如何在ArrayBlockingQueue中使用Predicate方法“removeIf”

Question

I have the following classes:

WorkerTask.java

   public interface WorkerTask extends Task {

   // Constants
   public static final short WORKERTASK_SPIDER = 1;
   public static final short WORKERTASK_PARSER = 2;
   public static final short WORKERTASK_PRODUCT = 3;

   public int getType();
}

WorkerPool.java

class workerPool {

     private ThreadPoolExecutor executorPool_;

     //----------------------------------------------------  

     public WorkerPool(int poolSize) 
     {
        executorPool_ = new ThreadPoolExecutor(
           poolSize,5,10,TimeUnit.SECONDS,
           new ArrayBlockingQueue<Runnable>(10000000,false),
           Executors.defaultThreadFactory()
     );

     //----------------------------------------------------        

     public void assign(WorkerTask workerTask) {
         executorPool_.execute(new WorkerThread(workerTask));
     }

     //----------------------------------------------------  

     public void removeTasks(int siteID) {
        executorPool_.getQueue().removeIf(...);     
     }
}

I want to call the method removeTasks to remove certain amount of pending tasks but I have no idea of how to use the method removeIf. It says: Removes all of the elements of this collection that satisfy the given predicate, but I have no idea how to create the parameter Predicate. Any idea?

Answer 1

If you had a Queue<WorkerTask> , you could do something like this:

queue.removeIf(task -> task.getSiteID() == siteID)

There are several problems. One problem is that the queue you get from getQueue() is BlockingQueue<Runnable> and not Queue<WorkerTask> . If you're submitting Runnable instances to the pool, the queue might contain references to your actual tasks; if so, you could downcast them to WorkerTask . However, this isn't guaranteed. Furthermore, the class doc for ThreadPoolExecutor says (under "Queue maintenance"):

Method getQueue() allows access to the work queue for purposes of monitoring and debugging. Use of this method for any other purpose is strongly discouraged. Two supplied methods, remove(Runnable) and purge() are available to assist in storage reclamation when large numbers of queued tasks become cancelled.

Looking at the remove(Runnable) method, its doc says

It may fail to remove tasks that have been converted into other forms before being placed on the internal queue.

This suggests that you should hang onto the Runnable instances that have been submitted in order to call remove() on them later. Or, call submit(Runnable) to get a Future and save those instances around in order to cancel them.

But there is also a second problem that probably renders this approach inadequate. Suppose you've found a way to remove or cancel the matching tasks from the queue. Another thread might have decided to submit a new task that matches, but hasn't submitted it yet. There's a race condition here. You might be able to cancel the enqueued tasks, but after you've done so, you can't guarantee that new matching tasks haven't been submitted.

Here's an alternative approach. Presumably, when you cancel (or whatever) a site ID, there's some logic somewhere to stop submitting new tasks that match that side ID. The problem is how to deal with matching tasks that are "in-flight," that is, that are in the queue or are about to be enqueued.

Instead of trying to cancel the matching tasks, change the task so that if its site ID has been canceled, the task turns into a no-op. You could record the cancellation of a site ID in, say, a ConcurrentHashMap . Any task would check this map before beginning its work, and if the site ID is present, it'd simply return. Adding a site ID to the map would have the immediate effect of ensuring that no new task on that site ID will commence. (Tasks that have already started will run to completion.) Any in-flight tasks will eventually drain from the queue without causing any actual work to occur.

Answer 2

A predicate is a function that receives an input and returns a boolean value.

If you are using java 8 you can use lambda expressions: (elem) -> return elem.id == siteID