Why there is no ConcurrentHashSet against ConcurrentHashMap

HashSet is based on HashMap.

If we look at HashSet<E> implementation, everything is been managed under HashMap<E,Object>.

<E> is used as a key of HashMap.

And we know that HashMap is not thread safe. That is why we have ConcurrentHashMap in Java.

Based on this, I am confused that why we don't have a ConcurrentHashSet which should be based on the ConcurrentHashMap?

Is there anything else that I am missing? I need to use Set in a multi-threaded environment.

Also, If I want to create my own ConcurrentHashSet can I achieve it by just replacing the HashMap to ConcurrentHashMap and leaving the rest as is?

There's no built in type for ConcurrentHashSet because you can always derive a set from a map. Since there are many types of maps, you use a method to produce a set from a given map (or map class).

Prior to Java 8, you produce a concurrent hash set backed by a concurrent hash map, by using Collections.newSetFromMap(map)

In Java 8 (pointed out by @Matt), you can get a concurrent hash set view via ConcurrentHashMap.newKeySet(). This is a bit simpler than the old newSetFromMap which required you to pass in an empty map object. But it is specific to ConcurrentHashMap.

Anyway, the Java designers could have created a new set interface every time a new map interface was created, but that pattern would be impossible to enforce when third parties create their own maps. It is better to have the static methods that derive new sets; that approach always works, even when you create your own map implementations.

Set<String> mySet = Collections.newSetFromMap(new ConcurrentHashMap<String, Boolean>());

With Guava 15 you can also simply use:

Set s = Sets.newConcurrentHashSet();

Like Ray Toal mentioned it is as easy as:

Set<String> myConcurrentSet = ConcurrentHashMap.newKeySet();

It looks like Java provides a concurrent Set implementation with its ConcurrentSkipListSet. A SkipList Set is just a special kind of set implementation. It still implements the Serializable, Cloneable, Iterable, Collection, NavigableSet, Set, SortedSet interfaces. This might work for you if you only need the Set interface.

As pointed by this the best way to obtain a concurrency-able HashSet is by means of Collections.synchronizedSet()

Set s = Collections.synchronizedSet(new HashSet(...));

This worked for me and I haven't seen anybody really pointing to it.

EDIT This is less efficient than the currently aproved solution, as Eugene points out, since it just wraps your set into a synchronized decorator, while a ConcurrentHashMap actually implements low-level concurrency and it can back your Set just as fine. So thanks to Mr. Stepanenkov for making that clear.

http://docs.oracle.com/javase/8/docs/api/java/util/Collections.html#synchronizedSet-java.util.Set-

You can use guava's Sets.newSetFromMap(map) to get one. Java 6 also has that method in java.util.Collections

import java.util.AbstractSet;
import java.util.Iterator;
import java.util.Set;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.ConcurrentMap;


public class ConcurrentHashSet<E> extends AbstractSet<E> implements Set<E>{
   private final ConcurrentMap<E, Object> theMap;

   private static final Object dummy = new Object();

   public ConcurrentHashSet(){
      theMap = new ConcurrentHashMap<E, Object>();
   }

   @Override
   public int size() {
      return theMap.size();
   }

   @Override
   public Iterator<E> iterator(){
      return theMap.keySet().iterator();
   }

   @Override
   public boolean isEmpty(){
      return theMap.isEmpty();
   }

   @Override
   public boolean add(final E o){
      return theMap.put(o, ConcurrentHashSet.dummy) == null;
   }

   @Override
   public boolean contains(final Object o){
      return theMap.containsKey(o);
   }

   @Override
   public void clear(){
      theMap.clear();
   }

   @Override
   public boolean remove(final Object o){
      return theMap.remove(o) == ConcurrentHashSet.dummy;
   }

   public boolean addIfAbsent(final E o){
      Object obj = theMap.putIfAbsent(o, ConcurrentHashSet.dummy);
      return obj == null;
   }
}

Why not use: CopyOnWriteArraySet from java.util.concurrent?