='ignore')) if verbose == 1: write_string('[debug] Batch file urls: ' + repr(batch_urls) + '\n') except OSError: _exit(f'ERROR: batch file {batchfile} could not be read') _enc = preferredencoding() return [ url.strip().decode(_enc, 'ignore') if isinstance(url, bytes) else url.strip() for url in batch_urls + urls] def print_extractor_information(opts, urls): out = '' if opts.list_extractors: # Importing GenericIE is currently slow since it imports YoutubeIE from .extractor.generic import GenericIE urls = dict.fromkeys(urls, False) for ie in list_extractor_classes(opts.age_limit): out += ie.IE_NAME + (' (CURRENTLY BROKEN)' if not ie.working() else '') + '\n' if ie == GenericIE: matched_urls = [url for url, matched in urls.items() if not matched] else: matched_urls = tuple(filter(ie.suitable, urls.keys())) urls.update(dict.fromkeys(matched_urls, True)) out += ''.join(f' {url}\n' for url in matched_urls) elif opts.list_extractor_descriptions: _SEARCHES = ('cute kittens', 'slithering pythons', 'falling cat', 'angry poodle', 'purple fish', 'running tortoise', 'sleeping bunny', 'burping cow') out = '\n'.join( ie.description(markdown=False, search_examples=_SEARCHES) for ie in list_extractor_classes(opts.age_limit) if ie.working() and ie.IE_DESC is not False) elif opts.ap_list_mso: out = 'Supported TV Providers:\n{}\n'.format(render_table( ['mso', 'mso name'], [[mso_id, mso_info['name']] for mso_id, mso_info in MSO_INFO.items()])) elif opts.list_extractors_json: from .extractor.generic import GenericIE dicts = [] e_index = 0 urls = dict.fromkeys(urls, False) if len(urls): for ie in gen_extractors(): if ie == GenericIE: matched_urls = [url for url, matched in urls.items() if not matched] else: matched_urls = tuple(filter(ie.suitable, urls.keys())) urls.update(dict.fromkeys(matched_urls, True)) # show only extractor with matched URL if len(matched_urls): data = {'index': e_index, 'name': ie.IE_NAME, 'desc': ie.IE_DESC if ie.IE_DESC else '', 'working': ie.working(), 'enabled': ie.is_enabled(), 'return_type': ie.return_type(), 'regex_urls': ie.list_regex_url(), 'matched_urls': matched_urls, } e_index += 1 dicts.append(data) else: # show all extractors for ie in gen_extractors(): data = {'index': e_index, 'name': ie.IE_NAME, 'desc': ie.IE_DESC if ie.IE_DESC else '', 'working': ie.working(), 'enabled': ie.is_enabled(), 'return_type': ie.return_type(), 'regex_urls': ie.list_regex_url(), } dicts.append(data) e_index += 1 out = json.dumps(dicts, indent=4) else: return False write_string(out, out=sys.stdout) return True def set_compat_opts(opts): def _unused_compat_opt(name): if name not in opts.compat_opts: return False opts.compat_opts.discard(name) opts.compat_opts.update([f'*{name}']) return True def set_default_compat(compat_name, opt_name, default=True, remove_compat=True): attr = getattr(opts, opt_name) if compat_name in opts.compat_opts: if attr is None: setattr(opts, opt_name, not default) return True else: if remove_compat: _unused_compat_opt(compat_name) return False elif attr is None: setattr(opts, opt_name, default) return None set_default_compat('abort-on-error', 'ignoreerrors', 'only_download') set_default_compat('no-playlist-metafiles', 'allow_playlist_files') set_default_compat('no-clean-infojson', 'clean_infojson') if 'no-attach-info-json' in opts.compat_opts: if opts.embed_infojson: _unused_compat_opt('no-attach-info-json') else: opts.embed_infojson = False if 'format-sort' in opts.compat_opts: opts.format_sort.extend(FormatSorter.ytdl_default) elif 'prefer-vp9-sort' in opts.compat_opts: opts.format_sort.extend(FormatSorter._prefer_vp9_sort) _video_multistreams_set = set_default_compat('multistreams', 'allow_multiple_video_streams', False, remove_compat=False) _audio_multistreams_set = set_default_compat('multistreams', 'allow_multiple_audio_streams', False, remove_compat=False) if _video_multistreams_set is False and _audio_multistreams_set is False: _unused_compat_opt('multistreams') if 'filename' in opts.compat_opts: if opts.outtmpl.get('default') is None: opts.outtmpl.update({'default': '%(title)s-%(id)s.%(ext)s'}) else: _unused_compat_opt('filename') def validate_options(opts): def validate(cndn, name, value=None, msg=None): if cndn: return True raise ValueError((msg or 'invalid {name} "{value}" given').format(name=name, value=value)) def validate_in(name, value, items, msg=None): return validate(value is None or value in items, name, value, msg) def validate_regex(name, value, regex): return validate(value is None or re.match(regex, value), name, value) def validate_positive(name, value, strict=False): return validate(value is None or value > 0 or (not strict and value == 0), name, value, '{name} "{value}" must be positive' + ('' if strict else ' or 0')) def validate_minmax(min_val, max_val, min_name, max_name=None): if max_val is None or min_val is None or max_val >= min_val: return if not max_name: min_name, max_name = f'min {min_name}', f'max {min_name}' raise ValueError(f'{max_name} "{max_val}" must be must be greater than or equal to {min_name} "{min_val}"') # Usernames and passwords validate(sum(map(bool, (opts.usenetrc, opts.netrc_cmd, opts.username))) <= 1, '.netrc', msg='{name}, netrc command and username/password are mutually exclusive options') validate(opts.password is None or opts.username is not None, 'account username', msg='{name} missing') validate(opts.ap_password is None or opts.ap_username is not None, 'TV Provider account username', msg='{name} missing') validate_in('TV Provider', opts.ap_mso, MSO_INFO, 'Unsupported {name} "{value}", use --ap-list-mso to get a list of supported TV Providers') # Numbers validate_positive('autonumber start', opts.autonumber_start) validate_positive('autonumber size', opts.autonumber_size, True) validate_positive('concurrent fragments', opts.concurrent_fragment_downloads, True) validate_positive('playlist start', opts.playliststart, True) if opts.playlistend != -1: validate_minmax(opts.playliststart, opts.playlistend, 'playlist start', 'playlist end') # Time ranges validate_positive('subtitles sleep interval', opts.sleep_interval_subtitles) validate_positive('requests sleep interval', opts.sleep_interval_requests) validate_positive('sleep interval', opts.sleep_interval) validate_positive('max sleep interval', opts.max_sleep_interval) if opts.sleep_interval is None: validate( opts.max_sleep_interval is None, 'min sleep interval', msg='{name} must be specified; use --min-sleep-interval') elif opts.max_sleep_interval is None: opts.max_sleep_interval = opts.sleep_interval else: validate_minmax(opts.sleep_interval, opts.max_sleep_interval, 'sleep interval') if opts.wait_for_video is not None: min_wait, max_wait, *_ = map(parse_duration, [*opts.wait_for_video.split('-', 1), None]) validate(min_wait is not None and not (max_wait is None and '-' in opts.wait_for_video), 'time range to wait for video', opts.wait_for_video) validate_minmax(min_wait, max_wait, 'time range to wait for video') opts.wait_for_video = (min_wait, max_wait) # Format sort for f in opts.format_sort: validate_regex('format sorting', f, FormatSorter.regex) # Postprocessor formats if opts.convertsubtitles == 'none': opts.convertsubtitles = None if opts.convertthumbnails == 'none': opts.convertthumbnails = None validate_regex('merge output format', opts.merge_output_format, r'({0})(/({0}))*'.format('|'.join(map(re.escape, FFmpegMergerPP.SUPPORTED_EXTS)))) validate_regex('audio format', opts.audioformat, FFmpegExtractAudioPP.FORMAT_RE) validate_in('subtitle format', opts.convertsubtitles, FFmpegSubtitlesConvertorPP.SUPPORTED_EXTS) validate_regex('thumbnail format', opts.convertthumbnails, FFmpegThumbnailsConvertorPP.FORMAT_RE) validate_regex('recode video format', opts.recodevideo, FFmpegVideoConvertorPP.FORMAT_RE) validate_regex('remux video format', opts.remuxvideo, FFmpegVideoRemuxerPP.FORMAT_RE) if opts.audioquality: opts.audioquality = opts.audioquality.strip('k').strip('K') # int_or_none prevents inf, nan validate_positive('audio quality', int_or_none(float_or_none(opts.audioquality), default=0)) # Retries def parse_retries(name, value): if value is None: return None elif value in ('inf', 'infinite'): return float('inf') try: return int(value) except (TypeError, ValueError): validate(False, f'{name} retry count', value) opts.retries = parse_retries('download', opts.retries) opts.fragment_retries = parse_retries('fragment', opts.fragment_retries) opts.extractor_retries = parse_retries('extractor', opts.extractor_retries) opts.file_access_retries = parse_retries('file access', opts.file_access_retries) # Retry sleep function def parse_sleep_func(expr): NUMBER_RE = r'\d+(?:\.\d+)?' op, start, limit, step, *_ = (*tuple(re.fullmatch( rf'(?:(linear|exp)=)?({NUMBER_RE})(?::({NUMBER_RE})?)?(?::({NUMBER_RE}))?', expr.strip()).groups()), None, None) if op == 'exp': return lambda n: min(float(start) * (float(step or 2) ** n), float(limit or 'inf')) else: default_step = start if op or limit else 0 return lambda n: min(float(start) + float(step or default_step) * n, float(limit or 'inf')) for key, expr in opts.retry_sleep.items(): if not expr: del opts.retry_sleep[key] continue try: opts.retry_sleep[key] = parse_sleep_func(expr) except AttributeError: raise ValueError(f'invalid {key} retry sleep expression {expr!r}') # Bytes def validate_bytes(name, value): if value is None: return None numeric_limit = parse_bytes(value) validate(numeric_limit is not None, 'rate limit', value) return numeric_limit opts.ratelimit = validate_bytes('rate limit', opts.ratelimit) opts.throttledratelimit = validate_bytes('throttled rate limit', opts.throttledratelimit) opts.min_filesize = validate_bytes('min filesize', opts.min_filesize) opts.max_filesize = validate_bytes('max filesize', opts.max_filesize) opts.buffersize = validate_bytes('buffer size', opts.buffersize) opts.http_chunk_size = validate_bytes('http chunk size', opts.http_chunk_size) # Output templates def validate_outtmpl(tmpl, msg): err = YoutubeDL.validate_outtmpl(tmpl) if err: raise ValueError(f'invalid {msg} "{tmpl}": {err}') for k, tmpl in opts.outtmpl.items(): validate_outtmpl(tmpl, f'{k} output template') for type_, tmpl_list in opts.forceprint.items(): for tmpl in tmpl_list: validate_outtmpl(tmpl, f'{type_} print template') for type_, tmpl_list in opts.print_to_file.items(): for tmpl, file in tmpl_list: validate_outtmpl(tmpl, f'{type_} print to file template') validate_outtmpl(file, f'{type_} print to file filename') validate_outtmpl(opts.sponsorblock_chapter_title, 'SponsorBlock chapter title') for k, tmpl in opts.progress_template.items(): k = f'{k[:-6]} console title' if '-title' in k else f'{k} progress' validate_outtmpl(tmpl, f'{k} template') outtmpl_default = opts.outtmpl.get('default') if outtmpl_default == '': opts.skip_download = None del opts.outtmpl['default'] def parse_chapters(name, value, advanced=False): parse_timestamp = lambda x: float('inf') if x in ('inf', 'infinite') else parse_duration(x) TIMESTAMP_RE = r'''(?x)(?: (?P-?)(?P[^-]+) )?\s*-\s*(?: (?P-?)(?P[^-]+) )?''' chapters, ranges, from_url = [], [], False for regex in value or []: if advanced and regex == '*from-url': from_url = True continue elif not regex.startswith('*'): try: chapters.append(re.compile(regex)) except re.error as err: raise ValueError(f'invalid {name} regex "{regex}" - {err}') continue for range_ in map(str.strip, regex[1:].split(',')): mobj = range_ != '-' and re.fullmatch(TIMESTAMP_RE, range_) dur = mobj and [parse_timestamp(mobj.group('start') or '0'), parse_timestamp(mobj.group('end') or 'inf')] signs = mobj and (mobj.group('start_sign'), mobj.group('end_sign')) err = None if None in (dur or [None]): err = 'Must be of the form "*start-end"' elif not advanced and any(signs): err = 'Negative timestamps are not allowed' else: dur[0] *= -1 if signs[0] else 1 dur[1] *= -1 if signs[1] else 1 if dur[1] == float('-inf'): err = '"-inf" is not a valid end' if err: raise ValueError(f'invalid {name} time range "{regex}". {err}') ranges.append(dur) return chapters, ranges, from_url opts.remove_chapters, opts.remove_ranges, _ = parse_chapters('--remove-chapters', opts.remove_chapters) opts.download_ranges = download_range_func(*parse_chapters('--download-sections', opts.download_ranges, True)) # Cookies from browser if opts.cookiesfrombrowser: container = None mobj = re.fullmatch(r'''(?x) (?P[^+:]+) (?:\s*\+\s*(?P[^:]+))? (?:\s*:\s*(?!:)(?P.+?))? (?:\s*::\s*(?P.+))? ''', opts.cookiesfrombrowser) if mobj is None: raise ValueError(f'invalid cookies from browser arguments: {opts.cookiesfrombrowser}') browser_name, keyring, profile, container = mobj.group('name', 'keyring', 'profile', 'container') browser_name = browser_name.lower() if browser_name not in SUPPORTED_BROWSERS: raise ValueError(f'unsupported browser specified for cookies: "{browser_name}". ' f'Supported browsers are: {", ".join(sorted(SUPPORTED_BROWSERS))}') if keyring is not None: keyring = keyring.upper() if keyring not in SUPPORTED_KEYRINGS: raise ValueError(f'unsupported keyring specified for cookies: "{keyring}". ' f'Supported keyrings are: {", ".join(sorted(SUPPORTED_KEYRINGS))}') opts.cookiesfrombrowser = (browser_name, profile, keyring, container) if opts.impersonate is not None: opts.impersonate = ImpersonateTarget.from_str(opts.impersonate.lower()) # MetadataParser def metadataparser_actions(f): if isinstance(f, str): cmd = f'--parse-metadata {shell_quote(f)}' try: actions = [MetadataFromFieldPP.to_action(f)] except Exception as err: raise ValueError(f'{cmd} is invalid; {err}') else: cmd = f'--replace-in-metadata {shell_quote(f)}' actions = ((MetadataParserPP.Actions.REPLACE, x, *f[1:]) for x in f[0].split(',')) for action in actions: try: MetadataParserPP.validate_action(*action) except Exception as err: raise ValueError(f'{cmd} is invalid; {err}') yield action if opts.metafromtitle is not None: opts.parse_metadata.setdefault('pre_process', []).append(f'title:{opts.metafromtitle}') opts.parse_metadata = { k: list(itertools.chain(*map(metadataparser_actions, v))) for k, v in opts.parse_metadata.items() } # Other options if opts.playlist_items is not None: try: tuple(PlaylistEntries.parse_playlist_items(opts.playlist_items)) except Exception as err: raise ValueError(f'Invalid playlist-items {opts.playlist_items!r}: {err}') opts.geo_bypass_country, opts.geo_bypass_ip_block = None, None if opts.geo_bypass.lower() not in ('default', 'never'): try: GeoUtils.random_ipv4(opts.geo_bypass) except Exception: raise ValueError(f'Unsupported --xff "{opts.geo_bypass}"') if len(opts.geo_bypass) == 2: opts.geo_bypass_country = opts.geo_bypass else: opts.geo_bypass_ip_block = opts.geo_bypass opts.geo_bypass = opts.geo_bypass.lower() != 'never' opts.match_filter = match_filter_func(opts.match_filter, opts.breaking_match_filter) if opts.download_archive is not None: opts.download_archive = expand_path(opts.download_archive) if opts.ffmpeg_location is not None: opts.ffmpeg_location = expand_path(opts.ffmpeg_location) if opts.user_agent is not None: opts.headers.setdefault('User-Agent', opts.user_agent) if opts.referer is not None: opts.headers.setdefault('Referer', opts.referer) if opts.no_sponsorblock: opts.sponsorblock_mark = opts.sponsorblock_remove = set() default_downloader = None for proto, path in opts.external_downloader.items(): if path == 'native': continue ed = get_external_downloader(path) if ed is None: raise ValueError( f'No such {format_field(proto, None, "%s ", ignore="default")}external downloader "{path}"') elif ed and proto == 'default': default_downloader = ed.get_basename() for policy in opts.color.values(): if policy not in ('always', 'auto', 'auto-tty', 'no_color', 'no_color-tty', 'never'): raise ValueError(f'"{policy}" is not a valid color policy') warnings, deprecation_warnings = [], [] # Common mistake: -f best if opts.format == 'best': warnings.append('.\n '.join(( '"-f best" selects the best pre-merged format which is often not the best option', 'To let yt-dlp download and merge the best available formats, simply do not pass any format selection', 'If you know what you are doing and want only the best pre-merged format, use "-f b" instead to suppress this warning'))) # --(postprocessor/downloader)-args without name def report_args_compat(name, value, key1, key2=None, where=None): if key1 in value and key2 not in value: warnings.append(f'{name.title()} arguments given without specifying name. ' f'The arguments will be given to {where or f"all {name}s"}') return True return False if report_args_compat('external downloader', opts.external_downloader_args, 'default', where=default_downloader) and default_downloader: # Compat with youtube-dl's behavior. See https://github.com/ytdl-org/youtube-dl/commit/49c5293014bc11ec8c009856cd63cffa6296c1e1 opts.external_downloader_args.setdefault(default_downloader, opts.external_downloader_args.pop('default')) if report_args_compat('post-processor', opts.postprocessor_args, 'default-compat', 'default'): opts.postprocessor_args['default'] = opts.postprocessor_args.pop('default-compat') opts.postprocessor_args.setdefault('sponskrub', []) def report_conflict(arg1, opt1, arg2='--allow-unplayable-formats', opt2='allow_unplayable_formats', val1=NO_DEFAULT, val2=NO_DEFAULT, default=False): if val2 is NO_DEFAULT: val2 = getattr(opts, opt2) if not val2: return if val1 is NO_DEFAULT: val1 = getattr(opts, opt1) if val1: warnings.append(f'{arg1} is ignored since {arg2} was given') setattr(opts, opt1, default) # Conflicting options report_conflict('--playlist-reverse', 'playlist_reverse', '--playlist-random', 'playlist_random') report_conflict('--playlist-reverse', 'playlist_reverse', '--lazy-playlist', 'lazy_playlist') report_conflict('--playlist-random', 'playlist_random', '--lazy-playlist', 'lazy_playlist') report_conflict('--dateafter', 'dateafter', '--date', 'date', default=None) report_conflict('--datebefore', 'datebefore', '--date', 'date', default=None) report_conflict('--exec-before-download', 'exec_before_dl_cmd', '"--exec before_dl:"', 'exec_cmd', val2=opts.exec_cmd.get('before_dl')) report_conflict('--id', 'useid', '--output', 'outtmpl', val2=opts.outtmpl.get('default')) report_conflict('--remux-video', 'remuxvideo', '--recode-video', 'recodevideo') report_conflict('--sponskrub', 'sponskrub', '--remove-chapters', 'remove_chapters') report_conflict('--sponskrub', 'sponskrub', '--sponsorblock-mark', 'sponsorblock_mark') report_conflict('--sponskrub', 'sponskrub', '--sponsorblock-remove', 'sponsorblock_remove') report_conflict('--sponskrub-cut', 'sponskrub_cut', '--split-chapter', 'split_chapters', val1=opts.sponskrub and opts.sponskrub_cut) # Conflicts with --allow-unplayable-formats report_conflict('--embed-metadata', 'addmetadata') report_conflict('--embed-chapters', 'addchapters') report_conflict('--embed-info-json', 'embed_infojson') report_conflict('--embed-subs', 'embedsubtitles') report_conflict('--embed-thumbnail', 'embedthumbnail') report_conflict('--extract-audio', 'extractaudio') report_conflict('--fixup', 'fixup', val1=opts.fixup not in (None, 'never', 'ignore'), default='never') report_conflict('--recode-video', 'recodevideo') report_conflict('--remove-chapters', 'remove_chapters', default=[]) report_conflict('--remux-video', 'remuxvideo') report_conflict('--sponskrub', 'sponskrub') report_conflict('--sponsorblock-remove', 'sponsorblock_remove', default=set()) report_conflict('--xattrs', 'xattrs') # Fully deprecated options def report_deprecation(val, old, new=None): if not val: return deprecation_warnings.append( f'{old} is deprecated and may be removed in a future version. Use {new} instead' if new else f'{old} is deprecated and may not work as expected') report_deprecation(opts.sponskrub, '--sponskrub', '--sponsorblock-mark or --sponsorblock-remove') report_deprecation(not opts.prefer_ffmpeg, '--prefer-avconv', 'ffmpeg') # report_deprecation(opts.include_ads, '--include-ads') # We may re-implement this in future # report_deprecation(opts.call_home, '--call-home') # We may re-implement this in future # report_deprecation(opts.writeannotations, '--write-annotations') # It's just that no website has it # Dependent options opts.date = DateRange.day(opts.date) if opts.date else DateRange(opts.dateafter, opts.datebefore) if opts.exec_before_dl_cmd: opts.exec_cmd['before_dl'] = opts.exec_before_dl_cmd if opts.useid: # --id is not deprecated in youtube-dl opts.outtmpl['default'] = '%(id)s.%(ext)s' if opts.overwrites: # --force-overwrites implies --no-continue opts.continue_dl = False if (opts.addmetadata or opts.sponsorblock_mark) and opts.addchapters is None: # Add chapters when adding metadata or marking sponsors opts.addchapters = True if opts.extractaudio and not opts.keepvideo and opts.format is None: # Do not unnecessarily download audio opts.format = 'bestaudio/best' if opts.getcomments and opts.writeinfojson is None and not opts.embed_infojson: # If JSON is not printed anywhere, but comments are requested, save it to file if not opts.dumpjson or opts.print_json or opts.dump_single_json: opts.writeinfojson = True if opts.allsubtitles and not (opts.embedsubtitles or opts.writeautomaticsub): # --all-sub automatically sets --write-sub if --write-auto-sub is not given opts.writesubtitles = True if opts.addmetadata and opts.embed_infojson is None: # If embedding metadata and infojson is present, embed it opts.embed_infojson = 'if_exists' # Ask for passwords if opts.username is not None and opts.password is None: opts.password = getpass.getpass('Type account password and press [Return]: ') if opts.ap_username is not None and opts.ap_password is None: opts.ap_password = getpass.getpass('Type TV provider account password and press [Return]: ') # compat option changes global state destructively; only allow from cli if 'allow-unsafe-ext' in opts.compat_opts: warnings.append( 'Using allow-unsafe-ext opens you up to potential attacks. ' 'Use with great care!') _UnsafeExtensionError.sanitize_extension = lambda x, prepend=False: x return warnings, deprecation_warnings def get_postprocessors(opts): yield from opts.add_postprocessors for when, actions in opts.parse_metadata.items(): yield { 'key': 'MetadataParser', 'actions': actions, 'when': when, } sponsorblock_query = opts.sponsorblock_mark | opts.sponsorblock_remove if sponsorblock_query: yield { 'key': 'SponsorBlock', 'categories': sponsorblock_query, 'api': opts.sponsorblock_api, 'when': 'after_filter', } if opts.convertsubtitles: yield { 'key': 'FFmpegSubtitlesConvertor', 'format': opts.convertsubtitles, 'when': 'before_dl', } if opts.convertthumbnails: yield { 'key': 'FFmpegThumbnailsConvertor', 'format': opts.convertthumbnails, 'when': 'before_dl', } if opts.extractaudio: yield { 'key': 'FFmpegExtractAudio', 'preferredcodec': opts.audioformat, 'preferredquality': opts.audioquality, 'nopostoverwrites': opts.nopostoverwrites, } if opts.remuxvideo: yield { 'key': 'FFmpegVideoRemuxer', 'preferedformat': opts.remuxvideo, } if opts.recodevideo: yield { 'key': 'FFmpegVideoConvertor', 'preferedformat': opts.recodevideo, } # If ModifyChapters is going to remove chapters, subtitles must already be in the container. if opts.embedsubtitles: keep_subs = 'no-keep-subs' not in opts.compat_opts yield { 'key': 'FFmpegEmbedSubtitle', # already_have_subtitle = True prevents the file from being deleted after embedding 'already_have_subtitle': opts.writesubtitles and keep_subs, } if not opts.writeautomaticsub and keep_subs: opts.writesubtitles = True # ModifyChapters must run before FFmpegMetadataPP if opts.remove_chapters or sponsorblock_query: yield { 'key': 'ModifyChapters', 'remove_chapters_patterns': opts.remove_chapters, 'remove_sponsor_segments': opts.sponsorblock_remove, 'remove_ranges': opts.remove_ranges, 'sponsorblock_chapter_title': opts.sponsorblock_chapter_title, 'force_keyframes': opts.force_keyframes_at_cuts, } # FFmpegMetadataPP should be run after FFmpegVideoConvertorPP and # FFmpegExtractAudioPP as containers before conversion may not support # metadata (3gp, webm, etc.) # By default ffmpeg preserves metadata applicable for both # source and target containers. From this point the container won't change, # so metadata can be added here. if opts.addmetadata or opts.addchapters or opts.embed_infojson: yield { 'key': 'FFmpegMetadata', 'add_chapters': opts.addchapters, 'add_metadata': opts.addmetadata, 'add_infojson': opts.embed_infojson, } # Deprecated # This should be above EmbedThumbnail since sponskrub removes the thumbnail attachment # but must be below EmbedSubtitle and FFmpegMetadata # See https://github.com/yt-dlp/yt-dlp/issues/204 , https://github.com/faissaloo/SponSkrub/issues/29 # If opts.sponskrub is None, sponskrub is used, but it silently fails if the executable can't be found if opts.sponskrub is not False: yield { 'key': 'SponSkrub', 'path': opts.sponskrub_path, 'args': opts.sponskrub_args, 'cut': opts.sponskrub_cut, 'force': opts.sponskrub_force, 'ignoreerror': opts.sponskrub is None, '_from_cli': True, } if opts.embedthumbnail: yield { 'key': 'EmbedThumbnail', # already_have_thumbnail = True prevents the file from being deleted after embedding 'already_have_thumbnail': opts.writethumbnail, } if not opts.writethumbnail: opts.writethumbnail = True opts.outtmpl['pl_thumbnail'] = '' if opts.split_chapters: yield { 'key': 'FFmpegSplitChapters', 'force_keyframes': opts.force_keyframes_at_cuts, } # XAttrMetadataPP should be run after post-processors that may change file contents if opts.xattrs: yield {'key': 'XAttrMetadata'} if opts.concat_playlist != 'never': yield { 'key': 'FFmpegConcat', 'only_multi_video': opts.concat_playlist != 'always', 'when': 'playlist', } # Exec must be the last PP of each category for when, exec_cmd in opts.exec_cmd.items(): yield { 'key': 'Exec', 'exec_cmd': exec_cmd, 'when': when, } ParsedOptions = collections.namedtuple('ParsedOptions', ('parser', 'options', 'urls', 'ydl_opts')) def parse_options(argv=None): """@returns ParsedOptions(parser, opts, urls, ydl_opts)""" parser, opts, urls = parseOpts(argv) urls = get_urls(urls, opts.batchfile, -1 if opts.quiet and not opts.verbose else opts.verbose) set_compat_opts(opts) try: warnings, deprecation_warnings = validate_options(opts) except ValueError as err: parser.error(f'{err}\n') postprocessors = list(get_postprocessors(opts)) print_only = bool(opts.forceprint) and all(k not in opts.forceprint for k in POSTPROCESS_WHEN[3:]) any_getting = any(getattr(opts, k) for k in ( 'dumpjson', 'dump_single_json', 'getdescription', 'getduration', 'getfilename', 'getformat', 'getid', 'getthumbnail', 'gettitle', 'geturl', )) if opts.quiet is None: opts.quiet = any_getting or opts.print_json or bool(opts.forceprint) playlist_pps = [pp for pp in postprocessors if pp.get('when') == 'playlist'] write_playlist_infojson = (opts.writeinfojson and not opts.clean_infojson and opts.allow_playlist_files and opts.outtmpl.get('pl_infojson') != '') if not any(( opts.extract_flat, opts.dump_single_json, opts.forceprint.get('playlist'), opts.print_to_file.get('playlist'), write_playlist_infojson, )): if not playlist_pps: opts.extract_flat = 'discard' elif playlist_pps == [{'key': 'FFmpegConcat', 'only_multi_video': True, 'when': 'playlist'}]: opts.extract_flat = 'discard_in_playlist' final_ext = ( opts.recodevideo if opts.recodevideo in FFmpegVideoConvertorPP.SUPPORTED_EXTS else opts.remuxvideo if opts.remuxvideo in FFmpegVideoRemuxerPP.SUPPORTED_EXTS else opts.audioformat if (opts.extractaudio and opts.audioformat in FFmpegExtractAudioPP.SUPPORTED_EXTS) else None) return ParsedOptions(parser, opts, urls, { 'usenetrc': opts.usenetrc, 'netrc_location': opts.netrc_location, 'netrc_cmd': opts.netrc_cmd, 'username': opts.username, 'password': opts.password, 'twofactor': opts.twofactor, 'videopassword': opts.videopassword, 'ap_mso': opts.ap_mso, 'ap_username': opts.ap_username, 'ap_password': opts.ap_password, 'client_certificate': opts.client_certificate, 'client_certificate_key': opts.client_certificate_key, 'client_certificate_password': opts.client_certificate_password, 'quiet': opts.quiet, 'no_warnings': opts.no_warnings, 'forceurl': opts.geturl, 'forcetitle': opts.gettitle, 'forceid': opts.getid, 'forcethumbnail': opts.getthumbnail, 'forcedescription': opts.getdescription, 'forceduration': opts.getduration, 'forcefilename': opts.getfilename, 'forceformat': opts.getformat, 'forceprint': opts.forceprint, 'print_to_file': opts.print_to_file, 'forcejson': opts.dumpjson or opts.print_json, 'dump_single_json': opts.dump_single_json, 'force_write_download_archive': opts.force_write_download_archive, 'simulate': (print_only or any_getting or None) if opts.simulate is None else opts.simulate, 'skip_download': opts.skip_download, 'format': opts.format, 'allow_unplayable_formats': opts.allow_unplayable_formats, 'ignore_no_formats_error': opts.ignore_no_formats_error, 'format_sort': opts.format_sort, 'format_sort_force': opts.format_sort_force, 'allow_multiple_video_streams': opts.allow_multiple_video_streams, 'allow_multiple_audio_streams': opts.allow_multiple_audio_streams, 'check_formats': opts.check_formats, 'listformats': opts.listformats, 'listformats_table': opts.listformats_table, 'outtmpl': opts.outtmpl, 'outtmpl_na_placeholder': opts.outtmpl_na_placeholder, 'paths': opts.paths, 'autonumber_size': opts.autonumber_size, 'autonumber_start': opts.autonumber_start, 'restrictfilenames': opts.restrictfilenames, 'windowsfilenames': opts.windowsfilenames, 'ignoreerrors': opts.ignoreerrors, 'force_generic_extractor': opts.force_generic_extractor, 'allowed_extractors': opts.allowed_extractors or ['default'], 'ratelimit': opts.ratelimit, 'throttledratelimit': opts.throttledratelimit, 'overwrites': opts.overwrites, 'retries': opts.retries, 'file_access_retries': opts.file_access_retries, 'fragment_retries': opts.fragment_retries, 'extractor_retries': opts.extractor_retries, 'retry_sleep_functions': opts.retry_sleep, 'skip_unavailable_fragments': opts.skip_unavailable_fragments, 'keep_fragments': opts.keep_fragments, 'concurrent_fragment_downloads': opts.concurrent_fragment_downloads, 'buffersize': opts.buffersize, 'noresizebuffer': opts.noresizebuffer, 'http_chunk_size': opts.http_chunk_size, 'continuedl': opts.continue_dl, 'noprogress': opts.quiet if opts.noprogress is None else opts.noprogress, 'progress_with_newline': opts.progress_with_newline, 'progress_template': opts.progress_template, 'progress_delta': opts.progress_delta, 'playliststart': opts.playliststart, 'playlistend': opts.playlistend, 'playlistreverse': opts.playlist_reverse, 'playlistrandom': opts.playlist_random, 'lazy_playlist': opts.lazy_playlist, 'noplaylist': opts.noplaylist, 'logtostderr': opts.outtmpl.get('default') == '-', 'consoletitle': opts.consoletitle, 'nopart': opts.nopart, 'updatetime': opts.updatetime, 'writedescription': opts.writedescription, 'writeannotations': opts.writeannotations, 'writeinfojson': opts.writeinfojson, 'allow_playlist_files': opts.allow_playlist_files, 'clean_infojson': opts.clean_infojson, 'getcomments': opts.getcomments, 'writethumbnail': opts.writethumbnail is True, 'write_all_thumbnails': opts.writethumbnail == 'all', 'writelink': opts.writelink, 'writeurllink': opts.writeurllink, 'writewebloclink': opts.writewebloclink, 'writedesktoplink': opts.writedesktoplink, 'writesubtitles': opts.writesubtitles, 'writeautomaticsub': opts.writeautomaticsub, 'allsubtitles': opts.allsubtitles, 'listsubtitles': opts.listsubtitles, 'subtitlesformat': opts.subtitlesformat, 'subtitleslangs': opts.subtitleslangs, 'matchtitle': opts.matchtitle, 'rejecttitle': opts.rejecttitle, 'max_downloads': opts.max_downloads, 'prefer_free_formats': opts.prefer_free_formats, 'trim_file_name': opts.trim_file_name, 'verbose': opts.verbose, 'dump_intermediate_pages': opts.dump_intermediate_pages, 'write_pages': opts.write_pages, 'load_pages': opts.load_pages, 'test': opts.test, 'keepvideo': opts.keepvideo, 'min_filesize': opts.min_filesize, 'max_filesize': opts.max_filesize, 'min_views': opts.min_views, 'max_views': opts.max_views, 'daterange': opts.date, 'cachedir': opts.cachedir, 'youtube_print_sig_code': opts.youtube_print_sig_code, 'age_limit': opts.age_limit, 'download_archive': opts.download_archive, 'break_on_existing': opts.break_on_existing, 'break_on_reject': opts.break_on_reject, 'break_per_url': opts.break_per_url, 'skip_playlist_after_errors': opts.skip_playlist_after_errors, 'cookiefile': opts.cookiefile, 'cookiesfrombrowser': opts.cookiesfrombrowser, 'legacyserverconnect': opts.legacy_server_connect, 'nocheckcertificate': opts.no_check_certificate, 'prefer_insecure': opts.prefer_insecure, 'enable_file_urls': opts.enable_file_urls, 'http_headers': opts.headers, 'proxy': opts.proxy, 'socket_timeout': opts.socket_timeout, 'bidi_workaround': opts.bidi_workaround, 'debug_printtraffic': opts.debug_printtraffic, 'prefer_ffmpeg': opts.prefer_ffmpeg, 'include_ads': opts.include_ads, 'default_search': opts.default_search, 'dynamic_mpd': opts.dynamic_mpd, 'extractor_args': opts.extractor_args, 'youtube_include_dash_manifest': opts.youtube_include_dash_manifest, 'youtube_include_hls_manifest': opts.youtube_include_hls_manifest, 'encoding': opts.encoding, 'extract_flat': opts.extract_flat, 'live_from_start': opts.live_from_start, 'wait_for_video': opts.wait_for_video, 'mark_watched': opts.mark_watched, 'merge_output_format': opts.merge_output_format, 'final_ext': final_ext, 'postprocessors': postprocessors, 'fixup': opts.fixup, 'source_address': opts.source_address, 'impersonate': opts.impersonate, 'call_home': opts.call_home, 'sleep_interval_requests': opts.sleep_interval_requests, 'sleep_interval': opts.sleep_interval, 'max_sleep_interval': opts.max_sleep_interval, 'sleep_interval_subtitles': opts.sleep_interval_subtitles, 'external_downloader': opts.external_downloader, 'download_ranges': opts.download_ranges, 'force_keyframes_at_cuts': opts.force_keyframes_at_cuts, 'list_thumbnails': opts.list_thumbnails, 'playlist_items': opts.playlist_items, 'xattr_set_filesize': opts.xattr_set_filesize, 'match_filter': opts.match_filter, 'color': opts.color, 'ffmpeg_location': opts.ffmpeg_location, 'hls_prefer_native': opts.hls_prefer_native, 'hls_use_mpegts': opts.hls_use_mpegts, 'hls_split_discontinuity': opts.hls_split_discontinuity, 'external_downloader_args': opts.external_downloader_args, 'postprocessor_args': opts.postprocessor_args, 'cn_verification_proxy': opts.cn_verification_proxy, 'geo_verification_proxy': opts.geo_verification_proxy, 'geo_bypass': opts.geo_bypass, 'geo_bypass_country': opts.geo_bypass_country, 'geo_bypass_ip_block': opts.geo_bypass_ip_block, '_warnings': warnings, '_deprecation_warnings': deprecation_warnings, 'compat_opts': opts.compat_opts, }) def _real_main(argv=None): setproctitle('yt-dlp') parser, opts, all_urls, ydl_opts = parse_options(argv) # HACK: Set the plugin dirs early on # TODO(coletdjnz): remove when plugin globals system is implemented if opts.plugin_dirs is not None: Config._plugin_dirs = list(map(expand_path, opts.plugin_dirs)) # Dump user agent if opts.dump_user_agent: ua = traverse_obj(opts.headers, 'User-Agent', casesense=False, default=std_headers['User-Agent']) write_string(f'{ua}\n', out=sys.stdout) return if print_extractor_information(opts, all_urls): return # We may need ffmpeg_location without having access to the YoutubeDL instance # See https://github.com/yt-dlp/yt-dlp/issues/2191 if opts.ffmpeg_location: FFmpegPostProcessor._ffmpeg_location.set(opts.ffmpeg_location) with YoutubeDL(ydl_opts) as ydl: pre_process = opts.update_self or opts.rm_cachedir actual_use = all_urls or opts.load_info_filename if opts.rm_cachedir: ydl.cache.remove() try: updater = Updater(ydl, opts.update_self) if opts.update_self and updater.update() and actual_use: if updater.cmd: return updater.restart() # This code is reachable only for zip variant in py < 3.10 # It makes sense to exit here, but the old behavior is to continue ydl.report_warning('Restart yt-dlp to use the updated version') # return 100, 'ERROR: The program must exit for the update to complete' except Exception: traceback.print_exc() ydl._download_retcode = 100 if opts.list_impersonate_targets: known_targets = [ # List of simplified targets we know are supported, # to help users know what dependencies may be required. (ImpersonateTarget('chrome'), 'curl_cffi'), (ImpersonateTarget('edge'), 'curl_cffi'), (ImpersonateTarget('safari'), 'curl_cffi'), ] available_targets = ydl._get_available_impersonate_targets() def make_row(target, handler): return [ join_nonempty(target.client.title(), target.version, delim='-') or '-', join_nonempty((target.os or '').title(), target.os_version, delim='-') or '-', handler, ] rows = [make_row(target, handler) for target, handler in available_targets] for known_target, known_handler in known_targets: if not any( known_target in target and handler == known_handler for target, handler in available_targets ): rows.append([ ydl._format_out(text, ydl.Styles.SUPPRESS) for text in make_row(known_target, f'{known_handler} (not available)') ]) ydl.to_screen('[info] Available impersonate targets') ydl.to_stdout(render_table(['Client', 'OS', 'Source'], rows, extra_gap=2, delim='-')) return if not actual_use: if pre_process: return ydl._download_retcode args = sys.argv[1:] if argv is None else argv ydl.warn_if_short_id(args) # Show a useful error message and wait for keypress if not launched from shell on Windows if not args and os.name == 'nt' and getattr(sys, 'frozen', False): import ctypes.wintypes import msvcrt kernel32 = ctypes.WinDLL('Kernel32') buffer = (1 * ctypes.wintypes.DWORD)() attached_processes = kernel32.GetConsoleProcessList(buffer, 1) # If we only have a single process attached, then the executable was double clicked # When using `pyinstaller` with `--onefile`, two processes get attached is_onefile = hasattr(sys, '_MEIPASS') and os.path.basename(sys._MEIPASS).startswith('_MEI') if attached_processes == 1 or (is_onefile and attached_processes == 2): print(parser._generate_error_message( 'Do not double-click the executable, instead call it from a command line.\n' 'Please read the README for further information on how to use yt-dlp: ' 'https://github.com/yt-dlp/yt-dlp#readme')) msvcrt.getch() _exit(2) parser.error( 'You must provide at least one URL.\n' 'Type yt-dlp --help to see a list of all options.') parser.destroy() try: if opts.load_info_filename is not None: if all_urls: ydl.report_warning('URLs are ignored due to --load-info-json') return ydl.download_with_info_file(expand_path(opts.load_info_filename)) else: return ydl.download(all_urls) except DownloadCancelled: ydl.to_screen('Aborting remaining downloads') return 101 def main(argv=None): global _IN_CLI _IN_CLI = True try: _exit(*variadic(_real_main(argv))) except (CookieLoadError, DownloadError): _exit(1) except SameFileError as e: _exit(f'ERROR: {e}') except KeyboardInterrupt: _exit('\nERROR: Interrupted by user') except BrokenPipeError as e: # https://docs.python.org/3/library/signal.html#note-on-sigpipe devnull = os.open(os.devnull, os.O_WRONLY) os.dup2(devnull, sys.stdout.fileno()) _exit(f'\nERROR: {e}') except optparse.OptParseError as e: _exit(2, f'\n{e}') from .extractor import gen_extractors, list_extractors __all__ = [ 'YoutubeDL', 'gen_extractors', 'list_extractors', 'main', 'parse_options', ]