在Integer上同步会导致NullPointerException

Question

This question is based on Synchronizing on an Integer value .

The solution there seems excellent only there is small problem it does not address the concern how to delete values from ConcurrentHashMap .

So to address that I did below program

import java.util.concurrent.ConcurrentHashMap;

public class Example {

    private final ConcurrentHashMap<Integer, Integer> concurrentHashMap = new ConcurrentHashMap<Integer, Integer>();

    public void doSomething(int i) {
        synchronized (getLockForId(i)) {
            concurrentHashMap.remove(i);
        }
    }

    public Integer getLockForId(int id) {
        concurrentHashMap.putIfAbsent(id, id); // I want to replace these two
                                                // operation with single one
                                                // since it seems the cause of
                                                // NPE
        return concurrentHashMap.get(id);
    }

    public static void main(String[] args) {
        final Example example = new Example();
        new Thread(new Runnable() {
            @Override
            public void run() {
                int i = 0;
                while (true) {
                    example.doSomething(++i);
                }
            }
        }).start();
        new Thread(new Runnable() {
            @Override
            public void run() {
                int i = 0;
                while (true) {
                    example.doSomething(++i);
                }
            }
        }).start();
    }
}

Problem is that it always results in NullPointerException . My first analysis was because I am deleting the value it gets assigned to null so it is causing NullPointerException . So I did below

    Object obj = new Object();
    synchronized (obj) {
        obj = null;
    }

But above does not result in NullPointerException . So My question is why it is throwing NullPointerException in above case?

Even if I do

public Integer getLockForId(int id) {
   return concurrentHashMap.putIfAbsent(id, id); 
}

It still results in NullPointerException because it only returns value when there is one else return null

Answer 1

Well yes, that would throw a NullPointerException . Consider this pattern:

 Thread 1                     Thread 2

 putIfAbsent (puts)
 get (returns non-null)
 acquire monitor
                              putIfAbsent (doesn't put)
 remove (removes value)
                              get (returns null)
                              acquire monitor (bang!)

It's not that the "value get assigned to null" - it's that Map.get returns null if there's no entry for the given key.

It's hard to know what to recommend, as your code really doesn't do anything useful . If you can say what you're trying to achieve in your real code, we can give you better suggestions, potentially.

EDIT: As noted by Nikita, just returning the value of putIfAbsent doesn't work, as that returns the previous value, or null if it was absent - whereas you want the new value for the entry.

I suspect you'll have to synchronize access to the map, basically, to make your getLockId method atomic with respect to the remove operation.

Answer 2

You can try to make all accesses to concurrentHashMap synchronous. So when you get value from map you synchronize on concurrentHashMap and when you remove it. Something like this:

public void doSomething(int i) {
    synchronized (getLockForId(i)) {
        // do stuff
        synchronized (concurrentHashMap) {
            concurrentHashMap.remove(i);
        }
    }
}


public Integer getLockForId(int id) {
    synchronized (concurrentHashMap) {
        concurrentHashMap.putIfAbsent(id, id);
        return concurrentHashMap.get(id);
    }
}

Answer 3

Consider using google's LoadingCache instead. It uses weak references, so garbage collection is left to the JVM; you don't have to delete references you no longer need. And it handles the concurrency issues involved with creating a new entry. The code for a value synchronizer would look like this:

import com.google.common.cache.CacheBuilder;
import com.google.common.cache.CacheLoader;
import com.google.common.cache.LoadingCache;

import java.util.concurrent.locks.Lock;
import java.util.concurrent.locks.ReentrantLock;

public class ValueSynchronizer<T> {
  private final LoadingCache<T, Lock> valueLocks;

  public ValueSynchronizer() {
    valueLocks = CacheBuilder.newBuilder().build(
      new CacheLoader<T, Lock>() {
        public Lock load(final T id) {
          return new ReentrantLock();
        }
      });
  }

  public void sync(final T onValue, final Runnable toDo)
  {
    final Lock lock = valueLocks.getUnchecked(onValue);
    lock.lock();

    try {
      toDo.run();
    }
    finally {
      lock.unlock();
    }
  }
}

Actually syncing on a value would look like this:

private final ValueSynchronizer<Long> retrySynchronizer 
    = new ValueSynchronizer<>();

@Override
public void onApplicationEvent(final EventThatCanBeSentMultipleTimes event)
{
  retrySynchronizer.sync(event.getEventId(), () ->
  {
    //...Your code here.
  });
}

(licensed under Apache 2.0 license: http://www.apache.org/licenses/LICENSE-2.0 )