123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196219721982199220022012202220322042205220622072208220922102211221222132214221522162217221822192220222122222223222422252226222722282229223022312232223322342235223622372238223922402241224222432244224522462247224822492250225122522253225422552256225722582259226022612262226322642265226622672268226922702271227222732274227522762277227822792280228122822283228422852286228722882289229022912292229322942295229622972298229923002301230223032304230523062307230823092310231123122313231423152316231723182319232023212322232323242325232623272328232923302331233223332334233523362337233823392340234123422343234423452346234723482349235023512352235323542355235623572358235923602361236223632364236523662367236823692370237123722373237423752376237723782379238023812382238323842385238623872388238923902391239223932394239523962397239823992400240124022403240424052406240724082409241024112412241324142415241624172418241924202421242224232424242524262427242824292430243124322433243424352436243724382439244024412442244324442445244624472448244924502451245224532454245524562457245824592460246124622463246424652466246724682469247024712472247324742475247624772478247924802481248224832484248524862487248824892490249124922493249424952496249724982499250025012502250325042505250625072508250925102511251225132514251525162517251825192520252125222523252425252526252725282529253025312532253325342535253625372538253925402541254225432544254525462547254825492550255125522553255425552556255725582559256025612562256325642565256625672568256925702571257225732574257525762577257825792580258125822583258425852586258725882589259025912592259325942595259625972598259926002601260226032604260526062607260826092610261126122613261426152616261726182619262026212622262326242625262626272628262926302631263226332634263526362637263826392640264126422643264426452646264726482649265026512652265326542655265626572658265926602661266226632664266526662667266826692670267126722673267426752676267726782679268026812682268326842685268626872688268926902691269226932694269526962697269826992700270127022703270427052706270727082709271027112712271327142715271627172718271927202721272227232724272527262727272827292730273127322733273427352736273727382739274027412742274327442745274627472748274927502751275227532754275527562757275827592760276127622763276427652766276727682769277027712772277327742775277627772778277927802781278227832784278527862787278827892790279127922793279427952796279727982799280028012802280328042805280628072808280928102811281228132814281528162817281828192820282128222823282428252826282728282829283028312832283328342835283628372838283928402841284228432844284528462847284828492850285128522853285428552856285728582859286028612862286328642865286628672868286928702871287228732874287528762877287828792880288128822883288428852886288728882889289028912892289328942895289628972898289929002901290229032904290529062907290829092910291129122913291429152916291729182919292029212922292329242925292629272928292929302931293229332934293529362937293829392940294129422943294429452946294729482949295029512952295329542955295629572958295929602961296229632964296529662967296829692970297129722973297429752976297729782979298029812982298329842985298629872988298929902991299229932994299529962997299829993000300130023003300430053006300730083009301030113012301330143015301630173018301930203021302230233024302530263027302830293030303130323033303430353036303730383039304030413042304330443045304630473048304930503051305230533054305530563057305830593060306130623063306430653066306730683069307030713072307330743075307630773078307930803081308230833084308530863087308830893090309130923093309430953096309730983099310031013102310331043105310631073108310931103111311231133114311531163117311831193120312131223123312431253126312731283129313031313132313331343135313631373138313931403141314231433144314531463147314831493150315131523153315431553156315731583159316031613162316331643165316631673168316931703171317231733174317531763177317831793180318131823183318431853186318731883189319031913192319331943195319631973198319932003201320232033204320532063207320832093210321132123213321432153216321732183219322032213222322332243225322632273228322932303231323232333234323532363237323832393240324132423243324432453246324732483249325032513252325332543255325632573258325932603261326232633264326532663267326832693270327132723273327432753276327732783279328032813282328332843285328632873288328932903291329232933294329532963297329832993300330133023303330433053306330733083309331033113312331333143315331633173318331933203321332233233324332533263327332833293330333133323333333433353336333733383339334033413342334333443345334633473348334933503351335233533354335533563357335833593360336133623363336433653366336733683369337033713372337333743375337633773378337933803381338233833384338533863387338833893390339133923393339433953396339733983399340034013402340334043405340634073408340934103411341234133414341534163417341834193420342134223423342434253426342734283429343034313432343334343435343634373438343934403441344234433444344534463447344834493450345134523453345434553456345734583459346034613462346334643465346634673468346934703471347234733474347534763477347834793480348134823483348434853486348734883489349034913492349334943495349634973498349935003501350235033504350535063507350835093510351135123513351435153516351735183519352035213522352335243525352635273528352935303531353235333534353535363537353835393540354135423543354435453546354735483549355035513552355335543555355635573558355935603561356235633564356535663567356835693570357135723573357435753576357735783579358035813582358335843585358635873588358935903591359235933594359535963597359835993600360136023603360436053606360736083609361036113612361336143615361636173618361936203621362236233624362536263627362836293630363136323633363436353636363736383639364036413642364336443645364636473648364936503651365236533654365536563657365836593660366136623663366436653666366736683669367036713672367336743675367636773678367936803681368236833684368536863687368836893690369136923693369436953696369736983699370037013702370337043705370637073708370937103711371237133714371537163717371837193720372137223723372437253726372737283729373037313732373337343735373637373738373937403741374237433744374537463747374837493750375137523753375437553756375737583759376037613762376337643765376637673768376937703771377237733774377537763777377837793780378137823783378437853786378737883789379037913792379337943795379637973798379938003801380238033804380538063807380838093810381138123813381438153816381738183819382038213822382338243825382638273828382938303831383238333834383538363837383838393840384138423843384438453846384738483849385038513852385338543855385638573858385938603861386238633864386538663867386838693870387138723873387438753876387738783879388038813882388338843885388638873888388938903891389238933894389538963897389838993900390139023903390439053906390739083909391039113912391339143915391639173918391939203921392239233924392539263927392839293930393139323933393439353936393739383939394039413942394339443945394639473948394939503951395239533954395539563957395839593960396139623963396439653966396739683969397039713972397339743975397639773978397939803981398239833984398539863987398839893990399139923993399439953996399739983999400040014002400340044005400640074008400940104011401240134014401540164017401840194020402140224023402440254026402740284029403040314032403340344035403640374038403940404041404240434044404540464047404840494050405140524053405440554056405740584059406040614062406340644065406640674068406940704071407240734074407540764077407840794080408140824083408440854086408740884089409040914092409340944095409640974098409941004101410241034104410541064107410841094110411141124113411441154116411741184119412041214122412341244125412641274128412941304131413241334134413541364137413841394140414141424143414441454146414741484149415041514152415341544155415641574158415941604161416241634164416541664167416841694170417141724173417441754176417741784179418041814182418341844185418641874188418941904191419241934194419541964197419841994200420142024203420442054206420742084209421042114212421342144215421642174218421942204221422242234224422542264227422842294230423142324233423442354236423742384239424042414242424342444245424642474248424942504251425242534254425542564257425842594260426142624263426442654266426742684269427042714272427342744275427642774278427942804281428242834284428542864287428842894290429142924293429442954296429742984299430043014302430343044305430643074308430943104311431243134314431543164317431843194320432143224323432443254326432743284329433043314332433343344335433643374338433943404341434243434344434543464347434843494350435143524353435443554356435743584359436043614362436343644365436643674368436943704371437243734374437543764377437843794380438143824383438443854386438743884389439043914392439343944395439643974398439944004401440244034404440544064407440844094410441144124413441444154416441744184419442044214422 |
- import collections
- import contextlib
- import copy
- import datetime as dt
- import errno
- import fileinput
- import functools
- import http.cookiejar
- import io
- import itertools
- import json
- import locale
- import operator
- import os
- import random
- import re
- import shutil
- import string
- import subprocess
- import sys
- import tempfile
- import time
- import tokenize
- import traceback
- import unicodedata
- from .cache import Cache
- from .compat import urllib # isort: split
- from .compat import compat_os_name, urllib_req_to_req
- from .cookies import CookieLoadError, LenientSimpleCookie, load_cookies
- from .downloader import FFmpegFD, get_suitable_downloader, shorten_protocol_name
- from .downloader.rtmp import rtmpdump_version
- from .extractor import gen_extractor_classes, get_info_extractor
- from .extractor.common import UnsupportedURLIE
- from .extractor.openload import PhantomJSwrapper
- from .minicurses import format_text
- from .networking import HEADRequest, Request, RequestDirector
- from .networking.common import _REQUEST_HANDLERS, _RH_PREFERENCES
- from .networking.exceptions import (
- HTTPError,
- NoSupportingHandlers,
- RequestError,
- SSLError,
- network_exceptions,
- )
- from .networking.impersonate import ImpersonateRequestHandler
- from .plugins import directories as plugin_directories
- from .postprocessor import _PLUGIN_CLASSES as plugin_pps
- from .postprocessor import (
- EmbedThumbnailPP,
- FFmpegFixupDuplicateMoovPP,
- FFmpegFixupDurationPP,
- FFmpegFixupM3u8PP,
- FFmpegFixupM4aPP,
- FFmpegFixupStretchedPP,
- FFmpegFixupTimestampPP,
- FFmpegMergerPP,
- FFmpegPostProcessor,
- FFmpegVideoConvertorPP,
- MoveFilesAfterDownloadPP,
- get_postprocessor,
- )
- from .postprocessor.ffmpeg import resolve_mapping as resolve_recode_mapping
- from .update import (
- REPOSITORY,
- _get_system_deprecation,
- _make_label,
- current_git_head,
- detect_variant,
- )
- from .utils import (
- DEFAULT_OUTTMPL,
- IDENTITY,
- LINK_TEMPLATES,
- MEDIA_EXTENSIONS,
- NO_DEFAULT,
- NUMBER_RE,
- OUTTMPL_TYPES,
- POSTPROCESS_WHEN,
- STR_FORMAT_RE_TMPL,
- STR_FORMAT_TYPES,
- ContentTooShortError,
- DateRange,
- DownloadCancelled,
- DownloadError,
- EntryNotInPlaylist,
- ExistingVideoReached,
- ExtractorError,
- FormatSorter,
- GeoRestrictedError,
- ISO3166Utils,
- LazyList,
- MaxDownloadsReached,
- Namespace,
- PagedList,
- PlaylistEntries,
- Popen,
- PostProcessingError,
- ReExtractInfo,
- RejectedVideoReached,
- SameFileError,
- UnavailableVideoError,
- UserNotLive,
- YoutubeDLError,
- age_restricted,
- bug_reports_message,
- date_from_str,
- deprecation_warning,
- determine_ext,
- determine_protocol,
- encode_compat_str,
- encodeFilename,
- escapeHTML,
- expand_path,
- extract_basic_auth,
- filter_dict,
- float_or_none,
- format_bytes,
- format_decimal_suffix,
- format_field,
- formatSeconds,
- get_compatible_ext,
- get_domain,
- int_or_none,
- iri_to_uri,
- is_path_like,
- join_nonempty,
- locked_file,
- make_archive_id,
- make_dir,
- number_of_digits,
- orderedSet,
- orderedSet_from_options,
- parse_filesize,
- preferredencoding,
- prepend_extension,
- remove_terminal_sequences,
- render_table,
- replace_extension,
- sanitize_filename,
- sanitize_path,
- sanitize_url,
- shell_quote,
- str_or_none,
- strftime_or_none,
- subtitles_filename,
- supports_terminal_sequences,
- system_identifier,
- filesize_from_tbr,
- timetuple_from_msec,
- to_high_limit_path,
- traverse_obj,
- try_call,
- try_get,
- url_basename,
- variadic,
- version_tuple,
- windows_enable_vt_mode,
- write_json_file,
- write_string,
- )
- from .utils._utils import _UnsafeExtensionError, _YDLLogger
- from .utils.networking import (
- HTTPHeaderDict,
- clean_headers,
- clean_proxies,
- std_headers,
- )
- from .version import CHANNEL, ORIGIN, RELEASE_GIT_HEAD, VARIANT, __version__
- if compat_os_name == 'nt':
- import ctypes
- def _catch_unsafe_extension_error(func):
- @functools.wraps(func)
- def wrapper(self, *args, **kwargs):
- try:
- return func(self, *args, **kwargs)
- except _UnsafeExtensionError as error:
- self.report_error(
- f'The extracted extension ({error.extension!r}) is unusual '
- 'and will be skipped for safety reasons. '
- f'If you believe this is an error{bug_reports_message(",")}')
- return wrapper
- class YoutubeDL:
- """YoutubeDL class.
- YoutubeDL objects are the ones responsible of downloading the
- actual video file and writing it to disk if the user has requested
- it, among some other tasks. In most cases there should be one per
- program. As, given a video URL, the downloader doesn't know how to
- extract all the needed information, task that InfoExtractors do, it
- has to pass the URL to one of them.
- For this, YoutubeDL objects have a method that allows
- InfoExtractors to be registered in a given order. When it is passed
- a URL, the YoutubeDL object handles it to the first InfoExtractor it
- finds that reports being able to handle it. The InfoExtractor extracts
- all the information about the video or videos the URL refers to, and
- YoutubeDL process the extracted information, possibly using a File
- Downloader to download the video.
- YoutubeDL objects accept a lot of parameters. In order not to saturate
- the object constructor with arguments, it receives a dictionary of
- options instead. These options are available through the params
- attribute for the InfoExtractors to use. The YoutubeDL also
- registers itself as the downloader in charge for the InfoExtractors
- that are added to it, so this is a "mutual registration".
- Available options:
- username: Username for authentication purposes.
- password: Password for authentication purposes.
- videopassword: Password for accessing a video.
- ap_mso: Adobe Pass multiple-system operator identifier.
- ap_username: Multiple-system operator account username.
- ap_password: Multiple-system operator account password.
- usenetrc: Use netrc for authentication instead.
- netrc_location: Location of the netrc file. Defaults to ~/.netrc.
- netrc_cmd: Use a shell command to get credentials
- verbose: Print additional info to stdout.
- quiet: Do not print messages to stdout.
- no_warnings: Do not print out anything for warnings.
- forceprint: A dict with keys WHEN mapped to a list of templates to
- print to stdout. The allowed keys are video or any of the
- items in utils.POSTPROCESS_WHEN.
- For compatibility, a single list is also accepted
- print_to_file: A dict with keys WHEN (same as forceprint) mapped to
- a list of tuples with (template, filename)
- forcejson: Force printing info_dict as JSON.
- dump_single_json: Force printing the info_dict of the whole playlist
- (or video) as a single JSON line.
- force_write_download_archive: Force writing download archive regardless
- of 'skip_download' or 'simulate'.
- simulate: Do not download the video files. If unset (or None),
- simulate only if listsubtitles, listformats or list_thumbnails is used
- format: Video format code. see "FORMAT SELECTION" for more details.
- You can also pass a function. The function takes 'ctx' as
- argument and returns the formats to download.
- See "build_format_selector" for an implementation
- allow_unplayable_formats: Allow unplayable formats to be extracted and downloaded.
- ignore_no_formats_error: Ignore "No video formats" error. Usefull for
- extracting metadata even if the video is not actually
- available for download (experimental)
- format_sort: A list of fields by which to sort the video formats.
- See "Sorting Formats" for more details.
- format_sort_force: Force the given format_sort. see "Sorting Formats"
- for more details.
- prefer_free_formats: Whether to prefer video formats with free containers
- over non-free ones of same quality.
- allow_multiple_video_streams: Allow multiple video streams to be merged
- into a single file
- allow_multiple_audio_streams: Allow multiple audio streams to be merged
- into a single file
- check_formats Whether to test if the formats are downloadable.
- Can be True (check all), False (check none),
- 'selected' (check selected formats),
- or None (check only if requested by extractor)
- paths: Dictionary of output paths. The allowed keys are 'home'
- 'temp' and the keys of OUTTMPL_TYPES (in utils/_utils.py)
- outtmpl: Dictionary of templates for output names. Allowed keys
- are 'default' and the keys of OUTTMPL_TYPES (in utils/_utils.py).
- For compatibility with youtube-dl, a single string can also be used
- outtmpl_na_placeholder: Placeholder for unavailable meta fields.
- restrictfilenames: Do not allow "&" and spaces in file names
- trim_file_name: Limit length of filename (extension excluded)
- windowsfilenames: Force the filenames to be windows compatible
- ignoreerrors: Do not stop on download/postprocessing errors.
- Can be 'only_download' to ignore only download errors.
- Default is 'only_download' for CLI, but False for API
- skip_playlist_after_errors: Number of allowed failures until the rest of
- the playlist is skipped
- allowed_extractors: List of regexes to match against extractor names that are allowed
- overwrites: Overwrite all video and metadata files if True,
- overwrite only non-video files if None
- and don't overwrite any file if False
- playlist_items: Specific indices of playlist to download.
- playlistrandom: Download playlist items in random order.
- lazy_playlist: Process playlist entries as they are received.
- matchtitle: Download only matching titles.
- rejecttitle: Reject downloads for matching titles.
- logger: Log messages to a logging.Logger instance.
- logtostderr: Print everything to stderr instead of stdout.
- consoletitle: Display progress in console window's titlebar.
- writedescription: Write the video description to a .description file
- writeinfojson: Write the video description to a .info.json file
- clean_infojson: Remove internal metadata from the infojson
- getcomments: Extract video comments. This will not be written to disk
- unless writeinfojson is also given
- writeannotations: Write the video annotations to a .annotations.xml file
- writethumbnail: Write the thumbnail image to a file
- allow_playlist_files: Whether to write playlists' description, infojson etc
- also to disk when using the 'write*' options
- write_all_thumbnails: Write all thumbnail formats to files
- writelink: Write an internet shortcut file, depending on the
- current platform (.url/.webloc/.desktop)
- writeurllink: Write a Windows internet shortcut file (.url)
- writewebloclink: Write a macOS internet shortcut file (.webloc)
- writedesktoplink: Write a Linux internet shortcut file (.desktop)
- writesubtitles: Write the video subtitles to a file
- writeautomaticsub: Write the automatically generated subtitles to a file
- listsubtitles: Lists all available subtitles for the video
- subtitlesformat: The format code for subtitles
- subtitleslangs: List of languages of the subtitles to download (can be regex).
- The list may contain "all" to refer to all the available
- subtitles. The language can be prefixed with a "-" to
- exclude it from the requested languages, e.g. ['all', '-live_chat']
- keepvideo: Keep the video file after post-processing
- daterange: A utils.DateRange object, download only if the upload_date is in the range.
- skip_download: Skip the actual download of the video file
- cachedir: Location of the cache files in the filesystem.
- False to disable filesystem cache.
- noplaylist: Download single video instead of a playlist if in doubt.
- age_limit: An integer representing the user's age in years.
- Unsuitable videos for the given age are skipped.
- min_views: An integer representing the minimum view count the video
- must have in order to not be skipped.
- Videos without view count information are always
- downloaded. None for no limit.
- max_views: An integer representing the maximum view count.
- Videos that are more popular than that are not
- downloaded.
- Videos without view count information are always
- downloaded. None for no limit.
- download_archive: A set, or the name of a file where all downloads are recorded.
- Videos already present in the file are not downloaded again.
- break_on_existing: Stop the download process after attempting to download a
- file that is in the archive.
- break_per_url: Whether break_on_reject and break_on_existing
- should act on each input URL as opposed to for the entire queue
- cookiefile: File name or text stream from where cookies should be read and dumped to
- cookiesfrombrowser: A tuple containing the name of the browser, the profile
- name/path from where cookies are loaded, the name of the keyring,
- and the container name, e.g. ('chrome', ) or
- ('vivaldi', 'default', 'BASICTEXT') or ('firefox', 'default', None, 'Meta')
- legacyserverconnect: Explicitly allow HTTPS connection to servers that do not
- support RFC 5746 secure renegotiation
- nocheckcertificate: Do not verify SSL certificates
- client_certificate: Path to client certificate file in PEM format. May include the private key
- client_certificate_key: Path to private key file for client certificate
- client_certificate_password: Password for client certificate private key, if encrypted.
- If not provided and the key is encrypted, yt-dlp will ask interactively
- prefer_insecure: Use HTTP instead of HTTPS to retrieve information.
- (Only supported by some extractors)
- enable_file_urls: Enable file:// URLs. This is disabled by default for security reasons.
- http_headers: A dictionary of custom headers to be used for all requests
- proxy: URL of the proxy server to use
- geo_verification_proxy: URL of the proxy to use for IP address verification
- on geo-restricted sites.
- socket_timeout: Time to wait for unresponsive hosts, in seconds
- bidi_workaround: Work around buggy terminals without bidirectional text
- support, using fridibi
- debug_printtraffic:Print out sent and received HTTP traffic
- default_search: Prepend this string if an input url is not valid.
- 'auto' for elaborate guessing
- encoding: Use this encoding instead of the system-specified.
- extract_flat: Whether to resolve and process url_results further
- * False: Always process. Default for API
- * True: Never process
- * 'in_playlist': Do not process inside playlist/multi_video
- * 'discard': Always process, but don't return the result
- from inside playlist/multi_video
- * 'discard_in_playlist': Same as "discard", but only for
- playlists (not multi_video). Default for CLI
- wait_for_video: If given, wait for scheduled streams to become available.
- The value should be a tuple containing the range
- (min_secs, max_secs) to wait between retries
- postprocessors: A list of dictionaries, each with an entry
- * key: The name of the postprocessor. See
- yt_dlp/postprocessor/__init__.py for a list.
- * when: When to run the postprocessor. Allowed values are
- the entries of utils.POSTPROCESS_WHEN
- Assumed to be 'post_process' if not given
- progress_hooks: A list of functions that get called on download
- progress, with a dictionary with the entries
- * status: One of "downloading", "error", or "finished".
- Check this first and ignore unknown values.
- * info_dict: The extracted info_dict
- If status is one of "downloading", or "finished", the
- following properties may also be present:
- * filename: The final filename (always present)
- * tmpfilename: The filename we're currently writing to
- * downloaded_bytes: Bytes on disk
- * total_bytes: Size of the whole file, None if unknown
- * total_bytes_estimate: Guess of the eventual file size,
- None if unavailable.
- * elapsed: The number of seconds since download started.
- * eta: The estimated time in seconds, None if unknown
- * speed: The download speed in bytes/second, None if
- unknown
- * fragment_index: The counter of the currently
- downloaded video fragment.
- * fragment_count: The number of fragments (= individual
- files that will be merged)
- Progress hooks are guaranteed to be called at least once
- (with status "finished") if the download is successful.
- postprocessor_hooks: A list of functions that get called on postprocessing
- progress, with a dictionary with the entries
- * status: One of "started", "processing", or "finished".
- Check this first and ignore unknown values.
- * postprocessor: Name of the postprocessor
- * info_dict: The extracted info_dict
- Progress hooks are guaranteed to be called at least twice
- (with status "started" and "finished") if the processing is successful.
- merge_output_format: "/" separated list of extensions to use when merging formats.
- final_ext: Expected final extension; used to detect when the file was
- already downloaded and converted
- fixup: Automatically correct known faults of the file.
- One of:
- - "never": do nothing
- - "warn": only emit a warning
- - "detect_or_warn": check whether we can do anything
- about it, warn otherwise (default)
- source_address: Client-side IP address to bind to.
- impersonate: Client to impersonate for requests.
- An ImpersonateTarget (from yt_dlp.networking.impersonate)
- sleep_interval_requests: Number of seconds to sleep between requests
- during extraction
- sleep_interval: Number of seconds to sleep before each download when
- used alone or a lower bound of a range for randomized
- sleep before each download (minimum possible number
- of seconds to sleep) when used along with
- max_sleep_interval.
- max_sleep_interval:Upper bound of a range for randomized sleep before each
- download (maximum possible number of seconds to sleep).
- Must only be used along with sleep_interval.
- Actual sleep time will be a random float from range
- [sleep_interval; max_sleep_interval].
- sleep_interval_subtitles: Number of seconds to sleep before each subtitle download
- listformats: Print an overview of available video formats and exit.
- list_thumbnails: Print a table of all thumbnails and exit.
- match_filter: A function that gets called for every video with the signature
- (info_dict, *, incomplete: bool) -> Optional[str]
- For backward compatibility with youtube-dl, the signature
- (info_dict) -> Optional[str] is also allowed.
- - If it returns a message, the video is ignored.
- - If it returns None, the video is downloaded.
- - If it returns utils.NO_DEFAULT, the user is interactively
- asked whether to download the video.
- - Raise utils.DownloadCancelled(msg) to abort remaining
- downloads when a video is rejected.
- match_filter_func in utils/_utils.py is one example for this.
- color: A Dictionary with output stream names as keys
- and their respective color policy as values.
- Can also just be a single color policy,
- in which case it applies to all outputs.
- Valid stream names are 'stdout' and 'stderr'.
- Valid color policies are one of 'always', 'auto',
- 'no_color', 'never', 'auto-tty' or 'no_color-tty'.
- geo_bypass: Bypass geographic restriction via faking X-Forwarded-For
- HTTP header
- geo_bypass_country:
- Two-letter ISO 3166-2 country code that will be used for
- explicit geographic restriction bypassing via faking
- X-Forwarded-For HTTP header
- geo_bypass_ip_block:
- IP range in CIDR notation that will be used similarly to
- geo_bypass_country
- external_downloader: A dictionary of protocol keys and the executable of the
- external downloader to use for it. The allowed protocols
- are default|http|ftp|m3u8|dash|rtsp|rtmp|mms.
- Set the value to 'native' to use the native downloader
- compat_opts: Compatibility options. See "Differences in default behavior".
- The following options do not work when used through the API:
- filename, abort-on-error, multistreams, no-live-chat,
- format-sort, no-clean-infojson, no-playlist-metafiles,
- no-keep-subs, no-attach-info-json, allow-unsafe-ext.
- Refer __init__.py for their implementation
- progress_template: Dictionary of templates for progress outputs.
- Allowed keys are 'download', 'postprocess',
- 'download-title' (console title) and 'postprocess-title'.
- The template is mapped on a dictionary with keys 'progress' and 'info'
- retry_sleep_functions: Dictionary of functions that takes the number of attempts
- as argument and returns the time to sleep in seconds.
- Allowed keys are 'http', 'fragment', 'file_access'
- download_ranges: A callback function that gets called for every video with
- the signature (info_dict, ydl) -> Iterable[Section].
- Only the returned sections will be downloaded.
- Each Section is a dict with the following keys:
- * start_time: Start time of the section in seconds
- * end_time: End time of the section in seconds
- * title: Section title (Optional)
- * index: Section number (Optional)
- force_keyframes_at_cuts: Re-encode the video when downloading ranges to get precise cuts
- noprogress: Do not print the progress bar
- live_from_start: Whether to download livestreams videos from the start
- The following parameters are not used by YoutubeDL itself, they are used by
- the downloader (see yt_dlp/downloader/common.py):
- nopart, updatetime, buffersize, ratelimit, throttledratelimit, min_filesize,
- max_filesize, test, noresizebuffer, retries, file_access_retries, fragment_retries,
- continuedl, xattr_set_filesize, hls_use_mpegts, http_chunk_size,
- external_downloader_args, concurrent_fragment_downloads, progress_delta.
- The following options are used by the post processors:
- ffmpeg_location: Location of the ffmpeg/avconv binary; either the path
- to the binary or its containing directory.
- postprocessor_args: A dictionary of postprocessor/executable keys (in lower case)
- and a list of additional command-line arguments for the
- postprocessor/executable. The dict can also have "PP+EXE" keys
- which are used when the given exe is used by the given PP.
- Use 'default' as the name for arguments to passed to all PP
- For compatibility with youtube-dl, a single list of args
- can also be used
- The following options are used by the extractors:
- extractor_retries: Number of times to retry for known errors (default: 3)
- dynamic_mpd: Whether to process dynamic DASH manifests (default: True)
- hls_split_discontinuity: Split HLS playlists to different formats at
- discontinuities such as ad breaks (default: False)
- extractor_args: A dictionary of arguments to be passed to the extractors.
- See "EXTRACTOR ARGUMENTS" for details.
- E.g. {'youtube': {'skip': ['dash', 'hls']}}
- mark_watched: Mark videos watched (even with --simulate). Only for YouTube
- The following options are deprecated and may be removed in the future:
- break_on_reject: Stop the download process when encountering a video that
- has been filtered out.
- - `raise DownloadCancelled(msg)` in match_filter instead
- force_generic_extractor: Force downloader to use the generic extractor
- - Use allowed_extractors = ['generic', 'default']
- playliststart: - Use playlist_items
- Playlist item to start at.
- playlistend: - Use playlist_items
- Playlist item to end at.
- playlistreverse: - Use playlist_items
- Download playlist items in reverse order.
- forceurl: - Use forceprint
- Force printing final URL.
- forcetitle: - Use forceprint
- Force printing title.
- forceid: - Use forceprint
- Force printing ID.
- forcethumbnail: - Use forceprint
- Force printing thumbnail URL.
- forcedescription: - Use forceprint
- Force printing description.
- forcefilename: - Use forceprint
- Force printing final filename.
- forceduration: - Use forceprint
- Force printing duration.
- allsubtitles: - Use subtitleslangs = ['all']
- Downloads all the subtitles of the video
- (requires writesubtitles or writeautomaticsub)
- include_ads: - Doesn't work
- Download ads as well
- call_home: - Not implemented
- Boolean, true iff we are allowed to contact the
- yt-dlp servers for debugging.
- post_hooks: - Register a custom postprocessor
- A list of functions that get called as the final step
- for each video file, after all postprocessors have been
- called. The filename will be passed as the only argument.
- hls_prefer_native: - Use external_downloader = {'m3u8': 'native'} or {'m3u8': 'ffmpeg'}.
- Use the native HLS downloader instead of ffmpeg/avconv
- if True, otherwise use ffmpeg/avconv if False, otherwise
- use downloader suggested by extractor if None.
- prefer_ffmpeg: - avconv support is deprecated
- If False, use avconv instead of ffmpeg if both are available,
- otherwise prefer ffmpeg.
- youtube_include_dash_manifest: - Use extractor_args
- If True (default), DASH manifests and related
- data will be downloaded and processed by extractor.
- You can reduce network I/O by disabling it if you don't
- care about DASH. (only for youtube)
- youtube_include_hls_manifest: - Use extractor_args
- If True (default), HLS manifests and related
- data will be downloaded and processed by extractor.
- You can reduce network I/O by disabling it if you don't
- care about HLS. (only for youtube)
- no_color: Same as `color='no_color'`
- no_overwrites: Same as `overwrites=False`
- """
- _NUMERIC_FIELDS = {
- 'width', 'height', 'asr', 'audio_channels', 'fps',
- 'tbr', 'abr', 'vbr', 'filesize', 'filesize_approx',
- 'timestamp', 'release_timestamp',
- 'duration', 'view_count', 'like_count', 'dislike_count', 'repost_count',
- 'average_rating', 'comment_count', 'age_limit',
- 'start_time', 'end_time',
- 'chapter_number', 'season_number', 'episode_number',
- 'track_number', 'disc_number', 'release_year',
- }
- _format_fields = {
- # NB: Keep in sync with the docstring of extractor/common.py
- 'url', 'manifest_url', 'manifest_stream_number', 'ext', 'format', 'format_id', 'format_note',
- 'width', 'height', 'aspect_ratio', 'resolution', 'dynamic_range', 'tbr', 'abr', 'acodec', 'asr', 'audio_channels',
- 'vbr', 'fps', 'vcodec', 'container', 'filesize', 'filesize_approx', 'rows', 'columns',
- 'player_url', 'protocol', 'fragment_base_url', 'fragments', 'is_from_start', 'is_dash_periods', 'request_data',
- 'preference', 'language', 'language_preference', 'quality', 'source_preference', 'cookies',
- 'http_headers', 'stretched_ratio', 'no_resume', 'has_drm', 'extra_param_to_segment_url', 'extra_param_to_key_url',
- 'hls_aes', 'downloader_options', 'page_url', 'app', 'play_path', 'tc_url', 'flash_version',
- 'rtmp_live', 'rtmp_conn', 'rtmp_protocol', 'rtmp_real_time',
- }
- _deprecated_multivalue_fields = {
- 'album_artist': 'album_artists',
- 'artist': 'artists',
- 'composer': 'composers',
- 'creator': 'creators',
- 'genre': 'genres',
- }
- _format_selection_exts = {
- 'audio': set(MEDIA_EXTENSIONS.common_audio),
- 'video': {*MEDIA_EXTENSIONS.common_video, '3gp'},
- 'storyboards': set(MEDIA_EXTENSIONS.storyboards),
- }
- def __init__(self, params=None, auto_init=True):
- """Create a FileDownloader object with the given options.
- @param auto_init Whether to load the default extractors and print header (if verbose).
- Set to 'no_verbose_header' to not print the header
- """
- if params is None:
- params = {}
- self.params = params
- self._ies = {}
- self._ies_instances = {}
- self._pps = {k: [] for k in POSTPROCESS_WHEN}
- self._printed_messages = set()
- self._first_webpage_request = True
- self._post_hooks = []
- self._progress_hooks = []
- self._postprocessor_hooks = []
- self._download_retcode = 0
- self._num_downloads = 0
- self._num_videos = 0
- self._playlist_level = 0
- self._playlist_urls = set()
- self.cache = Cache(self)
- self.__header_cookies = []
- stdout = sys.stderr if self.params.get('logtostderr') else sys.stdout
- self._out_files = Namespace(
- out=stdout,
- error=sys.stderr,
- screen=sys.stderr if self.params.get('quiet') else stdout,
- console=None if compat_os_name == 'nt' else next(
- filter(supports_terminal_sequences, (sys.stderr, sys.stdout)), None),
- )
- try:
- windows_enable_vt_mode()
- except Exception as e:
- self.write_debug(f'Failed to enable VT mode: {e}')
- if self.params.get('no_color'):
- if self.params.get('color') is not None:
- self.params.setdefault('_warnings', []).append(
- 'Overwriting params from "color" with "no_color"')
- self.params['color'] = 'no_color'
- term_allow_color = os.getenv('TERM', '').lower() != 'dumb'
- base_no_color = bool(os.getenv('NO_COLOR'))
- def process_color_policy(stream):
- stream_name = {sys.stdout: 'stdout', sys.stderr: 'stderr'}[stream]
- policy = traverse_obj(self.params, ('color', (stream_name, None), {str}, any)) or 'auto'
- if policy in ('auto', 'auto-tty', 'no_color-tty'):
- no_color = base_no_color
- if policy.endswith('tty'):
- no_color = policy.startswith('no_color')
- if term_allow_color and supports_terminal_sequences(stream):
- return 'no_color' if no_color else True
- return False
- assert policy in ('always', 'never', 'no_color'), policy
- return {'always': True, 'never': False}.get(policy, policy)
- self._allow_colors = Namespace(**{
- name: process_color_policy(stream)
- for name, stream in self._out_files.items_ if name != 'console'
- })
- system_deprecation = _get_system_deprecation()
- if system_deprecation:
- self.deprecated_feature(system_deprecation.replace('\n', '\n '))
- if self.params.get('allow_unplayable_formats'):
- self.report_warning(
- f'You have asked for {self._format_err("UNPLAYABLE", self.Styles.EMPHASIS)} formats to be listed/downloaded. '
- 'This is a developer option intended for debugging. \n'
- ' If you experience any issues while using this option, '
- f'{self._format_err("DO NOT", self.Styles.ERROR)} open a bug report')
- if self.params.get('bidi_workaround', False):
- try:
- import pty
- master, slave = pty.openpty()
- width = shutil.get_terminal_size().columns
- width_args = [] if width is None else ['-w', str(width)]
- sp_kwargs = {'stdin': subprocess.PIPE, 'stdout': slave, 'stderr': self._out_files.error}
- try:
- self._output_process = Popen(['bidiv', *width_args], **sp_kwargs)
- except OSError:
- self._output_process = Popen(['fribidi', '-c', 'UTF-8', *width_args], **sp_kwargs)
- self._output_channel = os.fdopen(master, 'rb')
- except OSError as ose:
- if ose.errno == errno.ENOENT:
- self.report_warning(
- 'Could not find fribidi executable, ignoring --bidi-workaround. '
- 'Make sure that fribidi is an executable file in one of the directories in your $PATH.')
- else:
- raise
- self.params['compat_opts'] = set(self.params.get('compat_opts', ()))
- self.params['http_headers'] = HTTPHeaderDict(std_headers, self.params.get('http_headers'))
- self._load_cookies(self.params['http_headers'].get('Cookie')) # compat
- self.params['http_headers'].pop('Cookie', None)
- if auto_init and auto_init != 'no_verbose_header':
- self.print_debug_header()
- def check_deprecated(param, option, suggestion):
- if self.params.get(param) is not None:
- self.report_warning(f'{option} is deprecated. Use {suggestion} instead')
- return True
- return False
- if check_deprecated('cn_verification_proxy', '--cn-verification-proxy', '--geo-verification-proxy'):
- if self.params.get('geo_verification_proxy') is None:
- self.params['geo_verification_proxy'] = self.params['cn_verification_proxy']
- check_deprecated('autonumber', '--auto-number', '-o "%(autonumber)s-%(title)s.%(ext)s"')
- check_deprecated('usetitle', '--title', '-o "%(title)s-%(id)s.%(ext)s"')
- check_deprecated('useid', '--id', '-o "%(id)s.%(ext)s"')
- for msg in self.params.get('_warnings', []):
- self.report_warning(msg)
- for msg in self.params.get('_deprecation_warnings', []):
- self.deprecated_feature(msg)
- if impersonate_target := self.params.get('impersonate'):
- if not self._impersonate_target_available(impersonate_target):
- raise YoutubeDLError(
- f'Impersonate target "{impersonate_target}" is not available. '
- f'Use --list-impersonate-targets to see available targets. '
- f'You may be missing dependencies required to support this target.')
- if 'list-formats' in self.params['compat_opts']:
- self.params['listformats_table'] = False
- if 'overwrites' not in self.params and self.params.get('nooverwrites') is not None:
- # nooverwrites was unnecessarily changed to overwrites
- # in 0c3d0f51778b153f65c21906031c2e091fcfb641
- # This ensures compatibility with both keys
- self.params['overwrites'] = not self.params['nooverwrites']
- elif self.params.get('overwrites') is None:
- self.params.pop('overwrites', None)
- else:
- self.params['nooverwrites'] = not self.params['overwrites']
- if self.params.get('simulate') is None and any((
- self.params.get('list_thumbnails'),
- self.params.get('listformats'),
- self.params.get('listsubtitles'),
- )):
- self.params['simulate'] = 'list_only'
- self.params.setdefault('forceprint', {})
- self.params.setdefault('print_to_file', {})
- # Compatibility with older syntax
- if not isinstance(params['forceprint'], dict):
- self.params['forceprint'] = {'video': params['forceprint']}
- if auto_init:
- self.add_default_info_extractors()
- if (sys.platform != 'win32'
- and sys.getfilesystemencoding() in ['ascii', 'ANSI_X3.4-1968']
- and not self.params.get('restrictfilenames', False)):
- # Unicode filesystem API will throw errors (#1474, #13027)
- self.report_warning(
- 'Assuming --restrict-filenames since file system encoding '
- 'cannot encode all characters. '
- 'Set the LC_ALL environment variable to fix this.')
- self.params['restrictfilenames'] = True
- self._parse_outtmpl()
- # Creating format selector here allows us to catch syntax errors before the extraction
- self.format_selector = (
- self.params.get('format') if self.params.get('format') in (None, '-')
- else self.params['format'] if callable(self.params['format'])
- else self.build_format_selector(self.params['format']))
- hooks = {
- 'post_hooks': self.add_post_hook,
- 'progress_hooks': self.add_progress_hook,
- 'postprocessor_hooks': self.add_postprocessor_hook,
- }
- for opt, fn in hooks.items():
- for ph in self.params.get(opt, []):
- fn(ph)
- for pp_def_raw in self.params.get('postprocessors', []):
- pp_def = dict(pp_def_raw)
- when = pp_def.pop('when', 'post_process')
- self.add_post_processor(
- get_postprocessor(pp_def.pop('key'))(self, **pp_def),
- when=when)
- def preload_download_archive(fn):
- """Preload the archive, if any is specified"""
- archive = set()
- if fn is None:
- return archive
- elif not is_path_like(fn):
- return fn
- self.write_debug(f'Loading archive file {fn!r}')
- try:
- with locked_file(fn, 'r', encoding='utf-8') as archive_file:
- for line in archive_file:
- archive.add(line.strip())
- except OSError as ioe:
- if ioe.errno != errno.ENOENT:
- raise
- return archive
- self.archive = preload_download_archive(self.params.get('download_archive'))
- def warn_if_short_id(self, argv):
- # short YouTube ID starting with dash?
- idxs = [
- i for i, a in enumerate(argv)
- if re.match(r'^-[0-9A-Za-z_-]{10}$', a)]
- if idxs:
- correct_argv = (
- ['yt-dlp']
- + [a for i, a in enumerate(argv) if i not in idxs]
- + ['--'] + [argv[i] for i in idxs]
- )
- self.report_warning(
- 'Long argument string detected. '
- f'Use -- to separate parameters and URLs, like this:\n{shell_quote(correct_argv)}')
- def add_info_extractor(self, ie):
- """Add an InfoExtractor object to the end of the list."""
- ie_key = ie.ie_key()
- self._ies[ie_key] = ie
- if not isinstance(ie, type):
- self._ies_instances[ie_key] = ie
- ie.set_downloader(self)
- def get_info_extractor(self, ie_key):
- """
- Get an instance of an IE with name ie_key, it will try to get one from
- the _ies list, if there's no instance it will create a new one and add
- it to the extractor list.
- """
- ie = self._ies_instances.get(ie_key)
- if ie is None:
- ie = get_info_extractor(ie_key)()
- self.add_info_extractor(ie)
- return ie
- def add_default_info_extractors(self):
- """
- Add the InfoExtractors returned by gen_extractors to the end of the list
- """
- all_ies = {ie.IE_NAME.lower(): ie for ie in gen_extractor_classes()}
- all_ies['end'] = UnsupportedURLIE()
- try:
- ie_names = orderedSet_from_options(
- self.params.get('allowed_extractors', ['default']), {
- 'all': list(all_ies),
- 'default': [name for name, ie in all_ies.items() if ie._ENABLED],
- }, use_regex=True)
- except re.error as e:
- raise ValueError(f'Wrong regex for allowed_extractors: {e.pattern}')
- for name in ie_names:
- self.add_info_extractor(all_ies[name])
- self.write_debug(f'Loaded {len(ie_names)} extractors')
- def add_post_processor(self, pp, when='post_process'):
- """Add a PostProcessor object to the end of the chain."""
- assert when in POSTPROCESS_WHEN, f'Invalid when={when}'
- self._pps[when].append(pp)
- pp.set_downloader(self)
- def add_post_hook(self, ph):
- """Add the post hook"""
- self._post_hooks.append(ph)
- def add_progress_hook(self, ph):
- """Add the download progress hook"""
- self._progress_hooks.append(ph)
- def add_postprocessor_hook(self, ph):
- """Add the postprocessing progress hook"""
- self._postprocessor_hooks.append(ph)
- for pps in self._pps.values():
- for pp in pps:
- pp.add_progress_hook(ph)
- def _bidi_workaround(self, message):
- if not hasattr(self, '_output_channel'):
- return message
- assert hasattr(self, '_output_process')
- assert isinstance(message, str)
- line_count = message.count('\n') + 1
- self._output_process.stdin.write((message + '\n').encode())
- self._output_process.stdin.flush()
- res = ''.join(self._output_channel.readline().decode()
- for _ in range(line_count))
- return res[:-len('\n')]
- def _write_string(self, message, out=None, only_once=False):
- if only_once:
- if message in self._printed_messages:
- return
- self._printed_messages.add(message)
- write_string(message, out=out, encoding=self.params.get('encoding'))
- def to_stdout(self, message, skip_eol=False, quiet=None):
- """Print message to stdout"""
- if quiet is not None:
- self.deprecation_warning('"YoutubeDL.to_stdout" no longer accepts the argument quiet. '
- 'Use "YoutubeDL.to_screen" instead')
- if skip_eol is not False:
- self.deprecation_warning('"YoutubeDL.to_stdout" no longer accepts the argument skip_eol. '
- 'Use "YoutubeDL.to_screen" instead')
- self._write_string(f'{self._bidi_workaround(message)}\n', self._out_files.out)
- def to_screen(self, message, skip_eol=False, quiet=None, only_once=False):
- """Print message to screen if not in quiet mode"""
- if self.params.get('logger'):
- self.params['logger'].debug(message)
- return
- if (self.params.get('quiet') if quiet is None else quiet) and not self.params.get('verbose'):
- return
- self._write_string(
- '{}{}'.format(self._bidi_workaround(message), ('' if skip_eol else '\n')),
- self._out_files.screen, only_once=only_once)
- def to_stderr(self, message, only_once=False):
- """Print message to stderr"""
- assert isinstance(message, str)
- if self.params.get('logger'):
- self.params['logger'].error(message)
- else:
- self._write_string(f'{self._bidi_workaround(message)}\n', self._out_files.error, only_once=only_once)
- def _send_console_code(self, code):
- if compat_os_name == 'nt' or not self._out_files.console:
- return
- self._write_string(code, self._out_files.console)
- def to_console_title(self, message):
- if not self.params.get('consoletitle', False):
- return
- message = remove_terminal_sequences(message)
- if compat_os_name == 'nt':
- if ctypes.windll.kernel32.GetConsoleWindow():
- # c_wchar_p() might not be necessary if `message` is
- # already of type unicode()
- ctypes.windll.kernel32.SetConsoleTitleW(ctypes.c_wchar_p(message))
- else:
- self._send_console_code(f'\033]0;{message}\007')
- def save_console_title(self):
- if not self.params.get('consoletitle') or self.params.get('simulate'):
- return
- self._send_console_code('\033[22;0t') # Save the title on stack
- def restore_console_title(self):
- if not self.params.get('consoletitle') or self.params.get('simulate'):
- return
- self._send_console_code('\033[23;0t') # Restore the title from stack
- def __enter__(self):
- self.save_console_title()
- return self
- def save_cookies(self):
- if self.params.get('cookiefile') is not None:
- self.cookiejar.save()
- def __exit__(self, *args):
- self.restore_console_title()
- self.close()
- def close(self):
- self.save_cookies()
- if '_request_director' in self.__dict__:
- self._request_director.close()
- del self._request_director
- def trouble(self, message=None, tb=None, is_error=True):
- """Determine action to take when a download problem appears.
- Depending on if the downloader has been configured to ignore
- download errors or not, this method may throw an exception or
- not when errors are found, after printing the message.
- @param tb If given, is additional traceback information
- @param is_error Whether to raise error according to ignorerrors
- """
- if message is not None:
- self.to_stderr(message)
- if self.params.get('verbose'):
- if tb is None:
- if sys.exc_info()[0]: # if .trouble has been called from an except block
- tb = ''
- if hasattr(sys.exc_info()[1], 'exc_info') and sys.exc_info()[1].exc_info[0]:
- tb += ''.join(traceback.format_exception(*sys.exc_info()[1].exc_info))
- tb += encode_compat_str(traceback.format_exc())
- else:
- tb_data = traceback.format_list(traceback.extract_stack())
- tb = ''.join(tb_data)
- if tb:
- self.to_stderr(tb)
- if not is_error:
- return
- if not self.params.get('ignoreerrors'):
- if sys.exc_info()[0] and hasattr(sys.exc_info()[1], 'exc_info') and sys.exc_info()[1].exc_info[0]:
- exc_info = sys.exc_info()[1].exc_info
- else:
- exc_info = sys.exc_info()
- raise DownloadError(message, exc_info)
- self._download_retcode = 1
- Styles = Namespace(
- HEADERS='yellow',
- EMPHASIS='light blue',
- FILENAME='green',
- ID='green',
- DELIM='blue',
- ERROR='red',
- BAD_FORMAT='light red',
- WARNING='yellow',
- SUPPRESS='light black',
- )
- def _format_text(self, handle, allow_colors, text, f, fallback=None, *, test_encoding=False):
- text = str(text)
- if test_encoding:
- original_text = text
- # handle.encoding can be None. See https://github.com/yt-dlp/yt-dlp/issues/2711
- encoding = self.params.get('encoding') or getattr(handle, 'encoding', None) or 'ascii'
- text = text.encode(encoding, 'ignore').decode(encoding)
- if fallback is not None and text != original_text:
- text = fallback
- return format_text(text, f) if allow_colors is True else text if fallback is None else fallback
- def _format_out(self, *args, **kwargs):
- return self._format_text(self._out_files.out, self._allow_colors.out, *args, **kwargs)
- def _format_screen(self, *args, **kwargs):
- return self._format_text(self._out_files.screen, self._allow_colors.screen, *args, **kwargs)
- def _format_err(self, *args, **kwargs):
- return self._format_text(self._out_files.error, self._allow_colors.error, *args, **kwargs)
- def report_warning(self, message, only_once=False):
- """
- Print the message to stderr, it will be prefixed with 'WARNING:'
- If stderr is a tty file the 'WARNING:' will be colored
- """
- if self.params.get('logger') is not None:
- self.params['logger'].warning(message)
- else:
- if self.params.get('no_warnings'):
- return
- self.to_stderr(f'{self._format_err("WARNING:", self.Styles.WARNING)} {message}', only_once)
- def deprecation_warning(self, message, *, stacklevel=0):
- deprecation_warning(
- message, stacklevel=stacklevel + 1, printer=self.report_error, is_error=False)
- def deprecated_feature(self, message):
- if self.params.get('logger') is not None:
- self.params['logger'].warning(f'Deprecated Feature: {message}')
- self.to_stderr(f'{self._format_err("Deprecated Feature:", self.Styles.ERROR)} {message}', True)
- def report_error(self, message, *args, **kwargs):
- """
- Do the same as trouble, but prefixes the message with 'ERROR:', colored
- in red if stderr is a tty file.
- """
- self.trouble(f'{self._format_err("ERROR:", self.Styles.ERROR)} {message}', *args, **kwargs)
- def write_debug(self, message, only_once=False):
- """Log debug message or Print message to stderr"""
- if not self.params.get('verbose', False):
- return
- message = f'[debug] {message}'
- if self.params.get('logger'):
- self.params['logger'].debug(message)
- else:
- self.to_stderr(message, only_once)
- def report_file_already_downloaded(self, file_name):
- """Report file has already been fully downloaded."""
- try:
- self.to_screen(f'[download] {file_name} has already been downloaded')
- except UnicodeEncodeError:
- self.to_screen('[download] The file has already been downloaded')
- def report_file_delete(self, file_name):
- """Report that existing file will be deleted."""
- try:
- self.to_screen(f'Deleting existing file {file_name}')
- except UnicodeEncodeError:
- self.to_screen('Deleting existing file')
- def raise_no_formats(self, info, forced=False, *, msg=None):
- has_drm = info.get('_has_drm')
- ignored, expected = self.params.get('ignore_no_formats_error'), bool(msg)
- msg = msg or has_drm and 'This video is DRM protected' or 'No video formats found!'
- if forced or not ignored:
- raise ExtractorError(msg, video_id=info['id'], ie=info['extractor'],
- expected=has_drm or ignored or expected)
- else:
- self.report_warning(msg)
- def parse_outtmpl(self):
- self.deprecation_warning('"YoutubeDL.parse_outtmpl" is deprecated and may be removed in a future version')
- self._parse_outtmpl()
- return self.params['outtmpl']
- def _parse_outtmpl(self):
- sanitize = IDENTITY
- if self.params.get('restrictfilenames'): # Remove spaces in the default template
- sanitize = lambda x: x.replace(' - ', ' ').replace(' ', '-')
- outtmpl = self.params.setdefault('outtmpl', {})
- if not isinstance(outtmpl, dict):
- self.params['outtmpl'] = outtmpl = {'default': outtmpl}
- outtmpl.update({k: sanitize(v) for k, v in DEFAULT_OUTTMPL.items() if outtmpl.get(k) is None})
- def get_output_path(self, dir_type='', filename=None):
- paths = self.params.get('paths', {})
- assert isinstance(paths, dict), '"paths" parameter must be a dictionary'
- path = os.path.join(
- expand_path(paths.get('home', '').strip()),
- expand_path(paths.get(dir_type, '').strip()) if dir_type else '',
- filename or '')
- return sanitize_path(path, force=self.params.get('windowsfilenames'))
- @staticmethod
- def _outtmpl_expandpath(outtmpl):
- # expand_path translates '%%' into '%' and '$$' into '$'
- # correspondingly that is not what we want since we need to keep
- # '%%' intact for template dict substitution step. Working around
- # with boundary-alike separator hack.
- sep = ''.join(random.choices(string.ascii_letters, k=32))
- outtmpl = outtmpl.replace('%%', f'%{sep}%').replace('$$', f'${sep}$')
- # outtmpl should be expand_path'ed before template dict substitution
- # because meta fields may contain env variables we don't want to
- # be expanded. E.g. for outtmpl "%(title)s.%(ext)s" and
- # title "Hello $PATH", we don't want `$PATH` to be expanded.
- return expand_path(outtmpl).replace(sep, '')
- @staticmethod
- def escape_outtmpl(outtmpl):
- """ Escape any remaining strings like %s, %abc% etc. """
- return re.sub(
- STR_FORMAT_RE_TMPL.format('', '(?![%(\0])'),
- lambda mobj: ('' if mobj.group('has_key') else '%') + mobj.group(0),
- outtmpl)
- @classmethod
- def validate_outtmpl(cls, outtmpl):
- """ @return None or Exception object """
- outtmpl = re.sub(
- STR_FORMAT_RE_TMPL.format('[^)]*', '[ljhqBUDS]'),
- lambda mobj: f'{mobj.group(0)[:-1]}s',
- cls._outtmpl_expandpath(outtmpl))
- try:
- cls.escape_outtmpl(outtmpl) % collections.defaultdict(int)
- return None
- except ValueError as err:
- return err
- @staticmethod
- def _copy_infodict(info_dict):
- info_dict = dict(info_dict)
- info_dict.pop('__postprocessors', None)
- info_dict.pop('__pending_error', None)
- return info_dict
- def prepare_outtmpl(self, outtmpl, info_dict, sanitize=False):
- """ Make the outtmpl and info_dict suitable for substitution: ydl.escape_outtmpl(outtmpl) % info_dict
- @param sanitize Whether to sanitize the output as a filename.
- For backward compatibility, a function can also be passed
- """
- info_dict.setdefault('epoch', int(time.time())) # keep epoch consistent once set
- info_dict = self._copy_infodict(info_dict)
- info_dict['duration_string'] = ( # %(duration>%H-%M-%S)s is wrong if duration > 24hrs
- formatSeconds(info_dict['duration'], '-' if sanitize else ':')
- if info_dict.get('duration', None) is not None
- else None)
- info_dict['autonumber'] = int(self.params.get('autonumber_start', 1) - 1 + self._num_downloads)
- info_dict['video_autonumber'] = self._num_videos
- if info_dict.get('resolution') is None:
- info_dict['resolution'] = self.format_resolution(info_dict, default=None)
- # For fields playlist_index, playlist_autonumber and autonumber convert all occurrences
- # of %(field)s to %(field)0Nd for backward compatibility
- field_size_compat_map = {
- 'playlist_index': number_of_digits(info_dict.get('__last_playlist_index') or 0),
- 'playlist_autonumber': number_of_digits(info_dict.get('n_entries') or 0),
- 'autonumber': self.params.get('autonumber_size') or 5,
- }
- TMPL_DICT = {}
- EXTERNAL_FORMAT_RE = re.compile(STR_FORMAT_RE_TMPL.format('[^)]*', f'[{STR_FORMAT_TYPES}ljhqBUDS]'))
- MATH_FUNCTIONS = {
- '+': float.__add__,
- '-': float.__sub__,
- '*': float.__mul__,
- }
- # Field is of the form key1.key2...
- # where keys (except first) can be string, int, slice or "{field, ...}"
- FIELD_INNER_RE = r'(?:\w+|%(num)s|%(num)s?(?::%(num)s?){1,2})' % {'num': r'(?:-?\d+)'} # noqa: UP031
- FIELD_RE = r'\w*(?:\.(?:%(inner)s|{%(field)s(?:,%(field)s)*}))*' % { # noqa: UP031
- 'inner': FIELD_INNER_RE,
- 'field': rf'\w*(?:\.{FIELD_INNER_RE})*',
- }
- MATH_FIELD_RE = rf'(?:{FIELD_RE}|-?{NUMBER_RE})'
- MATH_OPERATORS_RE = r'(?:{})'.format('|'.join(map(re.escape, MATH_FUNCTIONS.keys())))
- INTERNAL_FORMAT_RE = re.compile(rf'''(?xs)
- (?P<negate>-)?
- (?P<fields>{FIELD_RE})
- (?P<maths>(?:{MATH_OPERATORS_RE}{MATH_FIELD_RE})*)
- (?:>(?P<strf_format>.+?))?
- (?P<remaining>
- (?P<alternate>(?<!\\),[^|&)]+)?
- (?:&(?P<replacement>.*?))?
- (?:\|(?P<default>.*?))?
- )$''')
- def _from_user_input(field):
- if field == ':':
- return ...
- elif ':' in field:
- return slice(*map(int_or_none, field.split(':')))
- elif int_or_none(field) is not None:
- return int(field)
- return field
- def _traverse_infodict(fields):
- fields = [f for x in re.split(r'\.({.+?})\.?', fields)
- for f in ([x] if x.startswith('{') else x.split('.'))]
- for i in (0, -1):
- if fields and not fields[i]:
- fields.pop(i)
- for i, f in enumerate(fields):
- if not f.startswith('{'):
- fields[i] = _from_user_input(f)
- continue
- assert f.endswith('}'), f'No closing brace for {f} in {fields}'
- fields[i] = {k: list(map(_from_user_input, k.split('.'))) for k in f[1:-1].split(',')}
- return traverse_obj(info_dict, fields, traverse_string=True)
- def get_value(mdict):
- # Object traversal
- value = _traverse_infodict(mdict['fields'])
- # Negative
- if mdict['negate']:
- value = float_or_none(value)
- if value is not None:
- value *= -1
- # Do maths
- offset_key = mdict['maths']
- if offset_key:
- value = float_or_none(value)
- operator = None
- while offset_key:
- item = re.match(
- MATH_FIELD_RE if operator else MATH_OPERATORS_RE,
- offset_key).group(0)
- offset_key = offset_key[len(item):]
- if operator is None:
- operator = MATH_FUNCTIONS[item]
- continue
- item, multiplier = (item[1:], -1) if item[0] == '-' else (item, 1)
- offset = float_or_none(item)
- if offset is None:
- offset = float_or_none(_traverse_infodict(item))
- try:
- value = operator(value, multiplier * offset)
- except (TypeError, ZeroDivisionError):
- return None
- operator = None
- # Datetime formatting
- if mdict['strf_format']:
- value = strftime_or_none(value, mdict['strf_format'].replace('\\,', ','))
- # XXX: Workaround for https://github.com/yt-dlp/yt-dlp/issues/4485
- if sanitize and value == '':
- value = None
- return value
- na = self.params.get('outtmpl_na_placeholder', 'NA')
- def filename_sanitizer(key, value, restricted=self.params.get('restrictfilenames')):
- return sanitize_filename(str(value), restricted=restricted, is_id=(
- bool(re.search(r'(^|[_.])id(\.|$)', key))
- if 'filename-sanitization' in self.params['compat_opts']
- else NO_DEFAULT))
- sanitizer = sanitize if callable(sanitize) else filename_sanitizer
- sanitize = bool(sanitize)
- def _dumpjson_default(obj):
- if isinstance(obj, (set, LazyList)):
- return list(obj)
- return repr(obj)
- class _ReplacementFormatter(string.Formatter):
- def get_field(self, field_name, args, kwargs):
- if field_name.isdigit():
- return args[0], -1
- raise ValueError('Unsupported field')
- replacement_formatter = _ReplacementFormatter()
- def create_key(outer_mobj):
- if not outer_mobj.group('has_key'):
- return outer_mobj.group(0)
- key = outer_mobj.group('key')
- mobj = re.match(INTERNAL_FORMAT_RE, key)
- value, replacement, default, last_field = None, None, na, ''
- while mobj:
- mobj = mobj.groupdict()
- default = mobj['default'] if mobj['default'] is not None else default
- value = get_value(mobj)
- last_field, replacement = mobj['fields'], mobj['replacement']
- if value is None and mobj['alternate']:
- mobj = re.match(INTERNAL_FORMAT_RE, mobj['remaining'][1:])
- else:
- break
- if None not in (value, replacement):
- try:
- value = replacement_formatter.format(replacement, value)
- except ValueError:
- value, default = None, na
- fmt = outer_mobj.group('format')
- if fmt == 's' and last_field in field_size_compat_map and isinstance(value, int):
- fmt = f'0{field_size_compat_map[last_field]:d}d'
- flags = outer_mobj.group('conversion') or ''
- str_fmt = f'{fmt[:-1]}s'
- if value is None:
- value, fmt = default, 's'
- elif fmt[-1] == 'l': # list
- delim = '\n' if '#' in flags else ', '
- value, fmt = delim.join(map(str, variadic(value, allowed_types=(str, bytes)))), str_fmt
- elif fmt[-1] == 'j': # json
- value, fmt = json.dumps(
- value, default=_dumpjson_default,
- indent=4 if '#' in flags else None, ensure_ascii='+' not in flags), str_fmt
- elif fmt[-1] == 'h': # html
- value, fmt = escapeHTML(str(value)), str_fmt
- elif fmt[-1] == 'q': # quoted
- value = map(str, variadic(value) if '#' in flags else [value])
- value, fmt = shell_quote(value, shell=True), str_fmt
- elif fmt[-1] == 'B': # bytes
- value = f'%{str_fmt}'.encode() % str(value).encode()
- value, fmt = value.decode('utf-8', 'ignore'), 's'
- elif fmt[-1] == 'U': # unicode normalized
- value, fmt = unicodedata.normalize(
- # "+" = compatibility equivalence, "#" = NFD
- 'NF{}{}'.format('K' if '+' in flags else '', 'D' if '#' in flags else 'C'),
- value), str_fmt
- elif fmt[-1] == 'D': # decimal suffix
- num_fmt, fmt = fmt[:-1].replace('#', ''), 's'
- value = format_decimal_suffix(value, f'%{num_fmt}f%s' if num_fmt else '%d%s',
- factor=1024 if '#' in flags else 1000)
- elif fmt[-1] == 'S': # filename sanitization
- value, fmt = filename_sanitizer(last_field, value, restricted='#' in flags), str_fmt
- elif fmt[-1] == 'c':
- if value:
- value = str(value)[0]
- else:
- fmt = str_fmt
- elif fmt[-1] not in 'rsa': # numeric
- value = float_or_none(value)
- if value is None:
- value, fmt = default, 's'
- if sanitize:
- # If value is an object, sanitize might convert it to a string
- # So we convert it to repr first
- if fmt[-1] == 'r':
- value, fmt = repr(value), str_fmt
- elif fmt[-1] == 'a':
- value, fmt = ascii(value), str_fmt
- if fmt[-1] in 'csra':
- value = sanitizer(last_field, value)
- key = '{}\0{}'.format(key.replace('%', '%\0'), outer_mobj.group('format'))
- TMPL_DICT[key] = value
- return '{prefix}%({key}){fmt}'.format(key=key, fmt=fmt, prefix=outer_mobj.group('prefix'))
- return EXTERNAL_FORMAT_RE.sub(create_key, outtmpl), TMPL_DICT
- def evaluate_outtmpl(self, outtmpl, info_dict, *args, **kwargs):
- outtmpl, info_dict = self.prepare_outtmpl(outtmpl, info_dict, *args, **kwargs)
- return self.escape_outtmpl(outtmpl) % info_dict
- @_catch_unsafe_extension_error
- def _prepare_filename(self, info_dict, *, outtmpl=None, tmpl_type=None):
- assert None in (outtmpl, tmpl_type), 'outtmpl and tmpl_type are mutually exclusive'
- if outtmpl is None:
- outtmpl = self.params['outtmpl'].get(tmpl_type or 'default', self.params['outtmpl']['default'])
- try:
- outtmpl = self._outtmpl_expandpath(outtmpl)
- filename = self.evaluate_outtmpl(outtmpl, info_dict, True)
- if not filename:
- return None
- if tmpl_type in ('', 'temp'):
- final_ext, ext = self.params.get('final_ext'), info_dict.get('ext')
- if final_ext and ext and final_ext != ext and filename.endswith(f'.{final_ext}'):
- filename = replace_extension(filename, ext, final_ext)
- elif tmpl_type:
- force_ext = OUTTMPL_TYPES[tmpl_type]
- if force_ext:
- filename = replace_extension(filename, force_ext, info_dict.get('ext'))
- # https://github.com/blackjack4494/youtube-dlc/issues/85
- trim_file_name = self.params.get('trim_file_name', False)
- if trim_file_name:
- no_ext, *ext = filename.rsplit('.', 2)
- filename = join_nonempty(no_ext[:trim_file_name], *ext, delim='.')
- return filename
- except ValueError as err:
- self.report_error('Error in output template: ' + str(err) + ' (encoding: ' + repr(preferredencoding()) + ')')
- return None
- def prepare_filename(self, info_dict, dir_type='', *, outtmpl=None, warn=False):
- """Generate the output filename"""
- if outtmpl:
- assert not dir_type, 'outtmpl and dir_type are mutually exclusive'
- dir_type = None
- filename = self._prepare_filename(info_dict, tmpl_type=dir_type, outtmpl=outtmpl)
- if not filename and dir_type not in ('', 'temp'):
- return ''
- if warn:
- if not self.params.get('paths'):
- pass
- elif filename == '-':
- self.report_warning('--paths is ignored when an outputting to stdout', only_once=True)
- elif os.path.isabs(filename):
- self.report_warning('--paths is ignored since an absolute path is given in output template', only_once=True)
- if filename == '-' or not filename:
- return filename
- return self.get_output_path(dir_type, filename)
- def _match_entry(self, info_dict, incomplete=False, silent=False):
- """Returns None if the file should be downloaded"""
- _type = 'video' if 'playlist-match-filter' in self.params['compat_opts'] else info_dict.get('_type', 'video')
- assert incomplete or _type == 'video', 'Only video result can be considered complete'
- video_title = info_dict.get('title', info_dict.get('id', 'entry'))
- def check_filter():
- if _type in ('playlist', 'multi_video'):
- return
- elif _type in ('url', 'url_transparent') and not try_call(
- lambda: self.get_info_extractor(info_dict['ie_key']).is_single_video(info_dict['url'])):
- return
- if 'title' in info_dict:
- # This can happen when we're just evaluating the playlist
- title = info_dict['title']
- matchtitle = self.params.get('matchtitle', False)
- if matchtitle:
- if not re.search(matchtitle, title, re.IGNORECASE):
- return '"' + title + '" title did not match pattern "' + matchtitle + '"'
- rejecttitle = self.params.get('rejecttitle', False)
- if rejecttitle:
- if re.search(rejecttitle, title, re.IGNORECASE):
- return '"' + title + '" title matched reject pattern "' + rejecttitle + '"'
- date = info_dict.get('upload_date')
- if date is not None:
- date_range = self.params.get('daterange', DateRange())
- if date not in date_range:
- return f'{date_from_str(date).isoformat()} upload date is not in range {date_range}'
- view_count = info_dict.get('view_count')
- if view_count is not None:
- min_views = self.params.get('min_views')
- if min_views is not None and view_count < min_views:
- return 'Skipping %s, because it has not reached minimum view count (%d/%d)' % (video_title, view_count, min_views)
- max_views = self.params.get('max_views')
- if max_views is not None and view_count > max_views:
- return 'Skipping %s, because it has exceeded the maximum view count (%d/%d)' % (video_title, view_count, max_views)
- if age_restricted(info_dict.get('age_limit'), self.params.get('age_limit')):
- return f'Skipping "{video_title}" because it is age restricted'
- match_filter = self.params.get('match_filter')
- if match_filter is None:
- return None
- cancelled = None
- try:
- try:
- ret = match_filter(info_dict, incomplete=incomplete)
- except TypeError:
- # For backward compatibility
- ret = None if incomplete else match_filter(info_dict)
- except DownloadCancelled as err:
- if err.msg is not NO_DEFAULT:
- raise
- ret, cancelled = err.msg, err
- if ret is NO_DEFAULT:
- while True:
- filename = self._format_screen(self.prepare_filename(info_dict), self.Styles.FILENAME)
- reply = input(self._format_screen(
- f'Download "{filename}"? (Y/n): ', self.Styles.EMPHASIS)).lower().strip()
- if reply in {'y', ''}:
- return None
- elif reply == 'n':
- if cancelled:
- raise type(cancelled)(f'Skipping {video_title}')
- return f'Skipping {video_title}'
- return ret
- if self.in_download_archive(info_dict):
- reason = ''.join((
- format_field(info_dict, 'id', f'{self._format_screen("%s", self.Styles.ID)}: '),
- format_field(info_dict, 'title', f'{self._format_screen("%s", self.Styles.EMPHASIS)} '),
- 'has already been recorded in the archive'))
- break_opt, break_err = 'break_on_existing', ExistingVideoReached
- else:
- try:
- reason = check_filter()
- except DownloadCancelled as e:
- reason, break_opt, break_err = e.msg, 'match_filter', type(e)
- else:
- break_opt, break_err = 'break_on_reject', RejectedVideoReached
- if reason is not None:
- if not silent:
- self.to_screen('[download] ' + reason)
- if self.params.get(break_opt, False):
- raise break_err()
- return reason
- @staticmethod
- def add_extra_info(info_dict, extra_info):
- """Set the keys from extra_info in info dict if they are missing"""
- for key, value in extra_info.items():
- info_dict.setdefault(key, value)
- def extract_info(self, url, download=True, ie_key=None, extra_info=None,
- process=True, force_generic_extractor=False):
- """
- Extract and return the information dictionary of the URL
- Arguments:
- @param url URL to extract
- Keyword arguments:
- @param download Whether to download videos
- @param process Whether to resolve all unresolved references (URLs, playlist items).
- Must be True for download to work
- @param ie_key Use only the extractor with this key
- @param extra_info Dictionary containing the extra values to add to the info (For internal use only)
- @force_generic_extractor Force using the generic extractor (Deprecated; use ie_key='Generic')
- """
- if extra_info is None:
- extra_info = {}
- if not ie_key and force_generic_extractor:
- ie_key = 'Generic'
- if ie_key:
- ies = {ie_key: self._ies[ie_key]} if ie_key in self._ies else {}
- else:
- ies = self._ies
- for key, ie in ies.items():
- if not ie.suitable(url):
- continue
- if not ie.working():
- self.report_warning('The program functionality for this site has been marked as broken, '
- 'and will probably not work.')
- temp_id = ie.get_temp_id(url)
- if temp_id is not None and self.in_download_archive({'id': temp_id, 'ie_key': key}):
- self.to_screen(f'[download] {self._format_screen(temp_id, self.Styles.ID)}: '
- 'has already been recorded in the archive')
- if self.params.get('break_on_existing', False):
- raise ExistingVideoReached
- break
- return self.__extract_info(url, self.get_info_extractor(key), download, extra_info, process)
- else:
- extractors_restricted = self.params.get('allowed_extractors') not in (None, ['default'])
- self.report_error(f'No suitable extractor{format_field(ie_key, None, " (%s)")} found for URL {url}',
- tb=False if extractors_restricted else None)
- def _handle_extraction_exceptions(func):
- @functools.wraps(func)
- def wrapper(self, *args, **kwargs):
- while True:
- try:
- return func(self, *args, **kwargs)
- except (CookieLoadError, DownloadCancelled, LazyList.IndexError, PagedList.IndexError):
- raise
- except ReExtractInfo as e:
- if e.expected:
- self.to_screen(f'{e}; Re-extracting data')
- else:
- self.to_stderr('\r')
- self.report_warning(f'{e}; Re-extracting data')
- continue
- except GeoRestrictedError as e:
- msg = e.msg
- if e.countries:
- msg += '\nThis video is available in {}.'.format(', '.join(
- map(ISO3166Utils.short2full, e.countries)))
- msg += '\nYou might want to use a VPN or a proxy server (with --proxy) to workaround.'
- self.report_error(msg)
- except ExtractorError as e: # An error we somewhat expected
- self.report_error(str(e), e.format_traceback())
- except Exception as e:
- if self.params.get('ignoreerrors'):
- self.report_error(str(e), tb=encode_compat_str(traceback.format_exc()))
- else:
- raise
- break
- return wrapper
- def _wait_for_video(self, ie_result={}):
- if (not self.params.get('wait_for_video')
- or ie_result.get('_type', 'video') != 'video'
- or ie_result.get('formats') or ie_result.get('url')):
- return
- format_dur = lambda dur: '%02d:%02d:%02d' % timetuple_from_msec(dur * 1000)[:-1]
- last_msg = ''
- def progress(msg):
- nonlocal last_msg
- full_msg = f'{msg}\n'
- if not self.params.get('noprogress'):
- full_msg = msg + ' ' * (len(last_msg) - len(msg)) + '\r'
- elif last_msg:
- return
- self.to_screen(full_msg, skip_eol=True)
- last_msg = msg
- min_wait, max_wait = self.params.get('wait_for_video')
- diff = try_get(ie_result, lambda x: x['release_timestamp'] - time.time())
- if diff is None and ie_result.get('live_status') == 'is_upcoming':
- diff = round(random.uniform(min_wait, max_wait) if (max_wait and min_wait) else (max_wait or min_wait), 0)
- self.report_warning('Release time of video is not known')
- elif ie_result and (diff or 0) <= 0:
- self.report_warning('Video should already be available according to extracted info')
- diff = min(max(diff or 0, min_wait or 0), max_wait or float('inf'))
- self.to_screen(f'[wait] Waiting for {format_dur(diff)} - Press Ctrl+C to try now')
- wait_till = time.time() + diff
- try:
- while True:
- diff = wait_till - time.time()
- if diff <= 0:
- progress('')
- raise ReExtractInfo('[wait] Wait period ended', expected=True)
- progress(f'[wait] Remaining time until next attempt: {self._format_screen(format_dur(diff), self.Styles.EMPHASIS)}')
- time.sleep(1)
- except KeyboardInterrupt:
- progress('')
- raise ReExtractInfo('[wait] Interrupted by user', expected=True)
- except BaseException as e:
- if not isinstance(e, ReExtractInfo):
- self.to_screen('')
- raise
- def _load_cookies(self, data, *, autoscope=True):
- """Loads cookies from a `Cookie` header
- This tries to work around the security vulnerability of passing cookies to every domain.
- See: https://github.com/yt-dlp/yt-dlp/security/advisories/GHSA-v8mc-9377-rwjj
- @param data The Cookie header as string to load the cookies from
- @param autoscope If `False`, scope cookies using Set-Cookie syntax and error for cookie without domains
- If `True`, save cookies for later to be stored in the jar with a limited scope
- If a URL, save cookies in the jar with the domain of the URL
- """
- for cookie in LenientSimpleCookie(data).values():
- if autoscope and any(cookie.values()):
- raise ValueError('Invalid syntax in Cookie Header')
- domain = cookie.get('domain') or ''
- expiry = cookie.get('expires')
- if expiry == '': # 0 is valid
- expiry = None
- prepared_cookie = http.cookiejar.Cookie(
- cookie.get('version') or 0, cookie.key, cookie.value, None, False,
- domain, True, True, cookie.get('path') or '', bool(cookie.get('path')),
- cookie.get('secure') or False, expiry, False, None, None, {})
- if domain:
- self.cookiejar.set_cookie(prepared_cookie)
- elif autoscope is True:
- self.deprecated_feature(
- 'Passing cookies as a header is a potential security risk; '
- 'they will be scoped to the domain of the downloaded urls. '
- 'Please consider loading cookies from a file or browser instead.')
- self.__header_cookies.append(prepared_cookie)
- elif autoscope:
- self.report_warning(
- 'The extractor result contains an unscoped cookie as an HTTP header. '
- f'If you are using yt-dlp with an input URL{bug_reports_message(before=",")}',
- only_once=True)
- self._apply_header_cookies(autoscope, [prepared_cookie])
- else:
- self.report_error('Unscoped cookies are not allowed; please specify some sort of scoping',
- tb=False, is_error=False)
- def _apply_header_cookies(self, url, cookies=None):
- """Applies stray header cookies to the provided url
- This loads header cookies and scopes them to the domain provided in `url`.
- While this is not ideal, it helps reduce the risk of them being sent
- to an unintended destination while mostly maintaining compatibility.
- """
- parsed = urllib.parse.urlparse(url)
- if not parsed.hostname:
- return
- for cookie in map(copy.copy, cookies or self.__header_cookies):
- cookie.domain = f'.{parsed.hostname}'
- self.cookiejar.set_cookie(cookie)
- @_handle_extraction_exceptions
- def __extract_info(self, url, ie, download, extra_info, process):
- self._apply_header_cookies(url)
- try:
- ie_result = ie.extract(url)
- except UserNotLive as e:
- if process:
- if self.params.get('wait_for_video'):
- self.report_warning(e)
- self._wait_for_video()
- raise
- if ie_result is None: # Finished already (backwards compatibility; listformats and friends should be moved here)
- self.report_warning(f'Extractor {ie.IE_NAME} returned nothing{bug_reports_message()}')
- return
- if isinstance(ie_result, list):
- # Backwards compatibility: old IE result format
- ie_result = {
- '_type': 'compat_list',
- 'entries': ie_result,
- }
- if extra_info.get('original_url'):
- ie_result.setdefault('original_url', extra_info['original_url'])
- self.add_default_extra_info(ie_result, ie, url)
- if process:
- self._wait_for_video(ie_result)
- return self.process_ie_result(ie_result, download, extra_info)
- else:
- return ie_result
- def add_default_extra_info(self, ie_result, ie, url):
- if url is not None:
- self.add_extra_info(ie_result, {
- 'webpage_url': url,
- 'original_url': url,
- })
- webpage_url = ie_result.get('webpage_url')
- if webpage_url:
- self.add_extra_info(ie_result, {
- 'webpage_url_basename': url_basename(webpage_url),
- 'webpage_url_domain': get_domain(webpage_url),
- })
- if ie is not None:
- self.add_extra_info(ie_result, {
- 'extractor': ie.IE_NAME,
- 'extractor_key': ie.ie_key(),
- })
- def process_ie_result(self, ie_result, download=True, extra_info=None):
- """
- Take the result of the ie(may be modified) and resolve all unresolved
- references (URLs, playlist items).
- It will also download the videos if 'download'.
- Returns the resolved ie_result.
- """
- if extra_info is None:
- extra_info = {}
- result_type = ie_result.get('_type', 'video')
- if result_type in ('url', 'url_transparent'):
- ie_result['url'] = sanitize_url(
- ie_result['url'], scheme='http' if self.params.get('prefer_insecure') else 'https')
- if ie_result.get('original_url') and not extra_info.get('original_url'):
- extra_info = {'original_url': ie_result['original_url'], **extra_info}
- extract_flat = self.params.get('extract_flat', False)
- if ((extract_flat == 'in_playlist' and 'playlist' in extra_info)
- or extract_flat is True):
- info_copy = ie_result.copy()
- ie = try_get(ie_result.get('ie_key'), self.get_info_extractor)
- if ie and not ie_result.get('id'):
- info_copy['id'] = ie.get_temp_id(ie_result['url'])
- self.add_default_extra_info(info_copy, ie, ie_result['url'])
- self.add_extra_info(info_copy, extra_info)
- info_copy, _ = self.pre_process(info_copy)
- self._fill_common_fields(info_copy, False)
- self.__forced_printings(info_copy)
- self._raise_pending_errors(info_copy)
- if self.params.get('force_write_download_archive', False):
- self.record_download_archive(info_copy)
- return ie_result
- if result_type == 'video':
- self.add_extra_info(ie_result, extra_info)
- ie_result = self.process_video_result(ie_result, download=download)
- self._raise_pending_errors(ie_result)
- additional_urls = (ie_result or {}).get('additional_urls')
- if additional_urls:
- # TODO: Improve MetadataParserPP to allow setting a list
- if isinstance(additional_urls, str):
- additional_urls = [additional_urls]
- self.to_screen(
- '[info] {}: {} additional URL(s) requested'.format(ie_result['id'], len(additional_urls)))
- self.write_debug('Additional URLs: "{}"'.format('", "'.join(additional_urls)))
- ie_result['additional_entries'] = [
- self.extract_info(
- url, download, extra_info=extra_info,
- force_generic_extractor=self.params.get('force_generic_extractor'))
- for url in additional_urls
- ]
- return ie_result
- elif result_type == 'url':
- # We have to add extra_info to the results because it may be
- # contained in a playlist
- return self.extract_info(
- ie_result['url'], download,
- ie_key=ie_result.get('ie_key'),
- extra_info=extra_info)
- elif result_type == 'url_transparent':
- # Use the information from the embedding page
- info = self.extract_info(
- ie_result['url'], ie_key=ie_result.get('ie_key'),
- extra_info=extra_info, download=False, process=False)
- # extract_info may return None when ignoreerrors is enabled and
- # extraction failed with an error, don't crash and return early
- # in this case
- if not info:
- return info
- exempted_fields = {'_type', 'url', 'ie_key'}
- if not ie_result.get('section_end') and ie_result.get('section_start') is None:
- # For video clips, the id etc of the clip extractor should be used
- exempted_fields |= {'id', 'extractor', 'extractor_key'}
- new_result = info.copy()
- new_result.update(filter_dict(ie_result, lambda k, v: v is not None and k not in exempted_fields))
- # Extracted info may not be a video result (i.e.
- # info.get('_type', 'video') != video) but rather an url or
- # url_transparent. In such cases outer metadata (from ie_result)
- # should be propagated to inner one (info). For this to happen
- # _type of info should be overridden with url_transparent. This
- # fixes issue from https://github.com/ytdl-org/youtube-dl/pull/11163.
- if new_result.get('_type') == 'url':
- new_result['_type'] = 'url_transparent'
- return self.process_ie_result(
- new_result, download=download, extra_info=extra_info)
- elif result_type in ('playlist', 'multi_video'):
- # Protect from infinite recursion due to recursively nested playlists
- # (see https://github.com/ytdl-org/youtube-dl/issues/27833)
- webpage_url = ie_result.get('webpage_url') # Playlists maynot have webpage_url
- if webpage_url and webpage_url in self._playlist_urls:
- self.to_screen(
- '[download] Skipping already downloaded playlist: {}'.format(
- ie_result.get('title')) or ie_result.get('id'))
- return
- self._playlist_level += 1
- self._playlist_urls.add(webpage_url)
- self._fill_common_fields(ie_result, False)
- self._sanitize_thumbnails(ie_result)
- try:
- return self.__process_playlist(ie_result, download)
- finally:
- self._playlist_level -= 1
- if not self._playlist_level:
- self._playlist_urls.clear()
- elif result_type == 'compat_list':
- self.report_warning(
- 'Extractor {} returned a compat_list result. '
- 'It needs to be updated.'.format(ie_result.get('extractor')))
- def _fixup(r):
- self.add_extra_info(r, {
- 'extractor': ie_result['extractor'],
- 'webpage_url': ie_result['webpage_url'],
- 'webpage_url_basename': url_basename(ie_result['webpage_url']),
- 'webpage_url_domain': get_domain(ie_result['webpage_url']),
- 'extractor_key': ie_result['extractor_key'],
- })
- return r
- ie_result['entries'] = [
- self.process_ie_result(_fixup(r), download, extra_info)
- for r in ie_result['entries']
- ]
- return ie_result
- else:
- raise Exception(f'Invalid result type: {result_type}')
- def _ensure_dir_exists(self, path):
- return make_dir(path, self.report_error)
- @staticmethod
- def _playlist_infodict(ie_result, strict=False, **kwargs):
- info = {
- 'playlist_count': ie_result.get('playlist_count'),
- 'playlist': ie_result.get('title') or ie_result.get('id'),
- 'playlist_id': ie_result.get('id'),
- 'playlist_title': ie_result.get('title'),
- 'playlist_uploader': ie_result.get('uploader'),
- 'playlist_uploader_id': ie_result.get('uploader_id'),
- 'playlist_channel': ie_result.get('channel'),
- 'playlist_channel_id': ie_result.get('channel_id'),
- **kwargs,
- }
- if strict:
- return info
- if ie_result.get('webpage_url'):
- info.update({
- 'webpage_url': ie_result['webpage_url'],
- 'webpage_url_basename': url_basename(ie_result['webpage_url']),
- 'webpage_url_domain': get_domain(ie_result['webpage_url']),
- })
- return {
- **info,
- 'playlist_index': 0,
- '__last_playlist_index': max(ie_result.get('requested_entries') or (0, 0)),
- 'extractor': ie_result['extractor'],
- 'extractor_key': ie_result['extractor_key'],
- }
- def __process_playlist(self, ie_result, download):
- """Process each entry in the playlist"""
- assert ie_result['_type'] in ('playlist', 'multi_video')
- common_info = self._playlist_infodict(ie_result, strict=True)
- title = common_info.get('playlist') or '<Untitled>'
- if self._match_entry(common_info, incomplete=True) is not None:
- return
- self.to_screen(f'[download] Downloading {ie_result["_type"]}: {title}')
- all_entries = PlaylistEntries(self, ie_result)
- entries = orderedSet(all_entries.get_requested_items(), lazy=True)
- lazy = self.params.get('lazy_playlist')
- if lazy:
- resolved_entries, n_entries = [], 'N/A'
- ie_result['requested_entries'], ie_result['entries'] = None, None
- else:
- entries = resolved_entries = list(entries)
- n_entries = len(resolved_entries)
- ie_result['requested_entries'], ie_result['entries'] = tuple(zip(*resolved_entries)) or ([], [])
- if not ie_result.get('playlist_count'):
- # Better to do this after potentially exhausting entries
- ie_result['playlist_count'] = all_entries.get_full_count()
- extra = self._playlist_infodict(ie_result, n_entries=int_or_none(n_entries))
- ie_copy = collections.ChainMap(ie_result, extra)
- _infojson_written = False
- write_playlist_files = self.params.get('allow_playlist_files', True)
- if write_playlist_files and self.params.get('list_thumbnails'):
- self.list_thumbnails(ie_result)
- if write_playlist_files and not self.params.get('simulate'):
- _infojson_written = self._write_info_json(
- 'playlist', ie_result, self.prepare_filename(ie_copy, 'pl_infojson'))
- if _infojson_written is None:
- return
- if self._write_description('playlist', ie_result,
- self.prepare_filename(ie_copy, 'pl_description')) is None:
- return
- # TODO: This should be passed to ThumbnailsConvertor if necessary
- self._write_thumbnails('playlist', ie_result, self.prepare_filename(ie_copy, 'pl_thumbnail'))
- if lazy:
- if self.params.get('playlistreverse') or self.params.get('playlistrandom'):
- self.report_warning('playlistreverse and playlistrandom are not supported with lazy_playlist', only_once=True)
- elif self.params.get('playlistreverse'):
- entries.reverse()
- elif self.params.get('playlistrandom'):
- random.shuffle(entries)
- self.to_screen(f'[{ie_result["extractor"]}] Playlist {title}: Downloading {n_entries} items'
- f'{format_field(ie_result, "playlist_count", " of %s")}')
- keep_resolved_entries = self.params.get('extract_flat') != 'discard'
- if self.params.get('extract_flat') == 'discard_in_playlist':
- keep_resolved_entries = ie_result['_type'] != 'playlist'
- if keep_resolved_entries:
- self.write_debug('The information of all playlist entries will be held in memory')
- failures = 0
- max_failures = self.params.get('skip_playlist_after_errors') or float('inf')
- for i, (playlist_index, entry) in enumerate(entries):
- if lazy:
- resolved_entries.append((playlist_index, entry))
- if not entry:
- continue
- entry['__x_forwarded_for_ip'] = ie_result.get('__x_forwarded_for_ip')
- if not lazy and 'playlist-index' in self.params['compat_opts']:
- playlist_index = ie_result['requested_entries'][i]
- entry_copy = collections.ChainMap(entry, {
- **common_info,
- 'n_entries': int_or_none(n_entries),
- 'playlist_index': playlist_index,
- 'playlist_autonumber': i + 1,
- })
- if self._match_entry(entry_copy, incomplete=True) is not None:
- # For compatabilty with youtube-dl. See https://github.com/yt-dlp/yt-dlp/issues/4369
- resolved_entries[i] = (playlist_index, NO_DEFAULT)
- continue
- self.to_screen(
- f'[download] Downloading item {self._format_screen(i + 1, self.Styles.ID)} '
- f'of {self._format_screen(n_entries, self.Styles.EMPHASIS)}')
- entry_result = self.__process_iterable_entry(entry, download, collections.ChainMap({
- 'playlist_index': playlist_index,
- 'playlist_autonumber': i + 1,
- }, extra))
- if not entry_result:
- failures += 1
- if failures >= max_failures:
- self.report_error(
- f'Skipping the remaining entries in playlist "{title}" since {failures} items failed extraction')
- break
- if keep_resolved_entries:
- resolved_entries[i] = (playlist_index, entry_result)
- # Update with processed data
- ie_result['entries'] = [e for _, e in resolved_entries if e is not NO_DEFAULT]
- ie_result['requested_entries'] = [i for i, e in resolved_entries if e is not NO_DEFAULT]
- if ie_result['requested_entries'] == try_call(lambda: list(range(1, ie_result['playlist_count'] + 1))):
- # Do not set for full playlist
- ie_result.pop('requested_entries')
- # Write the updated info to json
- if _infojson_written is True and self._write_info_json(
- 'updated playlist', ie_result,
- self.prepare_filename(ie_copy, 'pl_infojson'), overwrite=True) is None:
- return
- ie_result = self.run_all_pps('playlist', ie_result)
- self.to_screen(f'[download] Finished downloading playlist: {title}')
- return ie_result
- @_handle_extraction_exceptions
- def __process_iterable_entry(self, entry, download, extra_info):
- return self.process_ie_result(
- entry, download=download, extra_info=extra_info)
- def _build_format_filter(self, filter_spec):
- " Returns a function to filter the formats according to the filter_spec "
- OPERATORS = {
- '<': operator.lt,
- '<=': operator.le,
- '>': operator.gt,
- '>=': operator.ge,
- '=': operator.eq,
- '!=': operator.ne,
- }
- operator_rex = re.compile(r'''(?x)\s*
- (?P<key>[\w.-]+)\s*
- (?P<op>{})(?P<none_inclusive>\s*\?)?\s*
- (?P<value>[0-9.]+(?:[kKmMgGtTpPeEzZyY]i?[Bb]?)?)\s*
- '''.format('|'.join(map(re.escape, OPERATORS.keys()))))
- m = operator_rex.fullmatch(filter_spec)
- if m:
- try:
- comparison_value = int(m.group('value'))
- except ValueError:
- comparison_value = parse_filesize(m.group('value'))
- if comparison_value is None:
- comparison_value = parse_filesize(m.group('value') + 'B')
- if comparison_value is None:
- raise ValueError(
- 'Invalid value {!r} in format specification {!r}'.format(
- m.group('value'), filter_spec))
- op = OPERATORS[m.group('op')]
- if not m:
- STR_OPERATORS = {
- '=': operator.eq,
- '^=': lambda attr, value: attr.startswith(value),
- '$=': lambda attr, value: attr.endswith(value),
- '*=': lambda attr, value: value in attr,
- '~=': lambda attr, value: value.search(attr) is not None,
- }
- str_operator_rex = re.compile(r'''(?x)\s*
- (?P<key>[a-zA-Z0-9._-]+)\s*
- (?P<negation>!\s*)?(?P<op>{})\s*(?P<none_inclusive>\?\s*)?
- (?P<quote>["'])?
- (?P<value>(?(quote)(?:(?!(?P=quote))[^\\]|\\.)+|[\w.-]+))
- (?(quote)(?P=quote))\s*
- '''.format('|'.join(map(re.escape, STR_OPERATORS.keys()))))
- m = str_operator_rex.fullmatch(filter_spec)
- if m:
- if m.group('op') == '~=':
- comparison_value = re.compile(m.group('value'))
- else:
- comparison_value = re.sub(r'''\\([\\"'])''', r'\1', m.group('value'))
- str_op = STR_OPERATORS[m.group('op')]
- if m.group('negation'):
- op = lambda attr, value: not str_op(attr, value)
- else:
- op = str_op
- if not m:
- raise SyntaxError(f'Invalid filter specification {filter_spec!r}')
- def _filter(f):
- actual_value = f.get(m.group('key'))
- if actual_value is None:
- return m.group('none_inclusive')
- return op(actual_value, comparison_value)
- return _filter
- def _check_formats(self, formats):
- for f in formats:
- working = f.get('__working')
- if working is not None:
- if working:
- yield f
- continue
- self.to_screen('[info] Testing format {}'.format(f['format_id']))
- path = self.get_output_path('temp')
- if not self._ensure_dir_exists(f'{path}/'):
- continue
- temp_file = tempfile.NamedTemporaryFile(suffix='.tmp', delete=False, dir=path or None)
- temp_file.close()
- try:
- success, _ = self.dl(temp_file.name, f, test=True)
- except (DownloadError, OSError, ValueError, *network_exceptions):
- success = False
- finally:
- if os.path.exists(temp_file.name):
- try:
- os.remove(temp_file.name)
- except OSError:
- self.report_warning(f'Unable to delete temporary file "{temp_file.name}"')
- f['__working'] = success
- if success:
- yield f
- else:
- self.to_screen('[info] Unable to download format {}. Skipping...'.format(f['format_id']))
- def _select_formats(self, formats, selector):
- return list(selector({
- 'formats': formats,
- 'has_merged_format': any('none' not in (f.get('acodec'), f.get('vcodec')) for f in formats),
- 'incomplete_formats': (all(f.get('vcodec') == 'none' for f in formats) # No formats with video
- or all(f.get('acodec') == 'none' for f in formats)), # OR, No formats with audio
- }))
- def _default_format_spec(self, info_dict):
- prefer_best = (
- self.params['outtmpl']['default'] == '-'
- or info_dict.get('is_live') and not self.params.get('live_from_start'))
- def can_merge():
- merger = FFmpegMergerPP(self)
- return merger.available and merger.can_merge()
- if not prefer_best and not can_merge():
- prefer_best = True
- formats = self._get_formats(info_dict)
- evaluate_formats = lambda spec: self._select_formats(formats, self.build_format_selector(spec))
- if evaluate_formats('b/bv+ba') != evaluate_formats('bv*+ba/b'):
- self.report_warning('ffmpeg not found. The downloaded format may not be the best available. '
- 'Installing ffmpeg is strongly recommended: https://github.com/yt-dlp/yt-dlp#dependencies')
- compat = (self.params.get('allow_multiple_audio_streams')
- or 'format-spec' in self.params['compat_opts'])
- return ('best/bestvideo+bestaudio' if prefer_best
- else 'bestvideo+bestaudio/best' if compat
- else 'bestvideo*+bestaudio/best')
- def build_format_selector(self, format_spec):
- def syntax_error(note, start):
- message = (
- 'Invalid format specification: '
- '{}\n\t{}\n\t{}^'.format(note, format_spec, ' ' * start[1]))
- return SyntaxError(message)
- PICKFIRST = 'PICKFIRST'
- MERGE = 'MERGE'
- SINGLE = 'SINGLE'
- GROUP = 'GROUP'
- FormatSelector = collections.namedtuple('FormatSelector', ['type', 'selector', 'filters'])
- allow_multiple_streams = {'audio': self.params.get('allow_multiple_audio_streams', False),
- 'video': self.params.get('allow_multiple_video_streams', False)}
- def _parse_filter(tokens):
- filter_parts = []
- for type_, string_, _start, _, _ in tokens:
- if type_ == tokenize.OP and string_ == ']':
- return ''.join(filter_parts)
- else:
- filter_parts.append(string_)
- def _remove_unused_ops(tokens):
- # Remove operators that we don't use and join them with the surrounding strings.
- # E.g. 'mp4' '-' 'baseline' '-' '16x9' is converted to 'mp4-baseline-16x9'
- ALLOWED_OPS = ('/', '+', ',', '(', ')')
- last_string, last_start, last_end, last_line = None, None, None, None
- for type_, string_, start, end, line in tokens:
- if type_ == tokenize.OP and string_ == '[':
- if last_string:
- yield tokenize.NAME, last_string, last_start, last_end, last_line
- last_string = None
- yield type_, string_, start, end, line
- # everything inside brackets will be handled by _parse_filter
- for type_, string_, start, end, line in tokens:
- yield type_, string_, start, end, line
- if type_ == tokenize.OP and string_ == ']':
- break
- elif type_ == tokenize.OP and string_ in ALLOWED_OPS:
- if last_string:
- yield tokenize.NAME, last_string, last_start, last_end, last_line
- last_string = None
- yield type_, string_, start, end, line
- elif type_ in [tokenize.NAME, tokenize.NUMBER, tokenize.OP]:
- if not last_string:
- last_string = string_
- last_start = start
- last_end = end
- else:
- last_string += string_
- if last_string:
- yield tokenize.NAME, last_string, last_start, last_end, last_line
- def _parse_format_selection(tokens, inside_merge=False, inside_choice=False, inside_group=False):
- selectors = []
- current_selector = None
- for type_, string_, start, _, _ in tokens:
- # ENCODING is only defined in Python 3.x
- if type_ == getattr(tokenize, 'ENCODING', None):
- continue
- elif type_ in [tokenize.NAME, tokenize.NUMBER]:
- current_selector = FormatSelector(SINGLE, string_, [])
- elif type_ == tokenize.OP:
- if string_ == ')':
- if not inside_group:
- # ')' will be handled by the parentheses group
- tokens.restore_last_token()
- break
- elif inside_merge and string_ in ['/', ',']:
- tokens.restore_last_token()
- break
- elif inside_choice and string_ == ',':
- tokens.restore_last_token()
- break
- elif string_ == ',':
- if not current_selector:
- raise syntax_error('"," must follow a format selector', start)
- selectors.append(current_selector)
- current_selector = None
- elif string_ == '/':
- if not current_selector:
- raise syntax_error('"/" must follow a format selector', start)
- first_choice = current_selector
- second_choice = _parse_format_selection(tokens, inside_choice=True)
- current_selector = FormatSelector(PICKFIRST, (first_choice, second_choice), [])
- elif string_ == '[':
- if not current_selector:
- current_selector = FormatSelector(SINGLE, 'best', [])
- format_filter = _parse_filter(tokens)
- current_selector.filters.append(format_filter)
- elif string_ == '(':
- if current_selector:
- raise syntax_error('Unexpected "("', start)
- group = _parse_format_selection(tokens, inside_group=True)
- current_selector = FormatSelector(GROUP, group, [])
- elif string_ == '+':
- if not current_selector:
- raise syntax_error('Unexpected "+"', start)
- selector_1 = current_selector
- selector_2 = _parse_format_selection(tokens, inside_merge=True)
- if not selector_2:
- raise syntax_error('Expected a selector', start)
- current_selector = FormatSelector(MERGE, (selector_1, selector_2), [])
- else:
- raise syntax_error(f'Operator not recognized: "{string_}"', start)
- elif type_ == tokenize.ENDMARKER:
- break
- if current_selector:
- selectors.append(current_selector)
- return selectors
- def _merge(formats_pair):
- format_1, format_2 = formats_pair
- formats_info = []
- formats_info.extend(format_1.get('requested_formats', (format_1,)))
- formats_info.extend(format_2.get('requested_formats', (format_2,)))
- if not allow_multiple_streams['video'] or not allow_multiple_streams['audio']:
- get_no_more = {'video': False, 'audio': False}
- for (i, fmt_info) in enumerate(formats_info):
- if fmt_info.get('acodec') == fmt_info.get('vcodec') == 'none':
- formats_info.pop(i)
- continue
- for aud_vid in ['audio', 'video']:
- if not allow_multiple_streams[aud_vid] and fmt_info.get(aud_vid[0] + 'codec') != 'none':
- if get_no_more[aud_vid]:
- formats_info.pop(i)
- break
- get_no_more[aud_vid] = True
- if len(formats_info) == 1:
- return formats_info[0]
- video_fmts = [fmt_info for fmt_info in formats_info if fmt_info.get('vcodec') != 'none']
- audio_fmts = [fmt_info for fmt_info in formats_info if fmt_info.get('acodec') != 'none']
- the_only_video = video_fmts[0] if len(video_fmts) == 1 else None
- the_only_audio = audio_fmts[0] if len(audio_fmts) == 1 else None
- output_ext = get_compatible_ext(
- vcodecs=[f.get('vcodec') for f in video_fmts],
- acodecs=[f.get('acodec') for f in audio_fmts],
- vexts=[f['ext'] for f in video_fmts],
- aexts=[f['ext'] for f in audio_fmts],
- preferences=(try_call(lambda: self.params['merge_output_format'].split('/'))
- or self.params.get('prefer_free_formats') and ('webm', 'mkv')))
- filtered = lambda *keys: filter(None, (traverse_obj(fmt, *keys) for fmt in formats_info))
- new_dict = {
- 'requested_formats': formats_info,
- 'format': '+'.join(filtered('format')),
- 'format_id': '+'.join(filtered('format_id')),
- 'ext': output_ext,
- 'protocol': '+'.join(map(determine_protocol, formats_info)),
- 'language': '+'.join(orderedSet(filtered('language'))) or None,
- 'format_note': '+'.join(orderedSet(filtered('format_note'))) or None,
- 'filesize_approx': sum(filtered('filesize', 'filesize_approx')) or None,
- 'tbr': sum(filtered('tbr', 'vbr', 'abr')),
- }
- if the_only_video:
- new_dict.update({
- 'width': the_only_video.get('width'),
- 'height': the_only_video.get('height'),
- 'resolution': the_only_video.get('resolution') or self.format_resolution(the_only_video),
- 'fps': the_only_video.get('fps'),
- 'dynamic_range': the_only_video.get('dynamic_range'),
- 'vcodec': the_only_video.get('vcodec'),
- 'vbr': the_only_video.get('vbr'),
- 'stretched_ratio': the_only_video.get('stretched_ratio'),
- 'aspect_ratio': the_only_video.get('aspect_ratio'),
- })
- if the_only_audio:
- new_dict.update({
- 'acodec': the_only_audio.get('acodec'),
- 'abr': the_only_audio.get('abr'),
- 'asr': the_only_audio.get('asr'),
- 'audio_channels': the_only_audio.get('audio_channels'),
- })
- return new_dict
- def _check_formats(formats):
- if self.params.get('check_formats') == 'selected':
- yield from self._check_formats(formats)
- return
- elif (self.params.get('check_formats') is not None
- or self.params.get('allow_unplayable_formats')):
- yield from formats
- return
- for f in formats:
- if f.get('has_drm') or f.get('__needs_testing'):
- yield from self._check_formats([f])
- else:
- yield f
- def _build_selector_function(selector):
- if isinstance(selector, list): # ,
- fs = [_build_selector_function(s) for s in selector]
- def selector_function(ctx):
- for f in fs:
- yield from f(ctx)
- return selector_function
- elif selector.type == GROUP: # ()
- selector_function = _build_selector_function(selector.selector)
- elif selector.type == PICKFIRST: # /
- fs = [_build_selector_function(s) for s in selector.selector]
- def selector_function(ctx):
- for f in fs:
- picked_formats = list(f(ctx))
- if picked_formats:
- return picked_formats
- return []
- elif selector.type == MERGE: # +
- selector_1, selector_2 = map(_build_selector_function, selector.selector)
- def selector_function(ctx):
- for pair in itertools.product(selector_1(ctx), selector_2(ctx)):
- yield _merge(pair)
- elif selector.type == SINGLE: # atom
- format_spec = selector.selector or 'best'
- # TODO: Add allvideo, allaudio etc by generalizing the code with best/worst selector
- if format_spec == 'all':
- def selector_function(ctx):
- yield from _check_formats(ctx['formats'][::-1])
- elif format_spec == 'mergeall':
- def selector_function(ctx):
- formats = list(_check_formats(
- f for f in ctx['formats'] if f.get('vcodec') != 'none' or f.get('acodec') != 'none'))
- if not formats:
- return
- merged_format = formats[-1]
- for f in formats[-2::-1]:
- merged_format = _merge((merged_format, f))
- yield merged_format
- else:
- format_fallback, seperate_fallback, format_reverse, format_idx = False, None, True, 1
- mobj = re.match(
- r'(?P<bw>best|worst|b|w)(?P<type>video|audio|v|a)?(?P<mod>\*)?(?:\.(?P<n>[1-9]\d*))?$',
- format_spec)
- if mobj is not None:
- format_idx = int_or_none(mobj.group('n'), default=1)
- format_reverse = mobj.group('bw')[0] == 'b'
- format_type = (mobj.group('type') or [None])[0]
- not_format_type = {'v': 'a', 'a': 'v'}.get(format_type)
- format_modified = mobj.group('mod') is not None
- format_fallback = not format_type and not format_modified # for b, w
- _filter_f = (
- (lambda f: f.get(f'{format_type}codec') != 'none')
- if format_type and format_modified # bv*, ba*, wv*, wa*
- else (lambda f: f.get(f'{not_format_type}codec') == 'none')
- if format_type # bv, ba, wv, wa
- else (lambda f: f.get('vcodec') != 'none' and f.get('acodec') != 'none')
- if not format_modified # b, w
- else lambda f: True) # b*, w*
- filter_f = lambda f: _filter_f(f) and (
- f.get('vcodec') != 'none' or f.get('acodec') != 'none')
- else:
- if format_spec in self._format_selection_exts['audio']:
- filter_f = lambda f: f.get('ext') == format_spec and f.get('acodec') != 'none'
- elif format_spec in self._format_selection_exts['video']:
- filter_f = lambda f: f.get('ext') == format_spec and f.get('acodec') != 'none' and f.get('vcodec') != 'none'
- seperate_fallback = lambda f: f.get('ext') == format_spec and f.get('vcodec') != 'none'
- elif format_spec in self._format_selection_exts['storyboards']:
- filter_f = lambda f: f.get('ext') == format_spec and f.get('acodec') == 'none' and f.get('vcodec') == 'none'
- else:
- filter_f = lambda f: f.get('format_id') == format_spec # id
- def selector_function(ctx):
- formats = list(ctx['formats'])
- matches = list(filter(filter_f, formats)) if filter_f is not None else formats
- if not matches:
- if format_fallback and ctx['incomplete_formats']:
- # for extractors with incomplete formats (audio only (soundcloud)
- # or video only (imgur)) best/worst will fallback to
- # best/worst {video,audio}-only format
- matches = list(filter(lambda f: f.get('vcodec') != 'none' or f.get('acodec') != 'none', formats))
- elif seperate_fallback and not ctx['has_merged_format']:
- # for compatibility with youtube-dl when there is no pre-merged format
- matches = list(filter(seperate_fallback, formats))
- matches = LazyList(_check_formats(matches[::-1 if format_reverse else 1]))
- try:
- yield matches[format_idx - 1]
- except LazyList.IndexError:
- return
- filters = [self._build_format_filter(f) for f in selector.filters]
- def final_selector(ctx):
- ctx_copy = dict(ctx)
- for _filter in filters:
- ctx_copy['formats'] = list(filter(_filter, ctx_copy['formats']))
- return selector_function(ctx_copy)
- return final_selector
- # HACK: Python 3.12 changed the underlying parser, rendering '7_a' invalid
- # Prefix numbers with random letters to avoid it being classified as a number
- # See: https://github.com/yt-dlp/yt-dlp/pulls/8797
- # TODO: Implement parser not reliant on tokenize.tokenize
- prefix = ''.join(random.choices(string.ascii_letters, k=32))
- stream = io.BytesIO(re.sub(r'\d[_\d]*', rf'{prefix}\g<0>', format_spec).encode())
- try:
- tokens = list(_remove_unused_ops(
- token._replace(string=token.string.replace(prefix, ''))
- for token in tokenize.tokenize(stream.readline)))
- except tokenize.TokenError:
- raise syntax_error('Missing closing/opening brackets or parenthesis', (0, len(format_spec)))
- class TokenIterator:
- def __init__(self, tokens):
- self.tokens = tokens
- self.counter = 0
- def __iter__(self):
- return self
- def __next__(self):
- if self.counter >= len(self.tokens):
- raise StopIteration
- value = self.tokens[self.counter]
- self.counter += 1
- return value
- next = __next__
- def restore_last_token(self):
- self.counter -= 1
- parsed_selector = _parse_format_selection(iter(TokenIterator(tokens)))
- return _build_selector_function(parsed_selector)
- def _calc_headers(self, info_dict, load_cookies=False):
- res = HTTPHeaderDict(self.params['http_headers'], info_dict.get('http_headers'))
- clean_headers(res)
- if load_cookies: # For --load-info-json
- self._load_cookies(res.get('Cookie'), autoscope=info_dict['url']) # compat
- self._load_cookies(info_dict.get('cookies'), autoscope=False)
- # The `Cookie` header is removed to prevent leaks and unscoped cookies.
- # See: https://github.com/yt-dlp/yt-dlp/security/advisories/GHSA-v8mc-9377-rwjj
- res.pop('Cookie', None)
- cookies = self.cookiejar.get_cookies_for_url(info_dict['url'])
- if cookies:
- encoder = LenientSimpleCookie()
- values = []
- for cookie in cookies:
- _, value = encoder.value_encode(cookie.value)
- values.append(f'{cookie.name}={value}')
- if cookie.domain:
- values.append(f'Domain={cookie.domain}')
- if cookie.path:
- values.append(f'Path={cookie.path}')
- if cookie.secure:
- values.append('Secure')
- if cookie.expires:
- values.append(f'Expires={cookie.expires}')
- if cookie.version:
- values.append(f'Version={cookie.version}')
- info_dict['cookies'] = '; '.join(values)
- if 'X-Forwarded-For' not in res:
- x_forwarded_for_ip = info_dict.get('__x_forwarded_for_ip')
- if x_forwarded_for_ip:
- res['X-Forwarded-For'] = x_forwarded_for_ip
- return res
- def _calc_cookies(self, url):
- self.deprecation_warning('"YoutubeDL._calc_cookies" is deprecated and may be removed in a future version')
- return self.cookiejar.get_cookie_header(url)
- def _sort_thumbnails(self, thumbnails):
- thumbnails.sort(key=lambda t: (
- t.get('preference') if t.get('preference') is not None else -1,
- t.get('width') if t.get('width') is not None else -1,
- t.get('height') if t.get('height') is not None else -1,
- t.get('id') if t.get('id') is not None else '',
- t.get('url')))
- def _sanitize_thumbnails(self, info_dict):
- thumbnails = info_dict.get('thumbnails')
- if thumbnails is None:
- thumbnail = info_dict.get('thumbnail')
- if thumbnail:
- info_dict['thumbnails'] = thumbnails = [{'url': thumbnail}]
- if not thumbnails:
- return
- def check_thumbnails(thumbnails):
- for t in thumbnails:
- self.to_screen(f'[info] Testing thumbnail {t["id"]}')
- try:
- self.urlopen(HEADRequest(t['url']))
- except network_exceptions as err:
- self.to_screen(f'[info] Unable to connect to thumbnail {t["id"]} URL {t["url"]!r} - {err}. Skipping...')
- continue
- yield t
- self._sort_thumbnails(thumbnails)
- for i, t in enumerate(thumbnails):
- if t.get('id') is None:
- t['id'] = str(i)
- if t.get('width') and t.get('height'):
- t['resolution'] = '%dx%d' % (t['width'], t['height'])
- t['url'] = sanitize_url(t['url'])
- if self.params.get('check_formats') is True:
- info_dict['thumbnails'] = LazyList(check_thumbnails(thumbnails[::-1]), reverse=True)
- else:
- info_dict['thumbnails'] = thumbnails
- def _fill_common_fields(self, info_dict, final=True):
- # TODO: move sanitization here
- if final:
- title = info_dict['fulltitle'] = info_dict.get('title')
- if not title:
- if title == '':
- self.write_debug('Extractor gave empty title. Creating a generic title')
- else:
- self.report_warning('Extractor failed to obtain "title". Creating a generic title instead')
- info_dict['title'] = f'{info_dict["extractor"].replace(":", "-")} video #{info_dict["id"]}'
- if info_dict.get('duration') is not None:
- info_dict['duration_string'] = formatSeconds(info_dict['duration'])
- for ts_key, date_key in (
- ('timestamp', 'upload_date'),
- ('release_timestamp', 'release_date'),
- ('modified_timestamp', 'modified_date'),
- ):
- if info_dict.get(date_key) is None and info_dict.get(ts_key) is not None:
- # Working around out-of-range timestamp values (e.g. negative ones on Windows,
- # see http://bugs.python.org/issue1646728)
- with contextlib.suppress(ValueError, OverflowError, OSError):
- upload_date = dt.datetime.fromtimestamp(info_dict[ts_key], dt.timezone.utc)
- info_dict[date_key] = upload_date.strftime('%Y%m%d')
- if not info_dict.get('release_year'):
- info_dict['release_year'] = traverse_obj(info_dict, ('release_date', {lambda x: int(x[:4])}))
- live_keys = ('is_live', 'was_live')
- live_status = info_dict.get('live_status')
- if live_status is None:
- for key in live_keys:
- if info_dict.get(key) is False:
- continue
- if info_dict.get(key):
- live_status = key
- break
- if all(info_dict.get(key) is False for key in live_keys):
- live_status = 'not_live'
- if live_status:
- info_dict['live_status'] = live_status
- for key in live_keys:
- if info_dict.get(key) is None:
- info_dict[key] = (live_status == key)
- if live_status == 'post_live':
- info_dict['was_live'] = True
- # Auto generate title fields corresponding to the *_number fields when missing
- # in order to always have clean titles. This is very common for TV series.
- for field in ('chapter', 'season', 'episode'):
- if final and info_dict.get(f'{field}_number') is not None and not info_dict.get(field):
- info_dict[field] = '%s %d' % (field.capitalize(), info_dict[f'{field}_number'])
- for old_key, new_key in self._deprecated_multivalue_fields.items():
- if new_key in info_dict and old_key in info_dict:
- if '_version' not in info_dict: # HACK: Do not warn when using --load-info-json
- self.deprecation_warning(f'Do not return {old_key!r} when {new_key!r} is present')
- elif old_value := info_dict.get(old_key):
- info_dict[new_key] = old_value.split(', ')
- elif new_value := info_dict.get(new_key):
- info_dict[old_key] = ', '.join(v.replace(',', '\N{FULLWIDTH COMMA}') for v in new_value)
- def _raise_pending_errors(self, info):
- err = info.pop('__pending_error', None)
- if err:
- self.report_error(err, tb=False)
- def sort_formats(self, info_dict):
- formats = self._get_formats(info_dict)
- formats.sort(key=FormatSorter(
- self, info_dict.get('_format_sort_fields') or []).calculate_preference)
- def process_video_result(self, info_dict, download=True):
- assert info_dict.get('_type', 'video') == 'video'
- self._num_videos += 1
- if 'id' not in info_dict:
- raise ExtractorError('Missing "id" field in extractor result', ie=info_dict['extractor'])
- elif not info_dict.get('id'):
- raise ExtractorError('Extractor failed to obtain "id"', ie=info_dict['extractor'])
- def report_force_conversion(field, field_not, conversion):
- self.report_warning(
- f'"{field}" field is not {field_not} - forcing {conversion} conversion, '
- 'there is an error in extractor')
- def sanitize_string_field(info, string_field):
- field = info.get(string_field)
- if field is None or isinstance(field, str):
- return
- report_force_conversion(string_field, 'a string', 'string')
- info[string_field] = str(field)
- def sanitize_numeric_fields(info):
- for numeric_field in self._NUMERIC_FIELDS:
- field = info.get(numeric_field)
- if field is None or isinstance(field, (int, float)):
- continue
- report_force_conversion(numeric_field, 'numeric', 'int')
- info[numeric_field] = int_or_none(field)
- sanitize_string_field(info_dict, 'id')
- sanitize_numeric_fields(info_dict)
- if info_dict.get('section_end') and info_dict.get('section_start') is not None:
- info_dict['duration'] = round(info_dict['section_end'] - info_dict['section_start'], 3)
- if (info_dict.get('duration') or 0) <= 0 and info_dict.pop('duration', None):
- self.report_warning('"duration" field is negative, there is an error in extractor')
- chapters = info_dict.get('chapters') or []
- if chapters and chapters[0].get('start_time'):
- chapters.insert(0, {'start_time': 0})
- dummy_chapter = {'end_time': 0, 'start_time': info_dict.get('duration')}
- for idx, (prev, current, next_) in enumerate(zip(
- (dummy_chapter, *chapters), chapters, (*chapters[1:], dummy_chapter)), 1):
- if current.get('start_time') is None:
- current['start_time'] = prev.get('end_time')
- if not current.get('end_time'):
- current['end_time'] = next_.get('start_time')
- if not current.get('title'):
- current['title'] = f'<Untitled Chapter {idx}>'
- if 'playlist' not in info_dict:
- # It isn't part of a playlist
- info_dict['playlist'] = None
- info_dict['playlist_index'] = None
- self._sanitize_thumbnails(info_dict)
- thumbnail = info_dict.get('thumbnail')
- thumbnails = info_dict.get('thumbnails')
- if thumbnail:
- info_dict['thumbnail'] = sanitize_url(thumbnail)
- elif thumbnails:
- info_dict['thumbnail'] = thumbnails[-1]['url']
- if info_dict.get('display_id') is None and 'id' in info_dict:
- info_dict['display_id'] = info_dict['id']
- self._fill_common_fields(info_dict)
- for cc_kind in ('subtitles', 'automatic_captions'):
- cc = info_dict.get(cc_kind)
- if cc:
- for _, subtitle in cc.items():
- for subtitle_format in subtitle:
- if subtitle_format.get('url'):
- subtitle_format['url'] = sanitize_url(subtitle_format['url'])
- if subtitle_format.get('ext') is None:
- subtitle_format['ext'] = determine_ext(subtitle_format['url']).lower()
- automatic_captions = info_dict.get('automatic_captions')
- subtitles = info_dict.get('subtitles')
- info_dict['requested_subtitles'] = self.process_subtitles(
- info_dict['id'], subtitles, automatic_captions)
- formats = self._get_formats(info_dict)
- # Backward compatibility with InfoExtractor._sort_formats
- field_preference = (formats or [{}])[0].pop('__sort_fields', None)
- if field_preference:
- info_dict['_format_sort_fields'] = field_preference
- info_dict['_has_drm'] = any( # or None ensures --clean-infojson removes it
- f.get('has_drm') and f['has_drm'] != 'maybe' for f in formats) or None
- if not self.params.get('allow_unplayable_formats'):
- formats = [f for f in formats if not f.get('has_drm') or f['has_drm'] == 'maybe']
- if formats and all(f.get('acodec') == f.get('vcodec') == 'none' for f in formats):
- self.report_warning(
- f'{"This video is DRM protected and " if info_dict["_has_drm"] else ""}'
- 'only images are available for download. Use --list-formats to see them'.capitalize())
- get_from_start = not info_dict.get('is_live') or bool(self.params.get('live_from_start'))
- if not get_from_start:
- info_dict['title'] += ' ' + dt.datetime.now().strftime('%Y-%m-%d %H:%M')
- if info_dict.get('is_live') and formats:
- formats = [f for f in formats if bool(f.get('is_from_start')) == get_from_start]
- if get_from_start and not formats:
- self.raise_no_formats(info_dict, msg=(
- '--live-from-start is passed, but there are no formats that can be downloaded from the start. '
- 'If you want to download from the current time, use --no-live-from-start'))
- def is_wellformed(f):
- url = f.get('url')
- if not url:
- self.report_warning(
- '"url" field is missing or empty - skipping format, '
- 'there is an error in extractor')
- return False
- if isinstance(url, bytes):
- sanitize_string_field(f, 'url')
- return True
- # Filter out malformed formats for better extraction robustness
- formats = list(filter(is_wellformed, formats or []))
- if not formats:
- self.raise_no_formats(info_dict)
- for fmt in formats:
- sanitize_string_field(fmt, 'format_id')
- sanitize_numeric_fields(fmt)
- fmt['url'] = sanitize_url(fmt['url'])
- if fmt.get('ext') is None:
- fmt['ext'] = determine_ext(fmt['url']).lower()
- if fmt['ext'] in ('aac', 'opus', 'mp3', 'flac', 'vorbis'):
- if fmt.get('acodec') is None:
- fmt['acodec'] = fmt['ext']
- if fmt.get('protocol') is None:
- fmt['protocol'] = determine_protocol(fmt)
- if fmt.get('resolution') is None:
- fmt['resolution'] = self.format_resolution(fmt, default=None)
- if fmt.get('dynamic_range') is None and fmt.get('vcodec') != 'none':
- fmt['dynamic_range'] = 'SDR'
- if fmt.get('aspect_ratio') is None:
- fmt['aspect_ratio'] = try_call(lambda: round(fmt['width'] / fmt['height'], 2))
- # For fragmented formats, "tbr" is often max bitrate and not average
- if (('manifest-filesize-approx' in self.params['compat_opts'] or not fmt.get('manifest_url'))
- and not fmt.get('filesize') and not fmt.get('filesize_approx')):
- fmt['filesize_approx'] = filesize_from_tbr(fmt.get('tbr'), info_dict.get('duration'))
- fmt['http_headers'] = self._calc_headers(collections.ChainMap(fmt, info_dict), load_cookies=True)
- # Safeguard against old/insecure infojson when using --load-info-json
- if info_dict.get('http_headers'):
- info_dict['http_headers'] = HTTPHeaderDict(info_dict['http_headers'])
- info_dict['http_headers'].pop('Cookie', None)
- # This is copied to http_headers by the above _calc_headers and can now be removed
- if '__x_forwarded_for_ip' in info_dict:
- del info_dict['__x_forwarded_for_ip']
- self.sort_formats({
- 'formats': formats,
- '_format_sort_fields': info_dict.get('_format_sort_fields'),
- })
- # Sanitize and group by format_id
- formats_dict = {}
- for i, fmt in enumerate(formats):
- if not fmt.get('format_id'):
- fmt['format_id'] = str(i)
- else:
- # Sanitize format_id from characters used in format selector expression
- fmt['format_id'] = re.sub(r'[\s,/+\[\]()]', '_', fmt['format_id'])
- formats_dict.setdefault(fmt['format_id'], []).append(fmt)
- # Make sure all formats have unique format_id
- common_exts = set(itertools.chain(*self._format_selection_exts.values()))
- for format_id, ambiguous_formats in formats_dict.items():
- ambigious_id = len(ambiguous_formats) > 1
- for i, fmt in enumerate(ambiguous_formats):
- if ambigious_id:
- fmt['format_id'] = f'{format_id}-{i}'
- # Ensure there is no conflict between id and ext in format selection
- # See https://github.com/yt-dlp/yt-dlp/issues/1282
- if fmt['format_id'] != fmt['ext'] and fmt['format_id'] in common_exts:
- fmt['format_id'] = 'f{}'.format(fmt['format_id'])
- if fmt.get('format') is None:
- fmt['format'] = '{id} - {res}{note}'.format(
- id=fmt['format_id'],
- res=self.format_resolution(fmt),
- note=format_field(fmt, 'format_note', ' (%s)'),
- )
- if self.params.get('check_formats') is True:
- formats = LazyList(self._check_formats(formats[::-1]), reverse=True)
- if not formats or formats[0] is not info_dict:
- # only set the 'formats' fields if the original info_dict list them
- # otherwise we end up with a circular reference, the first (and unique)
- # element in the 'formats' field in info_dict is info_dict itself,
- # which can't be exported to json
- info_dict['formats'] = formats
- info_dict, _ = self.pre_process(info_dict)
- if self._match_entry(info_dict, incomplete=self._format_fields) is not None:
- return info_dict
- self.post_extract(info_dict)
- info_dict, _ = self.pre_process(info_dict, 'after_filter')
- # The pre-processors may have modified the formats
- formats = self._get_formats(info_dict)
- list_only = self.params.get('simulate') == 'list_only'
- interactive_format_selection = not list_only and self.format_selector == '-'
- if self.params.get('list_thumbnails'):
- self.list_thumbnails(info_dict)
- if self.params.get('listsubtitles'):
- if 'automatic_captions' in info_dict:
- self.list_subtitles(
- info_dict['id'], automatic_captions, 'automatic captions')
- self.list_subtitles(info_dict['id'], subtitles, 'subtitles')
- if self.params.get('listformats') or interactive_format_selection:
- self.list_formats(info_dict)
- if list_only:
- # Without this printing, -F --print-json will not work
- self.__forced_printings(info_dict)
- return info_dict
- format_selector = self.format_selector
- while True:
- if interactive_format_selection:
- req_format = input(self._format_screen('\nEnter format selector ', self.Styles.EMPHASIS)
- + '(Press ENTER for default, or Ctrl+C to quit)'
- + self._format_screen(': ', self.Styles.EMPHASIS))
- try:
- format_selector = self.build_format_selector(req_format) if req_format else None
- except SyntaxError as err:
- self.report_error(err, tb=False, is_error=False)
- continue
- if format_selector is None:
- req_format = self._default_format_spec(info_dict)
- self.write_debug(f'Default format spec: {req_format}')
- format_selector = self.build_format_selector(req_format)
- formats_to_download = self._select_formats(formats, format_selector)
- if interactive_format_selection and not formats_to_download:
- self.report_error('Requested format is not available', tb=False, is_error=False)
- continue
- break
- if not formats_to_download:
- if not self.params.get('ignore_no_formats_error'):
- raise ExtractorError(
- 'Requested format is not available. Use --list-formats for a list of available formats',
- expected=True, video_id=info_dict['id'], ie=info_dict['extractor'])
- self.report_warning('Requested format is not available')
- # Process what we can, even without any available formats.
- formats_to_download = [{}]
- requested_ranges = tuple(self.params.get('download_ranges', lambda *_: [{}])(info_dict, self))
- best_format, downloaded_formats = formats_to_download[-1], []
- if download:
- if best_format and requested_ranges:
- def to_screen(*msg):
- self.to_screen(f'[info] {info_dict["id"]}: {" ".join(", ".join(variadic(m)) for m in msg)}')
- to_screen(f'Downloading {len(formats_to_download)} format(s):',
- (f['format_id'] for f in formats_to_download))
- if requested_ranges != ({}, ):
- to_screen(f'Downloading {len(requested_ranges)} time ranges:',
- (f'{c["start_time"]:.1f}-{c["end_time"]:.1f}' for c in requested_ranges))
- max_downloads_reached = False
- for fmt, chapter in itertools.product(formats_to_download, requested_ranges):
- new_info = self._copy_infodict(info_dict)
- new_info.update(fmt)
- offset, duration = info_dict.get('section_start') or 0, info_dict.get('duration') or float('inf')
- end_time = offset + min(chapter.get('end_time', duration), duration)
- # duration may not be accurate. So allow deviations <1sec
- if end_time == float('inf') or end_time > offset + duration + 1:
- end_time = None
- if chapter or offset:
- new_info.update({
- 'section_start': offset + chapter.get('start_time', 0),
- 'section_end': end_time,
- 'section_title': chapter.get('title'),
- 'section_number': chapter.get('index'),
- })
- downloaded_formats.append(new_info)
- try:
- self.process_info(new_info)
- except MaxDownloadsReached:
- max_downloads_reached = True
- self._raise_pending_errors(new_info)
- # Remove copied info
- for key, val in tuple(new_info.items()):
- if info_dict.get(key) == val:
- new_info.pop(key)
- if max_downloads_reached:
- break
- write_archive = {f.get('__write_download_archive', False) for f in downloaded_formats}
- assert write_archive.issubset({True, False, 'ignore'})
- if True in write_archive and False not in write_archive:
- self.record_download_archive(info_dict)
- info_dict['requested_downloads'] = downloaded_formats
- info_dict = self.run_all_pps('after_video', info_dict)
- if max_downloads_reached:
- raise MaxDownloadsReached
- # We update the info dict with the selected best quality format (backwards compatibility)
- info_dict.update(best_format)
- return info_dict
- def process_subtitles(self, video_id, normal_subtitles, automatic_captions):
- """Select the requested subtitles and their format"""
- available_subs, normal_sub_langs = {}, []
- if normal_subtitles and self.params.get('writesubtitles'):
- available_subs.update(normal_subtitles)
- normal_sub_langs = tuple(normal_subtitles.keys())
- if automatic_captions and self.params.get('writeautomaticsub'):
- for lang, cap_info in automatic_captions.items():
- if lang not in available_subs:
- available_subs[lang] = cap_info
- if not available_subs or (
- not self.params.get('writesubtitles')
- and not self.params.get('writeautomaticsub')):
- return None
- all_sub_langs = tuple(available_subs.keys())
- if self.params.get('allsubtitles', False):
- requested_langs = all_sub_langs
- elif self.params.get('subtitleslangs', False):
- try:
- requested_langs = orderedSet_from_options(
- self.params.get('subtitleslangs'), {'all': all_sub_langs}, use_regex=True)
- except re.error as e:
- raise ValueError(f'Wrong regex for subtitlelangs: {e.pattern}')
- else:
- requested_langs = LazyList(itertools.chain(
- ['en'] if 'en' in normal_sub_langs else [],
- filter(lambda f: f.startswith('en'), normal_sub_langs),
- ['en'] if 'en' in all_sub_langs else [],
- filter(lambda f: f.startswith('en'), all_sub_langs),
- normal_sub_langs, all_sub_langs,
- ))[:1]
- if requested_langs:
- self.to_screen(f'[info] {video_id}: Downloading subtitles: {", ".join(requested_langs)}')
- formats_query = self.params.get('subtitlesformat', 'best')
- formats_preference = formats_query.split('/') if formats_query else []
- subs = {}
- for lang in requested_langs:
- formats = available_subs.get(lang)
- if formats is None:
- self.report_warning(f'{lang} subtitles not available for {video_id}')
- continue
- for ext in formats_preference:
- if ext == 'best':
- f = formats[-1]
- break
- matches = list(filter(lambda f: f['ext'] == ext, formats))
- if matches:
- f = matches[-1]
- break
- else:
- f = formats[-1]
- self.report_warning(
- 'No subtitle format found matching "{}" for language {}, '
- 'using {}. Use --list-subs for a list of available subtitles'.format(formats_query, lang, f['ext']))
- subs[lang] = f
- return subs
- def _forceprint(self, key, info_dict):
- if info_dict is None:
- return
- info_copy = info_dict.copy()
- info_copy.setdefault('filename', self.prepare_filename(info_dict))
- if info_dict.get('requested_formats') is not None:
- # For RTMP URLs, also include the playpath
- info_copy['urls'] = '\n'.join(f['url'] + f.get('play_path', '') for f in info_dict['requested_formats'])
- elif info_dict.get('url'):
- info_copy['urls'] = info_dict['url'] + info_dict.get('play_path', '')
- info_copy['formats_table'] = self.render_formats_table(info_dict)
- info_copy['thumbnails_table'] = self.render_thumbnails_table(info_dict)
- info_copy['subtitles_table'] = self.render_subtitles_table(info_dict.get('id'), info_dict.get('subtitles'))
- info_copy['automatic_captions_table'] = self.render_subtitles_table(info_dict.get('id'), info_dict.get('automatic_captions'))
- def format_tmpl(tmpl):
- mobj = re.fullmatch(r'([\w.:,]|-\d|(?P<dict>{([\w.:,]|-\d)+}))+=?', tmpl)
- if not mobj:
- return tmpl
- fmt = '%({})s'
- if tmpl.startswith('{'):
- tmpl, fmt = f'.{tmpl}', '%({})j'
- if tmpl.endswith('='):
- tmpl, fmt = tmpl[:-1], '{0} = %({0})#j'
- return '\n'.join(map(fmt.format, [tmpl] if mobj.group('dict') else tmpl.split(',')))
- for tmpl in self.params['forceprint'].get(key, []):
- self.to_stdout(self.evaluate_outtmpl(format_tmpl(tmpl), info_copy))
- for tmpl, file_tmpl in self.params['print_to_file'].get(key, []):
- filename = self.prepare_filename(info_dict, outtmpl=file_tmpl)
- tmpl = format_tmpl(tmpl)
- self.to_screen(f'[info] Writing {tmpl!r} to: {filename}')
- if self._ensure_dir_exists(filename):
- with open(filename, 'a', encoding='utf-8', newline='') as f:
- f.write(self.evaluate_outtmpl(tmpl, info_copy) + os.linesep)
- return info_copy
- def __forced_printings(self, info_dict, filename=None, incomplete=True):
- if (self.params.get('forcejson')
- or self.params['forceprint'].get('video')
- or self.params['print_to_file'].get('video')):
- self.post_extract(info_dict)
- if filename:
- info_dict['filename'] = filename
- info_copy = self._forceprint('video', info_dict)
- def print_field(field, actual_field=None, optional=False):
- if actual_field is None:
- actual_field = field
- if self.params.get(f'force{field}') and (
- info_copy.get(field) is not None or (not optional and not incomplete)):
- self.to_stdout(info_copy[actual_field])
- print_field('title')
- print_field('id')
- print_field('url', 'urls')
- print_field('thumbnail', optional=True)
- print_field('description', optional=True)
- print_field('filename')
- if self.params.get('forceduration') and info_copy.get('duration') is not None:
- self.to_stdout(formatSeconds(info_copy['duration']))
- print_field('format')
- if self.params.get('forcejson'):
- self.to_stdout(json.dumps(self.sanitize_info(info_dict)))
- def dl(self, name, info, subtitle=False, test=False):
- if not info.get('url'):
- self.raise_no_formats(info, True)
- if test:
- verbose = self.params.get('verbose')
- quiet = self.params.get('quiet') or not verbose
- params = {
- 'test': True,
- 'quiet': quiet,
- 'verbose': verbose,
- 'noprogress': quiet,
- 'nopart': True,
- 'skip_unavailable_fragments': False,
- 'keep_fragments': False,
- 'overwrites': True,
- '_no_ytdl_file': True,
- }
- else:
- params = self.params
- fd = get_suitable_downloader(info, params, to_stdout=(name == '-'))(self, params)
- if not test:
- for ph in self._progress_hooks:
- fd.add_progress_hook(ph)
- urls = '", "'.join(
- (f['url'].split(',')[0] + ',<data>' if f['url'].startswith('data:') else f['url'])
- for f in info.get('requested_formats', []) or [info])
- self.write_debug(f'Invoking {fd.FD_NAME} downloader on "{urls}"')
- # Note: Ideally info should be a deep-copied so that hooks cannot modify it.
- # But it may contain objects that are not deep-copyable
- new_info = self._copy_infodict(info)
- if new_info.get('http_headers') is None:
- new_info['http_headers'] = self._calc_headers(new_info)
- return fd.download(name, new_info, subtitle)
- def existing_file(self, filepaths, *, default_overwrite=True):
- existing_files = list(filter(os.path.exists, orderedSet(filepaths)))
- if existing_files and not self.params.get('overwrites', default_overwrite):
- return existing_files[0]
- for file in existing_files:
- self.report_file_delete(file)
- os.remove(file)
- return None
- @_catch_unsafe_extension_error
- def process_info(self, info_dict):
- """Process a single resolved IE result. (Modifies it in-place)"""
- assert info_dict.get('_type', 'video') == 'video'
- original_infodict = info_dict
- if 'format' not in info_dict and 'ext' in info_dict:
- info_dict['format'] = info_dict['ext']
- if self._match_entry(info_dict) is not None:
- info_dict['__write_download_archive'] = 'ignore'
- return
- # Does nothing under normal operation - for backward compatibility of process_info
- self.post_extract(info_dict)
- def replace_info_dict(new_info):
- nonlocal info_dict
- if new_info == info_dict:
- return
- info_dict.clear()
- info_dict.update(new_info)
- new_info, _ = self.pre_process(info_dict, 'video')
- replace_info_dict(new_info)
- self._num_downloads += 1
- # info_dict['_filename'] needs to be set for backward compatibility
- info_dict['_filename'] = full_filename = self.prepare_filename(info_dict, warn=True)
- temp_filename = self.prepare_filename(info_dict, 'temp')
- files_to_move = {}
- # Forced printings
- self.__forced_printings(info_dict, full_filename, incomplete=('format' not in info_dict))
- def check_max_downloads():
- if self._num_downloads >= float(self.params.get('max_downloads') or 'inf'):
- raise MaxDownloadsReached
- if self.params.get('simulate'):
- info_dict['__write_download_archive'] = self.params.get('force_write_download_archive')
- check_max_downloads()
- return
- if full_filename is None:
- return
- if not self._ensure_dir_exists(encodeFilename(full_filename)):
- return
- if not self._ensure_dir_exists(encodeFilename(temp_filename)):
- return
- if self._write_description('video', info_dict,
- self.prepare_filename(info_dict, 'description')) is None:
- return
- sub_files = self._write_subtitles(info_dict, temp_filename)
- if sub_files is None:
- return
- files_to_move.update(dict(sub_files))
- thumb_files = self._write_thumbnails(
- 'video', info_dict, temp_filename, self.prepare_filename(info_dict, 'thumbnail'))
- if thumb_files is None:
- return
- files_to_move.update(dict(thumb_files))
- infofn = self.prepare_filename(info_dict, 'infojson')
- _infojson_written = self._write_info_json('video', info_dict, infofn)
- if _infojson_written:
- info_dict['infojson_filename'] = infofn
- # For backward compatibility, even though it was a private field
- info_dict['__infojson_filename'] = infofn
- elif _infojson_written is None:
- return
- # Note: Annotations are deprecated
- annofn = None
- if self.params.get('writeannotations', False):
- annofn = self.prepare_filename(info_dict, 'annotation')
- if annofn:
- if not self._ensure_dir_exists(encodeFilename(annofn)):
- return
- if not self.params.get('overwrites', True) and os.path.exists(encodeFilename(annofn)):
- self.to_screen('[info] Video annotations are already present')
- elif not info_dict.get('annotations'):
- self.report_warning('There are no annotations to write.')
- else:
- try:
- self.to_screen('[info] Writing video annotations to: ' + annofn)
- with open(encodeFilename(annofn), 'w', encoding='utf-8') as annofile:
- annofile.write(info_dict['annotations'])
- except (KeyError, TypeError):
- self.report_warning('There are no annotations to write.')
- except OSError:
- self.report_error('Cannot write annotations file: ' + annofn)
- return
- # Write internet shortcut files
- def _write_link_file(link_type):
- url = try_get(info_dict['webpage_url'], iri_to_uri)
- if not url:
- self.report_warning(
- f'Cannot write internet shortcut file because the actual URL of "{info_dict["webpage_url"]}" is unknown')
- return True
- linkfn = replace_extension(self.prepare_filename(info_dict, 'link'), link_type, info_dict.get('ext'))
- if not self._ensure_dir_exists(encodeFilename(linkfn)):
- return False
- if self.params.get('overwrites', True) and os.path.exists(encodeFilename(linkfn)):
- self.to_screen(f'[info] Internet shortcut (.{link_type}) is already present')
- return True
- try:
- self.to_screen(f'[info] Writing internet shortcut (.{link_type}) to: {linkfn}')
- with open(encodeFilename(to_high_limit_path(linkfn)), 'w', encoding='utf-8',
- newline='\r\n' if link_type == 'url' else '\n') as linkfile:
- template_vars = {'url': url}
- if link_type == 'desktop':
- template_vars['filename'] = linkfn[:-(len(link_type) + 1)]
- linkfile.write(LINK_TEMPLATES[link_type] % template_vars)
- except OSError:
- self.report_error(f'Cannot write internet shortcut {linkfn}')
- return False
- return True
- write_links = {
- 'url': self.params.get('writeurllink'),
- 'webloc': self.params.get('writewebloclink'),
- 'desktop': self.params.get('writedesktoplink'),
- }
- if self.params.get('writelink'):
- link_type = ('webloc' if sys.platform == 'darwin'
- else 'desktop' if sys.platform.startswith('linux')
- else 'url')
- write_links[link_type] = True
- if any(should_write and not _write_link_file(link_type)
- for link_type, should_write in write_links.items()):
- return
- new_info, files_to_move = self.pre_process(info_dict, 'before_dl', files_to_move)
- replace_info_dict(new_info)
- if self.params.get('skip_download'):
- info_dict['filepath'] = temp_filename
- info_dict['__finaldir'] = os.path.dirname(os.path.abspath(encodeFilename(full_filename)))
- info_dict['__files_to_move'] = files_to_move
- replace_info_dict(self.run_pp(MoveFilesAfterDownloadPP(self, False), info_dict))
- info_dict['__write_download_archive'] = self.params.get('force_write_download_archive')
- else:
- # Download
- info_dict.setdefault('__postprocessors', [])
- try:
- def existing_video_file(*filepaths):
- ext = info_dict.get('ext')
- converted = lambda file: replace_extension(file, self.params.get('final_ext') or ext, ext)
- file = self.existing_file(itertools.chain(*zip(map(converted, filepaths), filepaths)),
- default_overwrite=False)
- if file:
- info_dict['ext'] = os.path.splitext(file)[1][1:]
- return file
- fd, success = None, True
- if info_dict.get('protocol') or info_dict.get('url'):
- fd = get_suitable_downloader(info_dict, self.params, to_stdout=temp_filename == '-')
- if fd != FFmpegFD and 'no-direct-merge' not in self.params['compat_opts'] and (
- info_dict.get('section_start') or info_dict.get('section_end')):
- msg = ('This format cannot be partially downloaded' if FFmpegFD.available()
- else 'You have requested downloading the video partially, but ffmpeg is not installed')
- self.report_error(f'{msg}. Aborting')
- return
- if info_dict.get('requested_formats') is not None:
- old_ext = info_dict['ext']
- if self.params.get('merge_output_format') is None:
- if (info_dict['ext'] == 'webm'
- and info_dict.get('thumbnails')
- # check with type instead of pp_key, __name__, or isinstance
- # since we dont want any custom PPs to trigger this
- and any(type(pp) == EmbedThumbnailPP for pp in self._pps['post_process'])): # noqa: E721
- info_dict['ext'] = 'mkv'
- self.report_warning(
- 'webm doesn\'t support embedding a thumbnail, mkv will be used')
- new_ext = info_dict['ext']
- def correct_ext(filename, ext=new_ext):
- if filename == '-':
- return filename
- filename_real_ext = os.path.splitext(filename)[1][1:]
- filename_wo_ext = (
- os.path.splitext(filename)[0]
- if filename_real_ext in (old_ext, new_ext)
- else filename)
- return f'{filename_wo_ext}.{ext}'
- # Ensure filename always has a correct extension for successful merge
- full_filename = correct_ext(full_filename)
- temp_filename = correct_ext(temp_filename)
- dl_filename = existing_video_file(full_filename, temp_filename)
- info_dict['__real_download'] = False
- # NOTE: Copy so that original format dicts are not modified
- info_dict['requested_formats'] = list(map(dict, info_dict['requested_formats']))
- merger = FFmpegMergerPP(self)
- downloaded = []
- if dl_filename is not None:
- self.report_file_already_downloaded(dl_filename)
- elif fd:
- for f in info_dict['requested_formats'] if fd != FFmpegFD else []:
- f['filepath'] = fname = prepend_extension(
- correct_ext(temp_filename, info_dict['ext']),
- 'f{}'.format(f['format_id']), info_dict['ext'])
- downloaded.append(fname)
- info_dict['url'] = '\n'.join(f['url'] for f in info_dict['requested_formats'])
- success, real_download = self.dl(temp_filename, info_dict)
- info_dict['__real_download'] = real_download
- else:
- if self.params.get('allow_unplayable_formats'):
- self.report_warning(
- 'You have requested merging of multiple formats '
- 'while also allowing unplayable formats to be downloaded. '
- 'The formats won\'t be merged to prevent data corruption.')
- elif not merger.available:
- msg = 'You have requested merging of multiple formats but ffmpeg is not installed'
- if not self.params.get('ignoreerrors'):
- self.report_error(f'{msg}. Aborting due to --abort-on-error')
- return
- self.report_warning(f'{msg}. The formats won\'t be merged')
- if temp_filename == '-':
- reason = ('using a downloader other than ffmpeg' if FFmpegFD.can_merge_formats(info_dict, self.params)
- else 'but the formats are incompatible for simultaneous download' if merger.available
- else 'but ffmpeg is not installed')
- self.report_warning(
- f'You have requested downloading multiple formats to stdout {reason}. '
- 'The formats will be streamed one after the other')
- fname = temp_filename
- for f in info_dict['requested_formats']:
- new_info = dict(info_dict)
- del new_info['requested_formats']
- new_info.update(f)
- if temp_filename != '-':
- fname = prepend_extension(
- correct_ext(temp_filename, new_info['ext']),
- 'f{}'.format(f['format_id']), new_info['ext'])
- if not self._ensure_dir_exists(fname):
- return
- f['filepath'] = fname
- downloaded.append(fname)
- partial_success, real_download = self.dl(fname, new_info)
- info_dict['__real_download'] = info_dict['__real_download'] or real_download
- success = success and partial_success
- if downloaded and merger.available and not self.params.get('allow_unplayable_formats'):
- info_dict['__postprocessors'].append(merger)
- info_dict['__files_to_merge'] = downloaded
- # Even if there were no downloads, it is being merged only now
- info_dict['__real_download'] = True
- else:
- for file in downloaded:
- files_to_move[file] = None
- else:
- # Just a single file
- dl_filename = existing_video_file(full_filename, temp_filename)
- if dl_filename is None or dl_filename == temp_filename:
- # dl_filename == temp_filename could mean that the file was partially downloaded with --no-part.
- # So we should try to resume the download
- success, real_download = self.dl(temp_filename, info_dict)
- info_dict['__real_download'] = real_download
- else:
- self.report_file_already_downloaded(dl_filename)
- dl_filename = dl_filename or temp_filename
- info_dict['__finaldir'] = os.path.dirname(os.path.abspath(encodeFilename(full_filename)))
- except network_exceptions as err:
- self.report_error(f'unable to download video data: {err}')
- return
- except OSError as err:
- raise UnavailableVideoError(err)
- except ContentTooShortError as err:
- self.report_error(f'content too short (expected {err.expected} bytes and served {err.downloaded})')
- return
- self._raise_pending_errors(info_dict)
- if success and full_filename != '-':
- def fixup():
- do_fixup = True
- fixup_policy = self.params.get('fixup')
- vid = info_dict['id']
- if fixup_policy in ('ignore', 'never'):
- return
- elif fixup_policy == 'warn':
- do_fixup = 'warn'
- elif fixup_policy != 'force':
- assert fixup_policy in ('detect_or_warn', None)
- if not info_dict.get('__real_download'):
- do_fixup = False
- def ffmpeg_fixup(cndn, msg, cls):
- if not (do_fixup and cndn):
- return
- elif do_fixup == 'warn':
- self.report_warning(f'{vid}: {msg}')
- return
- pp = cls(self)
- if pp.available:
- info_dict['__postprocessors'].append(pp)
- else:
- self.report_warning(f'{vid}: {msg}. Install ffmpeg to fix this automatically')
- stretched_ratio = info_dict.get('stretched_ratio')
- ffmpeg_fixup(stretched_ratio not in (1, None),
- f'Non-uniform pixel ratio {stretched_ratio}',
- FFmpegFixupStretchedPP)
- downloader = get_suitable_downloader(info_dict, self.params) if 'protocol' in info_dict else None
- downloader = downloader.FD_NAME if downloader else None
- ext = info_dict.get('ext')
- postprocessed_by_ffmpeg = info_dict.get('requested_formats') or any((
- isinstance(pp, FFmpegVideoConvertorPP)
- and resolve_recode_mapping(ext, pp.mapping)[0] not in (ext, None)
- ) for pp in self._pps['post_process'])
- if not postprocessed_by_ffmpeg:
- ffmpeg_fixup(fd != FFmpegFD and ext == 'm4a'
- and info_dict.get('container') == 'm4a_dash',
- 'writing DASH m4a. Only some players support this container',
- FFmpegFixupM4aPP)
- ffmpeg_fixup(downloader == 'hlsnative' and not self.params.get('hls_use_mpegts')
- 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(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)
- ffmpeg_fixup(downloader == 'web_socket_fragment', 'Malformed duration detected', FFmpegFixupDurationPP)
- fixup()
- try:
- replace_info_dict(self.post_process(dl_filename, info_dict, files_to_move))
- except PostProcessingError as err:
- self.report_error(f'Postprocessing: {err}')
- return
- try:
- for ph in self._post_hooks:
- ph(info_dict['filepath'])
- except Exception as err:
- self.report_error(f'post hooks: {err}')
- return
- info_dict['__write_download_archive'] = True
- assert info_dict is original_infodict # Make sure the info_dict was modified in-place
- if self.params.get('force_write_download_archive'):
- info_dict['__write_download_archive'] = True
- check_max_downloads()
- def __download_wrapper(self, func):
- @functools.wraps(func)
- def wrapper(*args, **kwargs):
- try:
- res = func(*args, **kwargs)
- except CookieLoadError:
- raise
- except UnavailableVideoError as e:
- self.report_error(e)
- except DownloadCancelled as e:
- self.to_screen(f'[info] {e}')
- if not self.params.get('break_per_url'):
- raise
- self._num_downloads = 0
- else:
- if self.params.get('dump_single_json', False):
- self.post_extract(res)
- self.to_stdout(json.dumps(self.sanitize_info(res)))
- return wrapper
- def download(self, url_list):
- """Download a given list of URLs."""
- url_list = variadic(url_list) # Passing a single URL is a common mistake
- outtmpl = self.params['outtmpl']['default']
- if (len(url_list) > 1
- and outtmpl != '-'
- and '%' not in outtmpl
- and self.params.get('max_downloads') != 1):
- raise SameFileError(outtmpl)
- for url in url_list:
- self.__download_wrapper(self.extract_info)(
- url, force_generic_extractor=self.params.get('force_generic_extractor', False))
- return self._download_retcode
- def download_with_info_file(self, info_filename):
- with contextlib.closing(fileinput.FileInput(
- [info_filename], mode='r',
- openhook=fileinput.hook_encoded('utf-8'))) as f:
- # FileInput doesn't have a read method, we can't call json.load
- infos = [self.sanitize_info(info, self.params.get('clean_infojson', True))
- for info in variadic(json.loads('\n'.join(f)))]
- for info in infos:
- try:
- self.__download_wrapper(self.process_ie_result)(info, download=True)
- except (DownloadError, EntryNotInPlaylist, ReExtractInfo) as e:
- if not isinstance(e, EntryNotInPlaylist):
- self.to_stderr('\r')
- webpage_url = info.get('webpage_url')
- if webpage_url is None:
- raise
- self.report_warning(f'The info failed to download: {e}; trying with URL {webpage_url}')
- self.download([webpage_url])
- except ExtractorError as e:
- self.report_error(e)
- return self._download_retcode
- @staticmethod
- def sanitize_info(info_dict, remove_private_keys=False):
- """ Sanitize the infodict for converting to json """
- if info_dict is None:
- return info_dict
- info_dict.setdefault('epoch', int(time.time()))
- info_dict.setdefault('_type', 'video')
- info_dict.setdefault('_version', {
- 'version': __version__,
- 'current_git_head': current_git_head(),
- 'release_git_head': RELEASE_GIT_HEAD,
- 'repository': ORIGIN,
- })
- if remove_private_keys:
- reject = lambda k, v: v is None or k.startswith('__') or k in {
- 'requested_downloads', 'requested_formats', 'requested_subtitles', 'requested_entries',
- 'entries', 'filepath', '_filename', 'filename', 'infojson_filename', 'original_url',
- 'playlist_autonumber',
- }
- else:
- reject = lambda k, v: False
- def filter_fn(obj):
- if isinstance(obj, dict):
- return {k: filter_fn(v) for k, v in obj.items() if not reject(k, v)}
- elif isinstance(obj, (list, tuple, set, LazyList)):
- return list(map(filter_fn, obj))
- elif obj is None or isinstance(obj, (str, int, float, bool)):
- return obj
- else:
- return repr(obj)
- return filter_fn(info_dict)
- @staticmethod
- def filter_requested_info(info_dict, actually_filter=True):
- """ Alias of sanitize_info for backward compatibility """
- return YoutubeDL.sanitize_info(info_dict, actually_filter)
- def _delete_downloaded_files(self, *files_to_delete, info={}, msg=None):
- for filename in set(filter(None, files_to_delete)):
- if msg:
- self.to_screen(msg % filename)
- try:
- os.remove(filename)
- except OSError:
- self.report_warning(f'Unable to delete file {filename}')
- if filename in info.get('__files_to_move', []): # NB: Delete even if None
- del info['__files_to_move'][filename]
- @staticmethod
- def post_extract(info_dict):
- def actual_post_extract(info_dict):
- if info_dict.get('_type') in ('playlist', 'multi_video'):
- for video_dict in info_dict.get('entries', {}):
- actual_post_extract(video_dict or {})
- return
- post_extractor = info_dict.pop('__post_extractor', None) or dict
- info_dict.update(post_extractor())
- actual_post_extract(info_dict or {})
- def run_pp(self, pp, infodict):
- files_to_delete = []
- if '__files_to_move' not in infodict:
- infodict['__files_to_move'] = {}
- try:
- files_to_delete, infodict = pp.run(infodict)
- except PostProcessingError as e:
- # Must be True and not 'only_download'
- if self.params.get('ignoreerrors') is True:
- self.report_error(e)
- return infodict
- raise
- if not files_to_delete:
- return infodict
- if self.params.get('keepvideo', False):
- for f in files_to_delete:
- infodict['__files_to_move'].setdefault(f, '')
- else:
- self._delete_downloaded_files(
- *files_to_delete, info=infodict, msg='Deleting original file %s (pass -k to keep)')
- return infodict
- def run_all_pps(self, key, info, *, additional_pps=None):
- if key != 'video':
- self._forceprint(key, info)
- for pp in (additional_pps or []) + self._pps[key]:
- info = self.run_pp(pp, info)
- return info
- def pre_process(self, ie_info, key='pre_process', files_to_move=None):
- info = dict(ie_info)
- info['__files_to_move'] = files_to_move or {}
- try:
- info = self.run_all_pps(key, info)
- except PostProcessingError as err:
- msg = f'Preprocessing: {err}'
- info.setdefault('__pending_error', msg)
- self.report_error(msg, is_error=False)
- return info, info.pop('__files_to_move', None)
- def post_process(self, filename, info, files_to_move=None):
- """Run all the postprocessors on the given file."""
- info['filepath'] = filename
- info['__files_to_move'] = files_to_move or {}
- info = self.run_all_pps('post_process', info, additional_pps=info.get('__postprocessors'))
- info = self.run_pp(MoveFilesAfterDownloadPP(self), info)
- del info['__files_to_move']
- return self.run_all_pps('after_move', info)
- def _make_archive_id(self, info_dict):
- video_id = info_dict.get('id')
- if not video_id:
- return
- # Future-proof against any change in case
- # and backwards compatibility with prior versions
- extractor = info_dict.get('extractor_key') or info_dict.get('ie_key') # key in a playlist
- if extractor is None:
- url = str_or_none(info_dict.get('url'))
- if not url:
- return
- # Try to find matching extractor for the URL and take its ie_key
- for ie_key, ie in self._ies.items():
- if ie.suitable(url):
- extractor = ie_key
- break
- else:
- return
- return make_archive_id(extractor, video_id)
- def in_download_archive(self, info_dict):
- if not self.archive:
- return False
- vid_ids = [self._make_archive_id(info_dict)]
- vid_ids.extend(info_dict.get('_old_archive_ids') or [])
- return any(id_ in self.archive for id_ in vid_ids)
- def record_download_archive(self, info_dict):
- fn = self.params.get('download_archive')
- if fn is None:
- return
- vid_id = self._make_archive_id(info_dict)
- assert vid_id
- self.write_debug(f'Adding to archive: {vid_id}')
- if is_path_like(fn):
- with locked_file(fn, 'a', encoding='utf-8') as archive_file:
- archive_file.write(vid_id + '\n')
- self.archive.add(vid_id)
- @staticmethod
- def format_resolution(format, default='unknown'):
- if format.get('vcodec') == 'none' and format.get('acodec') != 'none':
- return 'audio only'
- if format.get('resolution') is not None:
- return format['resolution']
- if format.get('width') and format.get('height'):
- return '%dx%d' % (format['width'], format['height'])
- elif format.get('height'):
- return '{}p'.format(format['height'])
- elif format.get('width'):
- return '%dx?' % format['width']
- return default
- def _list_format_headers(self, *headers):
- if self.params.get('listformats_table', True) is not False:
- return [self._format_out(header, self.Styles.HEADERS) for header in headers]
- return headers
- def _format_note(self, fdict):
- res = ''
- if fdict.get('ext') in ['f4f', 'f4m']:
- res += '(unsupported)'
- if fdict.get('language'):
- if res:
- res += ' '
- res += '[{}]'.format(fdict['language'])
- if fdict.get('format_note') is not None:
- if res:
- res += ' '
- res += fdict['format_note']
- if fdict.get('tbr') is not None:
- if res:
- res += ', '
- res += '%4dk' % fdict['tbr']
- if fdict.get('container') is not None:
- if res:
- res += ', '
- res += '{} container'.format(fdict['container'])
- if (fdict.get('vcodec') is not None
- and fdict.get('vcodec') != 'none'):
- if res:
- res += ', '
- res += fdict['vcodec']
- if fdict.get('vbr') is not None:
- res += '@'
- elif fdict.get('vbr') is not None and fdict.get('abr') is not None:
- res += 'video@'
- if fdict.get('vbr') is not None:
- res += '%4dk' % fdict['vbr']
- if fdict.get('fps') is not None:
- if res:
- res += ', '
- res += '{}fps'.format(fdict['fps'])
- if fdict.get('acodec') is not None:
- if res:
- res += ', '
- if fdict['acodec'] == 'none':
- res += 'video only'
- else:
- res += '%-5s' % fdict['acodec']
- elif fdict.get('abr') is not None:
- if res:
- res += ', '
- res += 'audio'
- if fdict.get('abr') is not None:
- res += '@%3dk' % fdict['abr']
- if fdict.get('asr') is not None:
- res += ' (%5dHz)' % fdict['asr']
- if fdict.get('filesize') is not None:
- if res:
- res += ', '
- res += format_bytes(fdict['filesize'])
- elif fdict.get('filesize_approx') is not None:
- if res:
- res += ', '
- res += '~' + format_bytes(fdict['filesize_approx'])
- return res
- def _get_formats(self, info_dict):
- if info_dict.get('formats') is None:
- if info_dict.get('url') and info_dict.get('_type', 'video') == 'video':
- return [info_dict]
- return []
- return info_dict['formats']
- def render_formats_table(self, info_dict):
- formats = self._get_formats(info_dict)
- if not formats:
- return
- if not self.params.get('listformats_table', True) is not False:
- table = [
- [
- format_field(f, 'format_id'),
- format_field(f, 'ext'),
- self.format_resolution(f),
- self._format_note(f),
- ] for f in formats if (f.get('preference') or 0) >= -1000]
- return render_table(['format code', 'extension', 'resolution', 'note'], table, extra_gap=1)
- def simplified_codec(f, field):
- assert field in ('acodec', 'vcodec')
- codec = f.get(field)
- if not codec:
- return 'unknown'
- elif codec != 'none':
- return '.'.join(codec.split('.')[:4])
- if field == 'vcodec' and f.get('acodec') == 'none':
- return 'images'
- elif field == 'acodec' and f.get('vcodec') == 'none':
- return ''
- return self._format_out('audio only' if field == 'vcodec' else 'video only',
- self.Styles.SUPPRESS)
- delim = self._format_out('\u2502', self.Styles.DELIM, '|', test_encoding=True)
- table = [
- [
- self._format_out(format_field(f, 'format_id'), self.Styles.ID),
- format_field(f, 'ext'),
- format_field(f, func=self.format_resolution, ignore=('audio only', 'images')),
- format_field(f, 'fps', '\t%d', func=round),
- format_field(f, 'dynamic_range', '%s', ignore=(None, 'SDR')).replace('HDR', ''),
- format_field(f, 'audio_channels', '\t%s'),
- delim, (
- format_field(f, 'filesize', ' \t%s', func=format_bytes)
- or format_field(f, 'filesize_approx', '≈\t%s', func=format_bytes)
- or format_field(filesize_from_tbr(f.get('tbr'), info_dict.get('duration')), None,
- self._format_out('~\t%s', self.Styles.SUPPRESS), func=format_bytes)),
- format_field(f, 'tbr', '\t%dk', func=round),
- shorten_protocol_name(f.get('protocol', '')),
- delim,
- simplified_codec(f, 'vcodec'),
- format_field(f, 'vbr', '\t%dk', func=round),
- simplified_codec(f, 'acodec'),
- format_field(f, 'abr', '\t%dk', func=round),
- format_field(f, 'asr', '\t%s', func=format_decimal_suffix),
- join_nonempty(format_field(f, 'language', '[%s]'), join_nonempty(
- self._format_out('UNSUPPORTED', self.Styles.BAD_FORMAT) if f.get('ext') in ('f4f', 'f4m') else None,
- (self._format_out('Maybe DRM', self.Styles.WARNING) if f.get('has_drm') == 'maybe'
- else self._format_out('DRM', self.Styles.BAD_FORMAT) if f.get('has_drm') else None),
- format_field(f, 'format_note'),
- format_field(f, 'container', ignore=(None, f.get('ext'))),
- delim=', '), delim=' '),
- ] for f in formats if f.get('preference') is None or f['preference'] >= -1000]
- header_line = self._list_format_headers(
- 'ID', 'EXT', 'RESOLUTION', '\tFPS', 'HDR', 'CH', delim, '\tFILESIZE', '\tTBR', 'PROTO',
- delim, 'VCODEC', '\tVBR', 'ACODEC', '\tABR', '\tASR', 'MORE INFO')
- return render_table(
- header_line, table, hide_empty=True,
- delim=self._format_out('\u2500', self.Styles.DELIM, '-', test_encoding=True))
- def render_thumbnails_table(self, info_dict):
- thumbnails = list(info_dict.get('thumbnails') or [])
- if not thumbnails:
- return None
- return render_table(
- self._list_format_headers('ID', 'Width', 'Height', 'URL'),
- [[t.get('id'), t.get('width') or 'unknown', t.get('height') or 'unknown', t['url']] for t in thumbnails])
- def render_subtitles_table(self, video_id, subtitles):
- def _row(lang, formats):
- exts, names = zip(*((f['ext'], f.get('name') or 'unknown') for f in reversed(formats)))
- if len(set(names)) == 1:
- names = [] if names[0] == 'unknown' else names[:1]
- return [lang, ', '.join(names), ', '.join(exts)]
- if not subtitles:
- return None
- return render_table(
- self._list_format_headers('Language', 'Name', 'Formats'),
- [_row(lang, formats) for lang, formats in subtitles.items()],
- hide_empty=True)
- def __list_table(self, video_id, name, func, *args):
- table = func(*args)
- if not table:
- self.to_screen(f'{video_id} has no {name}')
- return
- self.to_screen(f'[info] Available {name} for {video_id}:')
- self.to_stdout(table)
- def list_formats(self, info_dict):
- self.__list_table(info_dict['id'], 'formats', self.render_formats_table, info_dict)
- def list_thumbnails(self, info_dict):
- self.__list_table(info_dict['id'], 'thumbnails', self.render_thumbnails_table, info_dict)
- def list_subtitles(self, video_id, subtitles, name='subtitles'):
- self.__list_table(video_id, name, self.render_subtitles_table, video_id, subtitles)
- def print_debug_header(self):
- if not self.params.get('verbose'):
- return
- from . import _IN_CLI # Must be delayed import
- # These imports can be slow. So import them only as needed
- from .extractor.extractors import _LAZY_LOADER
- from .extractor.extractors import (
- _PLUGIN_CLASSES as plugin_ies,
- _PLUGIN_OVERRIDES as plugin_ie_overrides,
- )
- def get_encoding(stream):
- ret = str(getattr(stream, 'encoding', f'missing ({type(stream).__name__})'))
- additional_info = []
- if os.environ.get('TERM', '').lower() == 'dumb':
- additional_info.append('dumb')
- if not supports_terminal_sequences(stream):
- from .utils import WINDOWS_VT_MODE # Must be imported locally
- additional_info.append('No VT' if WINDOWS_VT_MODE is False else 'No ANSI')
- if additional_info:
- ret = f'{ret} ({",".join(additional_info)})'
- return ret
- encoding_str = 'Encodings: locale {}, fs {}, pref {}, {}'.format(
- locale.getpreferredencoding(),
- sys.getfilesystemencoding(),
- self.get_encoding(),
- ', '.join(
- f'{key} {get_encoding(stream)}' for key, stream in self._out_files.items_
- if stream is not None and key != 'console'),
- )
- logger = self.params.get('logger')
- if logger:
- write_debug = lambda msg: logger.debug(f'[debug] {msg}')
- write_debug(encoding_str)
- else:
- write_string(f'[debug] {encoding_str}\n', encoding=None)
- write_debug = lambda msg: self._write_string(f'[debug] {msg}\n')
- source = detect_variant()
- if VARIANT not in (None, 'pip'):
- source += '*'
- klass = type(self)
- write_debug(join_nonempty(
- f'{REPOSITORY.rpartition("/")[2]} version',
- _make_label(ORIGIN, CHANNEL.partition('@')[2] or __version__, __version__),
- f'[{RELEASE_GIT_HEAD[:9]}]' if RELEASE_GIT_HEAD else '',
- '' if source == 'unknown' else f'({source})',
- '' if _IN_CLI else 'API' if klass == YoutubeDL else f'API:{self.__module__}.{klass.__qualname__}',
- delim=' '))
- if not _IN_CLI:
- write_debug(f'params: {self.params}')
- if not _LAZY_LOADER:
- if os.environ.get('YTDLP_NO_LAZY_EXTRACTORS'):
- write_debug('Lazy loading extractors is forcibly disabled')
- else:
- write_debug('Lazy loading extractors is disabled')
- if self.params['compat_opts']:
- write_debug('Compatibility options: {}'.format(', '.join(self.params['compat_opts'])))
- if current_git_head():
- write_debug(f'Git HEAD: {current_git_head()}')
- write_debug(system_identifier())
- exe_versions, ffmpeg_features = FFmpegPostProcessor.get_versions_and_features(self)
- ffmpeg_features = {key for key, val in ffmpeg_features.items() if val}
- if ffmpeg_features:
- exe_versions['ffmpeg'] += ' ({})'.format(','.join(sorted(ffmpeg_features)))
- exe_versions['rtmpdump'] = rtmpdump_version()
- exe_versions['phantomjs'] = PhantomJSwrapper._version()
- exe_str = ', '.join(
- f'{exe} {v}' for exe, v in sorted(exe_versions.items()) if v
- ) or 'none'
- write_debug(f'exe versions: {exe_str}')
- from .compat.compat_utils import get_package_info
- from .dependencies import available_dependencies
- write_debug('Optional libraries: %s' % (', '.join(sorted({
- join_nonempty(*get_package_info(m)) for m in available_dependencies.values()
- })) or 'none'))
- write_debug(f'Proxy map: {self.proxies}')
- write_debug(f'Request Handlers: {", ".join(rh.RH_NAME for rh in self._request_director.handlers.values())}')
- for plugin_type, plugins in {'Extractor': plugin_ies, 'Post-Processor': plugin_pps}.items():
- display_list = ['{}{}'.format(
- klass.__name__, '' if klass.__name__ == name else f' as {name}')
- for name, klass in plugins.items()]
- if plugin_type == 'Extractor':
- display_list.extend(f'{plugins[-1].IE_NAME.partition("+")[2]} ({parent.__name__})'
- for parent, plugins in plugin_ie_overrides.items())
- if not display_list:
- continue
- write_debug(f'{plugin_type} Plugins: {", ".join(sorted(display_list))}')
- plugin_dirs = plugin_directories()
- if plugin_dirs:
- write_debug(f'Plugin directories: {plugin_dirs}')
- # Not implemented
- if False and self.params.get('call_home'):
- ipaddr = self.urlopen('https://yt-dl.org/ip').read().decode()
- write_debug(f'Public IP address: {ipaddr}')
- latest_version = self.urlopen(
- 'https://yt-dl.org/latest/version').read().decode()
- if version_tuple(latest_version) > version_tuple(__version__):
- self.report_warning(
- f'You are using an outdated version (newest version: {latest_version})! '
- 'See https://yt-dl.org/update if you need help updating.')
- @functools.cached_property
- def proxies(self):
- """Global proxy configuration"""
- opts_proxy = self.params.get('proxy')
- if opts_proxy is not None:
- if opts_proxy == '':
- opts_proxy = '__noproxy__'
- proxies = {'all': opts_proxy}
- else:
- proxies = urllib.request.getproxies()
- # compat. Set HTTPS_PROXY to __noproxy__ to revert
- if 'http' in proxies and 'https' not in proxies:
- proxies['https'] = proxies['http']
- return proxies
- @functools.cached_property
- def cookiejar(self):
- """Global cookiejar instance"""
- try:
- return load_cookies(
- self.params.get('cookiefile'), self.params.get('cookiesfrombrowser'), self)
- except CookieLoadError as error:
- cause = error.__context__
- self.report_error(str(cause), tb=''.join(traceback.format_exception(cause)))
- raise
- @property
- def _opener(self):
- """
- Get a urllib OpenerDirector from the Urllib handler (deprecated).
- """
- self.deprecation_warning('YoutubeDL._opener is deprecated, use YoutubeDL.urlopen()')
- handler = self._request_director.handlers['Urllib']
- return handler._get_instance(cookiejar=self.cookiejar, proxies=self.proxies)
- def _get_available_impersonate_targets(self):
- # TODO(future): make available as public API
- return [
- (target, rh.RH_NAME)
- for rh in self._request_director.handlers.values()
- if isinstance(rh, ImpersonateRequestHandler)
- for target in rh.supported_targets
- ]
- def _impersonate_target_available(self, target):
- # TODO(future): make available as public API
- return any(
- rh.is_supported_target(target)
- for rh in self._request_director.handlers.values()
- if isinstance(rh, ImpersonateRequestHandler))
- def urlopen(self, req):
- """ Start an HTTP download """
- if isinstance(req, str):
- req = Request(req)
- elif isinstance(req, urllib.request.Request):
- self.deprecation_warning(
- 'Passing a urllib.request.Request object to YoutubeDL.urlopen() is deprecated. '
- 'Use yt_dlp.networking.common.Request instead.')
- req = urllib_req_to_req(req)
- assert isinstance(req, Request)
- # compat: Assume user:pass url params are basic auth
- url, basic_auth_header = extract_basic_auth(req.url)
- if basic_auth_header:
- req.headers['Authorization'] = basic_auth_header
- req.url = sanitize_url(url)
- clean_proxies(proxies=req.proxies, headers=req.headers)
- clean_headers(req.headers)
- try:
- return self._request_director.send(req)
- except NoSupportingHandlers as e:
- for ue in e.unsupported_errors:
- # FIXME: This depends on the order of errors.
- if not (ue.handler and ue.msg):
- continue
- if ue.handler.RH_KEY == 'Urllib' and 'unsupported url scheme: "file"' in ue.msg.lower():
- raise RequestError(
- 'file:// URLs are disabled by default in yt-dlp for security reasons. '
- 'Use --enable-file-urls to enable at your own risk.', cause=ue) from ue
- if (
- 'unsupported proxy type: "https"' in ue.msg.lower()
- and 'requests' not in self._request_director.handlers
- and 'curl_cffi' not in self._request_director.handlers
- ):
- raise RequestError(
- 'To use an HTTPS proxy for this request, one of the following dependencies needs to be installed: requests, curl_cffi')
- elif (
- re.match(r'unsupported url scheme: "wss?"', ue.msg.lower())
- and 'websockets' not in self._request_director.handlers
- ):
- raise RequestError(
- 'This request requires WebSocket support. '
- 'Ensure one of the following dependencies are installed: websockets',
- cause=ue) from ue
- elif re.match(r'unsupported (?:extensions: impersonate|impersonate target)', ue.msg.lower()):
- raise RequestError(
- f'Impersonate target "{req.extensions["impersonate"]}" is not available.'
- f' See --list-impersonate-targets for available targets.'
- f' This request requires browser impersonation, however you may be missing dependencies'
- f' required to support this target.')
- raise
- except SSLError as e:
- if 'UNSAFE_LEGACY_RENEGOTIATION_DISABLED' in str(e):
- raise RequestError('UNSAFE_LEGACY_RENEGOTIATION_DISABLED: Try using --legacy-server-connect', cause=e) from e
- elif 'SSLV3_ALERT_HANDSHAKE_FAILURE' in str(e):
- raise RequestError(
- 'SSLV3_ALERT_HANDSHAKE_FAILURE: The server may not support the current cipher list. '
- 'Try using --legacy-server-connect', cause=e) from e
- raise
- def build_request_director(self, handlers, preferences=None):
- logger = _YDLLogger(self)
- headers = self.params['http_headers'].copy()
- proxies = self.proxies.copy()
- clean_headers(headers)
- clean_proxies(proxies, headers)
- director = RequestDirector(logger=logger, verbose=self.params.get('debug_printtraffic'))
- for handler in handlers:
- director.add_handler(handler(
- logger=logger,
- headers=headers,
- cookiejar=self.cookiejar,
- proxies=proxies,
- prefer_system_certs='no-certifi' in self.params['compat_opts'],
- verify=not self.params.get('nocheckcertificate'),
- **traverse_obj(self.params, {
- 'verbose': 'debug_printtraffic',
- 'source_address': 'source_address',
- 'timeout': 'socket_timeout',
- 'legacy_ssl_support': 'legacyserverconnect',
- 'enable_file_urls': 'enable_file_urls',
- 'impersonate': 'impersonate',
- 'client_cert': {
- 'client_certificate': 'client_certificate',
- 'client_certificate_key': 'client_certificate_key',
- 'client_certificate_password': 'client_certificate_password',
- },
- }),
- ))
- director.preferences.update(preferences or [])
- if 'prefer-legacy-http-handler' in self.params['compat_opts']:
- director.preferences.add(lambda rh, _: 500 if rh.RH_KEY == 'Urllib' else 0)
- return director
- @functools.cached_property
- def _request_director(self):
- return self.build_request_director(_REQUEST_HANDLERS.values(), _RH_PREFERENCES)
- def encode(self, s):
- if isinstance(s, bytes):
- return s # Already encoded
- try:
- return s.encode(self.get_encoding())
- except UnicodeEncodeError as err:
- err.reason = err.reason + '. Check your system encoding configuration or use the --encoding option.'
- raise
- def get_encoding(self):
- encoding = self.params.get('encoding')
- if encoding is None:
- encoding = preferredencoding()
- return encoding
- def _write_info_json(self, label, ie_result, infofn, overwrite=None):
- """ Write infojson and returns True = written, 'exists' = Already exists, False = skip, None = error """
- if overwrite is None:
- overwrite = self.params.get('overwrites', True)
- if not self.params.get('writeinfojson'):
- return False
- elif not infofn:
- self.write_debug(f'Skipping writing {label} infojson')
- return False
- elif not self._ensure_dir_exists(infofn):
- return None
- elif not overwrite and os.path.exists(infofn):
- self.to_screen(f'[info] {label.title()} metadata is already present')
- return 'exists'
- self.to_screen(f'[info] Writing {label} metadata as JSON to: {infofn}')
- try:
- write_json_file(self.sanitize_info(ie_result, self.params.get('clean_infojson', True)), infofn)
- return True
- except OSError:
- self.report_error(f'Cannot write {label} metadata to JSON file {infofn}')
- return None
- def _write_description(self, label, ie_result, descfn):
- """ Write description and returns True = written, False = skip, None = error """
- if not self.params.get('writedescription'):
- return False
- elif not descfn:
- self.write_debug(f'Skipping writing {label} description')
- return False
- elif not self._ensure_dir_exists(descfn):
- return None
- elif not self.params.get('overwrites', True) and os.path.exists(descfn):
- self.to_screen(f'[info] {label.title()} description is already present')
- elif ie_result.get('description') is None:
- self.to_screen(f'[info] There\'s no {label} description to write')
- return False
- else:
- try:
- self.to_screen(f'[info] Writing {label} description to: {descfn}')
- with open(encodeFilename(descfn), 'w', encoding='utf-8') as descfile:
- descfile.write(ie_result['description'])
- except OSError:
- self.report_error(f'Cannot write {label} description file {descfn}')
- return None
- return True
- def _write_subtitles(self, info_dict, filename):
- """ Write subtitles to file and return list of (sub_filename, final_sub_filename); or None if error"""
- ret = []
- subtitles = info_dict.get('requested_subtitles')
- if not (self.params.get('writesubtitles') or self.params.get('writeautomaticsub')):
- # subtitles download errors are already managed as troubles in relevant IE
- # that way it will silently go on when used with unsupporting IE
- return ret
- elif not subtitles:
- self.to_screen('[info] There are no subtitles for the requested languages')
- return ret
- sub_filename_base = self.prepare_filename(info_dict, 'subtitle')
- if not sub_filename_base:
- self.to_screen('[info] Skipping writing video subtitles')
- return ret
- for sub_lang, sub_info in subtitles.items():
- sub_format = sub_info['ext']
- sub_filename = subtitles_filename(filename, sub_lang, sub_format, info_dict.get('ext'))
- sub_filename_final = subtitles_filename(sub_filename_base, sub_lang, sub_format, info_dict.get('ext'))
- existing_sub = self.existing_file((sub_filename_final, sub_filename))
- if existing_sub:
- self.to_screen(f'[info] Video subtitle {sub_lang}.{sub_format} is already present')
- sub_info['filepath'] = existing_sub
- ret.append((existing_sub, sub_filename_final))
- continue
- self.to_screen(f'[info] Writing video subtitles to: {sub_filename}')
- if sub_info.get('data') is not None:
- try:
- # Use newline='' to prevent conversion of newline characters
- # See https://github.com/ytdl-org/youtube-dl/issues/10268
- with open(sub_filename, 'w', encoding='utf-8', newline='') as subfile:
- subfile.write(sub_info['data'])
- sub_info['filepath'] = sub_filename
- ret.append((sub_filename, sub_filename_final))
- continue
- except OSError:
- self.report_error(f'Cannot write video subtitles file {sub_filename}')
- return None
- try:
- sub_copy = sub_info.copy()
- sub_copy.setdefault('http_headers', info_dict.get('http_headers'))
- self.dl(sub_filename, sub_copy, subtitle=True)
- sub_info['filepath'] = sub_filename
- ret.append((sub_filename, sub_filename_final))
- except (DownloadError, ExtractorError, OSError, ValueError, *network_exceptions) as err:
- msg = f'Unable to download video subtitles for {sub_lang!r}: {err}'
- if self.params.get('ignoreerrors') is not True: # False or 'only_download'
- if not self.params.get('ignoreerrors'):
- self.report_error(msg)
- raise DownloadError(msg)
- self.report_warning(msg)
- return ret
- def _write_thumbnails(self, label, info_dict, filename, thumb_filename_base=None):
- """ Write thumbnails to file and return list of (thumb_filename, final_thumb_filename); or None if error """
- write_all = self.params.get('write_all_thumbnails', False)
- thumbnails, ret = [], []
- if write_all or self.params.get('writethumbnail', False):
- thumbnails = info_dict.get('thumbnails') or []
- if not thumbnails:
- self.to_screen(f'[info] There are no {label} thumbnails to download')
- return ret
- multiple = write_all and len(thumbnails) > 1
- if thumb_filename_base is None:
- thumb_filename_base = filename
- if thumbnails and not thumb_filename_base:
- self.write_debug(f'Skipping writing {label} thumbnail')
- return ret
- if thumbnails and not self._ensure_dir_exists(filename):
- return None
- for idx, t in list(enumerate(thumbnails))[::-1]:
- thumb_ext = (f'{t["id"]}.' if multiple else '') + determine_ext(t['url'], 'jpg')
- thumb_display_id = f'{label} thumbnail {t["id"]}'
- thumb_filename = replace_extension(filename, thumb_ext, info_dict.get('ext'))
- thumb_filename_final = replace_extension(thumb_filename_base, thumb_ext, info_dict.get('ext'))
- existing_thumb = self.existing_file((thumb_filename_final, thumb_filename))
- if existing_thumb:
- self.to_screen('[info] {} is already present'.format((
- thumb_display_id if multiple else f'{label} thumbnail').capitalize()))
- t['filepath'] = existing_thumb
- ret.append((existing_thumb, thumb_filename_final))
- else:
- self.to_screen(f'[info] Downloading {thumb_display_id} ...')
- try:
- uf = self.urlopen(Request(t['url'], headers=t.get('http_headers', {})))
- self.to_screen(f'[info] Writing {thumb_display_id} to: {thumb_filename}')
- with open(encodeFilename(thumb_filename), 'wb') as thumbf:
- shutil.copyfileobj(uf, thumbf)
- ret.append((thumb_filename, thumb_filename_final))
- t['filepath'] = thumb_filename
- except network_exceptions as err:
- if isinstance(err, HTTPError) and err.status == 404:
- self.to_screen(f'[info] {thumb_display_id.title()} does not exist')
- else:
- self.report_warning(f'Unable to download {thumb_display_id}: {err}')
- thumbnails.pop(idx)
- if ret and not write_all:
- break
- return ret
|