When I read Android source code SkiaRecordingCanvas, I found that if a bitmap is marked as mutable, namely !isImmutable() == true, the bitmap will be cached in GPU by function SkiaPipeline::pinImages which calls the skia interface SKImage_pinAsTexture. But after I commented these lines in pinImages and re-compiled and pushed to the phone, I found the gifs displayed normally. The only difference is that the texture uploading is delayed from prepareTree to renderFrameImpl. So, why does Android use this method to cache the textures of mutable bitmaps?
bool SkiaPipeline::pinImages(std::vector<SkImage*>& mutableImages) {
// for (SkImage* image : mutableImages) {
// if (SkImage_pinAsTexture(image, mRenderThread.getGrContext())) {
// mPinnedImages.emplace_back(sk_ref_sp(image));
// } else {
// return false;
// }
// }
return true;
}
The pinning and unpinning is designed to work even if the underlying bitmap changes.
Pinning it causes the image to eagerly get uploaded to the GPU. Unpinning it allows the GPU related memory to be freed. See the docs at https://github.com/google/skia/blob/dc3332a07906872f37ec3a592db7831178886527/src/core/SkImagePriv.h#L62-L86
By commenting out this code, you're making it so that the image won't (or at least, may not) get uploaded to the GPU until closer to when it is drawn, which may lead to more work being done at unfortunate times, especially on lower end devices. It's also not clear to me what the consequence of calling unpin without a matching pin.
I am using Picasso to download and display images in views all accros my application. Those images are changing very rarely (they are considered valid for a few months).
Is there a simple way to ask Picasso (or the underlying okHttp) to keep those images on disc for this much time?
Disk caching happens "below" Picasso inside the HTTP client. In fact, this process is completely transparent. We never explicitly ask for a cached-version or an internet-version, the HTTP client will make the decision internally and do the right thing.
Because we opted to leverage the HTTP client for caching, we're offered very little control over how the caching actually happens. To answer your question, no, there is no way to tell Picasso (or OkHttp) to cache an image for longer than its headers allow.
I solved it with a Home-made cache, the trick is to add a parameter to the URL that is not used, but making each URL different every X minutes
Calendar cal2 = Calendar.getInstance();
long d = cal2.getTimeInMillis();
int extra = (int) Math.ceil(d/ (10*60*1000)); // 10 minutes cache
Picasso.with(getBaseContext())
.load("http://www.myurl.cat/myimage.png&extra=" + extra)
.placeholder(R.drawable.graphicLoading)
.error(R.drawable.graphicLoadingError)
.into(bottomGraphic);
Before thinking about HTTP behavior, make sure you set a large max size for the disk cache:
cache = Cache(File(application.filesDir, "photos"), Long.MAX_VALUE)
(MAX_VALUE is not recommended for production.) Don't store the cache in application.cacheDir, because android can clear that whenever it wants.
Add an interceptor to set max-stale, which tells the disk cache to use all old files:
val httpClient = OkHttpClient.Builder().cache(cache).addInterceptor { chain ->
// When offline, we always want to show old photos.
val neverExpireCacheControl = CacheControl.Builder().maxStale(Int.MAX_VALUE, TimeUnit.SECONDS).build()
val origRequest = chain.request()
val neverExpireRequest = origRequest.newBuilder().cacheControl(neverExpireCacheControl).build()
chain.proceed(neverExpireRequest)
}.build()
return Picasso.Builder(application).downloader(OkHttp3Downloader(httpClient)).loggingEnabled(true).build()
I discovered this solution by debugging CacheStrategy.getCandidate(). Take a look there if this doesn't solve your problem.
I am using Picasso to download and display images in views all accros my application. Those images are changing very rarely (they are considered valid for a few months).
Is there a simple way to ask Picasso (or the underlying okHttp) to keep those images on disc for this much time?
Disk caching happens "below" Picasso inside the HTTP client. In fact, this process is completely transparent. We never explicitly ask for a cached-version or an internet-version, the HTTP client will make the decision internally and do the right thing.
Because we opted to leverage the HTTP client for caching, we're offered very little control over how the caching actually happens. To answer your question, no, there is no way to tell Picasso (or OkHttp) to cache an image for longer than its headers allow.
I solved it with a Home-made cache, the trick is to add a parameter to the URL that is not used, but making each URL different every X minutes
Calendar cal2 = Calendar.getInstance();
long d = cal2.getTimeInMillis();
int extra = (int) Math.ceil(d/ (10*60*1000)); // 10 minutes cache
Picasso.with(getBaseContext())
.load("http://www.myurl.cat/myimage.png&extra=" + extra)
.placeholder(R.drawable.graphicLoading)
.error(R.drawable.graphicLoadingError)
.into(bottomGraphic);
Before thinking about HTTP behavior, make sure you set a large max size for the disk cache:
cache = Cache(File(application.filesDir, "photos"), Long.MAX_VALUE)
(MAX_VALUE is not recommended for production.) Don't store the cache in application.cacheDir, because android can clear that whenever it wants.
Add an interceptor to set max-stale, which tells the disk cache to use all old files:
val httpClient = OkHttpClient.Builder().cache(cache).addInterceptor { chain ->
// When offline, we always want to show old photos.
val neverExpireCacheControl = CacheControl.Builder().maxStale(Int.MAX_VALUE, TimeUnit.SECONDS).build()
val origRequest = chain.request()
val neverExpireRequest = origRequest.newBuilder().cacheControl(neverExpireCacheControl).build()
chain.proceed(neverExpireRequest)
}.build()
return Picasso.Builder(application).downloader(OkHttp3Downloader(httpClient)).loggingEnabled(true).build()
I discovered this solution by debugging CacheStrategy.getCandidate(). Take a look there if this doesn't solve your problem.
In a normal Android web app the maximum size for a WebSQL database is normally around 8MB. In a hybrid web app I am making I would like to increase this limit. How would I go about doing that?
It seems that WebStorage might have something to do with it, but the only method I can see there that seems to set the size, setQuotaForOrigin, is marked deprecated.
Sample code (that is not deprecated) is welcome :)
The quota for a web app seems to differ from that of a hybrid app (as in something running within a view). Regardless, by implementing the following in your android.app.Activity subclass you will double the quota until it finally stops at approximately 48MB.
#Override
public void onExceededDatabaseQuota(String url, String databaseIdentifier, long currentQuota, long estimatedSize, long totalUsedQuota, WebStorage.QuotaUpdater quotaUpdater) {
quotaUpdater.updateQuota(estimatedSize * 2);
}
The user will not be asked to interact when this happens.
You may want to look at this other question on how to increase quota limit.
Looks like changing the size is only a problem for users that already have a local DB created, this is because there is no way to change the size of the DB when running upgrade scripts for versioning. That said, you just only need to change the size in the initialization script (code in JavaScript):
var size = 10 * 1024 * 1024; // changed from 5 to 10MB
var db = openDatabase("oversized_db", "v1.1", "Ten MB DB", size);
Take into account that this will only affect new users. All users that have previously created a DB with a different size need to clear their cache (you have controll of the cache on a webView so its not that bad) but all previous data would be lost (unless you manually migrate it to the new DB with native code).
In my case I have to do the same but decreasing the size. That would take care of the issue.
Hope this helps.
I developed an application that uses lots of images on Android.
The app runs once, fills the information on the screen (Layouts, Listviews, Textviews, ImageViews, etc) and user reads the information.
There is no animation, no special effects or anything that can fill the memory.
Sometimes the drawables can change. Some are android resources and some are files saved in a folder in the SDCARD.
Then the user quits (the onDestroy method is executed and app stays in memory by the VM ) and then at some point the user enters again.
Each time the user enters to the app, I can see the memory growing more and more until user gets the java.lang.OutOfMemoryError.
So what is the best/correct way to handle many images?
Should I put them in static methods so they are not loaded all the time?
Do I have to clean the layout or the images used in the layout in a special way?
One of the most common errors that I found developing Android Apps is the “java.lang.OutOfMemoryError: Bitmap Size Exceeds VM Budget” error. I found this error frequently on activities using lots of bitmaps after changing orientation: the Activity is destroyed, created again and the layouts are “inflated” from the XML consuming the VM memory available for bitmaps.
Bitmaps on the previous activity layout are not properly de-allocated by the garbage collector because they have crossed references to their activity. After many experiments I found a quite good solution for this problem.
First, set the “id” attribute on the parent view of your XML layout:
<?xml version="1.0" encoding="utf-8"?>
<RelativeLayout xmlns:android="http://schemas.android.com/apk/res/android"
android:layout_width="fill_parent"
android:layout_height="fill_parent"
android:id="#+id/RootView"
>
...
Then, on the onDestroy() method of your Activity, call the unbindDrawables() method passing a reference to the parent View and then do a System.gc().
#Override
protected void onDestroy() {
super.onDestroy();
unbindDrawables(findViewById(R.id.RootView));
System.gc();
}
private void unbindDrawables(View view) {
if (view.getBackground() != null) {
view.getBackground().setCallback(null);
}
if (view instanceof ViewGroup) {
for (int i = 0; i < ((ViewGroup) view).getChildCount(); i++) {
unbindDrawables(((ViewGroup) view).getChildAt(i));
}
((ViewGroup) view).removeAllViews();
}
}
This unbindDrawables() method explores the view tree recursively and:
Removes callbacks on all the background drawables
Removes children on every viewgroup
It sounds like you have a memory leak. The problem isn't handling many images, it's that your images aren't getting deallocated when your activity is destroyed.
It's difficult to say why this is without looking at your code. However, this article has some tips that might help:
http://android-developers.blogspot.de/2009/01/avoiding-memory-leaks.html
In particular, using static variables is likely to make things worse, not better. You might need to add code that removes callbacks when your application redraws -- but again, there's not enough information here to say for sure.
To avoid this problem you can use native method Bitmap.recycle() before null-ing Bitmap object (or setting another value). Example:
public final void setMyBitmap(Bitmap bitmap) {
if (this.myBitmap != null) {
this.myBitmap.recycle();
}
this.myBitmap = bitmap;
}
And next you can change myBitmap w/o calling System.gc() like:
setMyBitmap(null);
setMyBitmap(anotherBitmap);
I've ran into this exact problem. The heap is pretty small so these images can get out of control rather quickly in regards to memory. One way is to give the garbage collector a hint to collect memory on a bitmap by calling its recycle method.
Also, the onDestroy method is not guaranteed to get called. You may want to move this logic/clean up into the onPause activity. Check out the Activity Lifecycle diagram/table on this page for more info.
This explanation might help:
http://code.google.com/p/android/issues/detail?id=8488#c80
"Fast Tips:
1) NEVER call System.gc() yourself. This has been propagated as a fix here, and it doesn't work. Do not do it. If you noticed in my explanation, before getting an OutOfMemoryError, the JVM already runs a garbage collection so there is no reason to do one again (its slowing your program down). Doing one at the end of your activity is just covering up the problem. It may causes the bitmap to be put on the finalizer queue faster, but there is no reason you couldn't have simply called recycle on each bitmap instead.
2) Always call recycle() on bitmaps you don't need anymore. At the very least, in the onDestroy of your activity go through and recycle all the bitmaps you were using. Also, if you want the bitmap instances to be collected from the dalvik heap faster, it doesn't hurt to clear any references to the bitmap.
3) Calling recycle() and then System.gc() still might not remove the bitmap from the Dalvik heap. DO NOT BE CONCERNED about this. recycle() did its job and freed the native memory, it will just take some time to go through the steps I outlined earlier to actually remove the bitmap from the Dalvik heap. This is NOT a big deal because the large chunk of native memory is already free!
4) Always assume there is a bug in the framework last. Dalvik is doing exactly what its supposed to do. It may not be what you expect or what you want, but its how it works. "
I had the exact same problem. After a few testing I found that this error is appearing for large image scaling. I reduced the image scaling and the problem disappeared.
P.S. At first I tried to reduce the image size without scaling the image down. That did not stop the error.
Following points really helped me a lot. There might be other points too, but these are very crucial:
Use application context(instead of activity.this) where ever possible.
Stop and release your threads in onPause() method of activity
Release your views / callbacks in onDestroy() method of activity
I suggest a convenient way to solve this problem.
Just assign the attribute "android:configChanges" value as followed in the Mainfest.xml for your errored activity.
like this:
<activity android:name=".main.MainActivity"
android:label="mainActivity"
android:configChanges="orientation|keyboardHidden|navigation">
</activity>
the first solution I gave out had really reduced the frequency of OOM error to a low level. But, it did not solve the problem totally. And then I will give out the 2nd solution:
As the OOM detailed, I have used too much runtime memory. So, I reduce the picture size in ~/res/drawable of my project. Such as an overqualified picture which has a resolution of 128X128, could be resized to 64x64 which would also be suitable for my application. And after I did so with a pile of pictures, the OOM error doesn't occur again.
I too am frustrated by the outofmemory bug. And yes, I too found that this error pops up a lot when scaling images. At first I tried creating image sizes for all densities, but I found this substantially increased the size of my app. So I'm now just using one image for all densities and scaling my images.
My application would throw an outofmemory error whenever the user went from one activity to another. Setting my drawables to null and calling System.gc() didn't work, neither did recycling my bitmapDrawables with getBitMap().recycle(). Android would continue to throw the outofmemory error with the first approach, and it would throw a canvas error message whenever it tried using a recycled bitmap with the second approach.
I took an even third approach. I set all views to null and the background to black. I do this cleanup in my onStop() method. This is the method that gets called as soon as the activity is no longer visible. If you do it in the onPause() method, users will see a black background. Not ideal. As for doing it in the onDestroy() method, there is no guarantee that it will get called.
To prevent a black screen from occurring if the user presses the back button on the device, I reload the activity in the onRestart() method by calling the startActivity(getIntent()) and then finish() methods.
Note: it's not really necessary to change the background to black.
The BitmapFactory.decode* methods, discussed in the Load Large Bitmaps Efficiently lesson, should not be executed on the main UI thread if the source data is read from disk or a network location (or really any source other than memory). The time this data takes to load is unpredictable and depends on a variety of factors (speed of reading from disk or network, size of image, power of CPU, etc.). If one of these tasks blocks the UI thread, the system flags your application as non-responsive and the user has the option of closing it (see Designing for Responsiveness for more information).
Well I've tried everything I found on the internet and none of them worked. Calling System.gc() only drags down the speed of app. Recycling bitmaps in onDestroy didn't work for me too.
The only thing that works now is to have a static list of all the bitmap so that the bitmaps survive after a restart. And just use the saved bitmaps instead of creating new ones every time the activity if restarted.
In my case the code looks like this:
private static BitmapDrawable currentBGDrawable;
if (new File(uriString).exists()) {
if (!uriString.equals(currentBGUri)) {
freeBackground();
bg = BitmapFactory.decodeFile(uriString);
currentBGUri = uriString;
bgDrawable = new BitmapDrawable(bg);
currentBGDrawable = bgDrawable;
} else {
bgDrawable = currentBGDrawable;
}
}
I had the same problem just with switching the background images with reasonable sizes. I got better results with setting the ImageView to null before putting in a new picture.
ImageView ivBg = (ImageView) findViewById(R.id.main_backgroundImage);
ivBg.setImageDrawable(null);
ivBg.setImageDrawable(getResources().getDrawable(R.drawable.new_picture));
FWIW, here's a lightweight bitmap-cache I coded and have used for a few months. It's not all-the-bells-and-whistles, so read the code before you use it.
/**
* Lightweight cache for Bitmap objects.
*
* There is no thread-safety built into this class.
*
* Note: you may wish to create bitmaps using the application-context, rather than the activity-context.
* I believe the activity-context has a reference to the Activity object.
* So for as long as the bitmap exists, it will have an indirect link to the activity,
* and prevent the garbaage collector from disposing the activity object, leading to memory leaks.
*/
public class BitmapCache {
private Hashtable<String,ArrayList<Bitmap>> hashtable = new Hashtable<String, ArrayList<Bitmap>>();
private StringBuilder sb = new StringBuilder();
public BitmapCache() {
}
/**
* A Bitmap with the given width and height will be returned.
* It is removed from the cache.
*
* An attempt is made to return the correct config, but for unusual configs (as at 30may13) this might not happen.
*
* Note that thread-safety is the caller's responsibility.
*/
public Bitmap get(int width, int height, Bitmap.Config config) {
String key = getKey(width, height, config);
ArrayList<Bitmap> list = getList(key);
int listSize = list.size();
if (listSize>0) {
return list.remove(listSize-1);
} else {
try {
return Bitmap.createBitmap(width, height, config);
} catch (RuntimeException e) {
// TODO: Test appendHockeyApp() works.
App.appendHockeyApp("BitmapCache has "+hashtable.size()+":"+listSize+" request "+width+"x"+height);
throw e ;
}
}
}
/**
* Puts a Bitmap object into the cache.
*
* Note that thread-safety is the caller's responsibility.
*/
public void put(Bitmap bitmap) {
if (bitmap==null) return ;
String key = getKey(bitmap);
ArrayList<Bitmap> list = getList(key);
list.add(bitmap);
}
private ArrayList<Bitmap> getList(String key) {
ArrayList<Bitmap> list = hashtable.get(key);
if (list==null) {
list = new ArrayList<Bitmap>();
hashtable.put(key, list);
}
return list;
}
private String getKey(Bitmap bitmap) {
int width = bitmap.getWidth();
int height = bitmap.getHeight();
Config config = bitmap.getConfig();
return getKey(width, height, config);
}
private String getKey(int width, int height, Config config) {
sb.setLength(0);
sb.append(width);
sb.append("x");
sb.append(height);
sb.append(" ");
switch (config) {
case ALPHA_8:
sb.append("ALPHA_8");
break;
case ARGB_4444:
sb.append("ARGB_4444");
break;
case ARGB_8888:
sb.append("ARGB_8888");
break;
case RGB_565:
sb.append("RGB_565");
break;
default:
sb.append("unknown");
break;
}
return sb.toString();
}
}