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. |
5 |
*/ |
6 |
|
7 |
package java.util.concurrent; |
8 |
|
9 |
import java.util.*; |
10 |
import java.io.Serializable; |
11 |
import java.io.IOException; |
12 |
import java.io.ObjectInputStream; |
13 |
import java.io.ObjectOutputStream; |
14 |
|
15 |
/** |
16 |
* A version of Hashtable supporting |
17 |
* concurrency for both retrievals and updates. |
18 |
* |
19 |
* <dl> |
20 |
* <dt> Retrievals |
21 |
* |
22 |
* <dd> Retrievals may overlap updates. Successful retrievals using |
23 |
* get(key) and containsKey(key) usually run without |
24 |
* locking. Unsuccessful retrievals (i.e., when the key is not |
25 |
* present) do involve brief locking. Because |
26 |
* retrieval operations can ordinarily overlap with update operations |
27 |
* (i.e., put, remove, and their derivatives), retrievals can only be |
28 |
* guaranteed to return the results of the most recently |
29 |
* <em>completed</em> operations holding upon their onset. Retrieval |
30 |
* operations may or may not return results reflecting in-progress |
31 |
* writing operations. However, the retrieval operations do always |
32 |
* return consistent results -- either those holding before any single |
33 |
* modification or after it, but never a nonsense result. For |
34 |
* aggregate operations such as putAll and clear, concurrent reads may |
35 |
* reflect insertion or removal of only some entries. <p> |
36 |
* |
37 |
* Iterators and Enumerations (i.e., those returned by |
38 |
* keySet().iterator(), entrySet().iterator(), values().iterator(), |
39 |
* keys(), and elements()) return elements reflecting the state of the |
40 |
* hash table at some point at or since the creation of the |
41 |
* iterator/enumeration. They will return at most one instance of |
42 |
* each element (via next()/nextElement()), but might or might not |
43 |
* reflect puts and removes that have been processed since |
44 |
* construction if the Iterator. They do <em>not</em> throw |
45 |
* ConcurrentModificationException. However, these iterators are |
46 |
* designed to be used by only one thread at a time. Passing an |
47 |
* iterator across multiple threads may lead to unpredictable traversal |
48 |
* if the table is being concurrently modified. <p> |
49 |
* |
50 |
* |
51 |
* <dt> Updates |
52 |
* |
53 |
* <dd> This class supports a hard-wired preset <em>concurrency |
54 |
* level</em> of 32. This allows a maximum of 32 put and/or remove |
55 |
* operations to proceed concurrently. This level is an upper bound on |
56 |
* concurrency, not a guarantee, since it interacts with how |
57 |
* well-strewn elements are across bins of the table. (The preset |
58 |
* value in part reflects the fact that even on large multiprocessors, |
59 |
* factors other than synchronization tend to be bottlenecks when more |
60 |
* than 32 threads concurrently attempt updates.) |
61 |
* Additionally, operations triggering internal resizing and clearing |
62 |
* do not execute concurrently with any operation. |
63 |
* <p> |
64 |
* |
65 |
* There is <em>NOT</em> any support for locking the entire table to |
66 |
* prevent updates. |
67 |
* |
68 |
* </dl> |
69 |
* |
70 |
* |
71 |
* This class may be used as a direct replacement for |
72 |
* java.util.Hashtable in any application that does not rely |
73 |
* on the ability to lock the entire table to prevent updates. |
74 |
* Like Hashtable but unlike java.util.HashMap, |
75 |
* this class does NOT allow <tt>null</tt> to be used as a key or |
76 |
* value. |
77 |
* <p> |
78 |
* |
79 |
**/ |
80 |
public class ConcurrentHashMap<K, V> extends AbstractMap<K, V> |
81 |
implements ConcurrentMap<K, V>, Cloneable, Serializable { |
82 |
|
83 |
/* |
84 |
The basic strategy is an optimistic-style scheme based on |
85 |
the guarantee that the hash table and its lists are always |
86 |
kept in a consistent enough state to be read without locking: |
87 |
|
88 |
* Read operations first proceed without locking, by traversing the |
89 |
apparently correct list of the apparently correct bin. If an |
90 |
entry is found, but not invalidated (value field null), it is |
91 |
returned. If not found, operations must recheck (after a memory |
92 |
barrier) to make sure they are using both the right list and |
93 |
the right table (which can change under resizes). If |
94 |
invalidated, reads must acquire main update lock to wait out |
95 |
the update, and then re-traverse. |
96 |
|
97 |
* All list additions are at the front of each bin, making it easy |
98 |
to check changes, and also fast to traverse. Entry next |
99 |
pointers are never assigned. Remove() builds new nodes when |
100 |
necessary to preserve this. |
101 |
|
102 |
* Remove() (also clear()) invalidates removed nodes to alert read |
103 |
operations that they must wait out the full modifications. |
104 |
|
105 |
* Locking for puts, removes (and, when necessary gets, etc) |
106 |
is controlled by Segments, each covering a portion of the |
107 |
table. During operations requiring global exclusivity (mainly |
108 |
resize and clear), ALL of these locks are acquired at once. |
109 |
Note that these segments are NOT contiguous -- they are based |
110 |
on the least 5 bits of hashcodes. This ensures that the same |
111 |
segment controls the same slots before and after resizing, which |
112 |
is necessary for supporting concurrent retrievals. This |
113 |
comes at the price of a mismatch of logical vs physical locality, |
114 |
but this seems not to be a performance problem in practice. |
115 |
|
116 |
*/ |
117 |
|
118 |
/** |
119 |
* The hash table data. |
120 |
*/ |
121 |
private transient Entry<K,V>[] table; |
122 |
|
123 |
|
124 |
/** |
125 |
* The number of concurrency control segments. |
126 |
* The value can be at most 32 since ints are used |
127 |
* as bitsets over segments. Emprically, it doesn't |
128 |
* seem to pay to decrease it either, so the value should be at least 32. |
129 |
* In other words, do not redefine this :-) |
130 |
**/ |
131 |
private static final int CONCURRENCY_LEVEL = 32; |
132 |
|
133 |
/** |
134 |
* Mask value for indexing into segments |
135 |
**/ |
136 |
private static final int SEGMENT_MASK = CONCURRENCY_LEVEL - 1; |
137 |
|
138 |
/** |
139 |
* Bookkeeping for each concurrency control segment. |
140 |
* Each segment contains a local count of the number of |
141 |
* elements in its region. |
142 |
* However, the main use of a Segment is for its lock. |
143 |
**/ |
144 |
private final static class Segment extends ReentrantLock { |
145 |
/** |
146 |
* The number of elements in this segment's region. |
147 |
**/ |
148 |
private int count; |
149 |
|
150 |
/** |
151 |
* Get the count under synch. |
152 |
**/ |
153 |
private int getCount() { |
154 |
lock(); |
155 |
try { |
156 |
return count; |
157 |
} |
158 |
finally { |
159 |
unlock(); |
160 |
} |
161 |
} |
162 |
|
163 |
} |
164 |
|
165 |
/** |
166 |
* The array of concurrency control segments. |
167 |
**/ |
168 |
private transient final Segment[] segments = new Segment[CONCURRENCY_LEVEL]; |
169 |
|
170 |
|
171 |
/** |
172 |
* The default initial number of table slots for this table (32). |
173 |
* Used when not otherwise specified in constructor. |
174 |
**/ |
175 |
public static int DEFAULT_INITIAL_CAPACITY = 32; |
176 |
|
177 |
|
178 |
/** |
179 |
* The minimum capacity, used if a lower value is implicitly specified |
180 |
* by either of the constructors with arguments. |
181 |
* MUST be a power of two. |
182 |
*/ |
183 |
private static final int MINIMUM_CAPACITY = 32; |
184 |
|
185 |
/** |
186 |
* The maximum capacity, used if a higher value is implicitly specified |
187 |
* by either of the constructors with arguments. |
188 |
* MUST be a power of two <= 1<<30. |
189 |
*/ |
190 |
private static final int MAXIMUM_CAPACITY = 1 << 30; |
191 |
|
192 |
/** |
193 |
* The default load factor for this table (0.75) |
194 |
* Used when not otherwise specified in constructor. |
195 |
**/ |
196 |
public static final float DEFAULT_LOAD_FACTOR = 0.75f; |
197 |
|
198 |
/** |
199 |
* The load factor for the hash table. |
200 |
* |
201 |
* @serial |
202 |
*/ |
203 |
private final float loadFactor; |
204 |
|
205 |
/** |
206 |
* Per-segment resize threshold. |
207 |
* |
208 |
* @serial |
209 |
*/ |
210 |
private int threshold; |
211 |
|
212 |
|
213 |
/** |
214 |
* Number of segments voting for resize. The table is |
215 |
* doubled when 1/4 of the segments reach threshold. |
216 |
* Volatile but updated without synch since this is just a heuristic. |
217 |
**/ |
218 |
private transient volatile int votesForResize; |
219 |
|
220 |
|
221 |
/** |
222 |
* Return the number of set bits in w. |
223 |
* For a derivation of this algorithm, see |
224 |
* "Algorithms and data structures with applications to |
225 |
* graphics and geometry", by Jurg Nievergelt and Klaus Hinrichs, |
226 |
* Prentice Hall, 1993. |
227 |
* See also notes by Torsten Sillke at |
228 |
* http://www.mathematik.uni-bielefeld.de/~sillke/PROBLEMS/bitcount |
229 |
**/ |
230 |
private static int bitcount(int w) { |
231 |
w -= (0xaaaaaaaa & w) >>> 1; |
232 |
w = (w & 0x33333333) + ((w >>> 2) & 0x33333333); |
233 |
w = (w + (w >>> 4)) & 0x0f0f0f0f; |
234 |
w += w >>> 8; |
235 |
w += w >>> 16; |
236 |
return w & 0xff; |
237 |
} |
238 |
|
239 |
/** |
240 |
* Returns the appropriate capacity (power of two) for the specified |
241 |
* initial capacity argument. |
242 |
*/ |
243 |
private int p2capacity(int initialCapacity) { |
244 |
int cap = initialCapacity; |
245 |
|
246 |
// Compute the appropriate capacity |
247 |
int result; |
248 |
if (cap > MAXIMUM_CAPACITY || cap < 0) { |
249 |
result = MAXIMUM_CAPACITY; |
250 |
} else { |
251 |
result = MINIMUM_CAPACITY; |
252 |
while (result < cap) |
253 |
result <<= 1; |
254 |
} |
255 |
return result; |
256 |
} |
257 |
|
258 |
/** |
259 |
* Return hash code for Object x. Since we are using power-of-two |
260 |
* tables, it is worth the effort to improve hashcode via |
261 |
* the same multiplicative scheme as used in IdentityHashMap. |
262 |
*/ |
263 |
private static int hash(Object x) { |
264 |
int h = x.hashCode(); |
265 |
// Multiply by 127 (quickly, via shifts), and mix in some high |
266 |
// bits to help guard against bunching of codes that are |
267 |
// consecutive or equally spaced. |
268 |
return ((h << 7) - h + (h >>> 9) + (h >>> 17)); |
269 |
} |
270 |
|
271 |
|
272 |
/** |
273 |
* Check for equality of non-null references x and y. |
274 |
**/ |
275 |
private boolean eq(Object x, Object y) { |
276 |
return x == y || x.equals(y); |
277 |
} |
278 |
|
279 |
/** Create table array and set the per-segment threshold **/ |
280 |
private Entry<K,V>[] newTable(int capacity) { |
281 |
threshold = (int)(capacity * loadFactor / CONCURRENCY_LEVEL) + 1; |
282 |
return new Entry<K,V>[capacity]; |
283 |
} |
284 |
|
285 |
/** |
286 |
* Constructs a new, empty map with the specified initial |
287 |
* capacity and the specified load factor. |
288 |
* |
289 |
* @param initialCapacity the initial capacity. |
290 |
* The actual initial capacity is rounded up to the nearest power of two. |
291 |
* @param loadFactor the load factor threshold, used to control resizing. |
292 |
* This value is used in an approximate way: When at least |
293 |
* a quarter of the segments of the table reach per-segment threshold, or |
294 |
* one of the segments itself exceeds overall threshold, |
295 |
* the table is doubled. |
296 |
* This will on average cause resizing when the table-wide |
297 |
* load factor is slightly less than the threshold. If you'd like |
298 |
* to avoid resizing, you can set this to a ridiculously large |
299 |
* value. |
300 |
* @throws IllegalArgumentException if the load factor is nonpositive. |
301 |
*/ |
302 |
public ConcurrentHashMap(int initialCapacity, float loadFactor) { |
303 |
if (!(loadFactor > 0)) |
304 |
throw new IllegalArgumentException("Illegal Load factor: "+ loadFactor); |
305 |
this.loadFactor = loadFactor; |
306 |
for (int i = 0; i < segments.length; ++i) |
307 |
segments[i] = new Segment(); |
308 |
int cap = p2capacity(initialCapacity); |
309 |
table = newTable(cap); |
310 |
} |
311 |
|
312 |
/** |
313 |
* Constructs a new, empty map with the specified initial |
314 |
* capacity and default load factor. |
315 |
* |
316 |
* @param initialCapacity the initial capacity of the |
317 |
* ConcurrentHashMap. |
318 |
* @throws IllegalArgumentException if the initial maximum number |
319 |
* of elements is less |
320 |
* than zero. |
321 |
*/ |
322 |
public ConcurrentHashMap(int initialCapacity) { |
323 |
this(initialCapacity, DEFAULT_LOAD_FACTOR); |
324 |
} |
325 |
|
326 |
/** |
327 |
* Constructs a new, empty map with a default initial capacity |
328 |
* and default load factor. |
329 |
*/ |
330 |
public ConcurrentHashMap() { |
331 |
this(DEFAULT_INITIAL_CAPACITY, DEFAULT_LOAD_FACTOR); |
332 |
} |
333 |
|
334 |
/** |
335 |
* Constructs a new map with the same mappings as the given map. The |
336 |
* map is created with a capacity of twice the number of mappings in |
337 |
* the given map or 32 (whichever is greater), and a default load factor. |
338 |
*/ |
339 |
public <A extends K, B extends V> ConcurrentHashMap(Map<A,B> t) { |
340 |
this(Math.max((int) (t.size() / DEFAULT_LOAD_FACTOR) + 1, |
341 |
MINIMUM_CAPACITY), |
342 |
DEFAULT_LOAD_FACTOR); |
343 |
putAll(t); |
344 |
} |
345 |
|
346 |
/** |
347 |
* Returns the number of key-value mappings in this map. |
348 |
* |
349 |
* @return the number of key-value mappings in this map. |
350 |
*/ |
351 |
public int size() { |
352 |
int c = 0; |
353 |
for (int i = 0; i < segments.length; ++i) |
354 |
c += segments[i].getCount(); |
355 |
return c; |
356 |
} |
357 |
|
358 |
/** |
359 |
* Returns <tt>true</tt> if this map contains no key-value mappings. |
360 |
* |
361 |
* @return <tt>true</tt> if this map contains no key-value mappings. |
362 |
*/ |
363 |
public boolean isEmpty() { |
364 |
for (int i = 0; i < segments.length; ++i) |
365 |
if (segments[i].getCount() != 0) |
366 |
return false; |
367 |
return true; |
368 |
} |
369 |
|
370 |
|
371 |
/** |
372 |
* Returns the value to which the specified key is mapped in this table. |
373 |
* |
374 |
* @param key a key in the table. |
375 |
* @return the value to which the key is mapped in this table; |
376 |
* <code>null</code> if the key is not mapped to any value in |
377 |
* this table. |
378 |
* @exception NullPointerException if the key is |
379 |
* <code>null</code>. |
380 |
* @see #put(Object, Object) |
381 |
*/ |
382 |
public V get(Object key) { |
383 |
int hash = hash(key); // throws null pointer exception if key null |
384 |
|
385 |
// Try first without locking... |
386 |
Entry<K,V>[] tab = table; |
387 |
int index = hash & (tab.length - 1); |
388 |
Entry<K,V> first = tab[index]; |
389 |
Entry<K,V> e; |
390 |
|
391 |
for (e = first; e != null; e = e.next) { |
392 |
if (e.hash == hash && eq(key, e.key)) { |
393 |
V value = e.value; |
394 |
if (value != null) |
395 |
return value; |
396 |
else |
397 |
break; |
398 |
} |
399 |
} |
400 |
|
401 |
// Recheck under synch if key apparently not there or interference |
402 |
Segment seg = segments[hash & SEGMENT_MASK]; |
403 |
seg.lock(); |
404 |
try { |
405 |
tab = table; |
406 |
index = hash & (tab.length - 1); |
407 |
Entry<K,V> newFirst = tab[index]; |
408 |
if (e != null || first != newFirst) { |
409 |
for (e = newFirst; e != null; e = e.next) { |
410 |
if (e.hash == hash && eq(key, e.key)) |
411 |
return e.value; |
412 |
} |
413 |
} |
414 |
return null; |
415 |
} |
416 |
finally { |
417 |
seg.unlock(); |
418 |
} |
419 |
} |
420 |
|
421 |
/** |
422 |
* Tests if the specified object is a key in this table. |
423 |
* |
424 |
* @param key possible key. |
425 |
* @return <code>true</code> if and only if the specified object |
426 |
* is a key in this table, as determined by the |
427 |
* <tt>equals</tt> method; <code>false</code> otherwise. |
428 |
* @exception NullPointerException if the key is |
429 |
* <code>null</code>. |
430 |
* @see #contains(Object) |
431 |
*/ |
432 |
public boolean containsKey(Object key) { |
433 |
return get(key) != null; |
434 |
} |
435 |
|
436 |
|
437 |
/** |
438 |
* Maps the specified <code>key</code> to the specified |
439 |
* <code>value</code> in this table. Neither the key nor the |
440 |
* value can be <code>null</code>. (Note that this policy is |
441 |
* the same as for java.util.Hashtable, but unlike java.util.HashMap, |
442 |
* which does accept nulls as valid keys and values.)<p> |
443 |
* |
444 |
* The value can be retrieved by calling the <code>get</code> method |
445 |
* with a key that is equal to the original key. |
446 |
* |
447 |
* @param key the table key. |
448 |
* @param value the value. |
449 |
* @return the previous value of the specified key in this table, |
450 |
* or <code>null</code> if it did not have one. |
451 |
* @exception NullPointerException if the key or value is |
452 |
* <code>null</code>. |
453 |
* @see Object#equals(Object) |
454 |
* @see #get(Object) |
455 |
*/ |
456 |
public V put(K key, V value) { |
457 |
if (value == null) |
458 |
throw new NullPointerException(); |
459 |
|
460 |
int hash = hash(key); |
461 |
Segment seg = segments[hash & SEGMENT_MASK]; |
462 |
int segcount; |
463 |
Entry<K,V>[] tab; |
464 |
int votes; |
465 |
|
466 |
seg.lock(); |
467 |
try { |
468 |
tab = table; |
469 |
int index = hash & (tab.length-1); |
470 |
Entry<K,V> first = tab[index]; |
471 |
|
472 |
for (Entry<K,V> e = first; e != null; e = e.next) { |
473 |
if (e.hash == hash && eq(key, e.key)) { |
474 |
V oldValue = e.value; |
475 |
e.value = value; |
476 |
return oldValue; |
477 |
} |
478 |
} |
479 |
|
480 |
// Add to front of list |
481 |
Entry<K,V> newEntry = new Entry<K,V>(hash, key, value, first); |
482 |
tab[index] = newEntry; |
483 |
|
484 |
if ((segcount = ++seg.count) < threshold) |
485 |
return null; |
486 |
|
487 |
int bit = (1 << (hash & SEGMENT_MASK)); |
488 |
votes = votesForResize; |
489 |
if ((votes & bit) == 0) |
490 |
votes = votesForResize |= bit; |
491 |
} |
492 |
finally { |
493 |
seg.unlock(); |
494 |
} |
495 |
|
496 |
// Attempt resize if 1/4 segs vote, |
497 |
// or if this seg itself reaches the overall threshold. |
498 |
// (The latter check is just a safeguard to avoid pathological cases.) |
499 |
if (bitcount(votes) >= CONCURRENCY_LEVEL / 4 || |
500 |
segcount > (threshold * CONCURRENCY_LEVEL)) |
501 |
resize(tab); |
502 |
|
503 |
return null; |
504 |
} |
505 |
|
506 |
public V putIfAbsent(K key, V value) { |
507 |
if (value == null) |
508 |
throw new NullPointerException(); |
509 |
|
510 |
int hash = hash(key); |
511 |
Segment seg = segments[hash & SEGMENT_MASK]; |
512 |
int segcount; |
513 |
Entry<K,V>[] tab; |
514 |
int votes; |
515 |
|
516 |
seg.lock(); |
517 |
try { |
518 |
tab = table; |
519 |
int index = hash & (tab.length-1); |
520 |
Entry<K,V> first = tab[index]; |
521 |
|
522 |
for (Entry<K,V> e = first; e != null; e = e.next) { |
523 |
if (e.hash == hash && eq(key, e.key)) { |
524 |
V oldValue = e.value; |
525 |
return oldValue; |
526 |
} |
527 |
} |
528 |
|
529 |
// Add to front of list |
530 |
Entry<K,V> newEntry = new Entry<K,V>(hash, key, value, first); |
531 |
tab[index] = newEntry; |
532 |
|
533 |
if ((segcount = ++seg.count) < threshold) |
534 |
return null; |
535 |
|
536 |
int bit = (1 << (hash & SEGMENT_MASK)); |
537 |
votes = votesForResize; |
538 |
if ((votes & bit) == 0) |
539 |
votes = votesForResize |= bit; |
540 |
} |
541 |
finally { |
542 |
seg.unlock(); |
543 |
} |
544 |
|
545 |
// Attempt resize if 1/4 segs vote, |
546 |
// or if this seg itself reaches the overall threshold. |
547 |
// (The latter check is just a safeguard to avoid pathological cases.) |
548 |
if (bitcount(votes) >= CONCURRENCY_LEVEL / 4 || |
549 |
segcount > (threshold * CONCURRENCY_LEVEL)) |
550 |
resize(tab); |
551 |
|
552 |
return value; |
553 |
} |
554 |
|
555 |
/** |
556 |
* Gather all locks in order to call rehash, by |
557 |
* recursing within synch blocks for each segment index. |
558 |
* @param index the current segment. initially call value must be 0 |
559 |
* @param assumedTab the state of table on first call to resize. If |
560 |
* this changes on any call, the attempt is aborted because the |
561 |
* table has already been resized by another thread. |
562 |
*/ |
563 |
private void resize(Entry<K,V>[] assumedTab) { |
564 |
boolean ok = true; |
565 |
int lastlocked = 0; |
566 |
for (int i = 0; i < segments.length; ++i) { |
567 |
segments[i].lock(); |
568 |
lastlocked = i; |
569 |
if (table != assumedTab) { |
570 |
ok = false; |
571 |
break; |
572 |
} |
573 |
} |
574 |
try { |
575 |
if (ok) |
576 |
rehash(); |
577 |
} |
578 |
finally { |
579 |
for (int i = lastlocked; i >= 0; --i) |
580 |
segments[i].unlock(); |
581 |
} |
582 |
} |
583 |
|
584 |
/** |
585 |
* Rehashes the contents of this map into a new table |
586 |
* with a larger capacity. |
587 |
*/ |
588 |
private void rehash() { |
589 |
votesForResize = 0; // reset |
590 |
|
591 |
Entry<K,V>[] oldTable = table; |
592 |
int oldCapacity = oldTable.length; |
593 |
|
594 |
if (oldCapacity >= MAXIMUM_CAPACITY) { |
595 |
threshold = Integer.MAX_VALUE; // avoid retriggering |
596 |
return; |
597 |
} |
598 |
|
599 |
int newCapacity = oldCapacity << 1; |
600 |
Entry<K,V>[] newTable = newTable(newCapacity); |
601 |
int mask = newCapacity - 1; |
602 |
|
603 |
/* |
604 |
* Reclassify nodes in each list to new Map. Because we are |
605 |
* using power-of-two expansion, the elements from each bin |
606 |
* must either stay at same index, or move to |
607 |
* oldCapacity+index. We also eliminate unnecessary node |
608 |
* creation by catching cases where old nodes can be reused |
609 |
* because their next fields won't change. Statistically, at |
610 |
* the default threshhold, only about one-sixth of them need |
611 |
* cloning. (The nodes they replace will be garbage |
612 |
* collectable as soon as they are no longer referenced by any |
613 |
* reader thread that may be in the midst of traversing table |
614 |
* right now.) |
615 |
*/ |
616 |
|
617 |
for (int i = 0; i < oldCapacity ; i++) { |
618 |
// We need to guarantee that any existing reads of old Map can |
619 |
// proceed. So we cannot yet null out each bin. |
620 |
Entry<K,V> e = oldTable[i]; |
621 |
|
622 |
if (e != null) { |
623 |
int idx = e.hash & mask; |
624 |
Entry<K,V> next = e.next; |
625 |
|
626 |
// Single node on list |
627 |
if (next == null) |
628 |
newTable[idx] = e; |
629 |
|
630 |
else { |
631 |
// Reuse trailing consecutive sequence of all same bit |
632 |
Entry<K,V> lastRun = e; |
633 |
int lastIdx = idx; |
634 |
for (Entry<K,V> last = next; last != null; last = last.next) { |
635 |
int k = last.hash & mask; |
636 |
if (k != lastIdx) { |
637 |
lastIdx = k; |
638 |
lastRun = last; |
639 |
} |
640 |
} |
641 |
newTable[lastIdx] = lastRun; |
642 |
|
643 |
// Clone all remaining nodes |
644 |
for (Entry<K,V> p = e; p != lastRun; p = p.next) { |
645 |
int k = p.hash & mask; |
646 |
newTable[k] = new Entry<K,V>(p.hash, p.key, |
647 |
p.value, newTable[k]); |
648 |
} |
649 |
} |
650 |
} |
651 |
} |
652 |
|
653 |
table = newTable; |
654 |
} |
655 |
|
656 |
|
657 |
/** |
658 |
* Removes the key (and its corresponding value) from this |
659 |
* table. This method does nothing if the key is not in the table. |
660 |
* |
661 |
* @param key the key that needs to be removed. |
662 |
* @return the value to which the key had been mapped in this table, |
663 |
* or <code>null</code> if the key did not have a mapping. |
664 |
* @exception NullPointerException if the key is |
665 |
* <code>null</code>. |
666 |
*/ |
667 |
public V remove(Object key) { |
668 |
return remove(key, null); |
669 |
} |
670 |
|
671 |
|
672 |
/** |
673 |
* Removes the (key, value) pair from this |
674 |
* table. This method does nothing if the key is not in the table, |
675 |
* or if the key is associated with a different value. This method |
676 |
* is needed by EntrySet. |
677 |
* |
678 |
* @param key the key that needs to be removed. |
679 |
* @param value the associated value. If the value is null, |
680 |
* it means "any value". |
681 |
* @return the value to which the key had been mapped in this table, |
682 |
* or <code>null</code> if the key did not have a mapping. |
683 |
* @exception NullPointerException if the key is |
684 |
* <code>null</code>. |
685 |
*/ |
686 |
private V remove(Object key, V value) { |
687 |
/* |
688 |
Find the entry, then |
689 |
1. Set value field to null, to force get() to retry |
690 |
2. Rebuild the list without this entry. |
691 |
All entries following removed node can stay in list, but |
692 |
all preceeding ones need to be cloned. Traversals rely |
693 |
on this strategy to ensure that elements will not be |
694 |
repeated during iteration. |
695 |
*/ |
696 |
|
697 |
int hash = hash(key); |
698 |
Segment seg = segments[hash & SEGMENT_MASK]; |
699 |
|
700 |
seg.lock(); |
701 |
try { |
702 |
Entry<K,V>[] tab = table; |
703 |
int index = hash & (tab.length-1); |
704 |
Entry<K,V> first = tab[index]; |
705 |
Entry<K,V> e = first; |
706 |
|
707 |
for (;;) { |
708 |
if (e == null) |
709 |
return null; |
710 |
if (e.hash == hash && eq(key, e.key)) |
711 |
break; |
712 |
e = e.next; |
713 |
} |
714 |
|
715 |
V oldValue = e.value; |
716 |
if (value != null && !value.equals(oldValue)) |
717 |
return null; |
718 |
|
719 |
e.value = null; |
720 |
|
721 |
Entry<K,V> head = e.next; |
722 |
for (Entry<K,V> p = first; p != e; p = p.next) |
723 |
head = new Entry<K,V>(p.hash, p.key, p.value, head); |
724 |
tab[index] = head; |
725 |
seg.count--; |
726 |
return oldValue; |
727 |
} |
728 |
finally { |
729 |
seg.unlock(); |
730 |
} |
731 |
} |
732 |
|
733 |
|
734 |
/** |
735 |
* Returns <tt>true</tt> if this map maps one or more keys to the |
736 |
* specified value. Note: This method requires a full internal |
737 |
* traversal of the hash table, and so is much slower than |
738 |
* method <tt>containsKey</tt>. |
739 |
* |
740 |
* @param value value whose presence in this map is to be tested. |
741 |
* @return <tt>true</tt> if this map maps one or more keys to the |
742 |
* specified value. |
743 |
* @exception NullPointerException if the value is <code>null</code>. |
744 |
*/ |
745 |
public boolean containsValue(Object value) { |
746 |
|
747 |
if (value == null) throw new NullPointerException(); |
748 |
|
749 |
for (int s = 0; s < segments.length; ++s) { |
750 |
Segment seg = segments[s]; |
751 |
Entry<K,V>[] tab; |
752 |
seg.lock(); |
753 |
try { |
754 |
tab = table; |
755 |
} |
756 |
finally { |
757 |
seg.unlock(); |
758 |
} |
759 |
for (int i = s; i < tab.length; i+= segments.length) { |
760 |
for (Entry<K,V> e = tab[i]; e != null; e = e.next) |
761 |
if (value.equals(e.value)) |
762 |
return true; |
763 |
} |
764 |
} |
765 |
return false; |
766 |
} |
767 |
|
768 |
/** |
769 |
* Tests if some key maps into the specified value in this table. |
770 |
* This operation is more expensive than the <code>containsKey</code> |
771 |
* method.<p> |
772 |
* |
773 |
* Note that this method is identical in functionality to containsValue, |
774 |
* (which is part of the Map interface in the collections framework). |
775 |
* |
776 |
* @param value a value to search for. |
777 |
* @return <code>true</code> if and only if some key maps to the |
778 |
* <code>value</code> argument in this table as |
779 |
* determined by the <tt>equals</tt> method; |
780 |
* <code>false</code> otherwise. |
781 |
* @exception NullPointerException if the value is <code>null</code>. |
782 |
* @see #containsKey(Object) |
783 |
* @see #containsValue(Object) |
784 |
* @see Map |
785 |
*/ |
786 |
public boolean contains(V value) { |
787 |
return containsValue(value); |
788 |
} |
789 |
|
790 |
/** |
791 |
* Copies all of the mappings from the specified map to this one. |
792 |
* |
793 |
* These mappings replace any mappings that this map had for any of the |
794 |
* keys currently in the specified Map. |
795 |
* |
796 |
* @param t Mappings to be stored in this map. |
797 |
*/ |
798 |
public <A extends K, B extends V> void putAll(Map<A, B> t) { |
799 |
int n = t.size(); |
800 |
if (n == 0) |
801 |
return; |
802 |
|
803 |
// Expand enough to hold at least n elements without resizing. |
804 |
// We can only resize table by factor of two at a time. |
805 |
// It is faster to rehash with fewer elements, so do it now. |
806 |
for(;;) { |
807 |
Entry<K,V>[] tab; |
808 |
int max; |
809 |
// must synch on some segment. pick 0. |
810 |
segments[0].lock(); |
811 |
try { |
812 |
tab = table; |
813 |
max = threshold * CONCURRENCY_LEVEL; |
814 |
} |
815 |
finally { |
816 |
segments[0].unlock(); |
817 |
} |
818 |
if (n < max) |
819 |
break; |
820 |
resize(tab); |
821 |
} |
822 |
|
823 |
for (Iterator<Map.Entry<A,B>> it = t.entrySet().iterator(); it.hasNext();) { |
824 |
Map.Entry<A,B> entry = (Map.Entry<A,B>) it.next(); |
825 |
put(entry.getKey(), entry.getValue()); |
826 |
} |
827 |
} |
828 |
|
829 |
/** |
830 |
* Removes all mappings from this map. |
831 |
*/ |
832 |
public void clear() { |
833 |
// We don't need all locks at once so long as locks |
834 |
// are obtained in low to high order |
835 |
for (int s = 0; s < segments.length; ++s) { |
836 |
Segment seg = segments[s]; |
837 |
seg.lock(); |
838 |
try { |
839 |
Entry<K,V>[] tab = table; |
840 |
for (int i = s; i < tab.length; i+= segments.length) { |
841 |
for (Entry<K,V> e = tab[i]; e != null; e = e.next) |
842 |
e.value = null; |
843 |
tab[i] = null; |
844 |
seg.count = 0; |
845 |
} |
846 |
} |
847 |
finally { |
848 |
seg.unlock(); |
849 |
} |
850 |
} |
851 |
} |
852 |
|
853 |
/** |
854 |
* Returns a shallow copy of this |
855 |
* <tt>ConcurrentHashMap</tt> instance: the keys and |
856 |
* values themselves are not cloned. |
857 |
* |
858 |
* @return a shallow copy of this map. |
859 |
*/ |
860 |
public Object clone() { |
861 |
// We cannot call super.clone, since it would share final segments array, |
862 |
// and there's no way to reassign finals. |
863 |
return new ConcurrentHashMap<K,V>(this); |
864 |
} |
865 |
|
866 |
// Views |
867 |
|
868 |
private transient Set<K> keySet = null; |
869 |
private transient Set<Map.Entry<K,V>> entrySet = null; |
870 |
private transient Collection<V> values = null; |
871 |
|
872 |
/** |
873 |
* Returns a set view of the keys contained in this map. The set is |
874 |
* backed by the map, so changes to the map are reflected in the set, and |
875 |
* vice-versa. The set supports element removal, which removes the |
876 |
* corresponding mapping from this map, via the <tt>Iterator.remove</tt>, |
877 |
* <tt>Set.remove</tt>, <tt>removeAll</tt>, <tt>retainAll</tt>, and |
878 |
* <tt>clear</tt> operations. It does not support the <tt>add</tt> or |
879 |
* <tt>addAll</tt> operations. |
880 |
* |
881 |
* @return a set view of the keys contained in this map. |
882 |
*/ |
883 |
public Set<K> keySet() { |
884 |
Set<K> ks = keySet; |
885 |
return (ks != null)? ks : (keySet = new KeySet()); |
886 |
} |
887 |
|
888 |
private class KeySet extends AbstractSet<K> { |
889 |
public Iterator<K> iterator() { |
890 |
return new KeyIterator(); |
891 |
} |
892 |
public int size() { |
893 |
return ConcurrentHashMap.this.size(); |
894 |
} |
895 |
public boolean contains(Object o) { |
896 |
return ConcurrentHashMap.this.containsKey(o); |
897 |
} |
898 |
public boolean remove(Object o) { |
899 |
return ConcurrentHashMap.this.remove(o) != null; |
900 |
} |
901 |
public void clear() { |
902 |
ConcurrentHashMap.this.clear(); |
903 |
} |
904 |
} |
905 |
|
906 |
/** |
907 |
* Returns a collection view of the values contained in this map. The |
908 |
* collection is backed by the map, so changes to the map are reflected in |
909 |
* the collection, and vice-versa. The collection supports element |
910 |
* removal, which removes the corresponding mapping from this map, via the |
911 |
* <tt>Iterator.remove</tt>, <tt>Collection.remove</tt>, |
912 |
* <tt>removeAll</tt>, <tt>retainAll</tt>, and <tt>clear</tt> operations. |
913 |
* It does not support the <tt>add</tt> or <tt>addAll</tt> operations. |
914 |
* |
915 |
* @return a collection view of the values contained in this map. |
916 |
*/ |
917 |
public Collection<V> values() { |
918 |
Collection<V> vs = values; |
919 |
return (vs != null)? vs : (values = new Values()); |
920 |
} |
921 |
|
922 |
private class Values extends AbstractCollection<V> { |
923 |
public Iterator<V> iterator() { |
924 |
return new ValueIterator(); |
925 |
} |
926 |
public int size() { |
927 |
return ConcurrentHashMap.this.size(); |
928 |
} |
929 |
public boolean contains(Object o) { |
930 |
return ConcurrentHashMap.this.containsValue(o); |
931 |
} |
932 |
public void clear() { |
933 |
ConcurrentHashMap.this.clear(); |
934 |
} |
935 |
} |
936 |
|
937 |
/** |
938 |
* Returns a collection view of the mappings contained in this map. Each |
939 |
* element in the returned collection is a <tt>Map.Entry</tt>. The |
940 |
* collection is backed by the map, so changes to the map are reflected in |
941 |
* the collection, and vice-versa. The collection supports element |
942 |
* removal, which removes the corresponding mapping from the map, via the |
943 |
* <tt>Iterator.remove</tt>, <tt>Collection.remove</tt>, |
944 |
* <tt>removeAll</tt>, <tt>retainAll</tt>, and <tt>clear</tt> operations. |
945 |
* It does not support the <tt>add</tt> or <tt>addAll</tt> operations. |
946 |
* |
947 |
* @return a collection view of the mappings contained in this map. |
948 |
*/ |
949 |
public Set<Map.Entry<K,V>> entrySet() { |
950 |
Set<Map.Entry<K,V>> es = entrySet; |
951 |
return (es != null) ? es : (entrySet = new EntrySet()); |
952 |
} |
953 |
|
954 |
private class EntrySet extends AbstractSet<Map.Entry<K,V>> { |
955 |
public Iterator<Map.Entry<K,V>> iterator() { |
956 |
return new EntryIterator(); |
957 |
} |
958 |
public boolean contains(Map.Entry<K,V> entry) { |
959 |
V v = ConcurrentHashMap.this.get(entry.getKey()); |
960 |
return v != null && v.equals(entry.getValue()); |
961 |
} |
962 |
public boolean remove(Map.Entry<K,V> e) { |
963 |
return ConcurrentHashMap.this.remove(e.getKey(), e.getValue()) != null; |
964 |
} |
965 |
public int size() { |
966 |
return ConcurrentHashMap.this.size(); |
967 |
} |
968 |
public void clear() { |
969 |
ConcurrentHashMap.this.clear(); |
970 |
} |
971 |
} |
972 |
|
973 |
/** |
974 |
* Returns an enumeration of the keys in this table. |
975 |
* |
976 |
* @return an enumeration of the keys in this table. |
977 |
* @see Enumeration |
978 |
* @see #elements() |
979 |
* @see #keySet() |
980 |
* @see Map |
981 |
*/ |
982 |
public Enumeration keys() { |
983 |
return new KeyIterator(); |
984 |
} |
985 |
|
986 |
/** |
987 |
* Returns an enumeration of the values in this table. |
988 |
* Use the Enumeration methods on the returned object to fetch the elements |
989 |
* sequentially. |
990 |
* |
991 |
* @return an enumeration of the values in this table. |
992 |
* @see java.util.Enumeration |
993 |
* @see #keys() |
994 |
* @see #values() |
995 |
* @see Map |
996 |
*/ |
997 |
public Enumeration elements() { |
998 |
return new ValueIterator(); |
999 |
} |
1000 |
|
1001 |
/** |
1002 |
* ConcurrentHashMap collision list entry. |
1003 |
*/ |
1004 |
private static class Entry<K,V> implements Map.Entry<K,V> { |
1005 |
/* |
1006 |
The use of volatile for value field ensures that |
1007 |
we can detect status changes without synchronization. |
1008 |
The other fields are never changed, and are |
1009 |
marked as final. |
1010 |
*/ |
1011 |
|
1012 |
private final K key; |
1013 |
private volatile V value; |
1014 |
private final int hash; |
1015 |
private final Entry<K,V> next; |
1016 |
|
1017 |
Entry(int hash, K key, V value, Entry<K,V> next) { |
1018 |
this.value = value; |
1019 |
this.hash = hash; |
1020 |
this.key = key; |
1021 |
this.next = next; |
1022 |
} |
1023 |
|
1024 |
// Map.Entry Ops |
1025 |
|
1026 |
public K getKey() { |
1027 |
return key; |
1028 |
} |
1029 |
|
1030 |
/** |
1031 |
* Get the value. Note: In an entrySet or entrySet.iterator, |
1032 |
* unless you can guarantee lack of concurrent modification, |
1033 |
* <tt>getValue</tt> <em>might</em> return null, reflecting the |
1034 |
* fact that the entry has been concurrently removed. However, |
1035 |
* there are no assurances that concurrent removals will be |
1036 |
* reflected using this method. |
1037 |
* |
1038 |
* @return the current value, or null if the entry has been |
1039 |
* detectably removed. |
1040 |
**/ |
1041 |
public V getValue() { |
1042 |
return value; |
1043 |
} |
1044 |
|
1045 |
/** |
1046 |
* Set the value of this entry. Note: In an entrySet or |
1047 |
* entrySet.iterator), unless you can guarantee lack of concurrent |
1048 |
* modification, <tt>setValue</tt> is not strictly guaranteed to |
1049 |
* actually replace the value field obtained via the <tt>get</tt> |
1050 |
* operation of the underlying hash table in multithreaded |
1051 |
* applications. If iterator-wide synchronization is not used, |
1052 |
* and any other concurrent <tt>put</tt> or <tt>remove</tt> |
1053 |
* operations occur, sometimes even to <em>other</em> entries, |
1054 |
* then this change is not guaranteed to be reflected in the hash |
1055 |
* table. (It might, or it might not. There are no assurances |
1056 |
* either way.) |
1057 |
* |
1058 |
* @param value the new value. |
1059 |
* @return the previous value, or null if entry has been detectably |
1060 |
* removed. |
1061 |
* @exception NullPointerException if the value is <code>null</code>. |
1062 |
* |
1063 |
**/ |
1064 |
public V setValue(V value) { |
1065 |
if (value == null) |
1066 |
throw new NullPointerException(); |
1067 |
V oldValue = this.value; |
1068 |
this.value = value; |
1069 |
return oldValue; |
1070 |
} |
1071 |
|
1072 |
public boolean equals(Object o) { |
1073 |
if (!(o instanceof Map.Entry)) |
1074 |
return false; |
1075 |
Map.Entry e = (Map.Entry)o; |
1076 |
return (key.equals(e.getKey()) && value.equals(e.getValue())); |
1077 |
} |
1078 |
|
1079 |
public int hashCode() { |
1080 |
return key.hashCode() ^ value.hashCode(); |
1081 |
} |
1082 |
|
1083 |
public String toString() { |
1084 |
return key + "=" + value; |
1085 |
} |
1086 |
|
1087 |
} |
1088 |
|
1089 |
private abstract class HashIterator<T> implements Iterator<T>, Enumeration { |
1090 |
private final Entry<K,V>[] tab; // snapshot of table |
1091 |
private int index; // current slot |
1092 |
Entry<K,V> entry = null; // current node of slot |
1093 |
K currentKey; // key for current node |
1094 |
V currentValue; // value for current node |
1095 |
private Entry lastReturned = null; // last node returned by next |
1096 |
|
1097 |
private HashIterator() { |
1098 |
// force all segments to synch |
1099 |
for (int i = 0; i < segments.length; ++i) { |
1100 |
segments[i].lock(); |
1101 |
segments[i].unlock(); |
1102 |
} |
1103 |
tab = table; |
1104 |
index = tab.length - 1; |
1105 |
} |
1106 |
|
1107 |
public boolean hasMoreElements() { return hasNext(); } |
1108 |
public Object nextElement() { return next(); } |
1109 |
|
1110 |
public boolean hasNext() { |
1111 |
/* |
1112 |
currentkey and currentValue are set here to ensure that next() |
1113 |
returns normally if hasNext() returns true. This avoids |
1114 |
surprises especially when final element is removed during |
1115 |
traversal -- instead, we just ignore the removal during |
1116 |
current traversal. |
1117 |
*/ |
1118 |
|
1119 |
while (true) { |
1120 |
if (entry != null) { |
1121 |
V v = entry.value; |
1122 |
if (v != null) { |
1123 |
currentKey = entry.key; |
1124 |
currentValue = v; |
1125 |
return true; |
1126 |
} |
1127 |
else |
1128 |
entry = entry.next; |
1129 |
} |
1130 |
|
1131 |
while (entry == null && index >= 0) |
1132 |
entry = tab[index--]; |
1133 |
|
1134 |
if (entry == null) { |
1135 |
currentKey = null; |
1136 |
currentValue = null; |
1137 |
return false; |
1138 |
} |
1139 |
} |
1140 |
} |
1141 |
|
1142 |
abstract T returnValueOfNext(); |
1143 |
|
1144 |
public T next() { |
1145 |
if (currentKey == null && !hasNext()) |
1146 |
throw new NoSuchElementException(); |
1147 |
|
1148 |
T result = returnValueOfNext(); |
1149 |
lastReturned = entry; |
1150 |
currentKey = null; |
1151 |
currentValue = null; |
1152 |
entry = entry.next; |
1153 |
return result; |
1154 |
} |
1155 |
|
1156 |
public void remove() { |
1157 |
if (lastReturned == null) |
1158 |
throw new IllegalStateException(); |
1159 |
ConcurrentHashMap.this.remove(lastReturned.key); |
1160 |
lastReturned = null; |
1161 |
} |
1162 |
|
1163 |
} |
1164 |
|
1165 |
private class KeyIterator extends HashIterator<K> { |
1166 |
K returnValueOfNext() { return currentKey; } |
1167 |
public K next() { return super.next(); } |
1168 |
} |
1169 |
|
1170 |
private class ValueIterator extends HashIterator<V> { |
1171 |
V returnValueOfNext() { return currentValue; } |
1172 |
public V next() { return super.next(); } |
1173 |
} |
1174 |
|
1175 |
private class EntryIterator extends HashIterator<Map.Entry<K,V>> { |
1176 |
Map.Entry<K,V> returnValueOfNext() { return entry; } |
1177 |
public Map.Entry<K,V> next() { return super.next(); } |
1178 |
} |
1179 |
|
1180 |
/** |
1181 |
* Save the state of the <tt>ConcurrentHashMap</tt> |
1182 |
* instance to a stream (i.e., |
1183 |
* serialize it). |
1184 |
* |
1185 |
* @serialData |
1186 |
* An estimate of the table size, followed by |
1187 |
* the key (Object) and value (Object) |
1188 |
* for each key-value mapping, followed by a null pair. |
1189 |
* The key-value mappings are emitted in no particular order. |
1190 |
*/ |
1191 |
private void writeObject(java.io.ObjectOutputStream s) throws IOException { |
1192 |
// Write out the loadfactor, and any hidden stuff |
1193 |
s.defaultWriteObject(); |
1194 |
|
1195 |
// Write out capacity estimate. It is OK if this |
1196 |
// changes during the write, since it is only used by |
1197 |
// readObject to set initial capacity, to avoid needless resizings. |
1198 |
|
1199 |
int cap; |
1200 |
segments[0].lock(); |
1201 |
try { |
1202 |
cap = table.length; |
1203 |
} |
1204 |
finally { |
1205 |
segments[0].unlock(); |
1206 |
} |
1207 |
s.writeInt(cap); |
1208 |
|
1209 |
// Write out keys and values (alternating) |
1210 |
for (int k = 0; k < segments.length; ++k) { |
1211 |
Segment seg = segments[k]; |
1212 |
Entry[] tab; |
1213 |
seg.lock(); |
1214 |
try { |
1215 |
tab = table; |
1216 |
} |
1217 |
finally { |
1218 |
seg.unlock(); |
1219 |
} |
1220 |
for (int i = k; i < tab.length; i+= segments.length) { |
1221 |
for (Entry e = tab[i]; e != null; e = e.next) { |
1222 |
s.writeObject(e.key); |
1223 |
s.writeObject(e.value); |
1224 |
} |
1225 |
} |
1226 |
} |
1227 |
|
1228 |
s.writeObject(null); |
1229 |
s.writeObject(null); |
1230 |
} |
1231 |
|
1232 |
/** |
1233 |
* Reconstitute the <tt>ConcurrentHashMap</tt> |
1234 |
* instance from a stream (i.e., |
1235 |
* deserialize it). |
1236 |
*/ |
1237 |
private void readObject(java.io.ObjectInputStream s) |
1238 |
throws IOException, ClassNotFoundException { |
1239 |
|
1240 |
// Read in the threshold, loadfactor, and any hidden stuff |
1241 |
s.defaultReadObject(); |
1242 |
|
1243 |
int cap = s.readInt(); |
1244 |
table = newTable(cap); |
1245 |
for (int i = 0; i < segments.length; ++i) |
1246 |
segments[i] = new Segment(); |
1247 |
|
1248 |
|
1249 |
// Read the keys and values, and put the mappings in the table |
1250 |
while (true) { |
1251 |
K key = (K) s.readObject(); |
1252 |
V value = (V) s.readObject(); |
1253 |
if (key == null) |
1254 |
break; |
1255 |
put(key, value); |
1256 |
} |
1257 |
} |
1258 |
} |