Is there a fixed sized queue which removes excessive elements?

Actually the LinkedHashMap does exactly what you want. You need to override the removeEldestEntry method.

Example for a queue with max 10 elements:

  queue = new LinkedHashMap<Integer, String>()
  {
     @Override
     protected boolean removeEldestEntry(Map.Entry<Integer, String> eldest)
     {
        return this.size() > 10;   
     }
  };

If the "removeEldestEntry" returns true, the eldest entry is removed from the map.

There is no existing implementation in the Java Language and Runtime. All Queues extend AbstractQueue, and its doc clearly states that adding an element to a full queue always ends with an exception. It would be best ( and quite simple ) to wrap a Queue into a class of your own for having the functionality you need.

Once again, because all queues are children of AbstractQueue, simply use that as your internal data type and you should have a flexible implementation running in virtually no time :-)

UPDATE:

As outlined below, there are two open implementations available (this answer is quite old, folks!), see this answer for details.

public class CircularQueue<E> extends LinkedList<E> {
    private int capacity = 10;

    public CircularQueue(int capacity){
        this.capacity = capacity;
    }

    @Override
    public boolean add(E e) {
        if(size() >= capacity)
            removeFirst();
        return super.add(e);
    }
}

Usage and test result:

public static void main(String[] args) {
    CircularQueue<String> queue = new CircularQueue<>(3);
    queue.add("a");
    queue.add("b");
    queue.add("c");
    System.out.println(queue.toString());   //[a, b, c]

    String first = queue.pollFirst();       //a
    System.out.println(queue.toString());   //[b,c]

    queue.add("d");
    queue.add("e");
    queue.add("f");
    System.out.println(queue.toString());   //[d, e, f]
}

I think what you're describing is a circular queue. Here is an example and here is a better one

Sounds like an ordinary List where the add method contains an extra snippet which truncates the list if it gets too long.

If that is too simple, then you probably need to edit your problem description.

Also see this SO question, or ArrayBlockingQueue (be careful about blocking, this might be unwanted in your case).

It is not quite clear what requirements you have that led you to ask this question. If you need a fixed size data structure, you might also want to look at different caching policies. However, since you have a queue, my best guess is that you're looking for some type of router functionality. In that case, I would go with a ring buffer: an array that has a first and last index. Whenever an element is added, you just increment the last element index, and when an element is removed, increment the first element index. In both cases, addition is performed modulo the array size, and make sure to increment the other index when needed, that is, when the queue is full or empty.

Also, if it is a router-type application, you might also want to experiment with an algorithm such as Random Early Dropping (RED), which drops elements from the queue randomly even before it gets filled up. In some cases, RED has been found to have better overall performance than the simple method of allowing the queue to fill up before dropping.

A Simple solution, below is a Queue of "String"

LinkedHashMap<Integer, String> queue;
int queueKeysCounter;

queue.put(queueKeysCounter++, "My String");
queueKeysCounter %= QUEUE_SIZE;

Note that this will not maintain the Order of the items in the Queue, but it will replace the oldest entry.

As it's advised in OOPs that we should prefer Composition over Inheritance

Here my solution keeping that in mind.

package com.choiceview;

import java.util.ArrayDeque;

class Ideone {
    public static void main(String[] args) {
        LimitedArrayDeque<Integer> q = new LimitedArrayDeque<>(3);
        q.add(1);
        q.add(2);
        q.add(3);
        System.out.println(q);

        q.add(4);
        // First entry ie 1 got pushed out
        System.out.println(q);
    }
}

class LimitedArrayDeque<T> {

    private int maxSize;
    private ArrayDeque<T> queue;

    private LimitedArrayDeque() {

    }

    public LimitedArrayDeque(int maxSize) {
        this.maxSize = maxSize;
        queue = new ArrayDeque<T>(maxSize);
    }

    public void add(T t) {
        if (queue.size() == maxSize) {
            queue.removeFirst();
        }
        queue.add(t);
    }

    public boolean remove(T t) {
        return queue.remove(t);
    }

    public boolean contains(T t) {
        return queue.contains(t);
    }

    @Override
    public String toString() {
        return queue.toString();
    }
}

Ok, I'll throw out my version too. :-) This is build to be very performant - for when that matters. It's not based on LinkedList - and is thread safe (should be at least). FIFO

static class FixedSizeCircularReference<T> {
    T[] entries

    FixedSizeCircularReference(int size) {
        this.entries = new Object[size] as T[]
        this.size = size
    }
    int cur = 0
    int size

    synchronized void add(T entry) {
        entries[cur++] = entry
        if (cur >= size) {
            cur = 0
        }
    }

    List<T> asList() {
        int c = cur
        int s = size
        T[] e = entries.collect() as T[]
        List<T> list = new ArrayList<>()
        int oldest = (c == s - 1) ? 0 : c
        for (int i = 0; i < e.length; i++) {
            def entry = e[oldest + i < s ? oldest + i : oldest + i - s]
            if (entry) list.add(entry)
        }
        return list
    }
}