Solved FutureTask.get() blocks with BukkitScheduler

Code:

Callable<String> task = new Callable<String>() {
    @Override
    public String call() {
        return "okay";
    }
};
FutureTask<String> ft = new FutureTask<String>(task);
Bukkit.getScheduler().runTask(this, ft);
try {
    System.out.println(ft.get());
} catch (InterruptedException e) {
    e.printStackTrace();
} catch (ExecutionException e) {
    e.printStackTrace();
}

I'm trying to schedule a FutureTask, but it blocks on get(), even if call() simply returns a string. When I try to schedule it without using the BukkitScheduler API, everything works properly. Any ideas how to solve this?

 

Have you tried the "done()" method yet to verify that it has completed yet? Since we're working with bukkit, everything is in ticks, so it would be slower than normal.

 

I don't need to check if it's done. I can wait for it to complete. And as You can see it simply returns a string, so it shouldn't block so long.

 

Yeah, but bukkit runs on ticks. Every 1/20th second (approx.) it will run tasks, and in your case it will start it after 1 tick meaning you don't have a return value yet.

 

There's a simple reason for why get() is blocking - the main thread is waiting for itself.

The scheduler will run a synchronous task (runTask) on the main thread during the same tick, but right after the current method finishes (onCommand(), onEnable(), a task, etc.). The problem here is that you're blocking the main thread with get(), which locks it in the current method, and the scheduler never gets a chance to run the task you scheduled. This is called a deadlock.

The only solution is to not call get() in the first place. If you need to process a value after it has been computed, you should schedule a new task inside current task instead. Also, if this task is blocking, remember to schedule it asynchronously:

Code:java

public class ExampleMod extends JavaPlugin {
    @Override
    public void onEnable() {
        getServer().getScheduler().runTaskAsynchronously(this, new Runnable() {
            @Override
            public void run() {
                final String value = "VALUE";
 
                // Get the value - for instance by retrieving it from an MySQL server
                try {
                    Thread.sleep(1000);
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
 
                getServer().getScheduler().runTask(ExampleMod.this, new Runnable() {
                    @Override
                    public void run() {
                        // Display it
                        getServer().broadcastMessage(value);
                    }
                });
            }
        });
    }
}

 

Comphenix, thank You. I didn't know the BukkitSchedulder works that way. Oh, and there's a note to callSynchMethod(), that get() shouldn't be called from the main server thread. I didn't notice it before, because I wanted to run this task asynchronously (I used runTask() instead of runTaskAsynchronously() in code from this thread, because it didn't work as well). Well, I'm going to have to schedule this task with the Executor, because I really need to call get().

 

You only need to call it if you absolutely must block the main thread. Otherwise, you can just as easily schedule a new task inside the first one, like in my previous example.

You really gain nothing by calling get() if all you're doing is scheduling a task on a single thread. You might as well do the work in the main thread in that case.

 

Comphenix, I'm trying to run some asynchronous computations early, so they don't block for long when I call get() in the main thread. I also store Future objects in a map, so they indicate running computations for their keys.