Synchronize Multiple Async Requests - java

Given an async service which should be closed upon completion. I would like to execute multiple instances of a requests. When all requests are done I want to close the service. I'm wondering what's the best way to achieve that. So here is the code demonstrating the problem but without actually closing the service:
class Service implements Closeable {
public Service() {/*...*/}
public ListenableFuture<Integer> processRequest(Integer param) {/*...*/}
#Override
public void close() {/*...*/}
}
public void proccessRequests(ArrayList<Integer> params) {
Service svc = new Service();
for (Integer param : params) {
final ListenableFuture<Integer> res = svc.processRequest(param);
}
}
I'm considering different options to close the service:
Use CountDownLatch this way:
public void processRequests(ArrayList<Integer> params) {
Service svc = new Service();
CountDownLatch latch = new CountDownLatch(params.size());
for (Integer param : params) {
final ListenableFuture<Integer> res = svc.processRequest(param);
Futures.addCallback(res, new FutureCallback<Integer>() {
#Override
public void onSuccess(Integer integer) {
latch.countDown();
if (latch.getCount() == 0) {
svc.close();
}
}
#Override
public void onFailure(Throwable throwable) {
latch.countDown();
if (latch.getCount() == 0) {
svc.close();
}
}
});
}
}
Use CountDownLatch this way:
public void processRequests(ArrayList<Integer> params) {
Service svc = new Service();
CountDownLatch latch = new CountDownLatch(params.size());
for (Integer param : params) {
final ListenableFuture<Integer> res = svc.processRequest(param);
Futures.addCallback(res, new FutureCallback<Integer>() {
#Override
public void onSuccess(Integer integer) {
latch.countDown();
}
#Override
public void onFailure(Throwable throwable) {
latch.countDown();
}
});
}
latch.await();
svc.close();
}
Similar to first option but using AtomicInteger.
What's the best way to achieve that? first, second, third, none of these?

Your first solution with CountDownLatch looks fine, but there are some other methods.
Starting from version 20.0 Futures class has a whenAllComplete method designed exactly for that purpose. Using it you can write:
Service svc = new Service();
Futures.whenAllComplete(params.stream().map(svc::processRequest).collect(Collectors.toList())).call(() -> {
svc.close();
return null;
}, ForkJoinPool.commonPool());
You can also use Java 8 CompletableFuture class that has a similar method allOf:
CompletableFuture.allOf(params.stream().map(svc::processRequest).toArray(CompletableFuture[]::new))
.thenAccept(v -> svc.close());
but in this case you have to make your Service return a CompletableFuture.

Related

The AsyncTask API is deprecated in Android 11. What are the alternatives?

Google is deprecating Android AsyncTask API in Android 11 and suggesting to use java.util.concurrent instead. you can check out the commit here
*
* #deprecated Use the standard <code>java.util.concurrent</code> or
* <a href="https://developer.android.com/topic/libraries/architecture/coroutines">
* Kotlin concurrency utilities</a> instead.
*/
#Deprecated
public abstract class AsyncTask<Params, Progress, Result> {
If you’re maintaining an older codebase with asynchronous tasks in Android, you’re likely going to have to change it in future. My question is that what should be proper replacement of the code snippet shown below using java.util.concurrent. It is a static inner class of an Activity. I am looking for something that will work with minSdkVersion 16
private static class LongRunningTask extends AsyncTask<String, Void, MyPojo> {
private static final String TAG = MyActivity.LongRunningTask.class.getSimpleName();
private WeakReference<MyActivity> activityReference;
LongRunningTask(MyActivity context) {
activityReference = new WeakReference<>(context);
}
#Override
protected MyPojo doInBackground(String... params) {
// Some long running task
}
#Override
protected void onPostExecute(MyPojo data) {
MyActivity activity = activityReference.get();
activity.progressBar.setVisibility(View.GONE);
populateData(activity, data) ;
}
}
You can directly use Executors from java.util.concurrent package.
I also searched about it and I found a solution in this Android Async API is Deprecated post.
Unfortunately, the post is using Kotlin, but after a little effort I have converted it into Java. So here is the solution.
ExecutorService executor = Executors.newSingleThreadExecutor();
Handler handler = new Handler(Looper.getMainLooper());
executor.execute(new Runnable() {
#Override
public void run() {
//Background work here
handler.post(new Runnable() {
#Override
public void run() {
//UI Thread work here
}
});
}
});
Pretty simple right? You can simplify it little more if you are using Java 8 in your project.
ExecutorService executor = Executors.newSingleThreadExecutor();
Handler handler = new Handler(Looper.getMainLooper());
executor.execute(() -> {
//Background work here
handler.post(() -> {
//UI Thread work here
});
});
Still, it cannot defeat kotlin terms of conciseness of the code, but better than the previous java version.
Hope this will help you. Thank You
private WeakReference<MyActivity> activityReference;
Good riddance that it's deprecated, because the WeakReference<Context> was always a hack, and not a proper solution.
Now people will have the opportunity to sanitize their code.
AsyncTask<String, Void, MyPojo>
Based on this code, Progress is actually not needed, and there is a String input + MyPojo output.
This is actually quite easy to accomplish without any use of AsyncTask.
public class TaskRunner {
private final Executor executor = Executors.newSingleThreadExecutor(); // change according to your requirements
private final Handler handler = new Handler(Looper.getMainLooper());
public interface Callback<R> {
void onComplete(R result);
}
public <R> void executeAsync(Callable<R> callable, Callback<R> callback) {
executor.execute(() -> {
final R result = callable.call();
handler.post(() -> {
callback.onComplete(result);
});
});
}
}
How to pass in the String? Like so:
class LongRunningTask implements Callable<MyPojo> {
private final String input;
public LongRunningTask(String input) {
this.input = input;
}
#Override
public MyPojo call() {
// Some long running task
return myPojo;
}
}
And
// in ViewModel
taskRunner.executeAsync(new LongRunningTask(input), (data) -> {
// MyActivity activity = activityReference.get();
// activity.progressBar.setVisibility(View.GONE);
// populateData(activity, data) ;
loadingLiveData.setValue(false);
dataLiveData.setValue(data);
});
// in Activity
#Override
protected void onCreate(Bundle savedInstanceState) {
super.onCreate(savedInstanceState);
setContentView(R.layout.main_activity);
viewModel = ViewModelProviders.of(this).get(MyViewModel.class);
viewModel.loadingLiveData.observe(this, (loading) -> {
if(loading) {
progressBar.setVisibility(View.VISIBLE);
} else {
progressBar.setVisibility(View.GONE);
}
});
viewModel.dataLiveData.observe(this, (data) -> {
populateData(data);
});
}
This example used a single-threaded pool which is good for DB writes (or serialized network requests), but if you want something for DB reads or multiple requests, you can consider the following Executor configuration:
private static final Executor THREAD_POOL_EXECUTOR =
new ThreadPoolExecutor(5, 128, 1,
TimeUnit.SECONDS, new LinkedBlockingQueue<Runnable>());
One of the simplest alternative is to use Thread
new Thread(new Runnable() {
#Override
public void run() {
// do your stuff
runOnUiThread(new Runnable() {
public void run() {
// do onPostExecute stuff
}
});
}
}).start();
If your project supports JAVA 8, you can use lambda:
new Thread(() -> {
// do background stuff here
runOnUiThread(()->{
// OnPostExecute stuff here
});
}).start();
According to the Android documentation AsyncTask was deprecated in API level 30 and it is suggested to use the standard java.util.concurrent or Kotlin concurrency utilities instead.
Using the latter it can be achieved pretty simple:
Create generic extension function on CoroutineScope:
fun <R> CoroutineScope.executeAsyncTask(
onPreExecute: () -> Unit,
doInBackground: () -> R,
onPostExecute: (R) -> Unit
) = launch {
onPreExecute() // runs in Main Thread
val result = withContext(Dispatchers.IO) {
doInBackground() // runs in background thread without blocking the Main Thread
}
onPostExecute(result) // runs in Main Thread
}
Use the function with any CoroutineScope which has Dispatchers.Main context:
In ViewModel:
class MyViewModel : ViewModel() {
fun someFun() {
viewModelScope.executeAsyncTask(onPreExecute = {
// ... runs in Main Thread
}, doInBackground = {
// ... runs in Worker(Background) Thread
"Result" // send data to "onPostExecute"
}, onPostExecute = {
// runs in Main Thread
// ... here "it" is the data returned from "doInBackground"
})
}
}
In Activity or Fragment:
lifecycleScope.executeAsyncTask(onPreExecute = {
// ... runs in Main Thread
}, doInBackground = {
// ... runs in Worker(Background) Thread
"Result" // send data to "onPostExecute"
}, onPostExecute = {
// runs in Main Thread
// ... here "it" is the data returned from "doInBackground"
})
To use viewModelScope or lifecycleScope add next line(s) to dependencies of the app's build.gradle file:
implementation "androidx.lifecycle:lifecycle-viewmodel-ktx:$LIFECYCLE_VERSION" // for viewModelScope
implementation "androidx.lifecycle:lifecycle-runtime-ktx:$LIFECYCLE_VERSION" // for lifecycleScope
At the time of writing final LIFECYCLE_VERSION = "2.3.0-alpha05"
UPDATE:
Also we can implement progress updating using onProgressUpdate function:
fun <P, R> CoroutineScope.executeAsyncTask(
onPreExecute: () -> Unit,
doInBackground: suspend (suspend (P) -> Unit) -> R,
onPostExecute: (R) -> Unit,
onProgressUpdate: (P) -> Unit
) = launch {
onPreExecute()
val result = withContext(Dispatchers.IO) {
doInBackground {
withContext(Dispatchers.Main) { onProgressUpdate(it) }
}
}
onPostExecute(result)
}
Using any CoroutineScope (viewModelScope/lifecycleScope, see implementations above) with Dispatchers.Main context we can call it:
someScope.executeAsyncTask(
onPreExecute = {
// ... runs in Main Thread
}, doInBackground = { publishProgress: suspend (progress: Int) -> Unit ->
// ... runs in Background Thread
// simulate progress update
publishProgress(50) // call `publishProgress` to update progress, `onProgressUpdate` will be called
delay(1000)
publishProgress(100)
"Result" // send data to "onPostExecute"
}, onPostExecute = {
// runs in Main Thread
// ... here "it" is a data returned from "doInBackground"
}, onProgressUpdate = {
// runs in Main Thread
// ... here "it" contains progress
}
)
Use this class to execute background task in Background Thread this class is work for all android API version include Android 11 also this code is same work like AsyncTask with doInBackground and onPostExecute methods
public abstract class BackgroundTask {
private Activity activity;
public BackgroundTask(Activity activity) {
this.activity = activity;
}
private void startBackground() {
new Thread(new Runnable() {
public void run() {
doInBackground();
activity.runOnUiThread(new Runnable() {
public void run() {
onPostExecute();
}
});
}
}).start();
}
public void execute(){
startBackground();
}
public abstract void doInBackground();
public abstract void onPostExecute();
}
After copying the above class, you can then use it with this:
new BackgroundTask(MainActivity.this) {
#Override
public void doInBackground() {
//put you background code
//same like doingBackground
//Background Thread
}
#Override
public void onPostExecute() {
//hear is result part same
//same like post execute
//UI Thread(update your UI widget)
}
}.execute();
Android deprecated AsyncTask API in Android 11 to get rid of a share of problems to begin with.
So, what's now?
Threads
Executers
RxJava
Listenable Futures
Coroutines 🔥
Why Coroutines?
Coroutines are the Kotlin way to do asynchronous programming. Compiler
support is stable since Kotlin 1.3, together with a
kotlinx.coroutines library -
Structured Concurrency
Non-blocking, sequential code
Cancellation propagation
Natural Exception Handling
Here I created a Alternative for AsyncTask using Coroutines which can be used same as AsyncTask without changing much code base in your project.
Create a new Abstract class AsyncTaskCoroutine which takes input parameter and output parameter datatypes of-course these parameters are optional :)
import kotlinx.coroutines.Dispatchers
import kotlinx.coroutines.GlobalScope
import kotlinx.coroutines.async
import kotlinx.coroutines.launch
abstract class AsyncTaskCoroutine<I, O> {
var result: O? = null
//private var result: O
open fun onPreExecute() {}
open fun onPostExecute(result: O?) {}
abstract fun doInBackground(vararg params: I): O
fun <T> execute(vararg input: I) {
GlobalScope.launch(Dispatchers.Main) {
onPreExecute()
callAsync(*input)
}
}
private suspend fun callAsync(vararg input: I) {
GlobalScope.async(Dispatchers.IO) {
result = doInBackground(*input)
}.await()
GlobalScope.launch(Dispatchers.Main) {
onPostExecute(result)
}
}
}
2 . Inside Activity use this as same as your old AsycnTask now
new AsyncTaskCoroutine() {
#Override
public Object doInBackground(Object[] params) {
return null;
}
#Override
public void onPostExecute(#Nullable Object result) {
}
#Override
public void onPreExecute() {
}
}.execute();
InCase if you need to send pass params
new AsyncTaskCoroutine<Integer, Boolean>() {
#Override
public Boolean doInBackground(Integer... params) {
return null;
}
#Override
public void onPostExecute(#Nullable Boolean result) {
}
#Override
public void onPreExecute() {
}
}.execute();
Google recommends using Java’s Concurrency framework or Kotlin Coroutines. but Rxjava end to have much more flexibility and features then java concurrency so gained quite a bit of popularity.
I actually wrote two Medium stories about it:
AsyncTas is deprecated now what
AsyncTas is deprecated now what part 2
The first one is with Java and a workaround with Runnable, the second is a Kotlin and coroutines solution.
Both are with code examples of course.
The accepted answer is good. But...
I didn't see cancel() method implementation
So my implementation with possibility to cancel the running task (simulating cancellation) is below.
Cancel is needed to not run postExecute() method in case of task interruption.
public abstract class AsyncTaskExecutor<Params> {
public static final String TAG = "AsyncTaskRunner";
private static final Executor THREAD_POOL_EXECUTOR =
new ThreadPoolExecutor(5, 128, 1,
TimeUnit.SECONDS, new LinkedBlockingQueue<Runnable>());
private final Handler mHandler = new Handler(Looper.getMainLooper());
private boolean mIsInterrupted = false;
protected void onPreExecute(){}
protected abstract Void doInBackground(Params... params);
protected void onPostExecute(){}
protected void onCancelled() {}
#SafeVarargs
public final void executeAsync(Params... params) {
THREAD_POOL_EXECUTOR.execute(() -> {
try {
checkInterrupted();
mHandler.post(this::onPreExecute);
checkInterrupted();
doInBackground(params);
checkInterrupted();
mHandler.post(this::onPostExecute);
} catch (InterruptedException ex) {
mHandler.post(this::onCancelled);
} catch (Exception ex) {
Log.e(TAG, "executeAsync: " + ex.getMessage() + "\n" + Debug.getStackTrace(ex));
}
});
}
private void checkInterrupted() throws InterruptedException {
if (isInterrupted()){
throw new InterruptedException();
}
}
public void cancel(boolean mayInterruptIfRunning){
setInterrupted(mayInterruptIfRunning);
}
public boolean isInterrupted() {
return mIsInterrupted;
}
public void setInterrupted(boolean interrupted) {
mIsInterrupted = interrupted;
}
}
Example of using this class:
public class MySearchTask extends AsyncTaskExecutor<String> {
public MySearchTask(){
}
#Override
protected Void doInBackground(String... params) {
// Your long running task
return null;
}
#Override
protected void onPostExecute() {
// update UI on task completed
}
#Override
protected void onCancelled() {
// update UI on task cancelled
}
}
MySearchTask searchTask = new MySearchTask();
searchTask.executeAsync("Test");
Here I also created an Alternative for AsyncTask using abstract class and it can be just copied as a class.
/app/src/main/java/../AsyncTasks.java
public abstract class AsyncTasks {
private final ExecutorService executors;
public AsyncTasks() {
this.executors = Executors.newSingleThreadExecutor();
}
private void startBackground() {
onPreExecute();
executors.execute(new Runnable() {
#Override
public void run() {
doInBackground();
new Handler(Looper.getMainLooper()).post(new Runnable() {
#Override
public void run() {
onPostExecute();
}
});
}
});
}
public void execute() {
startBackground();
}
public void shutdown() {
executors.shutdown();
}
public boolean isShutdown() {
return executors.isShutdown();
}
public abstract void onPreExecute();
public abstract void doInBackground();
public abstract void onPostExecute();
}
Implementation/ use of the above class
new AsyncTasks() {
#Override
public void onPreExecute() {
// before execution
}
#Override
public void doInBackground() {
// background task here
}
#Override
public void onPostExecute() {
// Ui task here
}
}.execute();
My custom replacement: https://github.com/JohnyDaDeveloper/AndroidAsync
It only works when the app is running (more specifically the activity which scheduled the task), but it's capable of updating the UI after the background task was completed
EDIT: My AsyncTask no longer reqires Activiy to function.
Just replace the whole class with this Thread and put it in a method to pass variables
new Thread(() -> {
// do background stuff here
runOnUiThread(()->{
// OnPostExecute stuff here
});
}).start();
and in Fragment add the Context to the runOnUiThread() methode:
new Thread(() -> {
// do background stuff here
context.runOnUiThread(()->{
// OnPostExecute stuff here
});
}).start();
You can use this custom class as an alternative of the AsyncTask<>, this is the same as AsyncTask so you not need to apply extra efforts for the same.
import android.os.Handler;
import android.os.Looper;
import androidx.annotation.NonNull;
import androidx.annotation.Nullable;
import java.util.concurrent.Callable;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.SynchronousQueue;
import java.util.concurrent.ThreadPoolExecutor;
import java.util.concurrent.TimeUnit;
public class TaskRunner {
private static final int CORE_THREADS = 3;
private static final long KEEP_ALIVE_SECONDS = 60L;
private static TaskRunner taskRunner = null;
private Handler handler = new Handler(Looper.getMainLooper());
private ThreadPoolExecutor executor;
private TaskRunner() {
executor = newThreadPoolExecutor();
}
public static TaskRunner getInstance() {
if (taskRunner == null) {
taskRunner = new TaskRunner();
}
return taskRunner;
}
public void shutdownService() {
if (executor != null) {
executor.shutdown();
}
}
public void execute(Runnable command) {
executor.execute(command);
}
public ExecutorService getExecutor() {
return executor;
}
public <R> void executeCallable(#NonNull Callable<R> callable, #NonNull OnCompletedCallback<R> callback) {
executor.execute(() -> {
R result = null;
try {
result = callable.call();
} catch (Exception e) {
e.printStackTrace(); // log this exception
} finally {
final R finalResult = result;
handler.post(() -> callback.onComplete(finalResult));
}
});
}
private ThreadPoolExecutor newThreadPoolExecutor() {
return new ThreadPoolExecutor(
CORE_THREADS,
Integer.MAX_VALUE,
KEEP_ALIVE_SECONDS,
TimeUnit.SECONDS,
new SynchronousQueue<>()
);
}
public interface OnCompletedCallback<R> {
void onComplete(#Nullable R result);
}
}
How to use it? Please follow the below examples.
With lambda expressions
TaskRunner.getInstance().executeCallable(() -> 1, result -> {
});
TaskRunner.getInstance().execute(() -> {
});
Without lambda expressions
TaskRunner.getInstance().executeCallable(new Callable<Integer>() {
#Override
public Integer call() throws Exception {
return 1;
}
}, new TaskRunner.OnCompletedCallback<Integer>() {
#Override
public void onComplete(#Nullable Integer result) {
}
});
TaskRunner.getInstance().execute(new Runnable() {
#Override
public void run() {
}
});
Note: Don't forget to shutdown executors service
TaskRunner.getInstance().shutdownService();
You can migrate to next approaches depends your needs
Thread + Handler
Executor
Future
IntentService
JobScheduler
RxJava
Coroutines (Kotlin)
[Android async variants]
My answer is similar to the others, but it is easier to read imo.
This is the class:
public class Async {
private static final ExecutorService executorService = Executors.newCachedThreadPool();
private static final Handler handler = new Handler(Looper.getMainLooper());
public static <T> void execute(Task<T> task) {
executorService.execute(() -> {
T t = task.doAsync();
handler.post(() -> {
task.doSync(t);
});
});
}
public interface Task<T> {
T doAsync();
void doSync(T t);
}
}
And here's an example on how to use it:
String url;
TextView responseCodeText;
Async.execute(new Async.Task<Integer>() {
#Override
public Integer doAsync() {
try {
HttpURLConnection connection = (HttpURLConnection) new URL(url).openConnection();
return connection.getResponseCode();
} catch (IOException e) {
return null;
}
}
#Override
public void doSync(Integer responseCode) {
responseCodeText.setText("responseCode=" + responseCode);
}
});
This is my code
import java.util.ArrayList;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import java.util.concurrent.Callable;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.Future;
public abstract class AsyncTaskRunner<T> {
private ExecutorService executorService = null;
private Set<Callable<T>> tasks = new HashSet<>();
public AsyncTaskRunner() {
this.executorService = Executors.newSingleThreadExecutor();
}
public AsyncTaskRunner(int threadNum) {
this.executorService = Executors.newFixedThreadPool(threadNum);
}
public void addTask(Callable<T> task) {
tasks.add(task);
}
public void execute() {
try {
List<Future<T>> features = executorService.invokeAll(tasks);
List<T> results = new ArrayList<>();
for (Future<T> feature : features) {
results.add(feature.get());
}
this.onPostExecute(results);
} catch (InterruptedException | ExecutionException e) {
e.printStackTrace();
this.onCancelled();
} finally {
executorService.shutdown();
}
}
protected abstract void onPostExecute(List<T> results);
protected void onCancelled() {
// stub
}
}
And usage example.
Extends the AsyncTaskRunner class,
class AsyncCalc extends AsyncTaskRunner<Integer> {
public void addRequest(final Integer int1, final Integer int2) {
this.addTask(new Callable<Integer>() {
#Override
public Integer call() throws Exception {
// Do something in background
return int1 + int2;
}
});
}
#Override
protected void onPostExecute(List<Integer> results) {
for (Integer answer: results) {
Log.d("AsyncCalc", answer.toString());
}
}
}
then use it!
AsyncCalc calc = new AsyncCalc();
calc.addRequest(1, 2);
calc.addRequest(2, 3);
calc.addRequest(3, 4);
calc.execute();
AsyncTask class does not seem to be removed any time soon, but we did simply un-deprecate it anyway, because:
We didn't want to add lots of suppress annotations.
The alternative solutions have too much boiler-plate, or in most cases, without any real advantage vs AsyncTask.
We did not want to re-invent the wheel.
We didn't want to fear the day it will finally be removed.
Refactoring takes too much time.
Example
Simply add below file to your project, then search for "android.os.AsyncTask" imports, and replase all to the packge you did choose for said file.
As you may already know, this is no big deal, and is basically what the well known AndroidX library does all the time.
Get AsyncTask.java file at: https://gist.github.com/top-master/0efddec3e2c35d77e30331e8c3bc725c
Docs says:
AsyncTask This class was deprecated in API level 30. Use the standard
java.util.concurrent or Kotlin concurrency utilities instead.
You need to use Handler or coroutines instead AsyncTask.
Use Handler for Java
new Handler(Looper.getMainLooper()).postDelayed(new Runnable() {
#Override
public void run() {
// Your Code
}
}, 3000);
Use Handler for Kotlin
Handler(Looper.getMainLooper()).postDelayed({
// Your Code
}, 3000)

Java, does the method with callback considered async?

I have created interface with one method:
public interface ResultCallback {
void onResult(String message);
}
And I have object with method that has interface as parameter:
public class Command() {
public void methodWithCallback(int param1, String param2, ResultCallback callback) {
// There are some calculations
callback.onResult(param2);
}
}
Then In my Main Java file I write this:
public class Main() {
public static void main(String[] args) {
Command c = new Command();
c.methodWithCallback(int 0, "SOME STRING", new ResultCallback() {
#Override
public void onResult(String str) {
// work with str
outsideMethod(str);
}
});
}
public void outsideMethod(String str) {
// some code
}
}
Does this code can be considered as async? And is it safe to call outsideMethod to handle params?
As said, it is not async. For the call to be async the method should execute on another thread.
You can't call outsideMethod since it's called from a static method. You need an instance of main to be able to call outsideMethod. If it is safe or not depends on what the code is doing.
One way to make it async is the following:
public class Command {
private ExecutorService iExecutor;
public Command(ExecutorService executor) {
iExecutor = executor;
}
public void methodWithCallback(final int param1, final String param2, final ResultCallback callback) {
iExecutor.execute(new Runnable() {
#Override
public void run() {
// There are some calculations
callback.onResult(param2);
}
});
}
}
You have to know what you're doing if using threads. Things have to be thread safe etc depending on how you are doing stuff. To run Command on a single thread create a single thread Executor and pass same Executor to all Commmand, like so:
ExecutorService executor = Executors.newSingleThreadExecutor();
Command command1 = new Command(executor);
Command command2 = new Command(executor);

Chain a series of asynchronous calls

I have a series of asynchronous operations
private void doSomething(){
get("/something", new Callback(){
void onComplete(String data){
updateUi(something, data);
doSomethingElse();
}
});
}
private void doSomethingElse(){
get("/something/else", new Callback(){
void onComplete(String data){
updateUi(somethingElse, data);
doYetAnotherThing();
}
});
}
private void doYetAnotherThing(){
get("/yet/another/thing", new Callback(){
void onComplete(String data){
updateUi(yetAnotherThing, data);
allDone();
}
});
}
This suffers from few problems:
Cannot reuse any of the callbacks elsewhere since each is intrinsically tied to the "next step"
Re-ordering operations or inserting another operation is non-intuitive and involves jumping all over the place.
I have looked at the following options to mitigate this:
ExecuterService#invokeAll - I don't see how this solution can be used without blocking.
RxJava - I would prefer to avoid such a paradigm shift in my application if I can!
Guava's ListenableFutures and its transform method. I saw this referred to in few places around the interwebs nut I honestly don't see how this would solve my problem.
So, the question is: What would be a good pattern to chain a series of asynchronous calls in Java? Looking for a solution that works with Java 7 since I need this for an Android app.
There certainly is some guessing involved, regarding the actual intention and use-case where you encountered this problem. Additionally, it is not entirely clear what something, somethingElse and yetAnotherThing are (where they come from and where they should go).
However, based on the information that you provided, and as an addition to (or rather extension or generalization of) the answer by slartidan: The difference between these dummy calls that you sketched there seem to be
The String argument that is passed to the get method
The Callback that is called
Which method is executed next
You could factor out these parts: The String argument and the Callback could be passed as parameters to a general method that creates a Callable. The sequence of the calls could simply be defined by placing these Callable objects into a list, in the appropriate order, and execute them all with a single threaded executor service.
As you can see in the main method of this example, the sequence of calls can then be configured rather easily:
import java.util.Arrays;
import java.util.concurrent.Callable;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
public class ChainedAsyncTest {
public static void main(String[] args) throws InterruptedException {
ChainedAsyncTest t = new ChainedAsyncTest();
ExecutorService e = Executors.newFixedThreadPool(1);
e.invokeAll(Arrays.asList(
t.call("/something", t.somethingCallback),
t.call("/something/else", t.somethingElseCallback),
t.call("/yet/another/thing", t.yetAnotherThingCallback),
t.allDone()));
}
private Callback somethingCallback = new Callback() {
#Override
public void onComplete(String data) {
updateUi("something", data);
}
};
private Callback somethingElseCallback = new Callback() {
#Override
public void onComplete(String data) {
updateUi("somethingElse", data);
}
};
private Callback yetAnotherThingCallback = new Callback() {
#Override
public void onComplete(String data) {
updateUi("yetAnotherThing", data);
}
};
private Callable<Void> call(
final String key, final Callback callback) {
return new Callable<Void>() {
#Override
public Void call() {
get(key, callback);
return null;
}
};
}
private Callable<Void> allDone() {
return new Callable<Void>() {
#Override
public Void call() {
System.out.println("allDone");
return null;
}
};
}
interface Callback
{
void onComplete(String data);
}
private void get(String string, Callback callback) {
System.out.println("Get "+string);
try {
Thread.sleep(500);
} catch (InterruptedException e) {
e.printStackTrace();
}
callback.onComplete("result of "+string);
}
private void updateUi(String string, String data) {
System.out.println("UpdateUI of "+string+" with "+data);
}
}
(The example uses invokeAll, which blocks until all tasks have been executed. This could be solved differently to be really non-blocking at the call site. The main idea is to create a list of the tasks, which are all created by the same method call)
Spontainious thought: You could define the chained calls as a method parameter to make your methods reusable. Here is my example code:
public class Scribble {
final Callback step1 = new Callback() {
void onComplete(String string) {
doSomethingElse(step2);
};
};
final Callback step2 = new Callback() {
void onComplete(String string) {
doYetAnotherThing(step3);
};
};
final Callback step3 = new Callback() {
void onComplete(String string) {
allDone();
}
};
private void start() {
doSomething(step1);
}
private void doSomething(final Callback externalCallback) {
get("/something", new Callback() {
void onComplete(String data) {
updateUi(something, data);
externalCallback.onComplete(data);
}
});
}
private void doSomethingElse(final Callback externalCallback) {
get("/something/else", new Callback() {
void onComplete(String data) {
updateUi(somethingElse, data);
externalCallback.onComplete(data);
}
});
}
private void doYetAnotherThing(final Callback externalCallback) {
get("/yet/another/thing", new Callback() {
void onComplete(String data) {
updateUi(yetAnotherThing, data);
externalCallback.onComplete(data);
}
});
}
// - the code below is only to make everything compilable -
public class Callback {
void onComplete(String string) {
}
}
private Object something;
protected Object somethingElse;
protected Object yetAnotherThing;
protected void allDone() {
System.out.println("Scribble.allDone()");
}
protected void updateUi(Object yetAnotherThing2, String data) {
System.out.println("Scribble.updateUi()"+data);
}
private void get(String string, Callback callback) {
System.out.println("get "+string);
callback.onComplete(string);
}
public static void main(String[] args) {
new Scribble().start();
}
}
I totally support the approved answer, but I'm also tossing in something I created for these types of problems that comes in handy when you start adding conditional logic within your chain of asynchronous actions. I recently fermented this into a simple library (jasync-driver).
Here is how you'd wire up your example. As you can see, each task has no knowledge of the task that follows. In contrast to the approved answer, the chaining of the tasks is done through a simple synchronous (...looking) method body instead of a list.
public void doChainedLogic() {
final AsyncTask<Void, Void> doSomething = new AsyncTask<Void, Void>() {
#Override
public void run(Void arg, final ResultHandler<Void> resultHandler) {
get("/something", new Callback() {
public void onComplete(String data) {
updateUi(something, data);
resultHandler.reportComplete();
}
});
}
};
final AsyncTask<Void, Void> doSomethingElse = new AsyncTask<Void, Void>() {
#Override
public void run(Void arg, final ResultHandler<Void> resultHandler) {
get("/something/else", new Callback() {
public void onComplete(String data) {
updateUi(somethingElse, data);
resultHandler.reportComplete();
}
});
}
};
final AsyncTask<Void, Void> doYetAnotherThing = new AsyncTask<Void, Void>() {
#Override
public void run(Void arg, final ResultHandler<Void> resultHandler) {
get("/yet/another/thing", new Callback() {
public void onComplete(String data) {
updateUi(yetAnotherThing, data);
resultHandler.reportComplete();
}
});
}
};
// This looks synchronous, but behind the scenes JasyncDriver is
// re-executing the body and skipping items already executed.
final JasyncDriver driver = new JasyncDriver();
driver.execute(new DriverBody() {
public void run() {
driver.execute(doSomething);
driver.execute(doSomethingElse);
driver.execute(doYetAnotherThing);
}
});
}
Now here's a tweak to the example that includes some conditional logic that depends upon an asynchronous result:
final AsyncTask<Void, String> checkSomething = new AsyncTask<Void, String>() {
#Override
public void run(Void arg, final ResultHandler<String> resultHandler) {
get("/check/something", new Callback() {
public void onComplete(String data) {
resultHandler.reportComplete(data);
}
});
}
};
final JasyncDriver driver = new JasyncDriver();
driver.execute(new DriverBody() {
public void run() {
driver.execute(doSomething);
if ("foobar".equals(driver.execute(checkSomething))) {
driver.execute(doSomethingElse);
}
driver.execute(doYetAnotherThing);
}
});
As you can see, asynchronous conditional logic is as simple as writing a standard if statement.

FutureTask with Callable that references the FutureTask

I have a situation where I need to create a FutureTask with a Callable that checks if it's owner has been cancelled. The code I have looks like this:
public static FutureTask<Result> makeFuture(final Call call, final TaskCompletionCallback completion) {
return new FutureTask<Result>(new Callable<Result>() {
#Override
public Result call() throws Exception {
Result result = CYLib.doNetworkRequest(call, new CarryOnCallback() {
#Override
public boolean shouldCarryOn() {
return !FutureTask.isDone();
}
});
return result;
}
});
}
Basically the doNetworkRequest asks the CarryOnCallback if it should continue at certain times during the operation. I would like for this callback to see if the FutureTask that is calling the doNetworkRequest was cancelled, which involves querying the actual FutureTask object.
Now I know that you can't really access 'this' because it hasn't been constructed yet. But is there a way around this, or a better design for my situation?
Cheers
EDIT:
Ok so I'm going about it like this now. Made a custom Callable and FutureTask. The Callable holds a reference to the FutureTask and this can be set manually after creating a new Callable:
public static MyTask makeMyTask(final Call call, final TaskCompletionCallback completion) {
MyTask task = null;
MyTask.InnerCallable innerCallable = new MyTask.InnerCallable(call, completion);
task = new MyTask(innerCallable);
innerCallable.setParent(task);
return task;
}
And just for reference, the InnerCallable looks like this:
public static class MyTask extends FutureTask<Result> {
InnerCallable callable;
public MyTask(InnerCallable callable) {
super(callable);
this.callable = callable;
}
private static class InnerCallable implements Callable<Result> {
private final Call call;
private final TaskCompletionCallback completion;
private WeakReference<MyTask> parentTask;
InnerCallable(Call call, TaskCompletionCallback completion) {
this.call = call;
this.completion = completion;
}
#Override
public Result call() {
Result result = CYLib.doNetworkRequest(this.call, new CarryOnCallback() {
#Override
public boolean shouldCarryOn() {
MyTask task = parentTask.get();
return !(task == null || task.isCancelled());
}
});
return result;
}
private void setParent(MyTask parentTask) {
this.parentTask = new WeakReference<MyTask>(parentTask);
}
}
}
So, your CYLib.doNetworkRequest() is working in another thread?
private static Map<Call,FutureTask> map=new HashMap();
public static FutureTask<Result> makeFuture(final Call call, final TaskCompletionCallback completion) {
FutureTask<Result> futureResult = new FutureTask<Result>(new Callable<Result>() {
#Override
public Result call() throws Exception {
Result result = CYLib.doNetworkRequest(call, new CarryOnCallback() {
#Override
public boolean shouldCarryOn() {
return !map.get(call).isCancelled();
}
});
return result;
}
});
map.put(call,futureResult);
return futureResult;
}

return String from a callback - Java

does anyone know how I can solve the following problem. I want to return a String from a callback, but I get only "The final local variable s cannot be assigned, since it is defined in an enclosing type", because of final.
public String getConstraint(int indexFdg) {
final String s;
AsyncCallback<String> callback = new AsyncCallback<String>() {
public void onFailure(Throwable caught) {
caught.printStackTrace();
}
public void onSuccess(String result) {
s = result;
}
};
SpeicherService.Util.getInstance().getConstraint(indexFdg, callback);
return s;
}
The whole point of an asynchronous callback is to notify you of something that happens asynchronously, at some time in the future. You can't return s from getConstraint if it's going to be set after the method has finished running.
When dealing with asynchronous callbacks you have to rethink the flow of your program. Instead of getConstraint returning a value, the code that would go on to use that value should be called as a result of the callback.
As a simple (incomplete) example, you would need to change this:
String s = getConstraint();
someGuiLabel.setText(s);
Into something like this:
myCallback = new AsyncCallback<String>() {
public void onSuccess(String result) {
someGuiLabel.setText(result);
}
}
fetchConstraintAsynchronously(myCallback);
Edit
A popular alternative is the concept of a future. A future is an object that you can return immediately but which will only have a value at some point in the future. It's a container where you only need to wait for the value at the point of asking for it.
You can think of holding a future as holding a ticket for your suit that is at the dry cleaning. You get the ticket immediately, can keep it in your wallet, give it to a friend... but as soon as you need to exchange it for the actual suit you need to wait until the suit is ready.
Java has such a class (Future<V>) that is used widely by the ExecutorService API.
An alternative workaround is to define a new class, called SyncResult
public class SyncResult {
private static final long TIMEOUT = 20000L;
private String result;
public String getResult() {
long startTimeMillis = System.currentTimeMillis();
while (result == null && System.currentTimeMillis() - startTimeMillis < TIMEOUT) {
synchronized (this) {
try {
wait(TIMEOUT);
} catch (Exception e) {
e.printStackTrace();
}
}
}
return result;
}
public void setResult(String result) {
this.result = result;
synchronized (this) {
notify();
}
}
}
Then change your code to this
public String getConstraint(int indexFdg) {
final SyncResult syncResult = new SyncResult();
AsyncCallback<String> callback = new AsyncCallback<String>() {
public void onFailure(Throwable caught) {
caught.printStackTrace();
}
public void onSuccess(String result) {
syncResult.setResult(result);
}
};
SpeicherService.Util.getInstance().getConstraint(indexFdg, callback);
return syncResult.getResult();
}
The getResult() method will be blocked until setResult(String) method been called or the TIMEOUT reached.

Categories

Resources