* improve space reserving, allows write better 4K/8K video data
* do not use cache dirs in the muxers, Android can force close NewPipe if the device is running out of storage. Is a aggressive cache cleaning >:/
* (for devs) webm & mkv are the same thing
* calculate the final file size inside of the mission, instead getting from the UI
* simplify ps algorithms constructors
* [missing old commit message] simplify the loading of pending downloads
* use Java I/O (classic way) on older android versions
* use Storage Access Framework on newer android versions (Android Lollipop or later)
* both changes have the external SD Card write permission
* add option to ask the save path on each download
* warn the user if the save paths are not defined, this only happens on the first NewPipe run (Android Lollipop or later)
* replace "In queue" to "Pending" in the downloads header to avoid confusions (all languages)
* use 29bits Clusters size to support huge video resolutions (fixes#2291) (WebmWriter.java)
* add missing changes to WebmMuxer.java (i forget select the audio track)
* full support for Directory API (Android Lollipop or later)
* best effort to handle any kind errors (missing file, revoked permissions, etc) and recover the download
* implemented directory choosing
* fix download database version upgrading
* misc. cleanup
* do not release permission on the old save path (if the user change the download directory) under SAF api
* re-work finished mission database
* re-work DownloadMission and bump it Serializable version
* keep the classic Java IO API
* SAF Tree API support on Android Lollipop or higher
* add wrapper for SAF stream opening
* implement Closeable in SharpStream to replace the dispose() method
* do required changes for this API:
** remove any file creation logic from DownloadInitializer
** make PostProcessing Serializable and reduce the number of iterations
** update all strings.xml files
** storage helpers: StoredDirectoryHelper & StoredFileHelper
** best effort to handle any kind of SAF errors/exceptions
Main changes:
* correctly check the available space (CircularFile.java)
* misc cleanup (CircularFile.java)
* use the "Error Reporter" for non-http errors
* rewrite network state checking and add better support for API 21 (Lollipop) or higher
* implement "metered networks"
* add buttons in "Downloads" activity to start/pause all pending downloads, ignoring the queue flag or if the network is "metered"
* add workaround for VPN connections and/or network switching. Example: switching WiFi to 3G
* rewrite DataReader ¡Webm muxer is now 57% more faster!
* rewrite CircularFile, use file buffers instead of memory buffers. Less troubles in low-end devices
* fix missing offset for KaxCluster (WebMWriter.java), manifested as no thumbnails on file explorers
Download queue:
* remember queue status, unless the user pause the download (un-queue)
* semi-automatic downloads, between networks. Effective if the user create a new download or the downloads activity is starts
* allow enqueue failed downloads
* new option, queue limit, enabled by default. Used to allow one or multiple downloads at same time
Miscellaneous:
* fix crash while selecting details/error menu (mistake on MissionFragment.java)
* misc serialize changes (DownloadMission.java)
* minor UI tweaks
* allow overwrite paused downloads
* fix wrong icons for grid/list button in downloads
* add share option
* implement #2006
* correct misspelled word in strings.xml (es) (cmn)
* fix MissionAdapter crash during device shutdown
New Mp4Muxer + required changes:
* new mp4 muxer (from dash only) with this, muxing on Android 7 is possible now!!!
* re-work in SharpStream
* drop mp4 dash muxer
* misc changes: add warning in SecondaryStreamHelper.java,
* strip m4a DASH files to normal m4a format (youtube only)
Fix storage issues:
* warn to the user if is choosing a "read only" download directory (for external SD Cards), useless is rooted :)
* "write proof" allow post-processing resuming only if the device ran out of space
* implement "insufficient storage" error for downloads
This pull requests complements pull request #2178 by reducing general computational time for the method getTimeString.
On my local machine (Desktop PC with Java) my tests with a sample size of 10000 calls to the method with param 86400001 showed a performance improvement of about 50%.
See sample code below to reproduce:
private static final StringBuilder stringBuilder = new StringBuilder();
private static final Formatter stringFormatter = new Formatter(stringBuilder, Locale.getDefault());
public static String getTimeString(int milliSeconds) {
int seconds = (milliSeconds % 60000) / 1000;
int minutes = (milliSeconds % 3600000) / 60000;
int hours = (milliSeconds % 86400000) / 3600000;
int days = (milliSeconds % (86400000 * 7)) / 86400000;
stringBuilder.setLength(0);
return days > 0 ? stringFormatter.format("%d:%02d:%02d:%02d", days, hours, minutes, seconds).toString()
: hours > 0 ? stringFormatter.format("%d:%02d:%02d", hours, minutes, seconds).toString()
: stringFormatter.format("%02d:%02d", minutes, seconds).toString();
}
public static String getTimeStringL(int milliSeconds) {
long seconds = (milliSeconds % 60000L) / 1000L;
long minutes = (milliSeconds % 3600000L) / 60000L;
long hours = (milliSeconds % 86400000L) / 3600000L;
long days = (milliSeconds % (86400000L * 7L)) / 86400000L;
stringBuilder.setLength(0);
return days > 0 ? stringFormatter.format("%d:%02d:%02d:%02d", days, hours, minutes, seconds).toString()
: hours > 0 ? stringFormatter.format("%d:%02d:%02d", hours, minutes, seconds).toString()
: stringFormatter.format("%02d:%02d", minutes, seconds).toString();
}
public static void main(String[] args) throws Exception {
final int SAMPLE_SIZE = 25000;
long[] results = new long[SAMPLE_SIZE];
for(int i = 0; i < SAMPLE_SIZE; i++) {
long now = System.nanoTime();
getTimeString(86400001);
results[i] = System.nanoTime() - now;
}
long sum = 0;
for(int i = 0; i < SAMPLE_SIZE; i++) {
sum += results[i];
}
System.out.println("Average execution time: " + (sum/SAMPLE_SIZE));
results = new long[SAMPLE_SIZE];
for(int i = 0; i < SAMPLE_SIZE; i++) {
long now = System.nanoTime();
getTimeStringL(86400001);
results[i] = System.nanoTime() - now;
}
sum = 0;
for(int i = 0; i < SAMPLE_SIZE; i++) {
sum += results[i];
}
System.out.println("Average execution time: " + (sum/SAMPLE_SIZE));
In VideoPlayer the Duration String is cached effectively by setting it to the playbackSeekBar. As the playbackSeekBar doesn't exist in BackgroundPlayer, using two addition variables will reduce performance impact of notification updates by almost 50% and thus perform similar to VideoPlayer.
This addresses issue #2170
