Monday, 3 December 2012

Java Cafe 3 : WeakReferences and the caveats of WeakHashMap

WeakReference

First of all, lets see what a WeakReference is. A normal reference to an object is a strong reference, e.g.:

List list = new LinkedList();
list.add(1)

list is a strong reference to a LinkedList object. It prevents the garbage-collector from removing the LinkedList from memory, as long as list is in the current scope of executing Java program. Now, you can also wrap it around a WeakReference:

WeakReference<List<Integer>> list = new WeakReference<List<Integer>>(new LinkedList<Integer>());

List myList = list.get();
myList.add(1);

You now have a WeakReference (list) to a LinkedList, and you can obtain the LinkedList using the WeakReference.get() method. The purpose of a WeakReference is to allow the garbage-collector to gc the referred object if necessary. Consider the code below:

WeakReference<List<Integer>> list = new WeakReference<List<Integer>>(new LinkedList());
list.get().add(1);

int i = 0;
while (i < 10) {
    System.gc();
    i++;
    if (list.get() == null) {
        System.out.println("GC-ed at iteration " + i);
        break;
    }
}

On my JVM, it prints out:

GC-ed at iteration 1

That is, the LinkedList has been garbage-collected after the first System.gc() call (note that this does not necessarily have to happen)! While this behaviour is not guaranteed, we have certainly shown that WeakReferences allows the referred object to be garbage-collected.

WeakHashMap

The java.util.WeakHashMap class is one of the least frequently used in the Collections package, and it uses WeakReferences internally. It is meant to serve as a caching map where the entries can be garbage-collected if the JVM is running low on memory. It accomplishes this by weak referencing the keys in the map, also called weak keys. For instance, if you have a WeakHashMap which maps Strings to Integers:

String key = new String("one");
Integer value = new Integer(1);

WeakHashMap<String,Integer> weakMap = new WeakHashMap<String,Integer>();
weakMap.put(key, value);

key = null;
value = null;

If there are no other references to key, then map entry (key, value) can (but non-necessarily) be garbage-collected the next time GC happens. If that happens, the size of the WeakHashMap will reduce from 1 to 0, and the object key will be garbage-collected:

String key = new String("one");
Integer value = new Integer(1);

WeakHashMap<String,Integer> weakMap = new WeakHashMap<String,Integer>();
weakMap.put(key, value);

key = null;

int i = 0;
while (i < 100) {
    System.gc();
    i++;
    if (weakMap.size() == 0) {
        System.out.println("GC-ed at iteration " + i);
        break;
    }
}

In my JVM, weakMap becomes empty at iteration 6. If there are no other references to the object value, it will be garbage-collected too.

One important fact to know about the WeakHashMap is that the value objects are held using strong references. Therefore, if you had mistakenly wrote:

WeakHashMap<String,String> weakMap = new WeakHashMap<String,String>();
weakMap.put(key, key);

The entry for key will never be garbage-collected, because key is also held as the value object using a strong reference. If you need to use the same object as the value, it is recommended that you wrap it around a WeakReference, like:

weakMap.put(key, new WeakReference(key));

This way, there will be no strong references to key, allowing it to be garbage-collected.