简而言之，我正在编写一个需要BlockingQueue实现的应用程序，它既提供FIFO添加/删除，也提供快速contains方法，因为我将其称为TON
LinkedBlockingQueue让我大部分都在那里，但似乎它的contains方法在线性时间内运行，因为它基于AbstractQueue的{{1}}方法。我没有看到Java API中的任何内容似乎宣传了具有快速contains开箱即用的LBQ。
让事情变得更加艰难的是，我的项目处于非常严峻的时间紧迫状态（不，这不是功课）。我可以做一个快速而又脏的LBQ扩展，下面有一个contains用于快速HashSet，但我仍然需要测试它，这可能会耗费大量的工时。我想知道是否有任何受信任/经过良好测试的库提供contains扩展，其中LinkedBlockingQueue方法在O（1）时间运行...？如果没有，欢迎任何其他建议。
答案 0 :(得分：10)
再次感谢所有提供意见的人。我最终编写了自己的HashedLinkedBlockingQueue实现代码。通过使用装饰器模式获得正确的同步证明比预期困难得多。向装饰器添加同步会导致在高负载下发生死锁（特别是，put(E element)和take()引入了先前不存在的保持和等待条件。当由于不必要的同步而导致性能下降时，很明显我需要花时间从头开始实现它。
add / remove / contains中的效果为O（1），但原始LinkedBlockingQueue的同步成本大约是其两倍 - 我说“大概加倍”因为LBQ使用两个锁 - 一个用于插入，一个用于在队列大小足够大以允许并发修改头部和尾部时删除。我的实现使用单个锁，因此remove必须等待添加完成，反之亦然。这是该类的源代码 - 我已经在多线程和单线程模式下测试了每个方法，但除了在我自己的应用程序中使用这个类之外，我还没有提出一个超级复杂的通用测试。使用风险自负！仅仅因为它在我的应用程序中有效并不意味着仍然没有错误：
import java.util.Arrays;
import java.util.Collection;
import java.util.HashMap;
import java.util.Iterator;
import java.util.LinkedList;
import java.util.NoSuchElementException;
import java.util.concurrent.BlockingQueue;
import java.util.concurrent.LinkedBlockingQueue;
import java.util.concurrent.TimeUnit;
/**
* Provides a single-lock queuing algorithm with fast contains(Object o) and
* remove(Object o), at the expense of higher synchronization cost when
* compared to {@link LinkedBlockingQueue}. This queue implementation does not
* allow for duplicate entries.
*
* <p>Use of this particular {@link BlockingQueue} implementation is encouraged
* when the cost of calling
* <code>{@link BlockingQueue#contains(Object o)}</code> or
* <code>{@link BlockingQueue#remove(Object o)}</code> outweighs the throughput
* benefit if using a {@link LinkedBlockingQueue}. This queue performs best
* when few threads require simultaneous access to it.
*
* <p>The basic operations this queue provides and their associated run times
* are as follows, where <i>n</i> is the number of elements in this queue and
* <i>m</i> is the number of elements in the specified collection, if any such
* collection is specified:
*
* <ul>
* <li><b>add(E element)</b> - <i>O(1)</i></li>
* <li><b>addAll(Collection<? extends E> c)</b> - <i>O(m)</i></li>
* <li><b>drainTo(Collection<? extends E> c, int maxElements)</b>
* - <i>O(maxElements*O(</i><code>c.add(Object o)</code><i>))</i></li>
* <li><b>contains(E element)</b> - <i>O(1)</i></li>
* <li><b>offer(E element)</b> - <i>O(1)</i></li>
* <li><b>poll()</b> - <i>O(1)</i></li>
* <li><b>remove(E element)</b> - <i>O(1)</i></li>
* <li><b>removeAll(Collection<? extends E> c)</b> - <i>O(m)</i></li>
* <li><b>retainAll(Collection<? extends E> c)</b> - <i>O(n*O(
* </i><code>c.contains(Object o)</code><i>))</i></li>
* </ul>
*
* @param <E> type of element this queue will handle. It is strongly
* recommended that the underlying element overrides <code>hashCode()</code>
* and <code>equals(Object o)</code> in an efficient manner.
*
* @author CodeBlind
*/
@SuppressWarnings("unused")
public class HashedLinkedBlockingQueue<E> implements BlockingQueue<E>{
/** Polling removes the head, offering adds to the tail. */
private Node head, tail;
/** Required for constant-time lookups and removals. */
private HashMap<E,Node> contents;
/** Allows the user to artificially limit the capacity of this queue. */
private final int maxCapacity;
//Constructors: -----------------------------------------------------------
/**
* Creates an empty queue with max capacity equal to
* {@link Integer#MAX_VALUE}.
*/
public HashedLinkedBlockingQueue(){ this(null,Integer.MAX_VALUE); }
/**
* Creates an empty queue with max capacity equals to the specified value.
* @param capacity (1 to {@link Integer#MAX_VALUE})
*/
public HashedLinkedBlockingQueue(int capacity){
this(null,Math.max(1,capacity));
}
/**
* Creates a new queue and initializes it to the contents of the specified
* collection, queued in the order returned by its iterator, with a max
* capacity of {@link Integer#MAX_VALUE}.
* @param c collection of elements to add
*/
public HashedLinkedBlockingQueue(Collection<? extends E> c){
this(c,Integer.MAX_VALUE);
}
/**
* Creates a new queue and initializes it to the contents of the specified
* collection, queued in the order returned by its iterator, with a max
* capacity equal to the specified value.
* @param c collection of elements to add
* @param capacity (1 to {@link Integer#MAX_VALUE})
*/
public HashedLinkedBlockingQueue(Collection<? extends E> c, int capacity){
maxCapacity = capacity;
contents = new HashMap<E,Node>();
if(c == null || c.isEmpty()){
head = null;
tail = null;
}
else for(E e : c) enqueue(e);
}
//Private helper methods: -------------------------------------------------
private E dequeue(){
if(contents.isEmpty()) return null;
Node n = head;
contents.remove(n.element);
if(contents.isEmpty()){
head = null;
tail = null;
}
else{
head.next.prev = null;
head = head.next;
n.next = null;
}
return n.element;
}
private void enqueue(E e){
if(contents.containsKey(e)) return;
Node n = new Node(e);
if(contents.isEmpty()){
head = n;
tail = n;
}
else{
tail.next = n;
n.prev = tail;
tail = n;
}
contents.put(e,n);
}
private void removeNode(Node n, boolean notify){
if(n == null) return;
if(n == head) dequeue();
else if(n == tail){
tail.prev.next = null;
tail = tail.prev;
n.prev = null;
}
else{
n.prev.next = n.next;
n.next.prev = n.prev;
n.prev = null;
n.next = null;
}
contents.remove(n.element);
if(notify) synchronized(this){ contents.notifyAll(); }
}
//Public instance methods: ------------------------------------------------
public void print(){
Node n = head;
int i = 1;
while(n != null){
System.out.println(i+": "+n);
n = n.next;
i++;
}
}
//Overridden methods: -----------------------------------------------------
@Override
public boolean add(E e){
synchronized(this){
if(remainingCapacity() < 1) throw new IllegalStateException();
enqueue(e);
contents.notifyAll();
}
return true;
}
@Override
public boolean addAll(Collection<? extends E> c){
boolean changed = true;
synchronized(this){
for(E e : c){
if(remainingCapacity() < 1) throw new IllegalStateException();
enqueue(e);
}
contents.notifyAll();
}
return changed;
}
@Override
public void clear(){
synchronized(this){
if(isEmpty()) return;
head = null;
tail = null;
contents.clear();
contents.notifyAll();
}
}
@Override
public boolean contains(Object o){
synchronized(this){ return contents.containsKey(o); }
}
@Override
public boolean containsAll(Collection<?> c) {
synchronized(this){
for(Object o : c) if(!contents.containsKey(o)) return false;
}
return true;
}
@Override
public int drainTo(Collection<? super E> c) {
return drainTo(c,maxCapacity);
}
@Override
public int drainTo(Collection<? super E> c, int maxElements) {
if(this == c) throw new IllegalArgumentException();
int transferred = 0;
synchronized(this){
while(!isEmpty() && transferred < maxElements)
if(c.add(dequeue())) transferred++;
if(transferred > 0) contents.notifyAll();
}
return transferred;
}
@Override
public E element(){
E e = peek();
if(e == null) throw new IllegalStateException();
return e;
}
@Override
public boolean isEmpty() {
synchronized(this){ return contents.isEmpty(); }
}
@Override
public Iterator<E> iterator(){ return new Itr(); }
@Override
public boolean offer(E e){
synchronized(this){
if(contents.containsKey(e)) return false;
enqueue(e);
contents.notifyAll();
}
return true;
}
@Override
public boolean offer(E e, long timeout, TimeUnit unit)
throws InterruptedException{
long remainingSleep = -1;
long millis = unit.toMillis(timeout);
long methodCalled = System.currentTimeMillis();
synchronized(this){
while((remainingSleep =
(methodCalled+millis)-System.currentTimeMillis()) > 0 &&
(remainingCapacity() < 1 || contents.containsKey(e))){
contents.wait(remainingSleep);
}
if(remainingSleep < 1) return false;
enqueue(e);
contents.notifyAll();
}
return true;
}
@Override
public E peek(){
synchronized(this){ return (head != null) ? head.element : null; }
}
@Override
public E poll(){
synchronized(this){
E e = dequeue();
if(e != null) contents.notifyAll();
return e;
}
}
@Override
public E poll(long timeout, TimeUnit unit) throws InterruptedException{
E e = null;
long remainingSleep = -1;
long millis = unit.toMillis(timeout);
long methodCalled = System.currentTimeMillis();
synchronized(this){
e = dequeue();
while(e == null && (remainingSleep = (methodCalled+millis)-
System.currentTimeMillis()) > 0){
contents.wait(remainingSleep);
e = dequeue();
}
if(e == null) e = dequeue();
if(e != null) contents.notifyAll();
}
return e;
}
@Override
public void put(E e) throws InterruptedException{
synchronized(this){
while(remainingCapacity() < 1) contents.wait();
enqueue(e);
contents.notifyAll();
}
}
@Override
public int remainingCapacity(){ return maxCapacity-size(); }
@Override
public E remove(){
E e = poll();
if(e == null) throw new IllegalStateException();
return e;
}
@Override
public boolean remove(Object o){
synchronized(this){
Node n = contents.get(o);
if(n == null) return false;
removeNode(n,true);
}
return true;
}
@Override
public boolean removeAll(Collection<?> c){
if(this == c){
synchronized(this){
if(isEmpty()){
clear();
return true;
}
}
return false;
}
boolean changed = false;
synchronized(this){
for(Object o : c){
Node n = contents.get(o);
if(n == null) continue;
removeNode(n,false);
changed = true;
}
if(changed) contents.notifyAll();
}
return changed;
}
@Override
public boolean retainAll(Collection<?> c){
boolean changed = false;
if(this == c) return changed;
synchronized(this){
for(E e : new LinkedList<E>(contents.keySet())){
if(!c.contains(e)){
Node n = contents.get(e);
if(n != null){
removeNode(n,false);
changed = true;
}
}
}
if(changed) contents.notifyAll();
}
return changed;
}
@Override
public int size(){ synchronized(this){ return contents.size(); }}
@Override
public E take() throws InterruptedException{
synchronized(this){
while(contents.isEmpty()) contents.wait();
return dequeue();
}
}
@Override
public Object[] toArray(){
synchronized(this){ return toArray(new Object[size()]); }
}
@SuppressWarnings("unchecked")
@Override
public <T> T[] toArray(T[] a) {
synchronized(this){
//Estimate size of array; be prepared to see more or fewer elements
int size = size();
T[] r = a.length >= size ? a :
(T[])java.lang.reflect.Array
.newInstance(a.getClass().getComponentType(), size);
Iterator<E> it = iterator();
for (int i = 0; i < r.length; i++) {
if (! it.hasNext()) { // fewer elements than expected
if (a != r)
return Arrays.copyOf(r, i);
r[i] = null; // null-terminate
return r;
}
r[i] = (T)it.next();
}
return it.hasNext() ? finishToArray(r, it) : r;
}
}
//Static helper methods: --------------------------------------------------
@SuppressWarnings("unchecked")
private static <T> T[] finishToArray(T[] r, Iterator<?> it) {
int i = r.length;
while (it.hasNext()) {
int cap = r.length;
if (i == cap) {
int newCap = ((cap / 2) + 1) * 3;
if (newCap <= cap) { // integer overflow
if (cap == Integer.MAX_VALUE)
throw new OutOfMemoryError
("Required array size too large");
newCap = Integer.MAX_VALUE;
}
r = Arrays.copyOf(r, newCap);
}
r[i++] = (T)it.next();
}
// trim if overallocated
return (i == r.length) ? r : Arrays.copyOf(r, i);
}
//Private inner classes: --------------------------------------------------
/**
* Provides a weak iterator that doesn't check for concurrent modification
* but also fails elegantly. A race condition exists when simultaneously
* iterating over the queue while the queue is being modified, but this is
* allowable per the Java specification for Iterators.
* @author CodeBlind
*/
private class Itr implements Iterator<E>{
private Node current;
private E currentElement;
private Itr(){
synchronized(HashedLinkedBlockingQueue.this){
current = head;
if(current != null) currentElement = current.element;
else currentElement = null;
}
}
@Override
public boolean hasNext(){
return currentElement != null;
}
@Override
public E next(){
if(currentElement == null) throw new NoSuchElementException();
synchronized(HashedLinkedBlockingQueue.this){
E e = currentElement;
current = current.next;
if(current == null || !contents.containsKey(current.element)){
current = null;
currentElement = null;
}
else currentElement = current.element;
return e;
}
}
@Override
public void remove(){
synchronized(HashedLinkedBlockingQueue.this){
if(current == null || !contents.containsKey(current.element))
throw new NoSuchElementException();
Node n = current;
current = current.next;
if(current != null && contents.containsKey(current.element))
currentElement = current.element;
else currentElement = null;
removeNode(n,true);
}
}
}
/**
* This class provides a simple implementation for a node in a double-
* linked list. It supports constant-time, in-place removals.
* @author CodeBlind
*/
private class Node{
private Node(E e){
element = e;
prev = null;
next = null;
}
private E element;
private Node prev, next;
@Override
public String toString(){
StringBuilder sb = new StringBuilder("Node[prev.element=");
if(prev == null) sb.append("null,element=");
else sb.append(prev.element+",element=");
sb.append(element+",next.element=");
if(next == null) sb.append("null]");
else sb.append(next.element+"]");
return sb.toString();
}
}
}
答案 1 :(得分：3)
正如其他人已经说过的那样，使用基于散列的结构就是解决方案。 我没有在Java Collection API上看到任何实现。
但是你可以使用Guava类ForwardingBlockingQueue轻松创建一个HashBlockingQueue装饰器