* Address build warnings and cleanup
* Actual name of who defined the protocol
* Remove uneeded detekt supression
* GraphQL Before-After cleanup
* Lint
* Cleanup unused functions
* Fix some discrepancies with the 1.5 source api and fix lang exception
* Switch to new Ktlint plugin
* Add ktlintCheck to PR builds
* Run formatter
* Put ktlint version in libs toml
* Fix lint
* Use Zip4Java from libs.toml
* Add "download ahead" mutation
Checks if the specified number of unread chapters, that should be downloaded, are available.
In case not enough chapters are downloaded, the number of missing unread chapters will get downloaded
* Optionally pass the latest read chapter id of a manga
In case a chapter will get marked as read, which also triggered the download ahead call, it's possible, that by the time the download ahead logic gets triggered, the chapter hasn't been marked as read yet.
This could then cause this chapter to be included in the chapters to get downloaded.
By providing the chapter id, this chapter will be used as the latest read chapter instead, and thus, not be included inn the chapters to download.
Chapters were added to the queue by database index order.
In case a chapters of different mangas got added to the queue, downloads got mingled instead of being group inserted per manga.
Also sort manga chapters by source order, to make sure, that, in case chapters of a manga are, for some reason, not in the correct order in the database, they will still get downloaded in the order of the source.
Flow::stateIn has "Strong equality-based conflation" (see documentation).
Thus, it omits every value in case it's equal to the previous one.
Since the DownloadManger::getStatus function returns a status with a queue, that contains all current "DownloadChapters" by reference, the equality check was always true.
Thus, progress changes of downloads were never sent to subscribers.
Subscriber were only notified about finished downloads (size of queue changed) or downloader status changes
In case a download was finished, but the downloader got stopped before it was able to remove the finished download from the queue, the downloader got stuck in an endless loop of starting and pausing downloads.
This was caused by selecting the next chapter to download and then recognizing in "Downloader::step", that there is another chapter to download before the current one in the queue.
However, since this recognized chapter is already downloaded, the downloader selected the next queued chapter again.
It was then stuck in this loop until the finished chapter was manually removed from the queue.
* Create manga download dir in case it's missing for cbz downloads
The directory, in which the cbz file should have been saved in, was never created.
* Correctly copy chapter download to final download location
"renameTo" does not include the content of a directory.
Thus, it just created an empty chapter folder int the final download directory
* Move chapter download logic to base class
* Do not reuse "FolderProvider" in "ArchiveProviders" download function
Due to reusing the "FolderProvider" to download a chapter as a cbz file, a normal chapter download folder was created.
In case the download was aborted before the cbz file got created and the folder deleted, the next time the chapter got downloaded, the wrong "FileProvider" was selected, causing the chapter not to get downloaded as a cbz file.
* Update chapter page refresh logic with logic from "ChapterMutation"
* Rename function to "getChapterDownloadReadyByIndex"
* Update "ChapterForDownload" to work with only "chapterId" being passed
* Return database chapter page list in case chapter is downloaded
In case the chapter is downloaded, fetching the chapter pages info should not be needed.
It should also currently break reading downloaded chapters while being offline, since the page request will always fail, since there is no internet connection
* Make server config value changes subscribable
* Make server config value changes subscribable - Update usage
* Add util functions to listen to server config value changes
* Listen to server config value changes - Auto backups
* Listen to server config value changes - Auto global update
* Listen to server config value changes - WebUI auto updates
* Listen to server config value changes - Javalin update ip and port
* Listen to server config value changes - Update socks proxy
* Listen to server config value changes - Update debug log level
* Listen to server config value changes - Update system tray icon
* Update config values one at a time
In case settings are changed in quick succession it's possible that each setting update reverts the change of the previous changed setting because the internal config hasn't been updated yet.
E.g.
1. settingA changed
2. settingB changed
3. settingA updates config file
4. settingB updates config file (internal config hasn't been updated yet with change from settingA)
5. settingA updates internal config (settingA updated)
6. settingB updates internal config (settingB updated, settingA outdated)
now settingA is unchanged because settingB reverted its change while updating the config with its new value
* Always add log interceptor to OkHttpClient
In case debug logs are disabled then the KotlinLogging log level will be set to level > debug and thus, these logs won't get logged
* Rename "maxParallelUpdateRequests" to "maxSourcesInParallel"
* Use server setting "maxSourcesInParallel" for downloads
* Listen to server config value changes - downloads
* Always use latest server settings - Browser
* Always use latest server settings - folders
* [Test] Fix type error
* Rename "DownloadedFilesProvider" to "ChaptersFilesProvider"
* Move files into sub packages
* Further abstract "DownloadedFilesProvider"
* Rename "getCachedImageResponse" to "getImageResponse"
* Extract getting cached image response into new function
* Decouple thumbnail cache and download
* Download and delete permanent thumbnails
When adding/removing manga from/to the library make sure the permanent thumbnail files will get handled properly
* Move thumbnail cache to actual temp folder
* Rename "mangaDownloadsRoot" to "downloadRoot"
* Move manga downloads into "mangas" subfolder
* Clear downloaded thumbnail
* Extract reorder logic into function
* Save download queue everytime a download was finished
The download queue was never saved after a download was finished.
This caused finished download to be restored on a server start, which caused unnecessary "downloads" which most of the time would just finish immediately since the pages were still in the cache
* Wait for download queue save process to be finished
Since multiple downloaders could be finished at the same time, the download queue should be saved synchronously
* Remove unnecessary download queue save trigger
This gets called everytime a downloader finished downloading all chapters of its source.
Since the queue is now saved everytime a download is finished, this is trigger is not needed anymore
* Change log level of download error
* Change type of sourceId in Downloader
Unclear why it was converted to Long since it just got converted back to String anyway when it was used in the Downloader
* Only stop downloads from source of the Downloader
The downloader just changed the state of all downloads, ignoring if they are from the source the Downloader is for or not
* Remove unnecessary DownloadManager::start calls
In case chapters were added to the queue the DownloadManager will start itself
* Extract download filtering into property
* Improve Downloader logging
* Notify clients only in case Downloader was started
In case nothing was done there is nothing to notify about
* Do not start Downloaders for failed downloads
In case there were failed chapter downloads in the queue the DownloadManager still created a Downloader and started it.
This Downloader would than immediately call "onComplete", since there is no available download, which then would refresh the Downloaders again which created an infinite loop until the failed download got removed from the queue
* Retry download in case it failed it gets re-added to the queue
In case a failed downloaded that was still in the queue was tried to get added to the queue again, nothing happened.
Instead of doing nothing, the download should get retried.
Thus, it also provides the logic to easily retry a failed download by just "adding" the chapter to the queue again.
Currently, to retry a failed download, the download has to be removed from the queue and then get re-added.
* Rename function "unqueue" to "dequeue"
* Move "dequeue" function
* Extract dequeue logic into function
* Improve DownloadManager logging
* Override "toString" of DownloadChapter
* Separate cache dir from download dir
* Move downloader logic outside of caching/image download logic
* remove unnecessary method duplication
* moved download logic inside download provider
* optimize and handle partial downloads
* made code review changes
* Add immediate updates to download queue manager for updates that always needs to be delivered
* Update server/src/main/kotlin/suwayomi/tachidesk/manga/impl/download/DownloadManager.kt
Co-authored-by: Mitchell Syer <Mitchellptbo@gmail.com>
* Revert change to make sure that data in status sent to client are actual
* Reduce number of immediate updates to clients
Co-authored-by: Mitchell Syer <Mitchellptbo@gmail.com>
* Downloader rewrite
- Rewrite downloader to use coroutines instead of a thread
- Remove unused Page functions
- Add page progress
- Add ProgressResponseBody
- Add support for canceling a download in the middle of downloading
- Fix clear download queue
* Minor fix
* Minor improvements
- notifyAllClients now launches in another thread and only sends new data every second
- Better handling of download queue checker in step()
- Minor improvements and fixes
* Reorder downloads
* Download in parallel by source
* Remove TODO
* Add POST /downloads endpoint for creating multiple
* Fix review notes
* Add chapter id to API endpoints
* Rewrite batch chapter download to use chapter id instead of mangaId+chapterIndex combination
* Change EnqueueInput format to be more futureproof
* Change endpoint path
* Change endpoint path
