ViewVC Help

View File | Revision Log | Show Annotations | Download File | Root Listing

root/jsr166/jsr166/src/main/java/util/concurrent/ConcurrentHashMap.java

Comparing jsr166/src/main/java/util/concurrent/ConcurrentHashMap.java (file contents):
Revision 1.23 by dl, Sat Sep 13 18:51:10 2003 UTC vs.
Revision 1.110 by jsr166, Wed Apr 27 14:06:30 2011 UTC

#	Line 1 | Line 1
1		/*
2		* Written by Doug Lea with assistance from members of JCP JSR-166
3	<	* Expert Group and released to the public domain. Use, modify, and
4	<	* redistribute this code in any way without acknowledgement.
3	>	* Expert Group and released to the public domain, as explained at
4	>	* http://creativecommons.org/publicdomain/zero/1.0/
5		*/
6
7		package java.util.concurrent;
#	Line 17 | Line 17 | import java.io.ObjectOutputStream;
17		* adjustable expected concurrency for updates. This class obeys the
18		* same functional specification as {@link java.util.Hashtable}, and
19		* includes versions of methods corresponding to each method of
20	<	* <tt>Hashtable</tt> . However, even though all operations are
20	>	* <tt>Hashtable</tt>. However, even though all operations are
21		* thread-safe, retrieval operations do <em>not</em> entail locking,
22		* and there is <em>not</em> any support for locking the entire table
23		* in a way that prevents all access.  This class is fully
24		* interoperable with <tt>Hashtable</tt> in programs that rely on its
25		* thread safety but not on its synchronization details.
26		*
27	<	* <p> Retrieval operations (including <tt>get</tt>) ordinarily
28	<	* overlap with update operations (including <tt>put</tt> and
29	<	* <tt>remove</tt>). Retrievals reflect the results of the most
30	<	* recently <em>completed</em> update operations holding upon their
31	<	* onset.  For aggregate operations such as <tt>putAll</tt> and
32	<	* <tt>clear</tt>, concurrent retrievals may reflect insertion or
27	>	* <p> Retrieval operations (including <tt>get</tt>) generally do not
28	>	* block, so may overlap with update operations (including
29	>	* <tt>put</tt> and <tt>remove</tt>). Retrievals reflect the results
30	>	* of the most recently <em>completed</em> update operations holding
31	>	* upon their onset.  For aggregate operations such as <tt>putAll</tt>
32	>	* and <tt>clear</tt>, concurrent retrievals may reflect insertion or
33		* removal of only some entries.  Similarly, Iterators and
34		* Enumerations return elements reflecting the state of the hash table
35		* at some point at or since the creation of the iterator/enumeration.
36	<	* They do <em>not</em> throw <tt>ConcurrentModificationException</tt>.
37	<	* However, Iterators are designed to be used by only one thread at a
38	<	* time.
36	>	* They do <em>not</em> throw {@link ConcurrentModificationException}.
37	>	* However, iterators are designed to be used by only one thread at a time.
38		*
39		* <p> The allowed concurrency among update operations is guided by
40		* the optional <tt>concurrencyLevel</tt> constructor argument
41	<	* (default 16), which is used as a hint for internal sizing.  The
41	>	* (default <tt>16</tt>), which is used as a hint for internal sizing.  The
42		* table is internally partitioned to try to permit the indicated
43		* number of concurrent updates without contention. Because placement
44		* in hash tables is essentially random, the actual concurrency will
45		* vary.  Ideally, you should choose a value to accommodate as many
46	<	* threads as will ever concurrently access the table. Using a
46	>	* threads as will ever concurrently modify the table. Using a
47		* significantly higher value than you need can waste space and time,
48		* and a significantly lower value can lead to thread contention. But
49		* overestimates and underestimates within an order of magnitude do
50	<	* not usually have much noticeable impact.
50	>	* not usually have much noticeable impact. A value of one is
51	>	* appropriate when it is known that only one thread will modify and
52	>	* all others will only read. Also, resizing this or any other kind of
53	>	* hash table is a relatively slow operation, so, when possible, it is
54	>	* a good idea to provide estimates of expected table sizes in
55	>	* constructors.
56		*
57	<	* <p>This class implements all of the <em>optional</em> methods
58	<	* of the {@link Map} and {@link Iterator} interfaces.
57	>	* <p>This class and its views and iterators implement all of the
58	>	* <em>optional</em> methods of the {@link Map} and {@link Iterator}
59	>	* interfaces.
60		*
61	<	* <p> Like {@link java.util.Hashtable} but unlike {@link
62	<	* java.util.HashMap}, this class does NOT allow <tt>null</tt> to be
63	<	* used as a key or value.
61	>	* <p> Like {@link Hashtable} but unlike {@link HashMap}, this class
62	>	* does <em>not</em> allow <tt>null</tt> to be used as a key or value.
63	>	*
64	>	* <p>This class is a member of the
65	>	* <a href="{@docRoot}/../technotes/guides/collections/index.html">
66	>	* Java Collections Framework</a>.
67		*
68		* @since 1.5
69		* @author Doug Lea
70	+	* @param <K> the type of keys maintained by this map
71	+	* @param <V> the type of mapped values
72		*/
73		public class ConcurrentHashMap<K, V> extends AbstractMap<K, V>
74	<	implements ConcurrentMap<K, V>, Cloneable, Serializable {
74	>	implements ConcurrentMap<K, V>, Serializable {
75		private static final long serialVersionUID = 7249069246763182397L;
76
77		/*
78		* The basic strategy is to subdivide the table among Segments,
79	<	* each of which itself is a concurrently readable hash table.
79	>	* each of which itself is a concurrently readable hash table.  To
80	>	* reduce footprint, all but one segments are constructed only
81	>	* when first needed (see ensureSegment). To maintain visibility
82	>	* in the presence of lazy construction, accesses to segments as
83	>	* well as elements of segment's table must use volatile access,
84	>	* which is done via Unsafe within methods segmentAt etc
85	>	* below. These provide the functionality of AtomicReferenceArrays
86	>	* but reduce the levels of indirection. Additionally,
87	>	* volatile-writes of table elements and entry "next" fields
88	>	* within locked operations use the cheaper "lazySet" forms of
89	>	* writes (via putOrderedObject) because these writes are always
90	>	* followed by lock releases that maintain sequential consistency
91	>	* of table updates.
92	>	*
93	>	* Historical note: The previous version of this class relied
94	>	* heavily on "final" fields, which avoided some volatile reads at
95	>	* the expense of a large initial footprint.  Some remnants of
96	>	* that design (including forced construction of segment 0) exist
97	>	* to ensure serialization compatibility.
98		*/
99
100		/* ---------------- Constants -------------- */
101
102		/**
103	<	* The default initial number of table slots for this table.
104	<	* Used when not otherwise specified in constructor.
103	>	* The default initial capacity for this table,
104	>	* used when not otherwise specified in a constructor.
105	>	*/
106	>	static final int DEFAULT_INITIAL_CAPACITY = 16;
107	>
108	>	/**
109	>	* The default load factor for this table, used when not
110	>	* otherwise specified in a constructor.
111	>	*/
112	>	static final float DEFAULT_LOAD_FACTOR = 0.75f;
113	>
114	>	/**
115	>	* The default concurrency level for this table, used when not
116	>	* otherwise specified in a constructor.
117		*/
118	<	private static int DEFAULT_INITIAL_CAPACITY = 16;
118	>	static final int DEFAULT_CONCURRENCY_LEVEL = 16;
119
120		/**
121		* The maximum capacity, used if a higher value is implicitly
122		* specified by either of the constructors with arguments.  MUST
123	<	* be a power of two <= 1<<30 to ensure that entries are indexible
123	>	* be a power of two <= 1<<30 to ensure that entries are indexable
124		* using ints.
125		*/
126	<	static final int MAXIMUM_CAPACITY = 1 << 30;
126	>	static final int MAXIMUM_CAPACITY = 1 << 30;
127
128		/**
129	<	* The default load factor for this table.  Used when not
130	<	* otherwise specified in constructor.
129	>	* The minimum capacity for per-segment tables.  Must be a power
130	>	* of two, at least two to avoid immediate resizing on next use
131	>	* after lazy construction.
132		*/
133	<	static final float DEFAULT_LOAD_FACTOR = 0.75f;
133	>	static final int MIN_SEGMENT_TABLE_CAPACITY = 2;
134
135		/**
136	<	* The default number of concurrency control segments.
137	<	**/
138	<	private static final int DEFAULT_SEGMENTS = 16;
136	>	* The maximum number of segments to allow; used to bound
137	>	* constructor arguments. Must be power of two less than 1 << 24.
138	>	*/
139	>	static final int MAX_SEGMENTS = 1 << 16; // slightly conservative
140
141		/**
142	<	* The maximum number of segments to allow; used to bound ctor arguments.
142	>	* Number of unsynchronized retries in size and containsValue
143	>	* methods before resorting to locking. This is used to avoid
144	>	* unbounded retries if tables undergo continuous modification
145	>	* which would make it impossible to obtain an accurate result.
146		*/
147	<	private static final int MAX_SEGMENTS = 1 << 16; // slightly conservative
147	>	static final int RETRIES_BEFORE_LOCK = 2;
148
149		/* ---------------- Fields -------------- */
150
151		/**
152		* Mask value for indexing into segments. The upper bits of a
153		* key's hash code are used to choose the segment.
154	<	**/
155	<	private final int segmentMask;
154	>	*/
155	>	final int segmentMask;
156
157		/**
158		* Shift value for indexing within segments.
159	<	**/
160	<	private final int segmentShift;
159	>	*/
160	>	final int segmentShift;
161
162		/**
163	<	* The segments, each of which is a specialized hash table
163	>	* The segments, each of which is a specialized hash table.
164		*/
165	<	private final Segment[] segments;
165	>	final Segment<K,V>[] segments;
166
167	<	private transient Set<K> keySet;
168	<	private transient Set<Map.Entry<K,V>> entrySet;
169	<	private transient Collection<V> values;
125	<
126	<	/* ---------------- Small Utilities -------------- */
167	>	transient Set<K> keySet;
168	>	transient Set<Map.Entry<K,V>> entrySet;
169	>	transient Collection<V> values;
170
171		/**
172	<	* Return a hash code for non-null Object x.
173	<	* Uses the same hash code spreader as most other j.u hash tables.
131	<	* @param x the object serving as a key
132	<	* @return the hash code
172	>	* ConcurrentHashMap list entry. Note that this is never exported
173	>	* out as a user-visible Map.Entry.
174		*/
175	<	private static int hash(Object x) {
176	<	int h = x.hashCode();
177	<	h += ~(h << 9);
178	<	h ^=  (h >>> 14);
179	<	h +=  (h << 4);
180	<	h ^=  (h >>> 10);
181	<	return h;
175	>	static final class HashEntry<K,V> {
176	>	final int hash;
177	>	final K key;
178	>	volatile V value;
179	>	volatile HashEntry<K,V> next;
180	>
181	>	HashEntry(int hash, K key, V value, HashEntry<K,V> next) {
182	>	this.hash = hash;
183	>	this.key = key;
184	>	this.value = value;
185	>	this.next = next;
186	>	}
187	>
188	>	/**
189	>	* Sets next field with volatile write semantics.  (See above
190	>	* about use of putOrderedObject.)
191	>	*/
192	>	final void setNext(HashEntry<K,V> n) {
193	>	UNSAFE.putOrderedObject(this, nextOffset, n);
194	>	}
195	>
196	>	// Unsafe mechanics
197	>	static final sun.misc.Unsafe UNSAFE;
198	>	static final long nextOffset;
199	>	static {
200	>	try {
201	>	UNSAFE = sun.misc.Unsafe.getUnsafe();
202	>	Class k = HashEntry.class;
203	>	nextOffset = UNSAFE.objectFieldOffset
204	>	(k.getDeclaredField("next"));
205	>	} catch (Exception e) {
206	>	throw new Error(e);
207	>	}
208	>	}
209		}
210
211		/**
212	<	* Return the segment that should be used for key with given hash
213	<	*/
214	<	private Segment<K,V> segmentFor(int hash) {
215	<	return (Segment<K,V>) segments[(hash >>> segmentShift) & segmentMask];
212	>	* Gets the ith element of given table (if nonnull) with volatile
213	>	* read semantics. Note: This is manually integrated into a few
214	>	* performance-sensitive methods to reduce call overhead.
215	>	*/
216	>	@SuppressWarnings("unchecked")
217	>	static final <K,V> HashEntry<K,V> entryAt(HashEntry<K,V>[] tab, int i) {
218	>	return (tab == null) ? null :
219	>	(HashEntry<K,V>) UNSAFE.getObjectVolatile
220	>	(tab, ((long)i << TSHIFT) + TBASE);
221		}
222
223	<	/* ---------------- Inner Classes -------------- */
223	>	/**
224	>	* Sets the ith element of given table, with volatile write
225	>	* semantics. (See above about use of putOrderedObject.)
226	>	*/
227	>	static final <K,V> void setEntryAt(HashEntry<K,V>[] tab, int i,
228	>	HashEntry<K,V> e) {
229	>	UNSAFE.putOrderedObject(tab, ((long)i << TSHIFT) + TBASE, e);
230	>	}
231	>
232	>	/**
233	>	* Applies a supplemental hash function to a given hashCode, which
234	>	* defends against poor quality hash functions.  This is critical
235	>	* because ConcurrentHashMap uses power-of-two length hash tables,
236	>	* that otherwise encounter collisions for hashCodes that do not
237	>	* differ in lower or upper bits.
238	>	*/
239	>	private static int hash(int h) {
240	>	// Spread bits to regularize both segment and index locations,
241	>	// using variant of single-word Wang/Jenkins hash.
242	>	h += (h <<  15) ^ 0xffffcd7d;
243	>	h ^= (h >>> 10);
244	>	h += (h <<   3);
245	>	h ^= (h >>>  6);
246	>	h += (h <<   2) + (h << 14);
247	>	return h ^ (h >>> 16);
248	>	}
249
250		/**
251		* Segments are specialized versions of hash tables.  This
252		* subclasses from ReentrantLock opportunistically, just to
253		* simplify some locking and avoid separate construction.
254	<	**/
255	<	private static final class Segment<K,V> extends ReentrantLock implements Serializable {
254	>	*/
255	>	static final class Segment<K,V> extends ReentrantLock implements Serializable {
256		/*
257	<	* Segments maintain a table of entry lists that are ALWAYS
258	<	* kept in a consistent state, so can be read without locking.
259	<	* Next fields of nodes are immutable (final).  All list
260	<	* additions are performed at the front of each bin. This
261	<	* makes it easy to check changes, and also fast to traverse.
262	<	* When nodes would otherwise be changed, new nodes are
165	<	* created to replace them. This works well for hash tables
166	<	* since the bin lists tend to be short. (The average length
167	<	* is less than two for the default load factor threshold.)
168	<	*
169	<	* Read operations can thus proceed without locking, but rely
170	<	* on a memory barrier to ensure that completed write
171	<	* operations performed by other threads are
172	<	* noticed. Conveniently, the "count" field, tracking the
173	<	* number of elements, can also serve as the volatile variable
174	<	* providing proper read/write barriers. This is convenient
175	<	* because this field needs to be read in many read operations
176	<	* anyway.
177	<	*
178	<	* Implementors note. The basic rules for all this are:
179	<	*
180	<	*   - All unsynchronized read operations must first read the
181	<	*     "count" field, and should not look at table entries if
182	<	*     it is 0.
257	>	* Segments maintain a table of entry lists that are always
258	>	* kept in a consistent state, so can be read (via volatile
259	>	* reads of segments and tables) without locking.  This
260	>	* requires replicating nodes when necessary during table
261	>	* resizing, so the old lists can be traversed by readers
262	>	* still using old version of table.
263		*
264	<	*   - All synchronized write operations should write to
265	<	*     the "count" field after updating. The operations must not
266	<	*     take any action that could even momentarily cause
267	<	*     a concurrent read operation to see inconsistent
268	<	*     data. This is made easier by the nature of the read
269	<	*     operations in Map. For example, no operation
270	<	*     can reveal that the table has grown but the threshold
271	<	*     has not yet been updated, so there are no atomicity
272	<	*     requirements for this with respect to reads.
273	<	*
274	<	* As a guide, all critical volatile reads and writes are marked
275	<	* in code comments.
264	>	* This class defines only mutative methods requiring locking.
265	>	* Except as noted, the methods of this class perform the
266	>	* per-segment versions of ConcurrentHashMap methods.  (Other
267	>	* methods are integrated directly into ConcurrentHashMap
268	>	* methods.) These mutative methods use a form of controlled
269	>	* spinning on contention via methods scanAndLock and
270	>	* scanAndLockForPut. These intersperse tryLocks with
271	>	* traversals to locate nodes.  The main benefit is to absorb
272	>	* cache misses (which are very common for hash tables) while
273	>	* obtaining locks so that traversal is faster once
274	>	* acquired. We do not actually use the found nodes since they
275	>	* must be re-acquired under lock anyway to ensure sequential
276	>	* consistency of updates (and in any case may be undetectably
277	>	* stale), but they will normally be much faster to re-locate.
278	>	* Also, scanAndLockForPut speculatively creates a fresh node
279	>	* to use in put if no node is found.
280		*/
281
282	+	private static final long serialVersionUID = 2249069246763182397L;
283	+
284		/**
285	<	* The number of elements in this segment's region.
286	<	**/
287	<	transient volatile int count;
285	>	* The maximum number of times to tryLock in a prescan before
286	>	* possibly blocking on acquire in preparation for a locked
287	>	* segment operation. On multiprocessors, using a bounded
288	>	* number of retries maintains cache acquired while locating
289	>	* nodes.
290	>	*/
291	>	static final int MAX_SCAN_RETRIES =
292	>	Runtime.getRuntime().availableProcessors() > 1 ? 64 : 1;
293
294		/**
295	<	* Number of updates; used for checking lack of modifications
296	<	* in bulk-read methods.
295	>	* The per-segment table. Elements are accessed via
296	>	* entryAt/setEntryAt providing volatile semantics.
297		*/
298	<	transient int modCount;
298	>	transient volatile HashEntry<K,V>[] table;
299
300		/**
301	<	* The table is rehashed when its size exceeds this threshold.
302	<	* (The value of this field is always (int)(capacity *
303	<	* loadFactor).)
301	>	* The number of elements. Accessed only either within locks
302	>	* or among other volatile reads that maintain visibility.
303	>	*/
304	>	transient int count;
305	>
306	>	/**
307	>	* The total number of mutative operations in this segment.
308	>	* Even though this may overflows 32 bits, it provides
309	>	* sufficient accuracy for stability checks in CHM isEmpty()
310	>	* and size() methods.  Accessed only either within locks or
311	>	* among other volatile reads that maintain visibility.
312		*/
313	<	private transient int threshold;
313	>	transient int modCount;
314
315		/**
316	<	* The per-segment table
316	>	* The table is rehashed when its size exceeds this threshold.
317	>	* (The value of this field is always <tt>(int)(capacity *
318	>	* loadFactor)</tt>.)
319		*/
320	<	transient HashEntry[] table;
320	>	transient int threshold;
321
322		/**
323		* The load factor for the hash table.  Even though this value
#	Line 224 | Line 325 | public class ConcurrentHashMap<K, V> ext
325		* links to outer object.
326		* @serial
327		*/
328	<	private final float loadFactor;
228	<
229	<	Segment(int initialCapacity, float lf) {
230	<	loadFactor = lf;
231	<	setTable(new HashEntry[initialCapacity]);
232	<	}
328	>	final float loadFactor;
329
330	<	/**
331	<	* Set table to new HashEntry array.
332	<	* Call only while holding lock or in constructor.
333	<	**/
238	<	private void setTable(HashEntry[] newTable) {
239	<	table = newTable;
240	<	threshold = (int)(newTable.length * loadFactor);
241	<	count = count; // write-volatile
242	<	}
243	<
244	<	/* Specialized implementations of map methods */
245	<
246	<	V get(K key, int hash) {
247	<	if (count != 0) { // read-volatile
248	<	HashEntry[] tab = table;
249	<	int index = hash & (tab.length - 1);
250	<	HashEntry<K,V> e = (HashEntry<K,V>) tab[index];
251	<	while (e != null) {
252	<	if (e.hash == hash && key.equals(e.key))
253	<	return e.value;
254	<	e = e.next;
255	<	}
256	<	}
257	<	return null;
258	<	}
259	<
260	<	boolean containsKey(Object key, int hash) {
261	<	if (count != 0) { // read-volatile
262	<	HashEntry[] tab = table;
263	<	int index = hash & (tab.length - 1);
264	<	HashEntry<K,V> e = (HashEntry<K,V>) tab[index];
265	<	while (e != null) {
266	<	if (e.hash == hash && key.equals(e.key))
267	<	return true;
268	<	e = e.next;
269	<	}
270	<	}
271	<	return false;
272	<	}
273	<
274	<	boolean containsValue(Object value) {
275	<	if (count != 0) { // read-volatile
276	<	HashEntry[] tab = table;
277	<	int len = tab.length;
278	<	for (int i = 0 ; i < len; i++)
279	<	for (HashEntry<K,V> e = (HashEntry<K,V>)tab[i] ; e != null ; e = e.next)
280	<	if (value.equals(e.value))
281	<	return true;
282	<	}
283	<	return false;
330	>	Segment(float lf, int threshold, HashEntry<K,V>[] tab) {
331	>	this.loadFactor = lf;
332	>	this.threshold = threshold;
333	>	this.table = tab;
334		}
335
336	<	V put(K key, int hash, V value, boolean onlyIfAbsent) {
337	<	lock();
336	>	final V put(K key, int hash, V value, boolean onlyIfAbsent) {
337	>	HashEntry<K,V> node = tryLock() ? null :
338	>	scanAndLockForPut(key, hash, value);
339	>	V oldValue;
340		try {
341	<	int c = count;
342	<	HashEntry[] tab = table;
343	<	int index = hash & (tab.length - 1);
344	<	HashEntry<K,V> first = (HashEntry<K,V>) tab[index];
345	<
346	<	for (HashEntry<K,V> e = first; e != null; e = (HashEntry<K,V>) e.next) {
347	<	if (e.hash == hash && key.equals(e.key)) {
348	<	V oldValue = e.value;
349	<	if (!onlyIfAbsent)
350	<	e.value = value;
341	>	HashEntry<K,V>[] tab = table;
342	>	int index = (tab.length - 1) & hash;
343	>	HashEntry<K,V> first = entryAt(tab, index);
344	>	for (HashEntry<K,V> e = first;;) {
345	>	if (e != null) {
346	>	K k;
347	>	if ((k = e.key) == key ||
348	>	(e.hash == hash && key.equals(k))) {
349	>	oldValue = e.value;
350	>	if (!onlyIfAbsent) {
351	>	e.value = value;
352	>	++modCount;
353	>	}
354	>	break;
355	>	}
356	>	e = e.next;
357	>	}
358	>	else {
359	>	if (node != null)
360	>	node.setNext(first);
361	>	else
362	>	node = new HashEntry<K,V>(hash, key, value, first);
363	>	int c = count + 1;
364	>	if (c > threshold && tab.length < MAXIMUM_CAPACITY)
365	>	rehash(node);
366	>	else
367	>	setEntryAt(tab, index, node);
368		++modCount;
369	<	count = c; // write-volatile
370	<	return oldValue;
369	>	count = c;
370	>	oldValue = null;
371	>	break;
372		}
373		}
304	–
305	–	tab[index] = new HashEntry<K,V>(hash, key, value, first);
306	–	++modCount;
307	–	++c;
308	–	count = c; // write-volatile
309	–	if (c > threshold)
310	–	setTable(rehash(tab));
311	–	return null;
374		} finally {
375		unlock();
376		}
377	+	return oldValue;
378		}
379
380	<	private HashEntry[] rehash(HashEntry[] oldTable) {
381	<	int oldCapacity = oldTable.length;
382	<	if (oldCapacity >= MAXIMUM_CAPACITY)
383	<	return oldTable;
384	<
380	>	/**
381	>	* Doubles size of table and repacks entries, also adding the
382	>	* given node to new table
383	>	*/
384	>	@SuppressWarnings("unchecked")
385	>	private void rehash(HashEntry<K,V> node) {
386		/*
387	<	* Reclassify nodes in each list to new Map.  Because we are
388	<	* using power-of-two expansion, the elements from each bin
389	<	* must either stay at same index, or move with a power of two
390	<	* offset. We eliminate unnecessary node creation by catching
391	<	* cases where old nodes can be reused because their next
392	<	* fields won't change. Statistically, at the default
393	<	* threshhold, only about one-sixth of them need cloning when
394	<	* a table doubles. The nodes they replace will be garbage
395	<	* collectable as soon as they are no longer referenced by any
396	<	* reader thread that may be in the midst of traversing table
397	<	* right now.
387	>	* Reclassify nodes in each list to new table.  Because we
388	>	* are using power-of-two expansion, the elements from
389	>	* each bin must either stay at same index, or move with a
390	>	* power of two offset. We eliminate unnecessary node
391	>	* creation by catching cases where old nodes can be
392	>	* reused because their next fields won't change.
393	>	* Statistically, at the default threshold, only about
394	>	* one-sixth of them need cloning when a table
395	>	* doubles. The nodes they replace will be garbage
396	>	* collectable as soon as they are no longer referenced by
397	>	* any reader thread that may be in the midst of
398	>	* concurrently traversing table. Entry accesses use plain
399	>	* array indexing because they are followed by volatile
400	>	* table write.
401		*/
402	<
403	<	HashEntry[] newTable = new HashEntry[oldCapacity << 1];
404	<	int sizeMask = newTable.length - 1;
402	>	HashEntry<K,V>[] oldTable = table;
403	>	int oldCapacity = oldTable.length;
404	>	int newCapacity = oldCapacity << 1;
405	>	threshold = (int)(newCapacity * loadFactor);
406	>	HashEntry<K,V>[] newTable =
407	>	(HashEntry<K,V>[]) new HashEntry[newCapacity];
408	>	int sizeMask = newCapacity - 1;
409		for (int i = 0; i < oldCapacity ; i++) {
410	<	// We need to guarantee that any existing reads of old Map can
340	<	//  proceed. So we cannot yet null out each bin.
341	<	HashEntry<K,V> e = (HashEntry<K,V>)oldTable[i];
342	<
410	>	HashEntry<K,V> e = oldTable[i];
411		if (e != null) {
412		HashEntry<K,V> next = e.next;
413		int idx = e.hash & sizeMask;
414	<
347	<	//  Single node on list
348	<	if (next == null)
414	>	if (next == null)   //  Single node on list
415		newTable[idx] = e;
416	<
351	<	else {
352	<	// Reuse trailing consecutive sequence at same slot
416	>	else { // Reuse consecutive sequence at same slot
417		HashEntry<K,V> lastRun = e;
418		int lastIdx = idx;
419		for (HashEntry<K,V> last = next;
#	Line 362 | Line 426 | public class ConcurrentHashMap<K, V> ext
426		}
427		}
428		newTable[lastIdx] = lastRun;
429	<
366	<	// Clone all remaining nodes
429	>	// Clone remaining nodes
430		for (HashEntry<K,V> p = e; p != lastRun; p = p.next) {
431	<	int k = p.hash & sizeMask;
432	<	newTable[k] = new HashEntry<K,V>(p.hash,
433	<	p.key,
434	<	p.value,
435	<	(HashEntry<K,V>) newTable[k]);
431	>	V v = p.value;
432	>	int h = p.hash;
433	>	int k = h & sizeMask;
434	>	HashEntry<K,V> n = newTable[k];
435	>	newTable[k] = new HashEntry<K,V>(h, p.key, v, n);
436		}
437		}
438		}
439		}
440	<	return newTable;
440	>	int nodeIndex = node.hash & sizeMask; // add the new node
441	>	node.setNext(newTable[nodeIndex]);
442	>	newTable[nodeIndex] = node;
443	>	table = newTable;
444	>	}
445	>
446	>	/**
447	>	* Scans for a node containing given key while trying to
448	>	* acquire lock, creating and returning one if not found. Upon
449	>	* return, guarantees that lock is held. Unlike in most
450	>	* methods, calls to method equals are not screened: Since
451	>	* traversal speed doesn't matter, we might as well help warm
452	>	* up the associated code and accesses as well.
453	>	*
454	>	* @return a new node if key not found, else null
455	>	*/
456	>	private HashEntry<K,V> scanAndLockForPut(K key, int hash, V value) {
457	>	HashEntry<K,V> first = entryForHash(this, hash);
458	>	HashEntry<K,V> e = first;
459	>	HashEntry<K,V> node = null;
460	>	int retries = -1; // negative while locating node
461	>	while (!tryLock()) {
462	>	HashEntry<K,V> f; // to recheck first below
463	>	if (retries < 0) {
464	>	if (e == null) {
465	>	if (node == null) // speculatively create node
466	>	node = new HashEntry<K,V>(hash, key, value, null);
467	>	retries = 0;
468	>	}
469	>	else if (key.equals(e.key))
470	>	retries = 0;
471	>	else
472	>	e = e.next;
473	>	}
474	>	else if (++retries > MAX_SCAN_RETRIES) {
475	>	lock();
476	>	break;
477	>	}
478	>	else if ((retries & 1) == 0 &&
479	>	(f = entryForHash(this, hash)) != first) {
480	>	e = first = f; // re-traverse if entry changed
481	>	retries = -1;
482	>	}
483	>	}
484	>	return node;
485	>	}
486	>
487	>	/**
488	>	* Scans for a node containing the given key while trying to
489	>	* acquire lock for a remove or replace operation. Upon
490	>	* return, guarantees that lock is held.  Note that we must
491	>	* lock even if the key is not found, to ensure sequential
492	>	* consistency of updates.
493	>	*/
494	>	private void scanAndLock(Object key, int hash) {
495	>	// similar to but simpler than scanAndLockForPut
496	>	HashEntry<K,V> first = entryForHash(this, hash);
497	>	HashEntry<K,V> e = first;
498	>	int retries = -1;
499	>	while (!tryLock()) {
500	>	HashEntry<K,V> f;
501	>	if (retries < 0) {
502	>	if (e == null || key.equals(e.key))
503	>	retries = 0;
504	>	else
505	>	e = e.next;
506	>	}
507	>	else if (++retries > MAX_SCAN_RETRIES) {
508	>	lock();
509	>	break;
510	>	}
511	>	else if ((retries & 1) == 0 &&
512	>	(f = entryForHash(this, hash)) != first) {
513	>	e = first = f;
514	>	retries = -1;
515	>	}
516	>	}
517		}
518
519		/**
520		* Remove; match on key only if value null, else match both.
521		*/
522	<	V remove(Object key, int hash, Object value) {
523	<	lock();
522	>	final V remove(Object key, int hash, Object value) {
523	>	if (!tryLock())
524	>	scanAndLock(key, hash);
525	>	V oldValue = null;
526		try {
527	<	int c = count;
528	<	HashEntry[] tab = table;
529	<	int index = hash & (tab.length - 1);
530	<	HashEntry<K,V> first = (HashEntry<K,V>)tab[index];
531	<
532	<	HashEntry<K,V> e = first;
533	<	for (;;) {
534	<	if (e == null)
535	<	return null;
536	<	if (e.hash == hash && key.equals(e.key))
527	>	HashEntry<K,V>[] tab = table;
528	>	int index = (tab.length - 1) & hash;
529	>	HashEntry<K,V> e = entryAt(tab, index);
530	>	HashEntry<K,V> pred = null;
531	>	while (e != null) {
532	>	K k;
533	>	HashEntry<K,V> next = e.next;
534	>	if ((k = e.key) == key ||
535	>	(e.hash == hash && key.equals(k))) {
536	>	V v = e.value;
537	>	if (value == null || value == v || value.equals(v)) {
538	>	if (pred == null)
539	>	setEntryAt(tab, index, next);
540	>	else
541	>	pred.setNext(next);
542	>	++modCount;
543	>	--count;
544	>	oldValue = v;
545	>	}
546		break;
547	<	e = e.next;
547	>	}
548	>	pred = e;
549	>	e = next;
550		}
551	+	} finally {
552	+	unlock();
553	+	}
554	+	return oldValue;
555	+	}
556
557	<	V oldValue = e.value;
558	<	if (value != null && !value.equals(oldValue))
559	<	return null;
560	<
561	<	// All entries following removed node can stay in list, but
562	<	// all preceeding ones need to be cloned.
563	<	HashEntry<K,V> newFirst = e.next;
564	<	for (HashEntry<K,V> p = first; p != e; p = p.next)
565	<	newFirst = new HashEntry<K,V>(p.hash, p.key,
566	<	p.value, newFirst);
567	<	tab[index] = newFirst;
568	<	++modCount;
569	<	count = c-1; // write-volatile
570	<	return oldValue;
557	>	final boolean replace(K key, int hash, V oldValue, V newValue) {
558	>	if (!tryLock())
559	>	scanAndLock(key, hash);
560	>	boolean replaced = false;
561	>	try {
562	>	HashEntry<K,V> e;
563	>	for (e = entryForHash(this, hash); e != null; e = e.next) {
564	>	K k;
565	>	if ((k = e.key) == key ||
566	>	(e.hash == hash && key.equals(k))) {
567	>	if (oldValue.equals(e.value)) {
568	>	e.value = newValue;
569	>	++modCount;
570	>	replaced = true;
571	>	}
572	>	break;
573	>	}
574	>	}
575	>	} finally {
576	>	unlock();
577	>	}
578	>	return replaced;
579	>	}
580	>
581	>	final V replace(K key, int hash, V value) {
582	>	if (!tryLock())
583	>	scanAndLock(key, hash);
584	>	V oldValue = null;
585	>	try {
586	>	HashEntry<K,V> e;
587	>	for (e = entryForHash(this, hash); e != null; e = e.next) {
588	>	K k;
589	>	if ((k = e.key) == key ||
590	>	(e.hash == hash && key.equals(k))) {
591	>	oldValue = e.value;
592	>	e.value = value;
593	>	++modCount;
594	>	break;
595	>	}
596	>	}
597		} finally {
598		unlock();
599		}
600	+	return oldValue;
601		}
602
603	<	void clear() {
603	>	final void clear() {
604		lock();
605		try {
606	<	HashEntry[] tab = table;
606	>	HashEntry<K,V>[] tab = table;
607		for (int i = 0; i < tab.length ; i++)
608	<	tab[i] = null;
608	>	setEntryAt(tab, i, null);
609		++modCount;
610	<	count = 0; // write-volatile
610	>	count = 0;
611		} finally {
612		unlock();
613		}
614		}
615		}
616
617	+	// Accessing segments
618	+
619		/**
620	<	* ConcurrentHashMap list entry.
620	>	* Gets the jth element of given segment array (if nonnull) with
621	>	* volatile element access semantics via Unsafe. (The null check
622	>	* can trigger harmlessly only during deserialization.) Note:
623	>	* because each element of segments array is set only once (using
624	>	* fully ordered writes), some performance-sensitive methods rely
625	>	* on this method only as a recheck upon null reads.
626		*/
627	<	private static class HashEntry<K,V> implements Entry<K,V> {
628	<	private final K key;
629	<	private V value;
630	<	private final int hash;
631	<	private final HashEntry<K,V> next;
632	<
442	<	HashEntry(int hash, K key, V value, HashEntry<K,V> next) {
443	<	this.value = value;
444	<	this.hash = hash;
445	<	this.key = key;
446	<	this.next = next;
447	<	}
448	<
449	<	public K getKey() {
450	<	return key;
451	<	}
452	<
453	<	public V getValue() {
454	<	return value;
455	<	}
456	<
457	<	public V setValue(V newValue) {
458	<	// We aren't required to, and don't provide any
459	<	// visibility barriers for setting value.
460	<	if (newValue == null)
461	<	throw new NullPointerException();
462	<	V oldValue = this.value;
463	<	this.value = newValue;
464	<	return oldValue;
465	<	}
627	>	@SuppressWarnings("unchecked")
628	>	static final <K,V> Segment<K,V> segmentAt(Segment<K,V>[] ss, int j) {
629	>	long u = (j << SSHIFT) + SBASE;
630	>	return ss == null ? null :
631	>	(Segment<K,V>) UNSAFE.getObjectVolatile(ss, u);
632	>	}
633
634	<	public boolean equals(Object o) {
635	<	if (!(o instanceof Entry))
636	<	return false;
637	<	Entry<K,V> e = (Entry<K,V>)o;
638	<	return (key.equals(e.getKey()) && value.equals(e.getValue()));
634	>	/**
635	>	* Returns the segment for the given index, creating it and
636	>	* recording in segment table (via CAS) if not already present.
637	>	*
638	>	* @param k the index
639	>	* @return the segment
640	>	*/
641	>	@SuppressWarnings("unchecked")
642	>	private Segment<K,V> ensureSegment(int k) {
643	>	final Segment<K,V>[] ss = this.segments;
644	>	long u = (k << SSHIFT) + SBASE; // raw offset
645	>	Segment<K,V> seg;
646	>	if ((seg = (Segment<K,V>)UNSAFE.getObjectVolatile(ss, u)) == null) {
647	>	Segment<K,V> proto = ss[0]; // use segment 0 as prototype
648	>	int cap = proto.table.length;
649	>	float lf = proto.loadFactor;
650	>	int threshold = (int)(cap * lf);
651	>	HashEntry<K,V>[] tab = (HashEntry<K,V>[])new HashEntry[cap];
652	>	if ((seg = (Segment<K,V>)UNSAFE.getObjectVolatile(ss, u))
653	>	== null) { // recheck
654	>	Segment<K,V> s = new Segment<K,V>(lf, threshold, tab);
655	>	while ((seg = (Segment<K,V>)UNSAFE.getObjectVolatile(ss, u))
656	>	== null) {
657	>	if (UNSAFE.compareAndSwapObject(ss, u, null, seg = s))
658	>	break;
659	>	}
660	>	}
661		}
662	+	return seg;
663	+	}
664
665	<	public int hashCode() {
475	<	return  key.hashCode() ^ value.hashCode();
476	<	}
665	>	// Hash-based segment and entry accesses
666
667	<	public String toString() {
668	<	return key + "=" + value;
669	<	}
667	>	/**
668	>	* Gets the segment for the given hash code.
669	>	*/
670	>	@SuppressWarnings("unchecked")
671	>	private Segment<K,V> segmentForHash(int h) {
672	>	long u = (((h >>> segmentShift) & segmentMask) << SSHIFT) + SBASE;
673	>	return (Segment<K,V>) UNSAFE.getObjectVolatile(segments, u);
674		}
675
676	+	/**
677	+	* Gets the table entry for the given segment and hash code.
678	+	*/
679	+	@SuppressWarnings("unchecked")
680	+	static final <K,V> HashEntry<K,V> entryForHash(Segment<K,V> seg, int h) {
681	+	HashEntry<K,V>[] tab;
682	+	return (seg == null || (tab = seg.table) == null) ? null :
683	+	(HashEntry<K,V>) UNSAFE.getObjectVolatile
684	+	(tab, ((long)(((tab.length - 1) & h)) << TSHIFT) + TBASE);
685	+	}
686
687		/* ---------------- Public operations -------------- */
688
689		/**
690	<	* Constructs a new, empty map with the specified initial
691	<	* capacity and the specified load factor.
690	>	* Creates a new, empty map with the specified initial
691	>	* capacity, load factor and concurrency level.
692		*
693		* @param initialCapacity the initial capacity. The implementation
694		* performs internal sizing to accommodate this many elements.
695		* @param loadFactor  the load factor threshold, used to control resizing.
696	+	* Resizing may be performed when the average number of elements per
697	+	* bin exceeds this threshold.
698		* @param concurrencyLevel the estimated number of concurrently
699		* updating threads. The implementation performs internal sizing
700	<	* to try to accommodate this many threads.
700	>	* to try to accommodate this many threads.
701		* @throws IllegalArgumentException if the initial capacity is
702		* negative or the load factor or concurrencyLevel are
703		* nonpositive.
704		*/
705	+	@SuppressWarnings("unchecked")
706		public ConcurrentHashMap(int initialCapacity,
707		float loadFactor, int concurrencyLevel) {
708		if (!(loadFactor > 0) || initialCapacity < 0 || concurrencyLevel <= 0)
709		throw new IllegalArgumentException();
504	–
710		if (concurrencyLevel > MAX_SEGMENTS)
711		concurrencyLevel = MAX_SEGMENTS;
507	–
712		// Find power-of-two sizes best matching arguments
713		int sshift = 0;
714		int ssize = 1;
#	Line 512 | Line 716 | public class ConcurrentHashMap<K, V> ext
716		++sshift;
717		ssize <<= 1;
718		}
719	<	segmentShift = 32 - sshift;
720	<	segmentMask = ssize - 1;
517	<	this.segments = new Segment[ssize];
518	<
719	>	this.segmentShift = 32 - sshift;
720	>	this.segmentMask = ssize - 1;
721		if (initialCapacity > MAXIMUM_CAPACITY)
722		initialCapacity = MAXIMUM_CAPACITY;
723		int c = initialCapacity / ssize;
724		if (c * ssize < initialCapacity)
725		++c;
726	<	int cap = 1;
726	>	int cap = MIN_SEGMENT_TABLE_CAPACITY;
727		while (cap < c)
728		cap <<= 1;
729	<
730	<	for (int i = 0; i < this.segments.length; ++i)
731	<	this.segments[i] = new Segment<K,V>(cap, loadFactor);
729	>	// create segments and segments[0]
730	>	Segment<K,V> s0 =
731	>	new Segment<K,V>(loadFactor, (int)(cap * loadFactor),
732	>	(HashEntry<K,V>[])new HashEntry[cap]);
733	>	Segment<K,V>[] ss = (Segment<K,V>[])new Segment[ssize];
734	>	UNSAFE.putOrderedObject(ss, SBASE, s0); // ordered write of segments[0]
735	>	this.segments = ss;
736		}
737
738		/**
739	<	* Constructs a new, empty map with the specified initial
740	<	* capacity,  and with default load factor and concurrencyLevel.
739	>	* Creates a new, empty map with the specified initial capacity
740	>	* and load factor and with the default concurrencyLevel (16).
741		*
742		* @param initialCapacity The implementation performs internal
743		* sizing to accommodate this many elements.
744	+	* @param loadFactor  the load factor threshold, used to control resizing.
745	+	* Resizing may be performed when the average number of elements per
746	+	* bin exceeds this threshold.
747	+	* @throws IllegalArgumentException if the initial capacity of
748	+	* elements is negative or the load factor is nonpositive
749	+	*
750	+	* @since 1.6
751	+	*/
752	+	public ConcurrentHashMap(int initialCapacity, float loadFactor) {
753	+	this(initialCapacity, loadFactor, DEFAULT_CONCURRENCY_LEVEL);
754	+	}
755	+
756	+	/**
757	+	* Creates a new, empty map with the specified initial capacity,
758	+	* and with default load factor (0.75) and concurrencyLevel (16).
759	+	*
760	+	* @param initialCapacity the initial capacity. The implementation
761	+	* performs internal sizing to accommodate this many elements.
762		* @throws IllegalArgumentException if the initial capacity of
763		* elements is negative.
764		*/
765		public ConcurrentHashMap(int initialCapacity) {
766	<	this(initialCapacity, DEFAULT_LOAD_FACTOR, DEFAULT_SEGMENTS);
766	>	this(initialCapacity, DEFAULT_LOAD_FACTOR, DEFAULT_CONCURRENCY_LEVEL);
767		}
768
769		/**
770	<	* Constructs a new, empty map with a default initial capacity,
771	<	* load factor, and concurrencyLevel.
770	>	* Creates a new, empty map with a default initial capacity (16),
771	>	* load factor (0.75) and concurrencyLevel (16).
772		*/
773		public ConcurrentHashMap() {
774	<	this(DEFAULT_INITIAL_CAPACITY, DEFAULT_LOAD_FACTOR, DEFAULT_SEGMENTS);
774	>	this(DEFAULT_INITIAL_CAPACITY, DEFAULT_LOAD_FACTOR, DEFAULT_CONCURRENCY_LEVEL);
775		}
776
777		/**
778	<	* Constructs a new map with the same mappings as the given map.  The
779	<	* map is created with a capacity of twice the number of mappings in
780	<	* the given map or 11 (whichever is greater), and a default load factor.
778	>	* Creates a new map with the same mappings as the given map.
779	>	* The map is created with a capacity of 1.5 times the number
780	>	* of mappings in the given map or 16 (whichever is greater),
781	>	* and a default load factor (0.75) and concurrencyLevel (16).
782	>	*
783	>	* @param m the map
784		*/
785	<	public <A extends K, B extends V> ConcurrentHashMap(Map<A,B> t) {
786	<	this(Math.max((int) (t.size() / DEFAULT_LOAD_FACTOR) + 1,
787	<	11),
788	<	DEFAULT_LOAD_FACTOR, DEFAULT_SEGMENTS);
789	<	putAll(t);
785	>	public ConcurrentHashMap(Map<? extends K, ? extends V> m) {
786	>	this(Math.max((int) (m.size() / DEFAULT_LOAD_FACTOR) + 1,
787	>	DEFAULT_INITIAL_CAPACITY),
788	>	DEFAULT_LOAD_FACTOR, DEFAULT_CONCURRENCY_LEVEL);
789	>	putAll(m);
790		}
791
792	<	// inherit Map javadoc
792	>	/**
793	>	* Returns <tt>true</tt> if this map contains no key-value mappings.
794	>	*
795	>	* @return <tt>true</tt> if this map contains no key-value mappings
796	>	*/
797		public boolean isEmpty() {
798		/*
799	<	* We need to keep track of per-segment modCounts to avoid ABA
800	<	* problems in which an element in one segment was added and
801	<	* in another removed during traversal, in which case the
802	<	* table was never actually empty at any point. Note the
803	<	* similar use of modCounts in the size() and containsValue()
804	<	* methods, which are the only other methods also susceptible
805	<	* to ABA problems.
799	>	* Sum per-segment modCounts to avoid mis-reporting when
800	>	* elements are concurrently added and removed in one segment
801	>	* while checking another, in which case the table was never
802	>	* actually empty at any point. (The sum ensures accuracy up
803	>	* through at least 1<<31 per-segment modifications before
804	>	* recheck.)  Methods size() and containsValue() use similar
805	>	* constructions for stability checks.
806		*/
807	<	int[] mc = new int[segments.length];
808	<	int mcsum = 0;
809	<	for (int i = 0; i < segments.length; ++i) {
810	<	if (segments[i].count != 0)
811	<	return false;
812	<	else
582	<	mcsum += mc[i] = segments[i].modCount;
583	<	}
584	<	// If mcsum happens to be zero, then we know we got a snapshot
585	<	// before any modifications at all were made.  This is
586	<	// probably common enough to bother tracking.
587	<	if (mcsum != 0) {
588	<	for (int i = 0; i < segments.length; ++i) {
589	<	if (segments[i].count != 0 ||
590	<	mc[i] != segments[i].modCount)
807	>	long sum = 0L;
808	>	final Segment<K,V>[] segments = this.segments;
809	>	for (int j = 0; j < segments.length; ++j) {
810	>	Segment<K,V> seg = segmentAt(segments, j);
811	>	if (seg != null) {
812	>	if (seg.count != 0)
813		return false;
814	+	sum += seg.modCount;
815		}
816		}
817	+	if (sum != 0L) { // recheck unless no modifications
818	+	for (int j = 0; j < segments.length; ++j) {
819	+	Segment<K,V> seg = segmentAt(segments, j);
820	+	if (seg != null) {
821	+	if (seg.count != 0)
822	+	return false;
823	+	sum -= seg.modCount;
824	+	}
825	+	}
826	+	if (sum != 0L)
827	+	return false;
828	+	}
829		return true;
830		}
831
832	<	// inherit Map javadoc
832	>	/**
833	>	* Returns the number of key-value mappings in this map.  If the
834	>	* map contains more than <tt>Integer.MAX_VALUE</tt> elements, returns
835	>	* <tt>Integer.MAX_VALUE</tt>.
836	>	*
837	>	* @return the number of key-value mappings in this map
838	>	*/
839		public int size() {
840	<	int[] mc = new int[segments.length];
841	<	for (;;) {
842	<	long sum = 0;
843	<	int mcsum = 0;
844	<	for (int i = 0; i < segments.length; ++i) {
845	<	sum += segments[i].count;
846	<	mcsum += mc[i] = segments[i].modCount;
847	<	}
848	<	int check = 0;
849	<	if (mcsum != 0) {
850	<	for (int i = 0; i < segments.length; ++i) {
851	<	check += segments[i].count;
852	<	if (mc[i] != segments[i].modCount) {
853	<	check = -1; // force retry
854	<	break;
840	>	// Try a few times to get accurate count. On failure due to
841	>	// continuous async changes in table, resort to locking.
842	>	final Segment<K,V>[] segments = this.segments;
843	>	int size;
844	>	boolean overflow; // true if size overflows 32 bits
845	>	long sum;         // sum of modCounts
846	>	long last = 0L;   // previous sum
847	>	int retries = -1; // first iteration isn't retry
848	>	try {
849	>	for (;;) {
850	>	if (retries++ == RETRIES_BEFORE_LOCK) {
851	>	for (int j = 0; j < segments.length; ++j)
852	>	ensureSegment(j).lock(); // force creation
853	>	}
854	>	sum = 0L;
855	>	size = 0;
856	>	overflow = false;
857	>	for (int j = 0; j < segments.length; ++j) {
858	>	Segment<K,V> seg = segmentAt(segments, j);
859	>	if (seg != null) {
860	>	sum += seg.modCount;
861	>	int c = seg.count;
862	>	if (c < 0 || (size += c) < 0)
863	>	overflow = true;
864		}
865		}
866	+	if (sum == last)
867	+	break;
868	+	last = sum;
869		}
870	<	if (check == sum) {
871	<	if (sum > Integer.MAX_VALUE)
872	<	return Integer.MAX_VALUE;
873	<	else
621	<	return (int)sum;
870	>	} finally {
871	>	if (retries > RETRIES_BEFORE_LOCK) {
872	>	for (int j = 0; j < segments.length; ++j)
873	>	segmentAt(segments, j).unlock();
874		}
875		}
876	+	return overflow ? Integer.MAX_VALUE : size;
877		}
878
626	–
879		/**
880	<	* Returns the value to which the specified key is mapped in this table.
880	>	* Returns the value to which the specified key is mapped,
881	>	* or {@code null} if this map contains no mapping for the key.
882	>	*
883	>	* <p>More formally, if this map contains a mapping from a key
884	>	* {@code k} to a value {@code v} such that {@code key.equals(k)},
885	>	* then this method returns {@code v}; otherwise it returns
886	>	* {@code null}.  (There can be at most one such mapping.)
887		*
888	<	* @param   key   a key in the table.
631	<	* @return  the value to which the key is mapped in this table;
632	<	*          <tt>null</tt> if the key is not mapped to any value in
633	<	*          this table.
634	<	* @throws  NullPointerException  if the key is
635	<	*               <tt>null</tt>.
888	>	* @throws NullPointerException if the specified key is null
889		*/
890		public V get(Object key) {
891	<	int hash = hash(key); // throws NullPointerException if key null
892	<	return segmentFor(hash).get((K) key, hash);
891	>	Segment<K,V> s; // manually integrate access methods to reduce overhead
892	>	HashEntry<K,V>[] tab;
893	>	int h = hash(key.hashCode());
894	>	long u = (((h >>> segmentShift) & segmentMask) << SSHIFT) + SBASE;
895	>	if ((s = (Segment<K,V>)UNSAFE.getObjectVolatile(segments, u)) != null &&
896	>	(tab = s.table) != null) {
897	>	for (HashEntry<K,V> e = (HashEntry<K,V>) UNSAFE.getObjectVolatile
898	>	(tab, ((long)(((tab.length - 1) & h)) << TSHIFT) + TBASE);
899	>	e != null; e = e.next) {
900	>	K k;
901	>	if ((k = e.key) == key || (e.hash == h && key.equals(k)))
902	>	return e.value;
903	>	}
904	>	}
905	>	return null;
906		}
907
908		/**
909		* Tests if the specified object is a key in this table.
910		*
911	<	* @param   key   possible key.
912	<	* @return  <tt>true</tt> if and only if the specified object
913	<	*          is a key in this table, as determined by the
914	<	*          <tt>equals</tt> method; <tt>false</tt> otherwise.
915	<	* @throws  NullPointerException  if the key is
650	<	*               <tt>null</tt>.
911	>	* @param  key   possible key
912	>	* @return <tt>true</tt> if and only if the specified object
913	>	*         is a key in this table, as determined by the
914	>	*         <tt>equals</tt> method; <tt>false</tt> otherwise.
915	>	* @throws NullPointerException if the specified key is null
916		*/
917	+	@SuppressWarnings("unchecked")
918		public boolean containsKey(Object key) {
919	<	int hash = hash(key); // throws NullPointerException if key null
920	<	return segmentFor(hash).containsKey(key, hash);
919	>	Segment<K,V> s; // same as get() except no need for volatile value read
920	>	HashEntry<K,V>[] tab;
921	>	int h = hash(key.hashCode());
922	>	long u = (((h >>> segmentShift) & segmentMask) << SSHIFT) + SBASE;
923	>	if ((s = (Segment<K,V>)UNSAFE.getObjectVolatile(segments, u)) != null &&
924	>	(tab = s.table) != null) {
925	>	for (HashEntry<K,V> e = (HashEntry<K,V>) UNSAFE.getObjectVolatile
926	>	(tab, ((long)(((tab.length - 1) & h)) << TSHIFT) + TBASE);
927	>	e != null; e = e.next) {
928	>	K k;
929	>	if ((k = e.key) == key || (e.hash == h && key.equals(k)))
930	>	return true;
931	>	}
932	>	}
933	>	return false;
934		}
935
936		/**
#	Line 660 | Line 939 | public class ConcurrentHashMap<K, V> ext
939		* traversal of the hash table, and so is much slower than
940		* method <tt>containsKey</tt>.
941		*
942	<	* @param value value whose presence in this map is to be tested.
942	>	* @param value value whose presence in this map is to be tested
943		* @return <tt>true</tt> if this map maps one or more keys to the
944	<	* specified value.
945	<	* @throws  NullPointerException  if the value is <tt>null</tt>.
944	>	*         specified value
945	>	* @throws NullPointerException if the specified value is null
946		*/
947		public boolean containsValue(Object value) {
948	+	// Same idea as size()
949		if (value == null)
950		throw new NullPointerException();
951	<
952	<	int[] mc = new int[segments.length];
953	<	for (;;) {
954	<	int sum = 0;
955	<	int mcsum = 0;
956	<	for (int i = 0; i < segments.length; ++i) {
957	<	int c = segments[i].count;
958	<	mcsum += mc[i] = segments[i].modCount;
959	<	if (segments[i].containsValue(value))
960	<	return true;
961	<	}
962	<	boolean cleanSweep = true;
963	<	if (mcsum != 0) {
964	<	for (int i = 0; i < segments.length; ++i) {
965	<	int c = segments[i].count;
966	<	if (mc[i] != segments[i].modCount) {
967	<	cleanSweep = false;
968	<	break;
951	>	final Segment<K,V>[] segments = this.segments;
952	>	boolean found = false;
953	>	long last = 0L;   // previous sum
954	>	int retries = -1;
955	>	try {
956	>	outer: for (;;) {
957	>	if (retries++ == RETRIES_BEFORE_LOCK) {
958	>	for (int j = 0; j < segments.length; ++j)
959	>	ensureSegment(j).lock(); // force creation
960	>	}
961	>	long sum = 0L;
962	>	for (int j = 0; j < segments.length; ++j) {
963	>	HashEntry<K,V>[] tab;
964	>	Segment<K,V> seg = segmentAt(segments, j);
965	>	if (seg != null && (tab = seg.table) != null) {
966	>	for (int i = 0 ; i < tab.length; i++) {
967	>	HashEntry<K,V> e;
968	>	for (e = entryAt(tab, i); e != null; e = e.next) {
969	>	V v = e.value;
970	>	if (v != null && value.equals(v)) {
971	>	found = true;
972	>	break outer;
973	>	}
974	>	}
975	>	}
976	>	sum += seg.modCount;
977		}
978		}
979	+	if (retries > 0 && sum == last)
980	+	break;
981	+	last = sum;
982	+	}
983	+	} finally {
984	+	if (retries > RETRIES_BEFORE_LOCK) {
985	+	for (int j = 0; j < segments.length; ++j)
986	+	segmentAt(segments, j).unlock();
987		}
692	–	if (cleanSweep)
693	–	return false;
988		}
989	+	return found;
990		}
991
992		/**
993		* Legacy method testing if some key maps into the specified value
994		* in this table.  This method is identical in functionality to
995	<	* {@link #containsValue}, and  exists solely to ensure
995	>	* {@link #containsValue}, and exists solely to ensure
996		* full compatibility with class {@link java.util.Hashtable},
997		* which supported this method prior to introduction of the
998		* Java Collections framework.
999	<
1000	<	* @param      value   a value to search for.
1001	<	* @return     <tt>true</tt> if and only if some key maps to the
1002	<	*             <tt>value</tt> argument in this table as
1003	<	*             determined by the <tt>equals</tt> method;
1004	<	*             <tt>false</tt> otherwise.
1005	<	* @throws  NullPointerException  if the value is <tt>null</tt>.
999	>	*
1000	>	* @param  value a value to search for
1001	>	* @return <tt>true</tt> if and only if some key maps to the
1002	>	*         <tt>value</tt> argument in this table as
1003	>	*         determined by the <tt>equals</tt> method;
1004	>	*         <tt>false</tt> otherwise
1005	>	* @throws NullPointerException if the specified value is null
1006		*/
1007		public boolean contains(Object value) {
1008		return containsValue(value);
1009		}
1010
1011		/**
1012	<	* Maps the specified <tt>key</tt> to the specified
1013	<	* <tt>value</tt> in this table. Neither the key nor the
719	<	* value can be <tt>null</tt>. <p>
1012	>	* Maps the specified key to the specified value in this table.
1013	>	* Neither the key nor the value can be null.
1014		*
1015	<	* The value can be retrieved by calling the <tt>get</tt> method
1015	>	* <p> The value can be retrieved by calling the <tt>get</tt> method
1016		* with a key that is equal to the original key.
1017		*
1018	<	* @param      key     the table key.
1019	<	* @param      value   the value.
1020	<	* @return     the previous value of the specified key in this table,
1021	<	*             or <tt>null</tt> if it did not have one.
1022	<	* @throws  NullPointerException  if the key or value is
729	<	*               <tt>null</tt>.
1018	>	* @param key key with which the specified value is to be associated
1019	>	* @param value value to be associated with the specified key
1020	>	* @return the previous value associated with <tt>key</tt>, or
1021	>	*         <tt>null</tt> if there was no mapping for <tt>key</tt>
1022	>	* @throws NullPointerException if the specified key or value is null
1023		*/
1024	+	@SuppressWarnings("unchecked")
1025		public V put(K key, V value) {
1026	+	Segment<K,V> s;
1027		if (value == null)
1028		throw new NullPointerException();
1029	<	int hash = hash(key);
1030	<	return segmentFor(hash).put(key, hash, value, false);
1029	>	int hash = hash(key.hashCode());
1030	>	int j = (hash >>> segmentShift) & segmentMask;
1031	>	if ((s = (Segment<K,V>)UNSAFE.getObject          // nonvolatile; recheck
1032	>	(segments, (j << SSHIFT) + SBASE)) == null) //  in ensureSegment
1033	>	s = ensureSegment(j);
1034	>	return s.put(key, hash, value, false);
1035		}
1036
1037		/**
1038	<	* If the specified key is not already associated
740	<	* with a value, associate it with the given value.
741	<	* This is equivalent to
742	<	* <pre>
743	<	*   if (!map.containsKey(key))
744	<	*      return map.put(key, value);
745	<	*   else
746	<	*      return map.get(key);
747	<	* </pre>
748	<	* Except that the action is performed atomically.
749	<	* @param key key with which the specified value is to be associated.
750	<	* @param value value to be associated with the specified key.
751	<	* @return previous value associated with specified key, or <tt>null</tt>
752	<	*         if there was no mapping for key.  A <tt>null</tt> return can
753	<	*         also indicate that the map previously associated <tt>null</tt>
754	<	*         with the specified key, if the implementation supports
755	<	*         <tt>null</tt> values.
756	<	*
757	<	* @throws UnsupportedOperationException if the <tt>put</tt> operation is
758	<	*            not supported by this map.
759	<	* @throws ClassCastException if the class of the specified key or value
760	<	*            prevents it from being stored in this map.
761	<	* @throws NullPointerException if the specified key or value is
762	<	*            <tt>null</tt>.
1038	>	* {@inheritDoc}
1039		*
1040	<	**/
1040	>	* @return the previous value associated with the specified key,
1041	>	*         or <tt>null</tt> if there was no mapping for the key
1042	>	* @throws NullPointerException if the specified key or value is null
1043	>	*/
1044	>	@SuppressWarnings("unchecked")
1045		public V putIfAbsent(K key, V value) {
1046	+	Segment<K,V> s;
1047		if (value == null)
1048		throw new NullPointerException();
1049	<	int hash = hash(key);
1050	<	return segmentFor(hash).put(key, hash, value, true);
1049	>	int hash = hash(key.hashCode());
1050	>	int j = (hash >>> segmentShift) & segmentMask;
1051	>	if ((s = (Segment<K,V>)UNSAFE.getObject
1052	>	(segments, (j << SSHIFT) + SBASE)) == null)
1053	>	s = ensureSegment(j);
1054	>	return s.put(key, hash, value, true);
1055		}
1056
772	–
1057		/**
1058		* Copies all of the mappings from the specified map to this one.
775	–	*
1059		* These mappings replace any mappings that this map had for any of the
1060	<	* keys currently in the specified Map.
1060	>	* keys currently in the specified map.
1061		*
1062	<	* @param t Mappings to be stored in this map.
1062	>	* @param m mappings to be stored in this map
1063		*/
1064	<	public void putAll(Map<? extends K, ? extends V> t) {
1065	<	for (Iterator<Map.Entry<? extends K, ? extends V>> it = (Iterator<Map.Entry<? extends K, ? extends V>>) t.entrySet().iterator(); it.hasNext(); ) {
783	<	Entry<? extends K, ? extends V> e = it.next();
1064	>	public void putAll(Map<? extends K, ? extends V> m) {
1065	>	for (Map.Entry<? extends K, ? extends V> e : m.entrySet())
1066		put(e.getKey(), e.getValue());
785	–	}
1067		}
1068
1069		/**
1070	<	* Removes the key (and its corresponding value) from this
1071	<	* table. This method does nothing if the key is not in the table.
1070	>	* Removes the key (and its corresponding value) from this map.
1071	>	* This method does nothing if the key is not in the map.
1072		*
1073	<	* @param   key   the key that needs to be removed.
1074	<	* @return  the value to which the key had been mapped in this table,
1075	<	*          or <tt>null</tt> if the key did not have a mapping.
1076	<	* @throws  NullPointerException  if the key is
796	<	*               <tt>null</tt>.
1073	>	* @param  key the key that needs to be removed
1074	>	* @return the previous value associated with <tt>key</tt>, or
1075	>	*         <tt>null</tt> if there was no mapping for <tt>key</tt>
1076	>	* @throws NullPointerException if the specified key is null
1077		*/
1078		public V remove(Object key) {
1079	<	int hash = hash(key);
1080	<	return segmentFor(hash).remove(key, hash, null);
1079	>	int hash = hash(key.hashCode());
1080	>	Segment<K,V> s = segmentForHash(hash);
1081	>	return s == null ? null : s.remove(key, hash, null);
1082		}
1083
1084		/**
1085	<	* Remove entry for key only if currently mapped to given value.
1086	<	* Acts as
1087	<	* <pre>
807	<	*  if (map.get(key).equals(value)) {
808	<	*     map.remove(key);
809	<	*     return true;
810	<	* } else return false;
811	<	* </pre>
812	<	* except that the action is performed atomically.
813	<	* @param key key with which the specified value is associated.
814	<	* @param value value associated with the specified key.
815	<	* @return true if the value was removed
816	<	* @throws NullPointerException if the specified key is
817	<	*            <tt>null</tt>.
1085	>	* {@inheritDoc}
1086	>	*
1087	>	* @throws NullPointerException if the specified key is null
1088		*/
1089		public boolean remove(Object key, Object value) {
1090	<	int hash = hash(key);
1091	<	return segmentFor(hash).remove(key, hash, value) != null;
1090	>	int hash = hash(key.hashCode());
1091	>	Segment<K,V> s;
1092	>	return value != null && (s = segmentForHash(hash)) != null &&
1093	>	s.remove(key, hash, value) != null;
1094		}
1095
1096		/**
1097	<	* Removes all mappings from this map.
1097	>	* {@inheritDoc}
1098	>	*
1099	>	* @throws NullPointerException if any of the arguments are null
1100		*/
1101	<	public void clear() {
1102	<	for (int i = 0; i < segments.length; ++i)
1103	<	segments[i].clear();
1101	>	public boolean replace(K key, V oldValue, V newValue) {
1102	>	int hash = hash(key.hashCode());
1103	>	if (oldValue == null || newValue == null)
1104	>	throw new NullPointerException();
1105	>	Segment<K,V> s = segmentForHash(hash);
1106	>	return s != null && s.replace(key, hash, oldValue, newValue);
1107		}
1108
832	–
1109		/**
1110	<	* Returns a shallow copy of this
835	<	* <tt>ConcurrentHashMap</tt> instance: the keys and
836	<	* values themselves are not cloned.
1110	>	* {@inheritDoc}
1111		*
1112	<	* @return a shallow copy of this map.
1112	>	* @return the previous value associated with the specified key,
1113	>	*         or <tt>null</tt> if there was no mapping for the key
1114	>	* @throws NullPointerException if the specified key or value is null
1115		*/
1116	<	public Object clone() {
1117	<	// We cannot call super.clone, since it would share final
1118	<	// segments array, and there's no way to reassign finals.
1116	>	public V replace(K key, V value) {
1117	>	int hash = hash(key.hashCode());
1118	>	if (value == null)
1119	>	throw new NullPointerException();
1120	>	Segment<K,V> s = segmentForHash(hash);
1121	>	return s == null ? null : s.replace(key, hash, value);
1122	>	}
1123
1124	<	float lf = segments[0].loadFactor;
1125	<	int segs = segments.length;
1126	<	int cap = (int)(size() / lf);
1127	<	if (cap < segs) cap = segs;
1128	<	ConcurrentHashMap<K,V> t = new ConcurrentHashMap<K,V>(cap, lf, segs);
1129	<	t.putAll(this);
1130	<	return t;
1124	>	/**
1125	>	* Removes all of the mappings from this map.
1126	>	*/
1127	>	public void clear() {
1128	>	final Segment<K,V>[] segments = this.segments;
1129	>	for (int j = 0; j < segments.length; ++j) {
1130	>	Segment<K,V> s = segmentAt(segments, j);
1131	>	if (s != null)
1132	>	s.clear();
1133	>	}
1134		}
1135
1136		/**
1137	<	* Returns a set view of the keys contained in this map.  The set is
1138	<	* backed by the map, so changes to the map are reflected in the set, and
1139	<	* vice-versa.  The set supports element removal, which removes the
1140	<	* corresponding mapping from this map, via the <tt>Iterator.remove</tt>,
1141	<	* <tt>Set.remove</tt>, <tt>removeAll</tt>, <tt>retainAll</tt>, and
1142	<	* <tt>clear</tt> operations.  It does not support the <tt>add</tt> or
1137	>	* Returns a {@link Set} view of the keys contained in this map.
1138	>	* The set is backed by the map, so changes to the map are
1139	>	* reflected in the set, and vice-versa.  The set supports element
1140	>	* removal, which removes the corresponding mapping from this map,
1141	>	* via the <tt>Iterator.remove</tt>, <tt>Set.remove</tt>,
1142	>	* <tt>removeAll</tt>, <tt>retainAll</tt>, and <tt>clear</tt>
1143	>	* operations.  It does not support the <tt>add</tt> or
1144		* <tt>addAll</tt> operations.
1145	<	* The returned <tt>iterator</tt> is a "weakly consistent" iterator that
1146	<	* will never throw {@link java.util.ConcurrentModificationException},
1145	>	*
1146	>	* <p>The view's <tt>iterator</tt> is a "weakly consistent" iterator
1147	>	* that will never throw {@link ConcurrentModificationException},
1148		* and guarantees to traverse elements as they existed upon
1149		* construction of the iterator, and may (but is not guaranteed to)
1150		* reflect any modifications subsequent to construction.
866	–	*
867	–	* @return a set view of the keys contained in this map.
1151		*/
1152		public Set<K> keySet() {
1153		Set<K> ks = keySet;
1154		return (ks != null) ? ks : (keySet = new KeySet());
1155		}
1156
874	–
1157		/**
1158	<	* Returns a collection view of the values contained in this map.  The
1159	<	* collection is backed by the map, so changes to the map are reflected in
1160	<	* the collection, and vice-versa.  The collection supports element
1161	<	* removal, which removes the corresponding mapping from this map, via the
1162	<	* <tt>Iterator.remove</tt>, <tt>Collection.remove</tt>,
1163	<	* <tt>removeAll</tt>, <tt>retainAll</tt>, and <tt>clear</tt> operations.
1164	<	* It does not support the <tt>add</tt> or <tt>addAll</tt> operations.
1165	<	* The returned <tt>iterator</tt> is a "weakly consistent" iterator that
1166	<	* will never throw {@link java.util.ConcurrentModificationException},
1158	>	* Returns a {@link Collection} view of the values contained in this map.
1159	>	* The collection is backed by the map, so changes to the map are
1160	>	* reflected in the collection, and vice-versa.  The collection
1161	>	* supports element removal, which removes the corresponding
1162	>	* mapping from this map, via the <tt>Iterator.remove</tt>,
1163	>	* <tt>Collection.remove</tt>, <tt>removeAll</tt>,
1164	>	* <tt>retainAll</tt>, and <tt>clear</tt> operations.  It does not
1165	>	* support the <tt>add</tt> or <tt>addAll</tt> operations.
1166	>	*
1167	>	* <p>The view's <tt>iterator</tt> is a "weakly consistent" iterator
1168	>	* that will never throw {@link ConcurrentModificationException},
1169		* and guarantees to traverse elements as they existed upon
1170		* construction of the iterator, and may (but is not guaranteed to)
1171		* reflect any modifications subsequent to construction.
888	–	*
889	–	* @return a collection view of the values contained in this map.
1172		*/
1173		public Collection<V> values() {
1174		Collection<V> vs = values;
1175		return (vs != null) ? vs : (values = new Values());
1176		}
1177
896	–
1178		/**
1179	<	* Returns a collection view of the mappings contained in this map.  Each
1180	<	* element in the returned collection is a <tt>Map.Entry</tt>.  The
1181	<	* collection is backed by the map, so changes to the map are reflected in
1182	<	* the collection, and vice-versa.  The collection supports element
1183	<	* removal, which removes the corresponding mapping from the map, via the
1184	<	* <tt>Iterator.remove</tt>, <tt>Collection.remove</tt>,
1185	<	* <tt>removeAll</tt>, <tt>retainAll</tt>, and <tt>clear</tt> operations.
1186	<	* It does not support the <tt>add</tt> or <tt>addAll</tt> operations.
1187	<	* The returned <tt>iterator</tt> is a "weakly consistent" iterator that
1188	<	* will never throw {@link java.util.ConcurrentModificationException},
1179	>	* Returns a {@link Set} view of the mappings contained in this map.
1180	>	* The set is backed by the map, so changes to the map are
1181	>	* reflected in the set, and vice-versa.  The set supports element
1182	>	* removal, which removes the corresponding mapping from the map,
1183	>	* via the <tt>Iterator.remove</tt>, <tt>Set.remove</tt>,
1184	>	* <tt>removeAll</tt>, <tt>retainAll</tt>, and <tt>clear</tt>
1185	>	* operations.  It does not support the <tt>add</tt> or
1186	>	* <tt>addAll</tt> operations.
1187	>	*
1188	>	* <p>The view's <tt>iterator</tt> is a "weakly consistent" iterator
1189	>	* that will never throw {@link ConcurrentModificationException},
1190		* and guarantees to traverse elements as they existed upon
1191		* construction of the iterator, and may (but is not guaranteed to)
1192		* reflect any modifications subsequent to construction.
911	–	*
912	–	* @return a collection view of the mappings contained in this map.
1193		*/
1194		public Set<Map.Entry<K,V>> entrySet() {
1195		Set<Map.Entry<K,V>> es = entrySet;
1196	<	return (es != null) ? es : (entrySet = (Set<Map.Entry<K,V>>) (Set) new EntrySet());
1196	>	return (es != null) ? es : (entrySet = new EntrySet());
1197		}
1198
919	–
1199		/**
1200		* Returns an enumeration of the keys in this table.
1201		*
1202	<	* @return  an enumeration of the keys in this table.
1203	<	* @see     #keySet
1202	>	* @return an enumeration of the keys in this table
1203	>	* @see #keySet()
1204		*/
1205		public Enumeration<K> keys() {
1206		return new KeyIterator();
#	Line 929 | Line 1208 | public class ConcurrentHashMap<K, V> ext
1208
1209		/**
1210		* Returns an enumeration of the values in this table.
932	–	* Use the Enumeration methods on the returned object to fetch the elements
933	–	* sequentially.
1211		*
1212	<	* @return  an enumeration of the values in this table.
1213	<	* @see     #values
1212	>	* @return an enumeration of the values in this table
1213	>	* @see #values()
1214		*/
1215		public Enumeration<V> elements() {
1216		return new ValueIterator();
#	Line 941 | Line 1218 | public class ConcurrentHashMap<K, V> ext
1218
1219		/* ---------------- Iterator Support -------------- */
1220
1221	<	private abstract class HashIterator {
1222	<	private int nextSegmentIndex;
1223	<	private int nextTableIndex;
1224	<	private HashEntry[] currentTable;
1225	<	private HashEntry<K, V> nextEntry;
1226	<	private HashEntry<K, V> lastReturned;
1221	>	abstract class HashIterator {
1222	>	int nextSegmentIndex;
1223	>	int nextTableIndex;
1224	>	HashEntry<K,V>[] currentTable;
1225	>	HashEntry<K, V> nextEntry;
1226	>	HashEntry<K, V> lastReturned;
1227
1228	<	private HashIterator() {
1228	>	HashIterator() {
1229		nextSegmentIndex = segments.length - 1;
1230		nextTableIndex = -1;
1231		advance();
1232		}
1233
1234	<	public boolean hasMoreElements() { return hasNext(); }
1235	<
1236	<	private void advance() {
1237	<	if (nextEntry != null && (nextEntry = nextEntry.next) != null)
1238	<	return;
1239	<
1240	<	while (nextTableIndex >= 0) {
1241	<	if ( (nextEntry = (HashEntry<K,V>)currentTable[nextTableIndex--]) != null)
1242	<	return;
1243	<	}
1244	<
1245	<	while (nextSegmentIndex >= 0) {
1246	<	Segment<K,V> seg = (Segment<K,V>)segments[nextSegmentIndex--];
1247	<	if (seg.count != 0) {
1248	<	currentTable = seg.table;
972	<	for (int j = currentTable.length - 1; j >= 0; --j) {
973	<	if ( (nextEntry = (HashEntry<K,V>)currentTable[j]) != null) {
974	<	nextTableIndex = j - 1;
975	<	return;
976	<	}
977	<	}
1234	>	/**
1235	>	* Sets nextEntry to first node of next non-empty table
1236	>	* (in backwards order, to simplify checks).
1237	>	*/
1238	>	final void advance() {
1239	>	for (;;) {
1240	>	if (nextTableIndex >= 0) {
1241	>	if ((nextEntry = entryAt(currentTable,
1242	>	nextTableIndex--)) != null)
1243	>	break;
1244	>	}
1245	>	else if (nextSegmentIndex >= 0) {
1246	>	Segment<K,V> seg = segmentAt(segments, nextSegmentIndex--);
1247	>	if (seg != null && (currentTable = seg.table) != null)
1248	>	nextTableIndex = currentTable.length - 1;
1249		}
1250	+	else
1251	+	break;
1252		}
1253		}
1254
1255	<	public boolean hasNext() { return nextEntry != null; }
1256	<
1257	<	HashEntry<K,V> nextEntry() {
985	<	if (nextEntry == null)
1255	>	final HashEntry<K,V> nextEntry() {
1256	>	HashEntry<K,V> e = nextEntry;
1257	>	if (e == null)
1258		throw new NoSuchElementException();
1259	<	lastReturned = nextEntry;
1260	<	advance();
1261	<	return lastReturned;
1259	>	lastReturned = e; // cannot assign until after null check
1260	>	if ((nextEntry = e.next) == null)
1261	>	advance();
1262	>	return e;
1263		}
1264
1265	<	public void remove() {
1265	>	public final boolean hasNext() { return nextEntry != null; }
1266	>	public final boolean hasMoreElements() { return nextEntry != null; }
1267	>
1268	>	public final void remove() {
1269		if (lastReturned == null)
1270		throw new IllegalStateException();
1271		ConcurrentHashMap.this.remove(lastReturned.key);
#	Line 997 | Line 1273 | public class ConcurrentHashMap<K, V> ext
1273		}
1274		}
1275
1276	<	private class KeyIterator extends HashIterator implements Iterator<K>, Enumeration<K> {
1277	<	public K next() { return super.nextEntry().key; }
1278	<	public K nextElement() { return super.nextEntry().key; }
1276	>	final class KeyIterator
1277	>	extends HashIterator
1278	>	implements Iterator<K>, Enumeration<K>
1279	>	{
1280	>	public final K next()        { return super.nextEntry().key; }
1281	>	public final K nextElement() { return super.nextEntry().key; }
1282	>	}
1283	>
1284	>	final class ValueIterator
1285	>	extends HashIterator
1286	>	implements Iterator<V>, Enumeration<V>
1287	>	{
1288	>	public final V next()        { return super.nextEntry().value; }
1289	>	public final V nextElement() { return super.nextEntry().value; }
1290		}
1291
1292	<	private class ValueIterator extends HashIterator implements Iterator<V>, Enumeration<V> {
1293	<	public V next() { return super.nextEntry().value; }
1294	<	public V nextElement() { return super.nextEntry().value; }
1292	>	/**
1293	>	* Custom Entry class used by EntryIterator.next(), that relays
1294	>	* setValue changes to the underlying map.
1295	>	*/
1296	>	final class WriteThroughEntry
1297	>	extends AbstractMap.SimpleEntry<K,V>
1298	>	{
1299	>	WriteThroughEntry(K k, V v) {
1300	>	super(k,v);
1301	>	}
1302	>
1303	>	/**
1304	>	* Sets our entry's value and writes through to the map. The
1305	>	* value to return is somewhat arbitrary here. Since a
1306	>	* WriteThroughEntry does not necessarily track asynchronous
1307	>	* changes, the most recent "previous" value could be
1308	>	* different from what we return (or could even have been
1309	>	* removed in which case the put will re-establish). We do not
1310	>	* and cannot guarantee more.
1311	>	*/
1312	>	public V setValue(V value) {
1313	>	if (value == null) throw new NullPointerException();
1314	>	V v = super.setValue(value);
1315	>	ConcurrentHashMap.this.put(getKey(), value);
1316	>	return v;
1317	>	}
1318		}
1319
1320	<	private class EntryIterator extends HashIterator implements Iterator<Entry<K,V>> {
1321	<	public Map.Entry<K,V> next() { return super.nextEntry(); }
1320	>	final class EntryIterator
1321	>	extends HashIterator
1322	>	implements Iterator<Entry<K,V>>
1323	>	{
1324	>	public Map.Entry<K,V> next() {
1325	>	HashEntry<K,V> e = super.nextEntry();
1326	>	return new WriteThroughEntry(e.key, e.value);
1327	>	}
1328		}
1329
1330	<	private class KeySet extends AbstractSet<K> {
1330	>	final class KeySet extends AbstractSet<K> {
1331		public Iterator<K> iterator() {
1332		return new KeyIterator();
1333		}
1334		public int size() {
1335		return ConcurrentHashMap.this.size();
1336		}
1337	+	public boolean isEmpty() {
1338	+	return ConcurrentHashMap.this.isEmpty();
1339	+	}
1340		public boolean contains(Object o) {
1341		return ConcurrentHashMap.this.containsKey(o);
1342		}
#	Line 1029 | Line 1348 | public class ConcurrentHashMap<K, V> ext
1348		}
1349		}
1350
1351	<	private class Values extends AbstractCollection<V> {
1351	>	final class Values extends AbstractCollection<V> {
1352		public Iterator<V> iterator() {
1353		return new ValueIterator();
1354		}
1355		public int size() {
1356		return ConcurrentHashMap.this.size();
1357		}
1358	+	public boolean isEmpty() {
1359	+	return ConcurrentHashMap.this.isEmpty();
1360	+	}
1361		public boolean contains(Object o) {
1362		return ConcurrentHashMap.this.containsValue(o);
1363		}
#	Line 1044 | Line 1366 | public class ConcurrentHashMap<K, V> ext
1366		}
1367		}
1368
1369	<	private class EntrySet extends AbstractSet<Map.Entry<K,V>> {
1369	>	final class EntrySet extends AbstractSet<Map.Entry<K,V>> {
1370		public Iterator<Map.Entry<K,V>> iterator() {
1371		return new EntryIterator();
1372		}
1373		public boolean contains(Object o) {
1374		if (!(o instanceof Map.Entry))
1375		return false;
1376	<	Map.Entry<K,V> e = (Map.Entry<K,V>)o;
1376	>	Map.Entry<?,?> e = (Map.Entry<?,?>)o;
1377		V v = ConcurrentHashMap.this.get(e.getKey());
1378		return v != null && v.equals(e.getValue());
1379		}
1380		public boolean remove(Object o) {
1381		if (!(o instanceof Map.Entry))
1382		return false;
1383	<	Map.Entry<K,V> e = (Map.Entry<K,V>)o;
1383	>	Map.Entry<?,?> e = (Map.Entry<?,?>)o;
1384		return ConcurrentHashMap.this.remove(e.getKey(), e.getValue());
1385		}
1386		public int size() {
1387		return ConcurrentHashMap.this.size();
1388		}
1389	+	public boolean isEmpty() {
1390	+	return ConcurrentHashMap.this.isEmpty();
1391	+	}
1392		public void clear() {
1393		ConcurrentHashMap.this.clear();
1394		}
#	Line 1072 | Line 1397 | public class ConcurrentHashMap<K, V> ext
1397		/* ---------------- Serialization Support -------------- */
1398
1399		/**
1400	<	* Save the state of the <tt>ConcurrentHashMap</tt>
1401	<	* instance to a stream (i.e.,
1077	<	* serialize it).
1400	>	* Saves the state of the <tt>ConcurrentHashMap</tt> instance to a
1401	>	* stream (i.e., serializes it).
1402		* @param s the stream
1403		* @serialData
1404		* the key (Object) and value (Object)
1405		* for each key-value mapping, followed by a null pair.
1406		* The key-value mappings are emitted in no particular order.
1407		*/
1408	<	private void writeObject(java.io.ObjectOutputStream s) throws IOException  {
1408	>	private void writeObject(java.io.ObjectOutputStream s) throws IOException {
1409	>	// force all segments for serialization compatibility
1410	>	for (int k = 0; k < segments.length; ++k)
1411	>	ensureSegment(k);
1412		s.defaultWriteObject();
1413
1414	+	final Segment<K,V>[] segments = this.segments;
1415		for (int k = 0; k < segments.length; ++k) {
1416	<	Segment<K,V> seg = (Segment<K,V>)segments[k];
1416	>	Segment<K,V> seg = segmentAt(segments, k);
1417		seg.lock();
1418		try {
1419	<	HashEntry[] tab = seg.table;
1419	>	HashEntry<K,V>[] tab = seg.table;
1420		for (int i = 0; i < tab.length; ++i) {
1421	<	for (HashEntry<K,V> e = (HashEntry<K,V>)tab[i]; e != null; e = e.next) {
1421	>	HashEntry<K,V> e;
1422	>	for (e = entryAt(tab, i); e != null; e = e.next) {
1423		s.writeObject(e.key);
1424		s.writeObject(e.value);
1425		}
#	Line 1104 | Line 1433 | public class ConcurrentHashMap<K, V> ext
1433		}
1434
1435		/**
1436	<	* Reconstitute the <tt>ConcurrentHashMap</tt>
1437	<	* instance from a stream (i.e.,
1109	<	* deserialize it).
1436	>	* Reconstitutes the <tt>ConcurrentHashMap</tt> instance from a
1437	>	* stream (i.e., deserializes it).
1438		* @param s the stream
1439		*/
1440	+	@SuppressWarnings("unchecked")
1441		private void readObject(java.io.ObjectInputStream s)
1442	<	throws IOException, ClassNotFoundException  {
1442	>	throws IOException, ClassNotFoundException {
1443		s.defaultReadObject();
1444
1445	<	// Initialize each segment to be minimally sized, and let grow.
1446	<	for (int i = 0; i < segments.length; ++i) {
1447	<	segments[i].setTable(new HashEntry[1]);
1445	>	// Re-initialize segments to be minimally sized, and let grow.
1446	>	int cap = MIN_SEGMENT_TABLE_CAPACITY;
1447	>	final Segment<K,V>[] segments = this.segments;
1448	>	for (int k = 0; k < segments.length; ++k) {
1449	>	Segment<K,V> seg = segments[k];
1450	>	if (seg != null) {
1451	>	seg.threshold = (int)(cap * seg.loadFactor);
1452	>	seg.table = (HashEntry<K,V>[]) new HashEntry[cap];
1453	>	}
1454		}
1455
1456		// Read the keys and values, and put the mappings in the table
#	Line 1127 | Line 1462 | public class ConcurrentHashMap<K, V> ext
1462		put(key, value);
1463		}
1464		}
1130	–	}
1465
1466	+	// Unsafe mechanics
1467	+	private static final sun.misc.Unsafe UNSAFE;
1468	+	private static final long SBASE;
1469	+	private static final int SSHIFT;
1470	+	private static final long TBASE;
1471	+	private static final int TSHIFT;
1472	+
1473	+	static {
1474	+	int ss, ts;
1475	+	try {
1476	+	UNSAFE = sun.misc.Unsafe.getUnsafe();
1477	+	Class tc = HashEntry[].class;
1478	+	Class sc = Segment[].class;
1479	+	TBASE = UNSAFE.arrayBaseOffset(tc);
1480	+	SBASE = UNSAFE.arrayBaseOffset(sc);
1481	+	ts = UNSAFE.arrayIndexScale(tc);
1482	+	ss = UNSAFE.arrayIndexScale(sc);
1483	+	} catch (Exception e) {
1484	+	throw new Error(e);
1485	+	}
1486	+	if ((ss & (ss-1)) != 0 || (ts & (ts-1)) != 0)
1487	+	throw new Error("data type scale not a power of two");
1488	+	SSHIFT = 31 - Integer.numberOfLeadingZeros(ss);
1489	+	TSHIFT = 31 - Integer.numberOfLeadingZeros(ts);
1490	+	}
1491	+
1492	+	}

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines