Can an Android AsyncTask doInBackground be synchronized to serialize the task execution?

Question

Is it possible to make AsyncTask.doInBackground synchronized - or achieve the same result in another way?

class SynchronizedTask extends AsyncTask {

    @Override
    protected synchronized Integer doInBackground(Object... params) {
        // do something that needs to be completed 
        // before another doInBackground can be called
    }
}

In my case, any AsyncTask.execute() can be started before a previous one has completed, but I need to execute the code in doInBackground only after the previous task has finished.

EDIT: As correctly pointed out, the synchronization works only on the same object instance. Unfortunately, it is not possible to create an AsyncTask and call execute() more than once on the same object instance, as specified in the "Threading rules" section of the AsyncTask documentation.

The solution is to use a custom Executor to serialize the tasks, or, if you use API 11 or above, AsyncTask.executeOnExecutor(), as suggested in the comments below.

I posted an answer showing an implementation of a SerialExecutor that can be used to queue tasks that will be executed sequentially.

Answer 1

Ideally, I'd like to be able to use AsyncTask.executeOnExecutor() with a SERIAL_EXECUTOR, but this is only available for API level 11 or above:

new AsyncTask().executeOnExecutor(AsyncTask.SERIAL_EXECUTOR, params);

To target the Android APIs below level 11, I ended up implementing a custom class which encapsulates an ExecutorService with a thread pool size of 1. The full code is open-sourced here.

Executors.newFixedThreadPool(int nThreads) creates a thread pool that reuses a fixed number of threads operating off a shared unbounded queue. At any point, at most nThreads threads will be active processing tasks. In my case, nThreads is 1, which means tasks can be queued, but only one task will be executed at any given time.

Here is the code:

public abstract class SerialExecutor {
    private final ExecutorService mExecutorService;

    public SerialExecutor() {
        mExecutorService = Executors.newFixedThreadPool(1);
    }

    public void queue(Context context, TaskParams params) {
        mExecutorService.submit(new SerialTask(context, params));
    }

    public void stop() {
        mExecutorService.shutdown();
    }

    public abstract void execute(TaskParams params);

    public static abstract class TaskParams { }

    private class SerialTask implements Runnable {
        private final Context mContext;
        private final TaskParams mParams;

        public SerialTask(Context context, TaskParams params) {
            mContext = context;
            mParams = params;
        }

        public void run() {
            execute(mParams);
            Activity a = (Activity) mContext;
            a.runOnUiThread(new OnPostExecute());
        }
    }

    /**
     * Used to notify the UI thread
     */
    private class OnPostExecute implements Runnable {

        public void run() {

        }
    }
}

This can be extended and used as a serial task executor in an Activity:

public class MyActivity extends Activity {
    private MySerialExecutor mSerialExecutor;

    @Override
    public void onCreate(Bundle savedInstanceState) {
        // ...
        mSerialExecutor = new MySerialExecutor();
    }

    @Override
    protected void onDestroy() {
        if (mSerialExecutor != null) {
            mSerialExecutor.stop();
        }
        super.onDestroy();
    }

    public void onTrigger(int param) {
        mSerialExecutor.queue(this, new MySerialExecutor.MyParams(param));
    }

    private static class MySerialExecutor extends SerialExecutor {

        public MySerialExecutor() {
            super();
        }

        @Override
        public void execute(TaskParams params) {
            MyParams myParams = (MyParams) params;
            // do something...
        }

        public static class MyParams extends TaskParams {
            // ... params definition

            public MyParams(int param) {
                // ... params init
            }
        }
    }
}

Answer 2

You may want to think about using IntentService instead. It seems like it may be a better fit for your process since it has built in features for queuing.