concurrenthashmap.java

Wicket一个开发Java Web应用程序框架。它使得开发web应用程序变得容易而轻松。 Wicket利用一个POJO data beans组件使得它可以与任何持久层技术相结合。

/*
  File: ConcurrentHashMap

  Written by Doug Lea. Adapted and released, under explicit
  permission, from JDK1.2 HashMap.java and Hashtable.java which
  carries the following copyright:

 * Copyright 1997 by Sun Microsystems, Inc.,
 * 901 San Antonio Road, Palo Alto, California, 94303, U.S.A.
 * All rights reserved.
 *
 * This software is the confidential and proprietary information
 * of Sun Microsystems, Inc. ("Confidential Information").  You
 * shall not disclose such Confidential Information and shall use
 * it only in accordance with the terms of the license agreement
 * you entered into with Sun.

  History:
  Date       Who                What
  26nov2000  dl               Created, based on ConcurrentReaderHashMap
  12jan2001  dl               public release
  17nov2001  dl               Minor tunings
  24oct2003  dl               Segment implements IClusterable
 */

package org.apache.wicket.util.concurrent;

import java.io.IOException;
import java.io.Serializable;
import java.util.AbstractCollection;
import java.util.AbstractMap;
import java.util.AbstractSet;
import java.util.Collection;
import java.util.Enumeration;
import java.util.Iterator;
import java.util.Map;
import java.util.NoSuchElementException;
import java.util.Set;

import org.apache.wicket.IClusterable;


/**
 * A version of Hashtable supporting concurrency for both retrievals and updates:
 * 
 * <dl>
 * <dt> Retrievals
 * 
 * <dd> Retrievals may overlap updates. (This is the same policy as ConcurrentReaderHashMap.)
 * Successful retrievals using get(key) and containsKey(key) usually run without locking.
 * Unsuccessful retrievals (i.e., when the key is not present) do involve brief synchronization
 * (locking). Because retrieval operations can ordinarily overlap with update operations (i.e., put,
 * remove, and their derivatives), retrievals can only be guaranteed to return the results of the
 * most recently <em>completed</em> operations holding upon their onset. Retrieval operations may
 * or may not return results reflecting in-progress writing operations. However, the retrieval
 * operations do always return consistent results -- either those holding before any single
 * modification or after it, but never a nonsense result. For aggregate operations such as putAll
 * and clear, concurrent reads may reflect insertion or removal of only some entries.
 * <p>
 * 
 * Iterators and Enumerations (i.e., those returned by keySet().iterator(), entrySet().iterator(),
 * values().iterator(), keys(), and elements()) return elements reflecting the state of the hash
 * table at some point at or since the creation of the iterator/enumeration. They will return at
 * most one instance of each element (via next()/nextElement()), but might or might not reflect puts
 * and removes that have been processed since they were created. They do <em>not</em> throw
 * ConcurrentModificationException. However, these iterators are designed to be used by only one
 * thread at a time. Passing an iterator across multiple threads may lead to unpredictable results
 * if the table is being concurrently modified.
 * <p>
 * 
 * 
 * <dt> Updates
 * 
 * <dd> This class supports a hard-wired preset <em>concurrency
 * level</em> of 32. This allows a
 * maximum of 32 put and/or remove operations to proceed concurrently. This level is an upper bound
 * on concurrency, not a guarantee, since it interacts with how well-strewn elements are across bins
 * of the table. (The preset value in part reflects the fact that even on large multiprocessors,
 * factors other than synchronization tend to be bottlenecks when more than 32 threads concurrently
 * attempt updates.) Additionally, operations triggering internal resizing and clearing do not
 * execute concurrently with any operation.
 * <p>
 * 
 * There is <em>NOT</em> any support for locking the entire table to prevent updates. This makes
 * it impossible, for example, to add an element only if it is not already present, since another
 * thread may be in the process of doing the same thing. If you need such capabilities, consider
 * instead using the ConcurrentReaderHashMap class.
 * 
 * </dl>
 * 
 * Because of how concurrency control is split up, the size() and isEmpty() methods require
 * accumulations across 32 control segments, and so might be slightly slower than you expect.
 * <p>
 * 
 * This class may be used as a direct replacement for java.util.Hashtable in any application that
 * does not rely on the ability to lock the entire table to prevent updates. As of this writing, it
 * performs much faster than Hashtable in typical multi-threaded applications with multiple readers
 * and writers. Like Hashtable but unlike java.util.HashMap, this class does NOT allow <tt>null</tt>
 * to be used as a key or value.
 * <p>
 * 
 * Implementation note: A slightly faster implementation of this class will be possible once planned
 * Java Memory Model revisions are in place.
 * 
 * <p>[<a href="http://gee.cs.oswego.edu/dl/classes/EDU/oswego/cs/dl/util/concurrent/intro.html">
 * Introduction to this package. </a>]
 * 
 */


public class ConcurrentHashMap extends AbstractMap implements Map, Cloneable, Serializable
{
	private static final long serialVersionUID = 1L;

	/*
	 * The basic strategy is an optimistic-style scheme based on the guarantee that the hash table
	 * and its lists are always kept in a consistent enough state to be read without locking:
	 * 
	 * Read operations first proceed without locking, by traversing the apparently correct list of
	 * the apparently correct bin. If an entry is found, but not invalidated (value field null), it
	 * is returned. If not found, operations must recheck (after a memory barrier) to make sure they
	 * are using both the right list and the right table (which can change under resizes). If
	 * invalidated, reads must acquire main update lock to wait out the update, and then
	 * re-traverse.
	 * 
	 * All list additions are at the front of each bin, making it easy to check changes, and also
	 * fast to traverse. Entry next pointers are never assigned. Remove() builds new nodes when
	 * necessary to preserve this.
	 * 
	 * Remove() (also clear()) invalidates removed nodes to alert read operations that they must
	 * wait out the full modifications.
	 * 
	 * Locking for puts, removes (and, when necessary gets, etc) is controlled by Segments, each
	 * covering a portion of the table. During operations requiring global exclusivity (mainly
	 * resize and clear), ALL of these locks are acquired at once. Note that these segments are NOT
	 * contiguous -- they are based on the least 5 bits of hashcodes. This ensures that the same
	 * segment controls the same slots before and after resizing, which is necessary for supporting
	 * concurrent retrievals. This comes at the price of a mismatch of logical vs physical locality,
	 * but this seems not to be a performance problem in practice.
	 * 
	 */

	/**
	 * The hash table data.
	 */
	protected transient Entry[] table;

	/**
	 * The number of concurrency control segments. The value can be at most 32 since ints are used
	 * as bitsets over segments. Empirically, it doesn't seem to pay to decrease it either, so the
	 * value should be at least 32. In other words, do not redefine this :-)
	 */
	protected static final int CONCURRENCY_LEVEL = 32;

	/**
	 * Mask value for indexing into segments
	 */
	protected static final int SEGMENT_MASK = CONCURRENCY_LEVEL - 1;

	/**
	 * Bookkeeping for each concurrency control segment. Each segment contains a local count of the
	 * number of elements in its region. However, the main use of a Segment is for its lock.
	 */
	protected final static class Segment implements IClusterable
	{
		private static final long serialVersionUID = 1L;

		/**
		 * The number of elements in this segment's region. It is always updated within synchronized
		 * blocks.
		 */
		protected int count;

		/**
		 * Get the count under synch.
		 * 
		 * @return count under sync
		 */
		protected synchronized int getCount()
		{
			return count;
		}

		/**
		 * Force a synchronization
		 */
		protected synchronized void synch()
		{
		}
	}

	/**
	 * The array of concurrency control segments.
	 */

	protected final Segment[] segments = new Segment[CONCURRENCY_LEVEL];

	/**
	 * The default initial number of table slots for this table (32). Used when not otherwise
	 * specified in constructor.
	 */
	public static final int DEFAULT_INITIAL_CAPACITY = 32;

	/**
	 * The minimum capacity, used if a lower value is implicitly specified by either of the
	 * constructors with arguments. MUST be a power of two.
	 */
	private static final int MINIMUM_CAPACITY = 32;

	/**
	 * The maximum capacity, used if a higher value is implicitly specified by either of the
	 * constructors with arguments. MUST be a power of two <= 1<<30.
	 */
	private static final int MAXIMUM_CAPACITY = 1 << 30;

	/**
	 * The default load factor for this table (0.75) Used when not otherwise specified in
	 * constructor.
	 */
	public static final float DEFAULT_LOAD_FACTOR = 0.75f;

	/**
	 * The load factor for the hash table.
	 * 
	 * @serial
	 */
	protected final float loadFactor;

	/**
	 * Per-segment resize threshold.
	 * 
	 * @serial
	 */
	protected int threshold;

	/**
	 * Number of segments voting for resize. The table is doubled when 1/4 of the segments reach
	 * threshold. Volatile but updated without synch since this is just a heuristic.
	 */
	protected transient volatile int votesForResize;

	/**
	 * Return the number of set bits in w. For a derivation of this algorithm, see "Algorithms and
	 * data structures with applications to graphics and geometry", by Jurg Nievergelt and Klaus
	 * Hinrichs, Prentice Hall, 1993. See also notes by Torsten Sillke at
	 * http://www.mathematik.uni-bielefeld.de/~sillke/PROBLEMS/bitcount
	 * 
	 * @param w
	 *            arg
	 * @return number of set bits
	 */
	protected static int bitcount(int w)
	{
		w -= (0xaaaaaaaa & w) >>> 1;
		w = (w & 0x33333333) + ((w >>> 2) & 0x33333333);
		w = (w + (w >>> 4)) & 0x0f0f0f0f;
		w += w >>> 8;
		w += w >>> 16;
		return w & 0xff;
	}

	/**
	 * Returns the appropriate capacity (power of two) for the specified initial capacity argument.
	 * 
	 * @param initialCapacity
	 *            the initial capacity
	 * @return appropriate capacity
	 */
	private int p2capacity(int initialCapacity)
	{
		int cap = initialCapacity;

		// Compute the appropriate capacity
		int result;
		if (cap > MAXIMUM_CAPACITY || cap < 0)
		{
			result = MAXIMUM_CAPACITY;
		}
		else
		{
			result = MINIMUM_CAPACITY;
			while (result < cap)
			{
				result <<= 1;
			}
		}
		return result;
	}

	/**
	 * Return hash code for Object x. Since we are using power-of-two tables, it is worth the effort
	 * to improve hashcode via the same multiplicative scheme as used in IdentityHashMap.
	 * 
	 * @param x
	 * @return hash code
	 */
	protected static int hash(Object x)
	{
		int h = x.hashCode();
		// Multiply by 127 (quickly, via shifts), and mix in some high
		// bits to help guard against bunching of codes that are
		// consecutive or equally spaced.
		return ((h << 7) - h + (h >>> 9) + (h >>> 17));
	}

	/**
	 * Check for equality of non-null references x and y.
	 * 
	 * @param x
	 *            ref
	 * @param y
	 *            ref
	 * @return is equal
	 */
	protected boolean eq(Object x, Object y)
	{
		return x == y || x.equals(y);
	}

	/**
	 * Create table array and set the per-segment threshold *
	 * 
	 * @param capacity
	 * @return table array
	 */
	protected Entry[] newTable(int capacity)
	{
		threshold = (int)(capacity * loadFactor / CONCURRENCY_LEVEL) + 1;
		return new Entry[capacity];
	}

	/**
	 * Constructs a new, empty map with the specified initial capacity and the specified load
	 * factor.
	 * 
	 * @param initialCapacity
	 *            the initial capacity. The actual initial capacity is rounded to the nearest power
	 *            of two.
	 * @param loadFactor
	 *            the load factor threshold, used to control resizing. This value is used in an
	 *            approximate way: When at least a quarter of the segments of the table reach
	 *            per-segment threshold, or one of the segments itself exceeds overall threshold,
	 *            the table is doubled. This will on average cause resizing when the table-wide load
	 *            factor is slightly less than the threshold. If you'd like to avoid resizing, you
	 *            can set this to a ridiculously large value.
	 * @throws IllegalArgumentException
	 *             if the load factor is nonpositive.
	 */
	public ConcurrentHashMap(int initialCapacity, float loadFactor)
	{
		if (!(loadFactor > 0))
		{
			throw new IllegalArgumentException("Illegal Load factor: " + loadFactor);
		}
		this.loadFactor = loadFactor;
		for (int i = 0; i < segments.length; ++i)
		{
			segments[i] = new Segment();
		}
		int cap = p2capacity(initialCapacity);
		table = newTable(cap);
	}

	/**
	 * Constructs a new, empty map with the specified initial capacity and default load factor.
	 * 
	 * @param initialCapacity
	 *            the initial capacity of the ConcurrentHashMap.
	 * @throws IllegalArgumentException
	 *             if the initial maximum number of elements is less than zero.
	 */
	public ConcurrentHashMap(int initialCapacity)
	{
		this(initialCapacity, DEFAULT_LOAD_FACTOR);
	}

	/**
	 * Constructs a new, empty map with a default initial capacity and default load factor.
	 */
	public ConcurrentHashMap()
	{
		this(DEFAULT_INITIAL_CAPACITY, DEFAULT_LOAD_FACTOR);
	}

	/**
	 * Constructs a new map with the same mappings as the given map. The map is created with a
	 * capacity of twice the number of mappings in the given map or 32 (whichever is greater), and a
	 * default load factor.
	 * 
	 * @param t
	 *            map to copy
	 */
	public ConcurrentHashMap(Map t)
	{
		this(Math.max((int)(t.size() / DEFAULT_LOAD_FACTOR) + 1, MINIMUM_CAPACITY),
				DEFAULT_LOAD_FACTOR);
		putAll(t);
	}

	/**
	 * Returns the number of key-value mappings in this map.
	 * 
	 * @return the number of key-value mappings in this map.
	 */
	public int size()
	{
		int c = 0;
		for (int i = 0; i < segments.length; ++i)
		{
			c += segments[i].getCount();
		}
		return c;
	}

	/**
	 * Returns <tt>true</tt> if this map contains no key-value mappings.
	 * 
	 * @return <tt>true</tt> if this map contains no key-value mappings.
	 */
	public boolean isEmpty()
	{
		for (int i = 0; i < segments.length; ++i)
		{
			if (segments[i].getCount() != 0)
			{
				return false;
			}
		}
		return true;
	}

	/**
	 * Returns the value to which the specified key is mapped in this table.
	 * 
	 * @param key
	 *            a key in the table.
	 * @return the value to which the key is mapped in this table; <code>null</code> if the key is
	 *         not mapped to any value in this table.
	 * @exception NullPointerException
	 *                if the key is <code>null</code>.
	 * @see #put(Object, Object)
	 */
	public Object get(Object key)
	{
		int hash = hash(key); // throws null pointer exception if key null

		// Try first without locking...
		Entry[] tab = table;
		int index = hash & (tab.length - 1);
		Entry first = tab[index];
		Entry e;

		for (e = first; e != null; e = e.next)
		{
			if (e.hash == hash && eq(key, e.key))
			{
				Object value = e.value;
				if (value != null)
				{
					return value;
				}
				else
				{
					break;
				}
			}
		}

		// Recheck under synch if key apparently not there or interference
		Segment seg = segments[hash & SEGMENT_MASK];
		synchronized (seg)
		{
			tab = table;
			index = hash & (tab.length - 1);
			Entry newFirst = tab[index];
			if (e != null || first != newFirst)
			{
				for (e = newFirst; e != null; e = e.next)
				{
					if (e.hash == hash && eq(key, e.key))
					{
						return e.value;
					}
				}