This header was not sent partially before and was added and guessed by OkHttp. This can create issues when using other HTTP clients than OkHttp, such as Cronet.
Some code in the modified classes has been improved and / or deduplicated, and usages of the UTF_8 constant of the Utils class has been replaced by StandardCharsets.UTF_8 where possible.
Note that this header has been not added in except in YoutubeDashManifestCreatorsUtils, as an empty body is sent in the POST requests made by this class.
This header was not sent before and was added and guessed by OkHttp. This can create issues when using other HTTP clients than OkHttp, such as Cronet.
Also make use of StandardCharsets.UTF_8 when getting bytes of bodies instead of the platform default's charset, to make sure to prevent some encoding issues on some JVMs.
The video "Makani’s first commercial-scale energy kite" (video ID:
An8vtD1FDqs), which has this behavior, is used for the new test,
NoVisualMetadataVideoTest, added in YoutubeStreamExtractorDefaultTest.
Tests of elements who throw an exception in this case (subscriber count, like
count, uploader avatar URL) test if the ParsingException exception is thrown by
YoutubeStreamExtractor.
More non-channel paths have been also added to the excluded custom name paths,
documentation and exception messages have been improved and fixed in some
places, and the licence header of YoutubeChannelLinkHandlerFactory has been
moved to its beginning and updated.
Also update the mock of the next InnerTube endpoint response of the
YoutubeStreamExtractorDefaultTest.DescriptionTestUnboxing test class with an
attributedDescription instead of a regular description
The video count is now returned for this playlist, so it isn't unknown.
The testStreamCount method of this test class asserts now that the stream count
is greater than 40.
This new layout doesn't provide author thumbnails and is completely different
for metadata, so the code to get them has been refactored.
The code of learning playlists video count check has been also removed, as it
seems to be not relevant anymore (the video count seems to be returned for
these playlists with both layouts).
Finally, unneeded overrides of subchannel methods, which don't apply to the
YouTube service, have been removed.
Using the player parameters used to get stories seems to fix the issue, which
affects currently only certain countries such as UK.
This is a workaround and should be fixed in a better way (by changing the
InnerTube additional client used for videos or finding what is now required in
Android player requests).
Use the same possible characters for variables everywhere, in order to avoid
potential future throttling parameter decryption function name parsing issues
related to the usage of other letter(s) than b.
- Quote the function name, as it may contain special regex symbols, such as
dollar;
- Support multiple lines;
- Use what looks like the end of the function for the end of the regex (this
part is inspired from yt-dlp throttling parameter decryption regex);
- Move the throttling function body regex into a private and static constant.
This new structure allow us to filter easily Trending shorts and Recently
trending sections.
On the previous one, this Recently trending section is now filtered, by
checking whether sections have a title, which isn't the case for normal trends
contrary to the other ones.
This makes that the extractor returns now only the real 50 "Now" YouTube
trends.
Elements inside arrays are now extracted dynamically instead of only the ones
of the first index, using Java 8's Stream API.
The getInitialPage() method of YoutubeTrendingExtractor can now throw a
ParsingException if no selected tab (corresponding to the one of the trends
type extracted) has been found.
Finally, the licence header has been moved to the top of the file and updated.
In the new data model currently A/B tested or deployed, the like count button
data is the same, but its path has been changed.
The extraction of the like count has been also improved, by using the multiple
accessibility data available instead of the only one which was used before.
This accessibility data has been also deprioritized, because it is language
dependent when there is no view.
Like count is always known, even for age-restricted videos, so returning -1 if
the like count extraction fails on this type of videos is not needed and hides
an extraction error: that's the reason why this return has been removed and the
exception will always be thrown, even if a video is age-restricted.
YoutubeThrottlingDecrypter is now a non-instantiable final class and non-static
attributes have been removed.
The static attributes of this class have been renamed, in order to respect the
naming specification used.
Added option to choose if you want to consent or not - currently this is done by a static variable in ``YoutubeParsingHelper`` - may not be the best long-term solution but for now the tests work again (in EU countries) 🥳
If YouTube detect that requests come from a third party client, they may
replace the real player response by another one of a video saying that this
content is not available on this app and to watch it on the latest version of
YouTube. We can detect this by checking whether the video ID of the player
response returned is the same as the one requested by the extractor.
Additional parameters have been added to the player requests of ANDROID and IOS
clients:
- for both clients: osName and osVersion: their respective values are:
- for the ANDROID one: Android and 12;
- for the IOS one: iOS and 15.6.0.19G71.
- for the ANDROID client: androidTargetSdkVersion, with the Android SDK version
corresponding to the Android version used in the player requests of this
client. This parameter is now required with this client to be sure to get a
correct player response, otherwise, the one of a video saying that this
content is not available in this app and to watch it with the latest version
of YouTube can be returned instead;
- for the IOS client: deviceMake, with Apple as its value.
The iOS version sent in the IOS client player requests has been also updated to
the version 15.6 of the OS.
Finally, a comment about the requirement to use the signature timestamp from
the player JavaScript base file for HTML5 player requests on videos with
obfuscated URLs has been added and replaces a previous one which may be not
true.
Common code in WEB client version HTML extraction has been deduplicated, usage of the Java 8 Stream API has been made and initial data fallback has been used as a last resort.
This means that the client version extraction from regexes will be used before this fallback, as it doesn't contain the full client version.
This can be used as a way to fingerprint the extractor, even if it seems to be not the case.
This will prevent any future extractor break due to decryption failure, like it was excepted to be the case before.
Some documentation about the throttling decryption has been also improved.
The ANDROID client was only fetched for video contents, where it can be useful on ended/post livestreams, if the n parameter of the WEB client cannot be decrypted, to avoid throttling issues (because the WEB client was only used before for ended/post livestreams).
It also provides an exclusive 48kbps M4A audio format in the adaptiveFormats array of the JSON player response, like other mobile clients (which can be also extracted from the response of the DASH manifest URL returned into the WEB client player's response, but the DASH manifest is not used by the extractor).
A note about non-fatality of fetching or parsing issues of the ANDROID and IOS clients has been added.
SoundCloud is currently removing this workaround completely, so there is no need to keep it, because it impacts the loading time (a HLS playlist was downloaded and parsed).
These values were only set before for video streams.
A fallback for the audio channels count has been added, in order to prevent exceptions when generating DASH manifests of audio streams: the fallback value is 2, because most audio streams on YouTube have 2 audio channels.
SoundCloud broke the workaround used to get a single file from HLS manifests for Opus manifests, but it still works for MP3 ones.
The code has been adapted to prevent an unneeded request (the one to the Opus HLS manifest) and the HLS delivery method is now used for SoundCloud MP3 and Opus streams, plus the progressive one (for tracks which have a progressive stream (MP3) and for the ones which doesn't have one, it is still used by trying to get a progressive stream, using the workaround).
Streams extraction has been also moved to Java 8 Stream's API and the relevant test has been also updated.
Move DASH manifests creation into a new subpackage of the YouTube package, dashmanifestcreators.
This subpackage contains:
- CreationException, exception extending Java's RuntimeException, thrown by manifest creators when something goes wrong;
- YoutubeDashManifestCreatorsUtils, class which contains all common methods and constants of all or a part of the manifest creators;
- a manifest creator has been added per delivery type of YouTube streams:
- YoutubeProgressiveDashManifestCreator, for progressive streams;
- YoutubeOtfDashManifestCreator, for OTF streams;
- YoutubePostLiveStreamDvrDashManifestCreator, for post-live DVR streams (which use the live delivery method).
Every DASH manifest creator has a getCache() static method, which returns the ManifestCreatorCache instance used to cache results.
DeliveryType has been also extracted from the YouTube DASH manifest creators part of the extractor and moved to the YouTube package.
YoutubeDashManifestCreatorTest has been updated and renamed to YoutubeDashManifestCreatorsTest, and YoutubeDashManifestCreator has been removed.
Finally, several documentation and exception messages fixes and improvements have been made.
This change allows to build DASH manifests using YoutubeDashManifestCreator with the real duration of streams and prevent potential cuts of the end of progressive streams, because the duration in YouTube's player response is in seconds and not milliseconds.
When moving opus-lo into a constant, opus-lo was renamed to opus_lo and was only used if no MP3 stream was available (which was not the case before the changes in BandcampRadioStreamExtractor related to the addition of the support of all delivery methods), so these changes removed the ability to get Opus streams of Bandcamp radios.
This commit reverts this unwanted change.
- Remove all of the methods used to access caches and replace them with three caches getters
- Rename caches to shorter and more meaningful names
- Remove redundant @throws tags that just say "if this method fails to do what it should do", which is obvious
This commits reverts a new behavior introduced in this branch, which only applied the decryption if needed on streams from the WEB client.
Also fix rebase issues and documentations style in YoutubeDashManifestCreator.
These methods don't need to be overriden, as they are not excepted to be used in collections.
Also improve the toString method of this class, which contains also now clearFactor and maximumSize attributes and for each operations.
This parameter is still used to get the initialization sequence of OTF and POST-live streams, but is not returned anymore in the manifests.
It has been removed in order to avoid fingerprinting based on the number sent (e.g. when starting to play a stream close to the end and using 123 as the request number where it should be 1) and should be added dynamically by clients in their requests.
The relevant test has been also updated.
Checkstyle issues in YoutubeDashManifestCreator have been fixed, and the changes in the resolution string returned for video streams in YoutubeStreamExtractor have been reverted, as they create issues on NewPipe right now.
To do so, a copy constructor has been added in the class.
This fixes, for instance, an issue in NewPipe, in which the ItagItem values where not the ones corresponsing to a stream but to another, when generating DASH manifests.
The test added in YoutubeDashManifestCreator uses a video of the Creative Commons channel, licenced under the Creative Commons Attribution licence (reuse allowed).
Also remove public keywords of tests in UtilsTest, as suggested by SonarLint, because they are not needed with Junit 5.
A new class has been added to do so: YoutubeDashManifestCreator.
It relies on a new class: ManifestCreatorCache, to cache the content, which relies on a new pair class named Pair.
Results are cached and there is a cache per delivery type, on which cache limit, clear factor, clearing and resetting can be applied to each cache and to all caches.
Look at code changes for more details.
Extract post live DVR streams as post live streams instead of live streams.
A new class has been in order to improve code: ItagInfo, which stores an itag, the content (URL) extracted and if its an URL or not.
A functional interface has been added in order to abstract the stream building: StreamBuilderHelper.
Also add the cver parameter added by the desktop web client on the corresponding streams (a new method has been added in YoutubeParsingHelper to check this and another for Android streams).
Some code in these classes has been also refactored/improved/optimized.
Some code in these classes has been also refactored/improved/optimized.
Also fix the extraction of PeerTube audio streams as video streams, which are now returned as audio streams.
These fields can be now replaced by a getter and a setter.
New fields have been added and will allow the creation of DASH manifests for OTF and ended livestreams. There are:
- contentLength;
- approxDurationMs;
- targetDurationSec;
- sampleRate;
- audioChannels.
DashMpdParser is only working with YouTube streams, as it uses the ItagItem class.
Also improve code and comments of StreamInfo (especially final use where possible).
Stream constructors are now private and streams can be constructed with new Builder classes per stream class. This change has been made to prevent creating and using several constructors in stream classes.
Some default cases have been also added in these Builder classes, so not everything has to be set, depending of the service and the content.
Google returns now the consent page of YouTube for YouTube Music in EU, which can be also avoided by adding the ucbcb parameter to the URL with the value 1 ("?ucbcb=1").
* Fixed obvious sonar(lint) warnings
* Abstracted some code (get*Streams)
* Used some new lines to make code better readable
* Chopped down brace-jungle in some methods
* Use StandardCharset (Java 8 4tw)
Mixes seems to be not given by YouTube anymore if you use a PENDING consent cookie value.
As mocks needs to updated, the test is always failing because of this change.
Use the TV embedded client technique to get streams of embeddable age-restricted videos.
This client doesn't provide the playerMicroFormatRenderer object in the player response, but it is still returned on the WEB player response, even for unavailable (but non-private) contents, so we need now to store it, as we are replacing the player response from the WEB client by the TV embedded one.
Otherwise, some metadata such as the unlisted property, category, the uploadDate and the publishDate properties.
The outdated code for these contents has been removed.
Add the racyCheckOk and contentCheckOk to player and next requests to the InnerTube API.
The first doesn't seem to make any difference when used anonymously, but the second one is needed to get streams of contents with a warning before they can be played.
Also apply some requested changes, fixes and improvements in YoutubeParsingHelper and YoutubeStreamExtractor.
Also move the iPhone device machine id to a constant, explain how it is used and move the licence in the header of the file, and fix missing imports in YoutubeStreamExtractor (due to a rebase issue).
InnerTube responses return pretty printed responses, which increase responses' size for nothing.
By using the prettyPrint parameter on requests and setting its value to false, responses are not pretty printed anymore, which reduces responses size, and so data transfer and processing times.
This usage has been recently deployed by YouTube on their websites.
In order to use still use mocks with the generation of random strings in player requests, we need to use YoutubeParsingHelper.setSeedForVideoTests() method in every stream test.
The iOS client is only enabled for livestreams and the Android client is now only enabled for videos, both by default.
A way to force, or not, the fetch of both clients have been added with two new static methods in YoutubeStreamExtractor.
This is done by fetching https://www.youtube.com/sw.js for YouTube and https://music.youtube.com/sw.js for YouTube Music.
Two new methods in Utils class have been added which allow to try to get a match of regular expressions in a string array, or a Pattern array, on a content, on a specific index or 0.
Also some code refactoring has been made in this class.
The cpn param, aka the content playback nonce param, is a parameter sent by YouTube web client in videoplayback requests, and for some of them, in the player request body. This PR adds it everywhere.
For the desktop/WEB client, some params were missing from the playbackContext object, which seemed (or not) to make YouTube throttle streams extracted from the WEB client. This PR adds them.
Fingerprinting on the WEB client basing on the client version used is not possible anymore, because the latest client version is extracted at the first time of a YouTube request on a session which require the extractor to fetch again the website (and this may come back the reCaptcha issues again unfortunately, but it seems there is no other way to get it).
For the Android client, the video id is now also sent as a query parameter, like a 12 characters string, in the t query parameter, in order to spoof better this client. Researches need to be done on this parameter, unique to each request, and how it is generated by clients.
This commit also fixes a small bug with the Android User-Agent string.
Some code improvements have been also made.
The boolean keyAndVersionExtracted in YoutubeParsingHelper was not set to false when resetting the client version and the key, which makes the extractor uses null on the next getting of the client version or the key if the clientVersion and the key were extracted before.
Also update client versions.
That is, basically where the overriding function was missing an annotation from the base method.
Also apply renaming of emptyDescription to EMPTY_DESCRIPTION
``contentFilters`` and ``sortfilter`` are get inside the ``ListLinkHandler`` and not the ``ListLinkHandlerFactory``
``ListLinkHandlerFactory`` only passes these values through when ``fromQuery`` is called
With respect to NewPipe's checkstyle.xml, checkstyle is disabled for javadoc comments. There is no need for strict rules over comments here in the extractor, as sometimes javadocs are just needed to clarify a small thing and having empty/meaningless @param or @throws is useless.
Replaces mix tests based on a strange mix type RDQM{videoId} (only reference I could find is https://github.com/ytdl-org/youtube-dl/issues/26228) and with an invalid video id of 13 characters (the first two characters were QM, but even after removing QM there still wasn't a video available at that id).
Also updates mocks.
Note: genre mixes already worked, now they are just considered as such in various video id extraction and in related items
Note 2: now extracting a mix id from a *normal* youtube mix id will fail if the video id wouldn't be exactly 11 characters long
It is a collector that can handle many extractor types, to be used when a list contains items of different types (e.g. search). It was renamed from InfoItemsSearchCollector so that it can now be used not just for search but for any extractor needing it. It supports, streams, channels, playlists and *mixes*.
ITEM_COUNT_UNKNOWN is returned when the JSON array which contains usally the number of videos is less than 3 items.
Also apply the same type of optimizations done in other PlaylistExtractors in YoutubePlaylistExtractor.
Also fix some issues in the extractors, remove uneeded overrides, use the Java 8 Stream API where possible and replace usages of Utils.UTF_8 with StandardCharsets.UTF_8 in these classes.
* Deprecated Utils#UTF-8; see StandardCharsets
* Added more helpful methods to ``ExtractorAsserts``
* Use parameterized (cool new) tests
* Restore functionality of some tests + updated mockdata
* Other code cleanups + Sonarlint improvements
csv:
Improved error messages
Exits early if it hasnt found any items in the first few lines
zip:
Now checks all CSV files instead of hard-coded paths
final qualifiers for immutable locals and parameters
Co-authored-by: litetex <40789489+litetex@users.noreply.github.com>
* Faster iframe api based player extraction.
Uses the IFrame API to reduce the required download to less than 1/50 of the size.
* Remove debug code.
* Extract to two methods.
* Add tests for player URL extraction.
* Add assertThat for tests.
Without removing RunWith and SuiteClasses annotations (and the corresponding imports) in YoutubePlaylistExtractorTest and YoutubeMixPlaylistExtractorTest, some mocks cannot be generated, so the CI fails because of the missing mocks. Mocks of workings tests have been also updated.
Migrate YouTube comments to the desktop version by using the `next` endpoint of the InnerTube internal API.
With the desktop version, we are able to get the exact like count of YouTube comments (by parsing the accessibility data) (the current extraction is used as a fallback). We are also now able to get if the uploader of the comment is verified or not.
Co-authored-by: TiA4f8R <74829229+TiA4f8R@users.noreply.github.com>
Here is now the requests which will be made by the `onFetchPage` method of `YoutubeStreamExtractor`:
- the desktop API is fetched.
If there is no streaming data, the desktop player API with the embed client screen will be fetched (and also the player code), then the Android mobile API.
- if there is no streaming data, a `ContentNotAvailableException` will be thrown by using the message provided in playability status
If the video is age restricted, a request to the next endpoint of the desktop player with the embed client screen will be sent.
Otherwise, the next endpoint will be fetched normally, if the content is available.
If the video is not age-restricted, a request to the player endpoint of the Android mobile API will be made.
We can get more streams by using the Android mobile API but some streams may be not available on this API, so the streaming data of the Android mobile API will be first used to get itags and then the streaming data of the desktop internal API will be used.
If the parsing of the Android mobile API went wrong, only the streams of the desktop API will be used.
Other code changes:
- `prepareJsonBuilder` in `YoutubeParsingHelper` was renamed to `prepareDesktopJsonBuilder`
- `prepareMobileJsonBuilder` in `YoutubeParsingHelper` was renamed to `prepareAndroidMobileJsonBuilder`
- two new methods in `YoutubeParsingHelper` were added: `prepareDesktopEmbedVideoJsonBuilder` and `prepareAndroidMobileEmbedVideoJsonBuilder`
- `createPlayerBodyWithSts` is now public and was moved to `YoutubeParsingHelper`
- a new method in `YoutubeJavaScriptExtractor` was added: `resetJavaScriptCode`, which was needed for the method `resetDebofuscationCode` of `YoutubeStreamExtractor`
- `areHardcodedClientVersionAndKeyValid` in `YoutubeParsingHelper` returns now a `boolean` instead of an `Optional<Boolean>`
- the `fetchVideoInfoPage` method of `YoutubeStreamExtractor` was removed because YouTube returns now 404 for every client with the `get_video_info` page
- some unused objects and some warnings in `YoutubeStreamExtractor` were removed and fixed
Co-authored-by: TiA4f8R <74829229+TiA4f8R@users.noreply.github.com>
Tobi
Stypox
Kavin
AudricV
ThetaDev
Isira Seneviratne
dependabot[bot]
Luigi Russo
chowder
xz-dev
TurtleArmyMc
litetex
Gábor Lipták
TiA4f8R
Fynn Godau
Nickoriginal
XiangRongLin
Abdu Ameen
Mohammed Anas
bopol
opusforlife2
talanc
magicfelix
gardenapple
TacoTheDank