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 1.92 by dl, Sat Dec 2 20:55:01 2006 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, as explained at
4	<	* http://creativecommons.org/licenses/publicdomain
4	>	* http://creativecommons.org/publicdomain/zero/1.0/
5		*/
6
7		package java.util.concurrent;
#	Line 76 | Line 76 | public class ConcurrentHashMap<K, V> ext
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 -------------- */
#	Line 108 | Line 126 | public class ConcurrentHashMap<K, V> ext
126		static final int MAXIMUM_CAPACITY = 1 << 30;
127
128		/**
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 int MIN_SEGMENT_TABLE_CAPACITY = 2;
134	+
135	+	/**
136		* The maximum number of segments to allow; used to bound
137	<	* constructor arguments.
137	>	* constructor arguments. Must be power of two less than 1 << 24.
138		*/
139		static final int MAX_SEGMENTS = 1 << 16; // slightly conservative
140
#	Line 135 | Line 160 | public class ConcurrentHashMap<K, V> ext
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		final Segment<K,V>[] segments;
166
#	Line 143 | Line 168 | public class ConcurrentHashMap<K, V> ext
168		transient Set<Map.Entry<K,V>> entrySet;
169		transient Collection<V> values;
170
146	–	/* ---------------- Small Utilities -------------- */
147	–
148	–	/**
149	–	* Applies a supplemental hash function to a given hashCode, which
150	–	* defends against poor quality hash functions.  This is critical
151	–	* because ConcurrentHashMap uses power-of-two length hash tables,
152	–	* that otherwise encounter collisions for hashCodes that do not
153	–	* differ in lower bits.
154	–	*/
155	–	private static int hash(int h) {
156	–	// Spread bits to regularize both segment and index locations,
157	–	// using variant of Jenkins's shift-based hash.
158	–	h += ~(h << 13);
159	–	h ^= h >>> 7;
160	–	h += h << 3;
161	–	h ^= h >>> 17;
162	–	h += h << 5;
163	–	return h;
164	–	}
165	–
166	–	/**
167	–	* Returns the segment that should be used for key with given hash
168	–	* @param hash the hash code for the key
169	–	* @return the segment
170	–	*/
171	–	final Segment<K,V> segmentFor(int hash) {
172	–	return segments[(hash >>> segmentShift) & segmentMask];
173	–	}
174	–
175	–	/* ---------------- Inner Classes -------------- */
176	–
171		/**
172		* ConcurrentHashMap list entry. Note that this is never exported
173		* out as a user-visible Map.Entry.
180	–	*
181	–	* Because the value field is volatile, not final, it is legal wrt
182	–	* the Java Memory Model for an unsynchronized reader to see null
183	–	* instead of initial value when read via a data race.  Although a
184	–	* reordering leading to this is not likely to ever actually
185	–	* occur, the Segment.readValueUnderLock method is used as a
186	–	* backup in case a null (pre-initialized) value is ever seen in
187	–	* an unsynchronized access method.
174		*/
175		static final class HashEntry<K,V> {
190	–	final K key;
176		final int hash;
177	+	final K key;
178		volatile V value;
179	<	final HashEntry<K,V> next;
179	>	volatile HashEntry<K,V> next;
180
181	<	HashEntry(K key, int hash, HashEntry<K,V> next, V value) {
196	<	this.key = key;
181	>	HashEntry(int hash, K key, V value, HashEntry<K,V> next) {
182		this.hash = hash;
183	<	this.next = next;
183	>	this.key = key;
184		this.value = value;
185	+	this.next = next;
186		}
187
188	<	@SuppressWarnings("unchecked")
189	<	static final <K,V> HashEntry<K,V>[] newArray(int i) {
190	<	return new HashEntry[i];
191	<	}
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	>	* 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	>	/**
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		/**
#	Line 212 | Line 254 | public class ConcurrentHashMap<K, V> ext
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
221	<	* created to replace them. This works well for hash tables
222	<	* since the bin lists tend to be short. (The average length
223	<	* is less than two for the default load factor threshold.)
224	<	*
225	<	* Read operations can thus proceed without locking, but rely
226	<	* on selected uses of volatiles to ensure that completed
227	<	* write operations performed by other threads are
228	<	* noticed. For most purposes, the "count" field, tracking the
229	<	* number of elements, serves as that volatile variable
230	<	* ensuring visibility.  This is convenient because this field
231	<	* needs to be read in many read operations anyway:
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 (unsynchronized) read operations must first read the
265	<	*     "count" field, and should not look at table entries if
266	<	*     it is 0.
267	<	*
268	<	*   - All (synchronized) write operations should write to
269	<	*     the "count" field after structurally changing any bin.
270	<	*     The operations must not take any action that could even
271	<	*     momentarily cause a concurrent read operation to see
272	<	*     inconsistent data. This is made easier by the nature of
273	<	*     the read operations in Map. For example, no operation
274	<	*     can reveal that the table has grown but the threshold
275	<	*     has not yet been updated, so there are no atomicity
276	<	*     requirements for this with respect to reads.
277	<	*
278	<	* As a guide, all critical volatile reads and writes to the
279	<	* count field are marked 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.
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	>	* The per-segment table. Elements are accessed via
296	>	* entryAt/setEntryAt providing volatile semantics.
297	>	*/
298	>	transient volatile HashEntry<K,V>[] table;
299	>
300	>	/**
301	>	* The number of elements. Accessed only either within locks
302	>	* or among other volatile reads that maintain visibility.
303		*/
304	<	transient volatile int count;
304	>	transient int count;
305
306		/**
307	<	* Number of updates that alter the size of the table. This is
308	<	* used during bulk-read methods to make sure they see a
309	<	* consistent snapshot: If modCounts change during a traversal
310	<	* of segments computing size or checking containsValue, then
311	<	* we might have an inconsistent view of state so (usually)
264	<	* must retry.
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		transient int modCount;
314
#	Line 273 | Line 320 | public class ConcurrentHashMap<K, V> ext
320		transient int threshold;
321
322		/**
276	–	* The per-segment table.
277	–	*/
278	–	transient volatile HashEntry<K,V>[] table;
279	–
280	–	/**
323		* The load factor for the hash table.  Even though this value
324		* is same for all segments, it is replicated to avoid needing
325		* links to outer object.
#	Line 285 | Line 327 | public class ConcurrentHashMap<K, V> ext
327		*/
328		final float loadFactor;
329
330	<	Segment(int initialCapacity, float lf) {
331	<	loadFactor = lf;
332	<	setTable(HashEntry.<K,V>newArray(initialCapacity));
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	<	@SuppressWarnings("unchecked")
337	<	static final <K,V> Segment<K,V>[] newArray(int i) {
338	<	return new Segment[i];
339	<	}
297	<
298	<	/**
299	<	* Sets table to new HashEntry array.
300	<	* Call only while holding lock or in constructor.
301	<	*/
302	<	void setTable(HashEntry<K,V>[] newTable) {
303	<	threshold = (int)(newTable.length * loadFactor);
304	<	table = newTable;
305	<	}
306	<
307	<	/**
308	<	* Returns properly casted first entry of bin for given hash.
309	<	*/
310	<	HashEntry<K,V> getFirst(int hash) {
311	<	HashEntry<K,V>[] tab = table;
312	<	return tab[hash & (tab.length - 1)];
313	<	}
314	<
315	<	/**
316	<	* Reads value field of an entry under lock. Called if value
317	<	* field ever appears to be null. This is possible only if a
318	<	* compiler happens to reorder a HashEntry initialization with
319	<	* its table assignment, which is legal under memory model
320	<	* but is not known to ever occur.
321	<	*/
322	<	V readValueUnderLock(HashEntry<K,V> e) {
323	<	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 {
325	–	return e.value;
326	–	} finally {
327	–	unlock();
328	–	}
329	–	}
330	–
331	–	/* Specialized implementations of map methods */
332	–
333	–	V get(Object key, int hash) {
334	–	if (count != 0) { // read-volatile
335	–	HashEntry<K,V> e = getFirst(hash);
336	–	while (e != null) {
337	–	if (e.hash == hash && key.equals(e.key)) {
338	–	V v = e.value;
339	–	if (v != null)
340	–	return v;
341	–	return readValueUnderLock(e); // recheck
342	–	}
343	–	e = e.next;
344	–	}
345	–	}
346	–	return null;
347	–	}
348	–
349	–	boolean containsKey(Object key, int hash) {
350	–	if (count != 0) { // read-volatile
351	–	HashEntry<K,V> e = getFirst(hash);
352	–	while (e != null) {
353	–	if (e.hash == hash && key.equals(e.key))
354	–	return true;
355	–	e = e.next;
356	–	}
357	–	}
358	–	return false;
359	–	}
360	–
361	–	boolean containsValue(Object value) {
362	–	if (count != 0) { // read-volatile
341		HashEntry<K,V>[] tab = table;
342	<	int len = tab.length;
343	<	for (int i = 0 ; i < len; i++) {
344	<	for (HashEntry<K,V> e = tab[i]; e != null; e = e.next) {
345	<	V v = e.value;
346	<	if (v == null) // recheck
347	<	v = readValueUnderLock(e);
348	<	if (value.equals(v))
349	<	return true;
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;
370	>	oldValue = null;
371	>	break;
372		}
373		}
374	–	}
375	–	return false;
376	–	}
377	–
378	–	boolean replace(K key, int hash, V oldValue, V newValue) {
379	–	lock();
380	–	try {
381	–	HashEntry<K,V> e = getFirst(hash);
382	–	while (e != null && (e.hash != hash || !key.equals(e.key)))
383	–	e = e.next;
384	–
385	–	boolean replaced = false;
386	–	if (e != null && oldValue.equals(e.value)) {
387	–	replaced = true;
388	–	e.value = newValue;
389	–	}
390	–	return replaced;
391	–	} finally {
392	–	unlock();
393	–	}
394	–	}
395	–
396	–	V replace(K key, int hash, V newValue) {
397	–	lock();
398	–	try {
399	–	HashEntry<K,V> e = getFirst(hash);
400	–	while (e != null && (e.hash != hash || !key.equals(e.key)))
401	–	e = e.next;
402	–
403	–	V oldValue = null;
404	–	if (e != null) {
405	–	oldValue = e.value;
406	–	e.value = newValue;
407	–	}
408	–	return oldValue;
409	–	} finally {
410	–	unlock();
411	–	}
412	–	}
413	–
414	–
415	–	V put(K key, int hash, V value, boolean onlyIfAbsent) {
416	–	lock();
417	–	try {
418	–	int c = count;
419	–	if (c++ > threshold) // ensure capacity
420	–	rehash();
421	–	HashEntry<K,V>[] tab = table;
422	–	int index = hash & (tab.length - 1);
423	–	HashEntry<K,V> first = tab[index];
424	–	HashEntry<K,V> e = first;
425	–	while (e != null && (e.hash != hash || !key.equals(e.key)))
426	–	e = e.next;
427	–
428	–	V oldValue;
429	–	if (e != null) {
430	–	oldValue = e.value;
431	–	if (!onlyIfAbsent)
432	–	e.value = value;
433	–	}
434	–	else {
435	–	oldValue = null;
436	–	++modCount;
437	–	tab[index] = new HashEntry<K,V>(key, hash, first, value);
438	–	count = c; // write-volatile
439	–	}
440	–	return oldValue;
374		} finally {
375		unlock();
376		}
377	+	return oldValue;
378		}
379
380	<	void rehash() {
381	<	HashEntry<K,V>[] oldTable = table;
382	<	int oldCapacity = oldTable.length;
383	<	if (oldCapacity >= MAXIMUM_CAPACITY)
384	<	return;
385	<
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	<	* threshold, 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<K,V>[] newTable = HashEntry.newArray(oldCapacity<<1);
404	<	threshold = (int)(newTable.length * loadFactor);
405	<	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++) {
470	–	// We need to guarantee that any existing reads of old Map can
471	–	//  proceed. So we cannot yet null out each bin.
410		HashEntry<K,V> e = oldTable[i];
473	–
411		if (e != null) {
412		HashEntry<K,V> next = e.next;
413		int idx = e.hash & sizeMask;
414	<
478	<	//  Single node on list
479	<	if (next == null)
414	>	if (next == null)   //  Single node on list
415		newTable[idx] = e;
416	<
482	<	else {
483	<	// 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 493 | Line 426 | public class ConcurrentHashMap<K, V> ext
426		}
427		}
428		newTable[lastIdx] = lastRun;
429	<
497	<	// 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;
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>(p.key, p.hash,
502	<	n, p.value);
435	>	newTable[k] = new HashEntry<K,V>(h, p.key, v, n);
436		}
437		}
438		}
439		}
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 {
516	–	int c = count - 1;
527		HashEntry<K,V>[] tab = table;
528	<	int index = hash & (tab.length - 1);
529	<	HashEntry<K,V> first = tab[index];
530	<	HashEntry<K,V> e = first;
531	<	while (e != null && (e.hash != hash || !key.equals(e.key)))
532	<	e = e.next;
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	>	}
548	>	pred = e;
549	>	e = next;
550	>	}
551	>	} finally {
552	>	unlock();
553	>	}
554	>	return oldValue;
555	>	}
556
557	<	V oldValue = null;
558	<	if (e != null) {
559	<	V v = e.value;
560	<	if (value == null || value.equals(v)) {
561	<	oldValue = v;
562	<	// All entries following removed node can stay
563	<	// in list, but all preceding ones need to be
564	<	// cloned.
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	<	HashEntry<K,V> newFirst = e.next;
534	<	for (HashEntry<K,V> p = first; p != e; p = p.next)
535	<	newFirst = new HashEntry<K,V>(p.key, p.hash,
536	<	newFirst, p.value);
537	<	tab[index] = newFirst;
538	<	count = c; // write-volatile
594	>	break;
595		}
596		}
541	–	return oldValue;
597		} finally {
598		unlock();
599		}
600	+	return oldValue;
601		}
602
603	<	void clear() {
604	<	if (count != 0) {
605	<	lock();
606	<	try {
607	<	HashEntry<K,V>[] tab = table;
608	<	for (int i = 0; i < tab.length ; i++)
609	<	tab[i] = null;
610	<	++modCount;
611	<	count = 0; // write-volatile
612	<	} finally {
613	<	unlock();
603	>	final void clear() {
604	>	lock();
605	>	try {
606	>	HashEntry<K,V>[] tab = table;
607	>	for (int i = 0; i < tab.length ; i++)
608	>	setEntryAt(tab, i, null);
609	>	++modCount;
610	>	count = 0;
611	>	} finally {
612	>	unlock();
613	>	}
614	>	}
615	>	}
616	>
617	>	// Accessing segments
618	>
619	>	/**
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	>	@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	>	/**
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	+	// Hash-based segment and entry accesses
666
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
#	Line 580 | Line 702 | public class ConcurrentHashMap<K, V> ext
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();
587	–
710		if (concurrencyLevel > MAX_SEGMENTS)
711		concurrencyLevel = MAX_SEGMENTS;
590	–
712		// Find power-of-two sizes best matching arguments
713		int sshift = 0;
714		int ssize = 1;
#	Line 595 | Line 716 | public class ConcurrentHashMap<K, V> ext
716		++sshift;
717		ssize <<= 1;
718		}
719	<	segmentShift = 32 - sshift;
720	<	segmentMask = ssize - 1;
600	<	this.segments = Segment.newArray(ssize);
601	<
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		/**
#	Line 672 | Line 795 | public class ConcurrentHashMap<K, V> ext
795		* @return <tt>true</tt> if this map contains no key-value mappings
796		*/
797		public boolean isEmpty() {
675	–	final Segment<K,V>[] segments = this.segments;
798		/*
799	<	* We 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
691	<	mcsum += mc[i] = segments[i].modCount;
692	<	}
693	<	// If mcsum happens to be zero, then we know we got a snapshot
694	<	// before any modifications at all were made.  This is
695	<	// probably common enough to bother tracking.
696	<	if (mcsum != 0) {
697	<	for (int i = 0; i < segments.length; ++i) {
698	<	if (segments[i].count != 0 ||
699	<	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
#	Line 711 | Line 837 | public class ConcurrentHashMap<K, V> ext
837		* @return the number of key-value mappings in this map
838		*/
839		public int size() {
714	–	final Segment<K,V>[] segments = this.segments;
715	–	long sum = 0;
716	–	long check = 0;
717	–	int[] mc = new int[segments.length];
840		// Try a few times to get accurate count. On failure due to
841		// continuous async changes in table, resort to locking.
842	<	for (int k = 0; k < RETRIES_BEFORE_LOCK; ++k) {
843	<	check = 0;
844	<	sum = 0;
845	<	int mcsum = 0;
846	<	for (int i = 0; i < segments.length; ++i) {
847	<	sum += segments[i].count;
848	<	mcsum += mc[i] = segments[i].modCount;
849	<	}
850	<	if (mcsum != 0) {
851	<	for (int i = 0; i < segments.length; ++i) {
852	<	check += segments[i].count;
853	<	if (mc[i] != segments[i].modCount) {
854	<	check = -1; // force retry
855	<	break;
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	+	} finally {
871	+	if (retries > RETRIES_BEFORE_LOCK) {
872	+	for (int j = 0; j < segments.length; ++j)
873	+	segmentAt(segments, j).unlock();
874		}
737	–	if (check == sum)
738	–	break;
875		}
876	<	if (check != sum) { // Resort to locking all segments
741	<	sum = 0;
742	<	for (int i = 0; i < segments.length; ++i)
743	<	segments[i].lock();
744	<	for (int i = 0; i < segments.length; ++i)
745	<	sum += segments[i].count;
746	<	for (int i = 0; i < segments.length; ++i)
747	<	segments[i].unlock();
748	<	}
749	<	if (sum > Integer.MAX_VALUE)
750	<	return Integer.MAX_VALUE;
751	<	else
752	<	return (int)sum;
876	>	return overflow ? Integer.MAX_VALUE : size;
877		}
878
879		/**
#	Line 764 | Line 888 | public class ConcurrentHashMap<K, V> ext
888		* @throws NullPointerException if the specified key is null
889		*/
890		public V get(Object key) {
891	<	int hash = hash(key.hashCode());
892	<	return segmentFor(hash).get(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		/**
#	Line 777 | Line 914 | public class ConcurrentHashMap<K, V> ext
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.hashCode());
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 794 | Line 945 | public class ConcurrentHashMap<K, V> ext
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();
799	–
800	–	// See explanation of modCount use above
801	–
951		final Segment<K,V>[] segments = this.segments;
803	–	int[] mc = new int[segments.length];
804	–
805	–	// Try a few times without locking
806	–	for (int k = 0; k < RETRIES_BEFORE_LOCK; ++k) {
807	–	int sum = 0;
808	–	int mcsum = 0;
809	–	for (int i = 0; i < segments.length; ++i) {
810	–	int c = segments[i].count;
811	–	mcsum += mc[i] = segments[i].modCount;
812	–	if (segments[i].containsValue(value))
813	–	return true;
814	–	}
815	–	boolean cleanSweep = true;
816	–	if (mcsum != 0) {
817	–	for (int i = 0; i < segments.length; ++i) {
818	–	int c = segments[i].count;
819	–	if (mc[i] != segments[i].modCount) {
820	–	cleanSweep = false;
821	–	break;
822	–	}
823	–	}
824	–	}
825	–	if (cleanSweep)
826	–	return false;
827	–	}
828	–	// Resort to locking all segments
829	–	for (int i = 0; i < segments.length; ++i)
830	–	segments[i].lock();
952		boolean found = false;
953	+	long last = 0L;   // previous sum
954	+	int retries = -1;
955		try {
956	<	for (int i = 0; i < segments.length; ++i) {
957	<	if (segments[i].containsValue(value)) {
958	<	found = true;
959	<	break;
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	<	for (int i = 0; i < segments.length; ++i)
985	<	segments[i].unlock();
984	>	if (retries > RETRIES_BEFORE_LOCK) {
985	>	for (int j = 0; j < segments.length; ++j)
986	>	segmentAt(segments, j).unlock();
987	>	}
988		}
989		return found;
990		}
#	Line 850 | Line 996 | public class ConcurrentHashMap<K, V> ext
996		* full compatibility with class {@link java.util.Hashtable},
997		* which supported this method prior to introduction of the
998		* Java Collections framework.
999	<
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
#	Line 875 | Line 1021 | public class ConcurrentHashMap<K, V> ext
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.hashCode());
1030	<	return segmentFor(hash).put(key, hash, value, false);
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		/**
#	Line 889 | Line 1041 | public class ConcurrentHashMap<K, V> ext
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.hashCode());
1050	<	return segmentFor(hash).put(key, hash, value, true);
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
1057		/**
#	Line 918 | Line 1076 | public class ConcurrentHashMap<K, V> ext
1076		* @throws NullPointerException if the specified key is null
1077		*/
1078		public V remove(Object key) {
1079	<	int hash = hash(key.hashCode());
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		/**
#	Line 929 | Line 1088 | public class ConcurrentHashMap<K, V> ext
1088		*/
1089		public boolean remove(Object key, Object value) {
1090		int hash = hash(key.hashCode());
1091	<	if (value == null)
1092	<	return false;
1093	<	return segmentFor(hash).remove(key, hash, value) != null;
1091	>	Segment<K,V> s;
1092	>	return value != null && (s = segmentForHash(hash)) != null &&
1093	>	s.remove(key, hash, value) != null;
1094		}
1095
1096		/**
#	Line 940 | Line 1099 | public class ConcurrentHashMap<K, V> ext
1099		* @throws NullPointerException if any of the arguments are null
1100		*/
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	<	int hash = hash(key.hashCode());
1106	<	return segmentFor(hash).replace(key, hash, oldValue, newValue);
1105	>	Segment<K,V> s = segmentForHash(hash);
1106	>	return s != null && s.replace(key, hash, oldValue, newValue);
1107		}
1108
1109		/**
#	Line 954 | Line 1114 | public class ConcurrentHashMap<K, V> ext
1114		* @throws NullPointerException if the specified key or value is null
1115		*/
1116		public V replace(K key, V value) {
1117	+	int hash = hash(key.hashCode());
1118		if (value == null)
1119		throw new NullPointerException();
1120	<	int hash = hash(key.hashCode());
1121	<	return segmentFor(hash).replace(key, hash, value);
1120	>	Segment<K,V> s = segmentForHash(hash);
1121	>	return s == null ? null : s.replace(key, hash, value);
1122		}
1123
1124		/**
1125		* Removes all of the mappings from this map.
1126		*/
1127		public void clear() {
1128	<	for (int i = 0; i < segments.length; ++i)
1129	<	segments[i].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		/**
#	Line 1035 | Line 1200 | public class ConcurrentHashMap<K, V> ext
1200		* Returns an enumeration of the keys in this table.
1201		*
1202		* @return an enumeration of the keys in this table
1203	<	* @see #keySet
1203	>	* @see #keySet()
1204		*/
1205		public Enumeration<K> keys() {
1206		return new KeyIterator();
#	Line 1045 | Line 1210 | public class ConcurrentHashMap<K, V> ext
1210		* Returns an enumeration of the values in this table.
1211		*
1212		* @return an enumeration of the values in this table
1213	<	* @see #values
1213	>	* @see #values()
1214		*/
1215		public Enumeration<V> elements() {
1216		return new ValueIterator();
#	Line 1066 | Line 1231 | public class ConcurrentHashMap<K, V> ext
1231		advance();
1232		}
1233
1234	<	public boolean hasMoreElements() { return hasNext(); }
1235	<
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	<	if (nextEntry != null && (nextEntry = nextEntry.next) != null)
1240	<	return;
1241	<
1242	<	while (nextTableIndex >= 0) {
1243	<	if ( (nextEntry = currentTable[nextTableIndex--]) != null)
1244	<	return;
1245	<	}
1246	<
1247	<	while (nextSegmentIndex >= 0) {
1248	<	Segment<K,V> seg = segments[nextSegmentIndex--];
1082	<	if (seg.count != 0) {
1083	<	currentTable = seg.table;
1084	<	for (int j = currentTable.length - 1; j >= 0; --j) {
1085	<	if ( (nextEntry = currentTable[j]) != null) {
1086	<	nextTableIndex = j - 1;
1087	<	return;
1088	<	}
1089	<	}
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() {
1097	<	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 1110 | Line 1274 | public class ConcurrentHashMap<K, V> ext
1274		}
1275
1276		final class KeyIterator
1277	<	extends HashIterator
1278	<	implements Iterator<K>, Enumeration<K>
1277	>	extends HashIterator
1278	>	implements Iterator<K>, Enumeration<K>
1279		{
1280	<	public K next()        { return super.nextEntry().key; }
1281	<	public K nextElement() { return super.nextEntry().key; }
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>
1285	>	extends HashIterator
1286	>	implements Iterator<V>, Enumeration<V>
1287		{
1288	<	public V next()        { return super.nextEntry().value; }
1289	<	public V nextElement() { return super.nextEntry().value; }
1288	>	public final V next()        { return super.nextEntry().value; }
1289	>	public final V nextElement() { return super.nextEntry().value; }
1290		}
1291
1292		/**
#	Line 1130 | Line 1294 | public class ConcurrentHashMap<K, V> ext
1294		* setValue changes to the underlying map.
1295		*/
1296		final class WriteThroughEntry
1297	<	extends AbstractMap.SimpleEntry<K,V>
1297	>	extends AbstractMap.SimpleEntry<K,V>
1298		{
1299		WriteThroughEntry(K k, V v) {
1300		super(k,v);
1301		}
1302
1303		/**
1304	<	* Set our entry's value and write through to the map. The
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
#	Line 1145 | Line 1309 | public class ConcurrentHashMap<K, V> ext
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) {
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);
#	Line 1154 | Line 1318 | public class ConcurrentHashMap<K, V> ext
1318		}
1319
1320		final class EntryIterator
1321	<	extends HashIterator
1322	<	implements Iterator<Entry<K,V>>
1321	>	extends HashIterator
1322	>	implements Iterator<Entry<K,V>>
1323		{
1324		public Map.Entry<K,V> next() {
1325		HashEntry<K,V> e = super.nextEntry();
#	Line 1170 | Line 1334 | public class ConcurrentHashMap<K, V> ext
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 1188 | Line 1355 | public class ConcurrentHashMap<K, V> ext
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 1216 | Line 1386 | public class ConcurrentHashMap<K, V> ext
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 1224 | Line 1397 | public class ConcurrentHashMap<K, V> ext
1397		/* ---------------- Serialization Support -------------- */
1398
1399		/**
1400	<	* Save the state of the <tt>ConcurrentHashMap</tt> instance to a
1401	<	* stream (i.e., 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 = segments[k];
1416	>	Segment<K,V> seg = segmentAt(segments, k);
1417		seg.lock();
1418		try {
1419		HashEntry<K,V>[] tab = seg.table;
1420		for (int i = 0; i < tab.length; ++i) {
1421	<	for (HashEntry<K,V> e = 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 1255 | Line 1433 | public class ConcurrentHashMap<K, V> ext
1433		}
1434
1435		/**
1436	<	* Reconstitute the <tt>ConcurrentHashMap</tt> instance from a
1437	<	* stream (i.e., 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 1277 | Line 1462 | public class ConcurrentHashMap<K, V> ext
1462		put(key, value);
1463		}
1464		}
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