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/publicdomain/zero/1.0/ |
5 |
*/ |
6 |
|
7 |
package java.util.concurrent; |
8 |
|
9 |
import java.io.ObjectStreamField; |
10 |
import java.io.Serializable; |
11 |
import java.lang.reflect.ParameterizedType; |
12 |
import java.lang.reflect.Type; |
13 |
import java.util.Arrays; |
14 |
import java.util.Collection; |
15 |
import java.util.Comparator; |
16 |
import java.util.ConcurrentModificationException; |
17 |
import java.util.Enumeration; |
18 |
import java.util.HashMap; |
19 |
import java.util.Hashtable; |
20 |
import java.util.Iterator; |
21 |
import java.util.Map; |
22 |
import java.util.NoSuchElementException; |
23 |
import java.util.Set; |
24 |
import java.util.concurrent.ConcurrentMap; |
25 |
import java.util.concurrent.ForkJoinPool; |
26 |
import java.util.concurrent.atomic.AtomicInteger; |
27 |
import java.util.concurrent.locks.LockSupport; |
28 |
import java.util.concurrent.locks.ReentrantLock; |
29 |
|
30 |
/** |
31 |
* A hash table supporting full concurrency of retrievals and |
32 |
* high expected concurrency for updates. This class obeys the |
33 |
* same functional specification as {@link java.util.Hashtable}, and |
34 |
* includes versions of methods corresponding to each method of |
35 |
* {@code Hashtable}. However, even though all operations are |
36 |
* thread-safe, retrieval operations do <em>not</em> entail locking, |
37 |
* and there is <em>not</em> any support for locking the entire table |
38 |
* in a way that prevents all access. This class is fully |
39 |
* interoperable with {@code Hashtable} in programs that rely on its |
40 |
* thread safety but not on its synchronization details. |
41 |
* |
42 |
* <p>Retrieval operations (including {@code get}) generally do not |
43 |
* block, so may overlap with update operations (including {@code put} |
44 |
* and {@code remove}). Retrievals reflect the results of the most |
45 |
* recently <em>completed</em> update operations holding upon their |
46 |
* onset. (More formally, an update operation for a given key bears a |
47 |
* <em>happens-before</em> relation with any (non-null) retrieval for |
48 |
* that key reporting the updated value.) For aggregate operations |
49 |
* such as {@code putAll} and {@code clear}, concurrent retrievals may |
50 |
* reflect insertion or removal of only some entries. Similarly, |
51 |
* Iterators and Enumerations return elements reflecting the state of |
52 |
* the hash table at some point at or since the creation of the |
53 |
* iterator/enumeration. They do <em>not</em> throw {@link |
54 |
* ConcurrentModificationException}. However, iterators are designed |
55 |
* to be used by only one thread at a time. Bear in mind that the |
56 |
* results of aggregate status methods including {@code size}, {@code |
57 |
* isEmpty}, and {@code containsValue} are typically useful only when |
58 |
* a map is not undergoing concurrent updates in other threads. |
59 |
* Otherwise the results of these methods reflect transient states |
60 |
* that may be adequate for monitoring or estimation purposes, but not |
61 |
* for program control. |
62 |
* |
63 |
* <p>The table is dynamically expanded when there are too many |
64 |
* collisions (i.e., keys that have distinct hash codes but fall into |
65 |
* the same slot modulo the table size), with the expected average |
66 |
* effect of maintaining roughly two bins per mapping (corresponding |
67 |
* to a 0.75 load factor threshold for resizing). There may be much |
68 |
* variance around this average as mappings are added and removed, but |
69 |
* overall, this maintains a commonly accepted time/space tradeoff for |
70 |
* hash tables. However, resizing this or any other kind of hash |
71 |
* table may be a relatively slow operation. When possible, it is a |
72 |
* good idea to provide a size estimate as an optional {@code |
73 |
* initialCapacity} constructor argument. An additional optional |
74 |
* {@code loadFactor} constructor argument provides a further means of |
75 |
* customizing initial table capacity by specifying the table density |
76 |
* to be used in calculating the amount of space to allocate for the |
77 |
* given number of elements. Also, for compatibility with previous |
78 |
* versions of this class, constructors may optionally specify an |
79 |
* expected {@code concurrencyLevel} as an additional hint for |
80 |
* internal sizing. Note that using many keys with exactly the same |
81 |
* {@code hashCode()} is a sure way to slow down performance of any |
82 |
* hash table. To ameliorate impact, when keys are {@link Comparable}, |
83 |
* this class may use comparison order among keys to help break ties. |
84 |
* |
85 |
* <p>A {@link Set} projection of a ConcurrentHashMap may be created |
86 |
* (using {@link #newKeySet()} or {@link #newKeySet(int)}), or viewed |
87 |
* (using {@link #keySet(Object)} when only keys are of interest, and the |
88 |
* mapped values are (perhaps transiently) not used or all take the |
89 |
* same mapping value. |
90 |
* |
91 |
* <p>This class and its views and iterators implement all of the |
92 |
* <em>optional</em> methods of the {@link Map} and {@link Iterator} |
93 |
* interfaces. |
94 |
* |
95 |
* <p>Like {@link Hashtable} but unlike {@link HashMap}, this class |
96 |
* does <em>not</em> allow {@code null} to be used as a key or value. |
97 |
* |
98 |
* <p>This class is a member of the |
99 |
* <a href="{@docRoot}/../technotes/guides/collections/index.html"> |
100 |
* Java Collections Framework</a>. |
101 |
* |
102 |
* @since 1.5 |
103 |
* @author Doug Lea |
104 |
* @param <K> the type of keys maintained by this map |
105 |
* @param <V> the type of mapped values |
106 |
*/ |
107 |
public class ConcurrentHashMap<K,V> implements ConcurrentMap<K,V>, Serializable { |
108 |
private static final long serialVersionUID = 7249069246763182397L; |
109 |
|
110 |
/* |
111 |
* Overview: |
112 |
* |
113 |
* The primary design goal of this hash table is to maintain |
114 |
* concurrent readability (typically method get(), but also |
115 |
* iterators and related methods) while minimizing update |
116 |
* contention. Secondary goals are to keep space consumption about |
117 |
* the same or better than java.util.HashMap, and to support high |
118 |
* initial insertion rates on an empty table by many threads. |
119 |
* |
120 |
* This map usually acts as a binned (bucketed) hash table. Each |
121 |
* key-value mapping is held in a Node. Most nodes are instances |
122 |
* of the basic Node class with hash, key, value, and next |
123 |
* fields. However, various subclasses exist: TreeNodes are |
124 |
* arranged in balanced trees, not lists. TreeBins hold the roots |
125 |
* of sets of TreeNodes. ForwardingNodes are placed at the heads |
126 |
* of bins during resizing. ReservationNodes are used as |
127 |
* placeholders while establishing values in computeIfAbsent and |
128 |
* related methods. The types TreeBin, ForwardingNode, and |
129 |
* ReservationNode do not hold normal user keys, values, or |
130 |
* hashes, and are readily distinguishable during search etc |
131 |
* because they have negative hash fields and null key and value |
132 |
* fields. (These special nodes are either uncommon or transient, |
133 |
* so the impact of carrying around some unused fields is |
134 |
* insignficant.) |
135 |
* |
136 |
* The table is lazily initialized to a power-of-two size upon the |
137 |
* first insertion. Each bin in the table normally contains a |
138 |
* list of Nodes (most often, the list has only zero or one Node). |
139 |
* Table accesses require volatile/atomic reads, writes, and |
140 |
* CASes. Because there is no other way to arrange this without |
141 |
* adding further indirections, we use intrinsics |
142 |
* (sun.misc.Unsafe) operations. |
143 |
* |
144 |
* We use the top (sign) bit of Node hash fields for control |
145 |
* purposes -- it is available anyway because of addressing |
146 |
* constraints. Nodes with negative hash fields are specially |
147 |
* handled or ignored in map methods. |
148 |
* |
149 |
* Insertion (via put or its variants) of the first node in an |
150 |
* empty bin is performed by just CASing it to the bin. This is |
151 |
* by far the most common case for put operations under most |
152 |
* key/hash distributions. Other update operations (insert, |
153 |
* delete, and replace) require locks. We do not want to waste |
154 |
* the space required to associate a distinct lock object with |
155 |
* each bin, so instead use the first node of a bin list itself as |
156 |
* a lock. Locking support for these locks relies on builtin |
157 |
* "synchronized" monitors. |
158 |
* |
159 |
* Using the first node of a list as a lock does not by itself |
160 |
* suffice though: When a node is locked, any update must first |
161 |
* validate that it is still the first node after locking it, and |
162 |
* retry if not. Because new nodes are always appended to lists, |
163 |
* once a node is first in a bin, it remains first until deleted |
164 |
* or the bin becomes invalidated (upon resizing). |
165 |
* |
166 |
* The main disadvantage of per-bin locks is that other update |
167 |
* operations on other nodes in a bin list protected by the same |
168 |
* lock can stall, for example when user equals() or mapping |
169 |
* functions take a long time. However, statistically, under |
170 |
* random hash codes, this is not a common problem. Ideally, the |
171 |
* frequency of nodes in bins follows a Poisson distribution |
172 |
* (http://en.wikipedia.org/wiki/Poisson_distribution) with a |
173 |
* parameter of about 0.5 on average, given the resizing threshold |
174 |
* of 0.75, although with a large variance because of resizing |
175 |
* granularity. Ignoring variance, the expected occurrences of |
176 |
* list size k are (exp(-0.5) * pow(0.5, k) / factorial(k)). The |
177 |
* first values are: |
178 |
* |
179 |
* 0: 0.60653066 |
180 |
* 1: 0.30326533 |
181 |
* 2: 0.07581633 |
182 |
* 3: 0.01263606 |
183 |
* 4: 0.00157952 |
184 |
* 5: 0.00015795 |
185 |
* 6: 0.00001316 |
186 |
* 7: 0.00000094 |
187 |
* 8: 0.00000006 |
188 |
* more: less than 1 in ten million |
189 |
* |
190 |
* Lock contention probability for two threads accessing distinct |
191 |
* elements is roughly 1 / (8 * #elements) under random hashes. |
192 |
* |
193 |
* Actual hash code distributions encountered in practice |
194 |
* sometimes deviate significantly from uniform randomness. This |
195 |
* includes the case when N > (1<<30), so some keys MUST collide. |
196 |
* Similarly for dumb or hostile usages in which multiple keys are |
197 |
* designed to have identical hash codes or ones that differs only |
198 |
* in masked-out high bits. So we use a secondary strategy that |
199 |
* applies when the number of nodes in a bin exceeds a |
200 |
* threshold. These TreeBins use a balanced tree to hold nodes (a |
201 |
* specialized form of red-black trees), bounding search time to |
202 |
* O(log N). Each search step in a TreeBin is at least twice as |
203 |
* slow as in a regular list, but given that N cannot exceed |
204 |
* (1<<64) (before running out of addresses) this bounds search |
205 |
* steps, lock hold times, etc, to reasonable constants (roughly |
206 |
* 100 nodes inspected per operation worst case) so long as keys |
207 |
* are Comparable (which is very common -- String, Long, etc). |
208 |
* TreeBin nodes (TreeNodes) also maintain the same "next" |
209 |
* traversal pointers as regular nodes, so can be traversed in |
210 |
* iterators in the same way. |
211 |
* |
212 |
* The table is resized when occupancy exceeds a percentage |
213 |
* threshold (nominally, 0.75, but see below). Any thread |
214 |
* noticing an overfull bin may assist in resizing after the |
215 |
* initiating thread allocates and sets up the replacement |
216 |
* array. However, rather than stalling, these other threads may |
217 |
* proceed with insertions etc. The use of TreeBins shields us |
218 |
* from the worst case effects of overfilling while resizes are in |
219 |
* progress. Resizing proceeds by transferring bins, one by one, |
220 |
* from the table to the next table. To enable concurrency, the |
221 |
* next table must be (incrementally) prefilled with place-holders |
222 |
* serving as reverse forwarders to the old table. Because we are |
223 |
* using power-of-two expansion, the elements from each bin must |
224 |
* either stay at same index, or move with a power of two |
225 |
* offset. We eliminate unnecessary node creation by catching |
226 |
* cases where old nodes can be reused because their next fields |
227 |
* won't change. On average, only about one-sixth of them need |
228 |
* cloning when a table doubles. The nodes they replace will be |
229 |
* garbage collectable as soon as they are no longer referenced by |
230 |
* any reader thread that may be in the midst of concurrently |
231 |
* traversing table. Upon transfer, the old table bin contains |
232 |
* only a special forwarding node (with hash field "MOVED") that |
233 |
* contains the next table as its key. On encountering a |
234 |
* forwarding node, access and update operations restart, using |
235 |
* the new table. |
236 |
* |
237 |
* Each bin transfer requires its bin lock, which can stall |
238 |
* waiting for locks while resizing. However, because other |
239 |
* threads can join in and help resize rather than contend for |
240 |
* locks, average aggregate waits become shorter as resizing |
241 |
* progresses. The transfer operation must also ensure that all |
242 |
* accessible bins in both the old and new table are usable by any |
243 |
* traversal. This is arranged by proceeding from the last bin |
244 |
* (table.length - 1) up towards the first. Upon seeing a |
245 |
* forwarding node, traversals (see class Traverser) arrange to |
246 |
* move to the new table without revisiting nodes. However, to |
247 |
* ensure that no intervening nodes are skipped, bin splitting can |
248 |
* only begin after the associated reverse-forwarders are in |
249 |
* place. |
250 |
* |
251 |
* The traversal scheme also applies to partial traversals of |
252 |
* ranges of bins (via an alternate Traverser constructor) |
253 |
* to support partitioned aggregate operations. Also, read-only |
254 |
* operations give up if ever forwarded to a null table, which |
255 |
* provides support for shutdown-style clearing, which is also not |
256 |
* currently implemented. |
257 |
* |
258 |
* Lazy table initialization minimizes footprint until first use, |
259 |
* and also avoids resizings when the first operation is from a |
260 |
* putAll, constructor with map argument, or deserialization. |
261 |
* These cases attempt to override the initial capacity settings, |
262 |
* but harmlessly fail to take effect in cases of races. |
263 |
* |
264 |
* The element count is maintained using a specialization of |
265 |
* LongAdder. We need to incorporate a specialization rather than |
266 |
* just use a LongAdder in order to access implicit |
267 |
* contention-sensing that leads to creation of multiple |
268 |
* CounterCells. The counter mechanics avoid contention on |
269 |
* updates but can encounter cache thrashing if read too |
270 |
* frequently during concurrent access. To avoid reading so often, |
271 |
* resizing under contention is attempted only upon adding to a |
272 |
* bin already holding two or more nodes. Under uniform hash |
273 |
* distributions, the probability of this occurring at threshold |
274 |
* is around 13%, meaning that only about 1 in 8 puts check |
275 |
* threshold (and after resizing, many fewer do so). |
276 |
* |
277 |
* TreeBins use a special form of comparison for search and |
278 |
* related operations (which is the main reason we cannot use |
279 |
* existing collections such as TreeMaps). TreeBins contain |
280 |
* Comparable elements, but may contain others, as well as |
281 |
* elements that are Comparable but not necessarily Comparable |
282 |
* for the same T, so we cannot invoke compareTo among them. To |
283 |
* handle this, the tree is ordered primarily by hash value, then |
284 |
* by Comparable.compareTo order if applicable. On lookup at a |
285 |
* node, if elements are not comparable or compare as 0 then both |
286 |
* left and right children may need to be searched in the case of |
287 |
* tied hash values. (This corresponds to the full list search |
288 |
* that would be necessary if all elements were non-Comparable and |
289 |
* had tied hashes.) The red-black balancing code is updated from |
290 |
* pre-jdk-collections |
291 |
* (http://gee.cs.oswego.edu/dl/classes/collections/RBCell.java) |
292 |
* based in turn on Cormen, Leiserson, and Rivest "Introduction to |
293 |
* Algorithms" (CLR). |
294 |
* |
295 |
* TreeBins also require an additional locking mechanism. While |
296 |
* list traversal is always possible by readers even during |
297 |
* updates, tree traversal is not, mainly beause of tree-rotations |
298 |
* that may change the root node and/or its linkages. TreeBins |
299 |
* include a simple read-write lock mechanism parasitic on the |
300 |
* main bin-synchronization strategy: Structural adjustments |
301 |
* associated with an insertion or removal are already bin-locked |
302 |
* (and so cannot conflict with other writers) but must wait for |
303 |
* ongoing readers to finish. Since there can be only one such |
304 |
* waiter, we use a simple scheme using a single "waiter" field to |
305 |
* block writers. However, readers need never block. If the root |
306 |
* lock is held, they proceed along the slow traversal path (via |
307 |
* next-pointers) until the lock becomes available or the list is |
308 |
* exhausted, whichever comes first. These cases are not fast, but |
309 |
* maximize aggregate expected throughput. |
310 |
* |
311 |
* Maintaining API and serialization compatibility with previous |
312 |
* versions of this class introduces several oddities. Mainly: We |
313 |
* leave untouched but unused constructor arguments refering to |
314 |
* concurrencyLevel. We accept a loadFactor constructor argument, |
315 |
* but apply it only to initial table capacity (which is the only |
316 |
* time that we can guarantee to honor it.) We also declare an |
317 |
* unused "Segment" class that is instantiated in minimal form |
318 |
* only when serializing. |
319 |
* |
320 |
* This file is organized to make things a little easier to follow |
321 |
* while reading than they might otherwise: First the main static |
322 |
* declarations and utilities, then fields, then main public |
323 |
* methods (with a few factorings of multiple public methods into |
324 |
* internal ones), then sizing methods, trees, traversers, and |
325 |
* bulk operations. |
326 |
*/ |
327 |
|
328 |
/* ---------------- Constants -------------- */ |
329 |
|
330 |
/** |
331 |
* The largest possible table capacity. This value must be |
332 |
* exactly 1<<30 to stay within Java array allocation and indexing |
333 |
* bounds for power of two table sizes, and is further required |
334 |
* because the top two bits of 32bit hash fields are used for |
335 |
* control purposes. |
336 |
*/ |
337 |
private static final int MAXIMUM_CAPACITY = 1 << 30; |
338 |
|
339 |
/** |
340 |
* The default initial table capacity. Must be a power of 2 |
341 |
* (i.e., at least 1) and at most MAXIMUM_CAPACITY. |
342 |
*/ |
343 |
private static final int DEFAULT_CAPACITY = 16; |
344 |
|
345 |
/** |
346 |
* The largest possible (non-power of two) array size. |
347 |
* Needed by toArray and related methods. |
348 |
*/ |
349 |
static final int MAX_ARRAY_SIZE = Integer.MAX_VALUE - 8; |
350 |
|
351 |
/** |
352 |
* The default concurrency level for this table. Unused but |
353 |
* defined for compatibility with previous versions of this class. |
354 |
*/ |
355 |
private static final int DEFAULT_CONCURRENCY_LEVEL = 16; |
356 |
|
357 |
/** |
358 |
* The load factor for this table. Overrides of this value in |
359 |
* constructors affect only the initial table capacity. The |
360 |
* actual floating point value isn't normally used -- it is |
361 |
* simpler to use expressions such as {@code n - (n >>> 2)} for |
362 |
* the associated resizing threshold. |
363 |
*/ |
364 |
private static final float LOAD_FACTOR = 0.75f; |
365 |
|
366 |
/** |
367 |
* The bin count threshold for using a tree rather than list for a |
368 |
* bin. Bins are converted to trees when adding an element to a |
369 |
* bin with at least this many nodes. The value must be greater |
370 |
* than 2, and should be at least 8 to mesh with assumptions in |
371 |
* tree removal about conversion back to plain bins upon |
372 |
* shrinkage. |
373 |
*/ |
374 |
static final int TREEIFY_THRESHOLD = 8; |
375 |
|
376 |
/** |
377 |
* The bin count threshold for untreeifying a (split) bin during a |
378 |
* resize operation. Should be less than TREEIFY_THRESHOLD, and at |
379 |
* most 6 to mesh with shrinkage detection under removal. |
380 |
*/ |
381 |
static final int UNTREEIFY_THRESHOLD = 6; |
382 |
|
383 |
/** |
384 |
* The smallest table capacity for which bins may be treeified. |
385 |
* (Otherwise the table is resized if too many nodes in a bin.) |
386 |
* The value should be at least 4 * TREEIFY_THRESHOLD to avoid |
387 |
* conflicts between resizing and treeification thresholds. |
388 |
*/ |
389 |
static final int MIN_TREEIFY_CAPACITY = 64; |
390 |
|
391 |
/** |
392 |
* Minimum number of rebinnings per transfer step. Ranges are |
393 |
* subdivided to allow multiple resizer threads. This value |
394 |
* serves as a lower bound to avoid resizers encountering |
395 |
* excessive memory contention. The value should be at least |
396 |
* DEFAULT_CAPACITY. |
397 |
*/ |
398 |
private static final int MIN_TRANSFER_STRIDE = 16; |
399 |
|
400 |
/* |
401 |
* Encodings for Node hash fields. See above for explanation. |
402 |
*/ |
403 |
static final int MOVED = 0x8fffffff; // (-1) hash for forwarding nodes |
404 |
static final int TREEBIN = 0x80000000; // hash for heads of treea |
405 |
static final int RESERVED = 0x80000001; // hash for transient reservations |
406 |
static final int HASH_BITS = 0x7fffffff; // usable bits of normal node hash |
407 |
|
408 |
/** Number of CPUS, to place bounds on some sizings */ |
409 |
static final int NCPU = Runtime.getRuntime().availableProcessors(); |
410 |
|
411 |
/** For serialization compatibility. */ |
412 |
private static final ObjectStreamField[] serialPersistentFields = { |
413 |
new ObjectStreamField("segments", Segment[].class), |
414 |
new ObjectStreamField("segmentMask", Integer.TYPE), |
415 |
new ObjectStreamField("segmentShift", Integer.TYPE) |
416 |
}; |
417 |
|
418 |
/* ---------------- Nodes -------------- */ |
419 |
|
420 |
/** |
421 |
* Key-value entry. This class is never exported out as a |
422 |
* user-mutable Map.Entry (i.e., one supporting setValue; see |
423 |
* MapEntry below), but can be used for read-only traversals used |
424 |
* in bulk tasks. Subclasses of Node with a negativehash field |
425 |
* are special, and contain null keys and values (but are never |
426 |
* exported). Otherwise, keys and vals are never null. |
427 |
*/ |
428 |
static class Node<K,V> implements Map.Entry<K,V> { |
429 |
final int hash; |
430 |
final K key; |
431 |
volatile V val; |
432 |
Node<K,V> next; |
433 |
|
434 |
Node(int hash, K key, V val, Node<K,V> next) { |
435 |
this.hash = hash; |
436 |
this.key = key; |
437 |
this.val = val; |
438 |
this.next = next; |
439 |
} |
440 |
|
441 |
public final K getKey() { return key; } |
442 |
public final V getValue() { return val; } |
443 |
public final int hashCode() { return key.hashCode() ^ val.hashCode(); } |
444 |
public final String toString(){ return key + "=" + val; } |
445 |
public final V setValue(V value) { |
446 |
throw new UnsupportedOperationException(); |
447 |
} |
448 |
|
449 |
public final boolean equals(Object o) { |
450 |
Object k, v, u; Map.Entry<?,?> e; |
451 |
return ((o instanceof Map.Entry) && |
452 |
(k = (e = (Map.Entry<?,?>)o).getKey()) != null && |
453 |
(v = e.getValue()) != null && |
454 |
(k == key || k.equals(key)) && |
455 |
(v == (u = val) || v.equals(u))); |
456 |
} |
457 |
|
458 |
/** |
459 |
* Virtualized support for map.get(); overridden in subclasses. |
460 |
*/ |
461 |
Node<K,V> find(int h, Object k) { |
462 |
Node<K,V> e = this; |
463 |
if (k != null) { |
464 |
do { |
465 |
K ek; |
466 |
if (e.hash == h && |
467 |
((ek = e.key) == k || (ek != null && k.equals(ek)))) |
468 |
return e; |
469 |
} while ((e = e.next) != null); |
470 |
} |
471 |
return null; |
472 |
} |
473 |
} |
474 |
|
475 |
/* ---------------- Static utilities -------------- */ |
476 |
|
477 |
/** |
478 |
* Spreads (XORs) higher bits of hash to lower and also forces top |
479 |
* bit to 0. Because the table uses power-of-two masking, sets of |
480 |
* hashes that vary only in bits above the current mask will |
481 |
* always collide. (Among known examples are sets of Float keys |
482 |
* holding consecutive whole numbers in small tables.) So we |
483 |
* apply a transform that spreads the impact of higher bits |
484 |
* downward. There is a tradeoff between speed, utility, and |
485 |
* quality of bit-spreading. Because many common sets of hashes |
486 |
* are already reasonably distributed (so don't benefit from |
487 |
* spreading), and because we use trees to handle large sets of |
488 |
* collisions in bins, we just XOR some shifted bits in the |
489 |
* cheapest possible way to reduce systematic lossage, as well as |
490 |
* to incorporate impact of the highest bits that would otherwise |
491 |
* never be used in index calculations because of table bounds. |
492 |
*/ |
493 |
static final int spread(int h) { |
494 |
return (h ^ (h >>> 16)) & HASH_BITS; |
495 |
} |
496 |
|
497 |
/** |
498 |
* Returns a power of two table size for the given desired capacity. |
499 |
* See Hackers Delight, sec 3.2 |
500 |
*/ |
501 |
private static final int tableSizeFor(int c) { |
502 |
int n = c - 1; |
503 |
n |= n >>> 1; |
504 |
n |= n >>> 2; |
505 |
n |= n >>> 4; |
506 |
n |= n >>> 8; |
507 |
n |= n >>> 16; |
508 |
return (n < 0) ? 1 : (n >= MAXIMUM_CAPACITY) ? MAXIMUM_CAPACITY : n + 1; |
509 |
} |
510 |
|
511 |
/** |
512 |
* Returns x's Class if it is of the form "class C implements |
513 |
* Comparable<C>", else null. |
514 |
*/ |
515 |
static Class<?> comparableClassFor(Object x) { |
516 |
if (x instanceof Comparable) { |
517 |
Class<?> c; Type[] ts, as; Type t; ParameterizedType p; |
518 |
if ((c = x.getClass()) == String.class) // bypass checks |
519 |
return c; |
520 |
if ((ts = c.getGenericInterfaces()) != null) { |
521 |
for (int i = 0; i < ts.length; ++i) { |
522 |
if (((t = ts[i]) instanceof ParameterizedType) && |
523 |
((p = (ParameterizedType)t).getRawType() == |
524 |
Comparable.class) && |
525 |
(as = p.getActualTypeArguments()) != null && |
526 |
as.length == 1 && as[0] == c) // type arg is c |
527 |
return c; |
528 |
} |
529 |
} |
530 |
} |
531 |
return null; |
532 |
} |
533 |
|
534 |
/** |
535 |
* Returns k.compareTo(x) if x matches kc (k's screened comparable |
536 |
* class), else 0. |
537 |
*/ |
538 |
@SuppressWarnings({"rawtypes","unchecked"}) // for cast to Comparable |
539 |
static int compareComparables(Class<?> kc, Object k, Object x) { |
540 |
return (x == null || x.getClass() != kc ? 0 : |
541 |
((Comparable)k).compareTo(x)); |
542 |
} |
543 |
|
544 |
/* ---------------- Table element access -------------- */ |
545 |
|
546 |
/* |
547 |
* Volatile access methods are used for table elements as well as |
548 |
* elements of in-progress next table while resizing. All uses of |
549 |
* the tab arguments must be null checked by callers. All callers |
550 |
* also paranoically precheck that tab's length is not zero (or an |
551 |
* equivalent check), thus ensuring that any index argument taking |
552 |
* the form of a hash value anded with (length - 1) is a valid |
553 |
* index. Note that, to be correct wrt arbitrary concurrency |
554 |
* errors by users, these checks must operate on local variables, |
555 |
* which accounts for some odd-looking inline assignments below. |
556 |
* Note that calls to setTabAt always occur within locked regions, |
557 |
* and so do not need full volatile semantics, but still require |
558 |
* ordering to maintain concurrent readability. |
559 |
*/ |
560 |
|
561 |
@SuppressWarnings("unchecked") |
562 |
static final <K,V> Node<K,V> tabAt(Node<K,V>[] tab, int i) { |
563 |
return (Node<K,V>)U.getObjectVolatile(tab, ((long)i << ASHIFT) + ABASE); |
564 |
} |
565 |
|
566 |
static final <K,V> boolean casTabAt(Node<K,V>[] tab, int i, |
567 |
Node<K,V> c, Node<K,V> v) { |
568 |
return U.compareAndSwapObject(tab, ((long)i << ASHIFT) + ABASE, c, v); |
569 |
} |
570 |
|
571 |
static final <K,V> void setTabAt(Node<K,V>[] tab, int i, Node<K,V> v) { |
572 |
U.putOrderedObject(tab, ((long)i << ASHIFT) + ABASE, v); |
573 |
} |
574 |
|
575 |
/* ---------------- Fields -------------- */ |
576 |
|
577 |
/** |
578 |
* The array of bins. Lazily initialized upon first insertion. |
579 |
* Size is always a power of two. Accessed directly by iterators. |
580 |
*/ |
581 |
transient volatile Node<K,V>[] table; |
582 |
|
583 |
/** |
584 |
* The next table to use; non-null only while resizing. |
585 |
*/ |
586 |
private transient volatile Node<K,V>[] nextTable; |
587 |
|
588 |
/** |
589 |
* Base counter value, used mainly when there is no contention, |
590 |
* but also as a fallback during table initialization |
591 |
* races. Updated via CAS. |
592 |
*/ |
593 |
private transient volatile long baseCount; |
594 |
|
595 |
/** |
596 |
* Table initialization and resizing control. When negative, the |
597 |
* table is being initialized or resized: -1 for initialization, |
598 |
* else -(1 + the number of active resizing threads). Otherwise, |
599 |
* when table is null, holds the initial table size to use upon |
600 |
* creation, or 0 for default. After initialization, holds the |
601 |
* next element count value upon which to resize the table. |
602 |
*/ |
603 |
private transient volatile int sizeCtl; |
604 |
|
605 |
/** |
606 |
* The next table index (plus one) to split while resizing. |
607 |
*/ |
608 |
private transient volatile int transferIndex; |
609 |
|
610 |
/** |
611 |
* The least available table index to split while resizing. |
612 |
*/ |
613 |
private transient volatile int transferOrigin; |
614 |
|
615 |
/** |
616 |
* Spinlock (locked via CAS) used when resizing and/or creating CounterCells. |
617 |
*/ |
618 |
private transient volatile int cellsBusy; |
619 |
|
620 |
/** |
621 |
* Table of counter cells. When non-null, size is a power of 2. |
622 |
*/ |
623 |
private transient volatile CounterCell[] counterCells; |
624 |
|
625 |
// views |
626 |
private transient KeySetView<K,V> keySet; |
627 |
private transient ValuesView<K,V> values; |
628 |
private transient EntrySetView<K,V> entrySet; |
629 |
|
630 |
|
631 |
/* ---------------- Public operations -------------- */ |
632 |
|
633 |
/** |
634 |
* Creates a new, empty map with the default initial table size (16). |
635 |
*/ |
636 |
public ConcurrentHashMap() { |
637 |
} |
638 |
|
639 |
/** |
640 |
* Creates a new, empty map with an initial table size |
641 |
* accommodating the specified number of elements without the need |
642 |
* to dynamically resize. |
643 |
* |
644 |
* @param initialCapacity The implementation performs internal |
645 |
* sizing to accommodate this many elements. |
646 |
* @throws IllegalArgumentException if the initial capacity of |
647 |
* elements is negative |
648 |
*/ |
649 |
public ConcurrentHashMap(int initialCapacity) { |
650 |
if (initialCapacity < 0) |
651 |
throw new IllegalArgumentException(); |
652 |
int cap = ((initialCapacity >= (MAXIMUM_CAPACITY >>> 1)) ? |
653 |
MAXIMUM_CAPACITY : |
654 |
tableSizeFor(initialCapacity + (initialCapacity >>> 1) + 1)); |
655 |
this.sizeCtl = cap; |
656 |
} |
657 |
|
658 |
/** |
659 |
* Creates a new map with the same mappings as the given map. |
660 |
* |
661 |
* @param m the map |
662 |
*/ |
663 |
public ConcurrentHashMap(Map<? extends K, ? extends V> m) { |
664 |
this.sizeCtl = DEFAULT_CAPACITY; |
665 |
putAll(m); |
666 |
} |
667 |
|
668 |
/** |
669 |
* Creates a new, empty map with an initial table size based on |
670 |
* the given number of elements ({@code initialCapacity}) and |
671 |
* initial table density ({@code loadFactor}). |
672 |
* |
673 |
* @param initialCapacity the initial capacity. The implementation |
674 |
* performs internal sizing to accommodate this many elements, |
675 |
* given the specified load factor. |
676 |
* @param loadFactor the load factor (table density) for |
677 |
* establishing the initial table size |
678 |
* @throws IllegalArgumentException if the initial capacity of |
679 |
* elements is negative or the load factor is nonpositive |
680 |
* |
681 |
* @since 1.6 |
682 |
*/ |
683 |
public ConcurrentHashMap(int initialCapacity, float loadFactor) { |
684 |
this(initialCapacity, loadFactor, 1); |
685 |
} |
686 |
|
687 |
/** |
688 |
* Creates a new, empty map with an initial table size based on |
689 |
* the given number of elements ({@code initialCapacity}), table |
690 |
* density ({@code loadFactor}), and number of concurrently |
691 |
* updating threads ({@code concurrencyLevel}). |
692 |
* |
693 |
* @param initialCapacity the initial capacity. The implementation |
694 |
* performs internal sizing to accommodate this many elements, |
695 |
* given the specified load factor. |
696 |
* @param loadFactor the load factor (table density) for |
697 |
* establishing the initial table size |
698 |
* @param concurrencyLevel the estimated number of concurrently |
699 |
* updating threads. The implementation may use this value as |
700 |
* a sizing hint. |
701 |
* @throws IllegalArgumentException if the initial capacity is |
702 |
* negative or the load factor or concurrencyLevel are |
703 |
* nonpositive |
704 |
*/ |
705 |
public ConcurrentHashMap(int initialCapacity, |
706 |
float loadFactor, int concurrencyLevel) { |
707 |
if (!(loadFactor > 0.0f) || initialCapacity < 0 || concurrencyLevel <= 0) |
708 |
throw new IllegalArgumentException(); |
709 |
if (initialCapacity < concurrencyLevel) // Use at least as many bins |
710 |
initialCapacity = concurrencyLevel; // as estimated threads |
711 |
long size = (long)(1.0 + (long)initialCapacity / loadFactor); |
712 |
int cap = (size >= (long)MAXIMUM_CAPACITY) ? |
713 |
MAXIMUM_CAPACITY : tableSizeFor((int)size); |
714 |
this.sizeCtl = cap; |
715 |
} |
716 |
|
717 |
// Original (since JDK1.2) Map methods |
718 |
|
719 |
/** |
720 |
* {@inheritDoc} |
721 |
*/ |
722 |
public int size() { |
723 |
long n = sumCount(); |
724 |
return ((n < 0L) ? 0 : |
725 |
(n > (long)Integer.MAX_VALUE) ? Integer.MAX_VALUE : |
726 |
(int)n); |
727 |
} |
728 |
|
729 |
/** |
730 |
* {@inheritDoc} |
731 |
*/ |
732 |
public boolean isEmpty() { |
733 |
return sumCount() <= 0L; // ignore transient negative values |
734 |
} |
735 |
|
736 |
/** |
737 |
* Returns the value to which the specified key is mapped, |
738 |
* or {@code null} if this map contains no mapping for the key. |
739 |
* |
740 |
* <p>More formally, if this map contains a mapping from a key |
741 |
* {@code k} to a value {@code v} such that {@code key.equals(k)}, |
742 |
* then this method returns {@code v}; otherwise it returns |
743 |
* {@code null}. (There can be at most one such mapping.) |
744 |
* |
745 |
* @throws NullPointerException if the specified key is null |
746 |
*/ |
747 |
public V get(Object key) { |
748 |
Node<K,V>[] tab; Node<K,V> e, p; int n, eh; K ek; |
749 |
int h = spread(key.hashCode()); |
750 |
if ((tab = table) != null && (n = tab.length) > 0 && |
751 |
(e = tabAt(tab, (n - 1) & h)) != null) { |
752 |
if ((eh = e.hash) == h) { |
753 |
if ((ek = e.key) == key || (ek != null && key.equals(ek))) |
754 |
return e.val; |
755 |
} |
756 |
else if (eh < 0) |
757 |
return (p = e.find(h, key)) != null ? p.val : null; |
758 |
while ((e = e.next) != null) { |
759 |
if (e.hash == h && |
760 |
((ek = e.key) == key || (ek != null && key.equals(ek)))) |
761 |
return e.val; |
762 |
} |
763 |
} |
764 |
return null; |
765 |
} |
766 |
|
767 |
/** |
768 |
* Tests if the specified object is a key in this table. |
769 |
* |
770 |
* @param key possible key |
771 |
* @return {@code true} if and only if the specified object |
772 |
* is a key in this table, as determined by the |
773 |
* {@code equals} method; {@code false} otherwise |
774 |
* @throws NullPointerException if the specified key is null |
775 |
*/ |
776 |
public boolean containsKey(Object key) { |
777 |
return get(key) != null; |
778 |
} |
779 |
|
780 |
/** |
781 |
* Returns {@code true} if this map maps one or more keys to the |
782 |
* specified value. Note: This method may require a full traversal |
783 |
* of the map, and is much slower than method {@code containsKey}. |
784 |
* |
785 |
* @param value value whose presence in this map is to be tested |
786 |
* @return {@code true} if this map maps one or more keys to the |
787 |
* specified value |
788 |
* @throws NullPointerException if the specified value is null |
789 |
*/ |
790 |
public boolean containsValue(Object value) { |
791 |
if (value == null) |
792 |
throw new NullPointerException(); |
793 |
Node<K,V>[] t; |
794 |
if ((t = table) != null) { |
795 |
Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length); |
796 |
for (Node<K,V> p; (p = it.advance()) != null; ) { |
797 |
V v; |
798 |
if ((v = p.val) == value || (v != null && value.equals(v))) |
799 |
return true; |
800 |
} |
801 |
} |
802 |
return false; |
803 |
} |
804 |
|
805 |
/** |
806 |
* Maps the specified key to the specified value in this table. |
807 |
* Neither the key nor the value can be null. |
808 |
* |
809 |
* <p>The value can be retrieved by calling the {@code get} method |
810 |
* with a key that is equal to the original key. |
811 |
* |
812 |
* @param key key with which the specified value is to be associated |
813 |
* @param value value to be associated with the specified key |
814 |
* @return the previous value associated with {@code key}, or |
815 |
* {@code null} if there was no mapping for {@code key} |
816 |
* @throws NullPointerException if the specified key or value is null |
817 |
*/ |
818 |
public V put(K key, V value) { |
819 |
return putVal(key, value, false); |
820 |
} |
821 |
|
822 |
/** Implementation for put and putIfAbsent */ |
823 |
final V putVal(K key, V value, boolean onlyIfAbsent) { |
824 |
if (key == null || value == null) throw new NullPointerException(); |
825 |
int hash = spread(key.hashCode()); |
826 |
int binCount = 0; |
827 |
for (Node<K,V>[] tab = table;;) { |
828 |
Node<K,V> f; int n, i, fh; |
829 |
if (tab == null || (n = tab.length) == 0) |
830 |
tab = initTable(); |
831 |
else if ((f = tabAt(tab, i = (n - 1) & hash)) == null) { |
832 |
if (casTabAt(tab, i, null, |
833 |
new Node<K,V>(hash, key, value, null))) |
834 |
break; // no lock when adding to empty bin |
835 |
} |
836 |
else if ((fh = f.hash) == MOVED) |
837 |
tab = helpTransfer(tab, f); |
838 |
else { |
839 |
V oldVal = null; |
840 |
synchronized (f) { |
841 |
if (tabAt(tab, i) == f) { |
842 |
if (fh >= 0) { |
843 |
binCount = 1; |
844 |
for (Node<K,V> e = f;; ++binCount) { |
845 |
K ek; |
846 |
if (e.hash == hash && |
847 |
((ek = e.key) == key || |
848 |
(ek != null && key.equals(ek)))) { |
849 |
oldVal = e.val; |
850 |
if (!onlyIfAbsent) |
851 |
e.val = value; |
852 |
break; |
853 |
} |
854 |
Node<K,V> pred = e; |
855 |
if ((e = e.next) == null) { |
856 |
pred.next = new Node<K,V>(hash, key, |
857 |
value, null); |
858 |
break; |
859 |
} |
860 |
} |
861 |
} |
862 |
else if (f instanceof TreeBin) { |
863 |
Node<K,V> p; |
864 |
binCount = 2; |
865 |
if ((p = ((TreeBin<K,V>)f).putTreeVal(hash, key, |
866 |
value)) != null) { |
867 |
oldVal = p.val; |
868 |
if (!onlyIfAbsent) |
869 |
p.val = value; |
870 |
} |
871 |
} |
872 |
} |
873 |
} |
874 |
if (binCount != 0) { |
875 |
if (binCount >= TREEIFY_THRESHOLD) |
876 |
treeifyBin(tab, i); |
877 |
if (oldVal != null) |
878 |
return oldVal; |
879 |
break; |
880 |
} |
881 |
} |
882 |
} |
883 |
addCount(1L, binCount); |
884 |
return null; |
885 |
} |
886 |
|
887 |
/** |
888 |
* Copies all of the mappings from the specified map to this one. |
889 |
* These mappings replace any mappings that this map had for any of the |
890 |
* keys currently in the specified map. |
891 |
* |
892 |
* @param m mappings to be stored in this map |
893 |
*/ |
894 |
public void putAll(Map<? extends K, ? extends V> m) { |
895 |
tryPresize(m.size()); |
896 |
for (Map.Entry<? extends K, ? extends V> e : m.entrySet()) |
897 |
putVal(e.getKey(), e.getValue(), false); |
898 |
} |
899 |
|
900 |
/** |
901 |
* Removes the key (and its corresponding value) from this map. |
902 |
* This method does nothing if the key is not in the map. |
903 |
* |
904 |
* @param key the key that needs to be removed |
905 |
* @return the previous value associated with {@code key}, or |
906 |
* {@code null} if there was no mapping for {@code key} |
907 |
* @throws NullPointerException if the specified key is null |
908 |
*/ |
909 |
public V remove(Object key) { |
910 |
return replaceNode(key, null, null); |
911 |
} |
912 |
|
913 |
/** |
914 |
* Implementation for the four public remove/replace methods: |
915 |
* Replaces node value with v, conditional upon match of cv if |
916 |
* non-null. If resulting value is null, delete. |
917 |
*/ |
918 |
final V replaceNode(Object key, V value, Object cv) { |
919 |
int hash = spread(key.hashCode()); |
920 |
for (Node<K,V>[] tab = table;;) { |
921 |
Node<K,V> f; int n, i, fh; |
922 |
if (tab == null || (n = tab.length) == 0 || |
923 |
(f = tabAt(tab, i = (n - 1) & hash)) == null) |
924 |
break; |
925 |
else if ((fh = f.hash) == MOVED) |
926 |
tab = helpTransfer(tab, f); |
927 |
else { |
928 |
V oldVal = null; |
929 |
boolean validated = false; |
930 |
synchronized (f) { |
931 |
if (tabAt(tab, i) == f) { |
932 |
if (fh >= 0) { |
933 |
validated = true; |
934 |
for (Node<K,V> e = f, pred = null;;) { |
935 |
K ek; |
936 |
if (e.hash == hash && |
937 |
((ek = e.key) == key || |
938 |
(ek != null && key.equals(ek)))) { |
939 |
V ev = e.val; |
940 |
if (cv == null || cv == ev || |
941 |
(ev != null && cv.equals(ev))) { |
942 |
oldVal = ev; |
943 |
if (value != null) |
944 |
e.val = value; |
945 |
else if (pred != null) |
946 |
pred.next = e.next; |
947 |
else |
948 |
setTabAt(tab, i, e.next); |
949 |
} |
950 |
break; |
951 |
} |
952 |
pred = e; |
953 |
if ((e = e.next) == null) |
954 |
break; |
955 |
} |
956 |
} |
957 |
else if (f instanceof TreeBin) { |
958 |
validated = true; |
959 |
TreeBin<K,V> t = (TreeBin<K,V>)f; |
960 |
TreeNode<K,V> r, p; |
961 |
if ((r = t.root) != null && |
962 |
(p = r.findTreeNode(hash, key, null)) != null) { |
963 |
V pv = p.val; |
964 |
if (cv == null || cv == pv || |
965 |
(pv != null && cv.equals(pv))) { |
966 |
oldVal = pv; |
967 |
if (value != null) |
968 |
p.val = value; |
969 |
else if (t.removeTreeNode(p)) |
970 |
setTabAt(tab, i, untreeify(t.first)); |
971 |
} |
972 |
} |
973 |
} |
974 |
} |
975 |
} |
976 |
if (validated) { |
977 |
if (oldVal != null) { |
978 |
if (value == null) |
979 |
addCount(-1L, -1); |
980 |
return oldVal; |
981 |
} |
982 |
break; |
983 |
} |
984 |
} |
985 |
} |
986 |
return null; |
987 |
} |
988 |
|
989 |
/** |
990 |
* Removes all of the mappings from this map. |
991 |
*/ |
992 |
public void clear() { |
993 |
long delta = 0L; // negative number of deletions |
994 |
int i = 0; |
995 |
Node<K,V>[] tab = table; |
996 |
while (tab != null && i < tab.length) { |
997 |
int fh; |
998 |
Node<K,V> f = tabAt(tab, i); |
999 |
if (f == null) |
1000 |
++i; |
1001 |
else if ((fh = f.hash) == MOVED) { |
1002 |
tab = helpTransfer(tab, f); |
1003 |
i = 0; // restart |
1004 |
} |
1005 |
else { |
1006 |
synchronized (f) { |
1007 |
if (tabAt(tab, i) == f) { |
1008 |
Node<K,V> p = (fh >= 0 ? f : |
1009 |
(f instanceof TreeBin) ? |
1010 |
((TreeBin<K,V>)f).first : null); |
1011 |
while (p != null) { |
1012 |
--delta; |
1013 |
p = p.next; |
1014 |
} |
1015 |
setTabAt(tab, i++, null); |
1016 |
} |
1017 |
} |
1018 |
} |
1019 |
} |
1020 |
if (delta != 0L) |
1021 |
addCount(delta, -1); |
1022 |
} |
1023 |
|
1024 |
/** |
1025 |
* Returns a {@link Set} view of the keys contained in this map. |
1026 |
* The set is backed by the map, so changes to the map are |
1027 |
* reflected in the set, and vice-versa. The set supports element |
1028 |
* removal, which removes the corresponding mapping from this map, |
1029 |
* via the {@code Iterator.remove}, {@code Set.remove}, |
1030 |
* {@code removeAll}, {@code retainAll}, and {@code clear} |
1031 |
* operations. It does not support the {@code add} or |
1032 |
* {@code addAll} operations. |
1033 |
* |
1034 |
* <p>The view's {@code iterator} is a "weakly consistent" iterator |
1035 |
* that will never throw {@link ConcurrentModificationException}, |
1036 |
* and guarantees to traverse elements as they existed upon |
1037 |
* construction of the iterator, and may (but is not guaranteed to) |
1038 |
* reflect any modifications subsequent to construction. |
1039 |
* |
1040 |
* @return the set view |
1041 |
*/ |
1042 |
public KeySetView<K,V> keySet() { |
1043 |
KeySetView<K,V> ks; |
1044 |
return (ks = keySet) != null ? ks : (keySet = new KeySetView<K,V>(this, null)); |
1045 |
} |
1046 |
|
1047 |
/** |
1048 |
* Returns a {@link Collection} view of the values contained in this map. |
1049 |
* The collection is backed by the map, so changes to the map are |
1050 |
* reflected in the collection, and vice-versa. The collection |
1051 |
* supports element removal, which removes the corresponding |
1052 |
* mapping from this map, via the {@code Iterator.remove}, |
1053 |
* {@code Collection.remove}, {@code removeAll}, |
1054 |
* {@code retainAll}, and {@code clear} operations. It does not |
1055 |
* support the {@code add} or {@code addAll} operations. |
1056 |
* |
1057 |
* <p>The view's {@code iterator} is a "weakly consistent" iterator |
1058 |
* that will never throw {@link ConcurrentModificationException}, |
1059 |
* and guarantees to traverse elements as they existed upon |
1060 |
* construction of the iterator, and may (but is not guaranteed to) |
1061 |
* reflect any modifications subsequent to construction. |
1062 |
* |
1063 |
* @return the collection view |
1064 |
*/ |
1065 |
public Collection<V> values() { |
1066 |
ValuesView<K,V> vs; |
1067 |
return (vs = values) != null ? vs : (values = new ValuesView<K,V>(this)); |
1068 |
} |
1069 |
|
1070 |
/** |
1071 |
* Returns a {@link Set} view of the mappings contained in this map. |
1072 |
* The set is backed by the map, so changes to the map are |
1073 |
* reflected in the set, and vice-versa. The set supports element |
1074 |
* removal, which removes the corresponding mapping from the map, |
1075 |
* via the {@code Iterator.remove}, {@code Set.remove}, |
1076 |
* {@code removeAll}, {@code retainAll}, and {@code clear} |
1077 |
* operations. |
1078 |
* |
1079 |
* <p>The view's {@code iterator} is a "weakly consistent" iterator |
1080 |
* that will never throw {@link ConcurrentModificationException}, |
1081 |
* and guarantees to traverse elements as they existed upon |
1082 |
* construction of the iterator, and may (but is not guaranteed to) |
1083 |
* reflect any modifications subsequent to construction. |
1084 |
* |
1085 |
* @return the set view |
1086 |
*/ |
1087 |
public Set<Map.Entry<K,V>> entrySet() { |
1088 |
EntrySetView<K,V> es; |
1089 |
return (es = entrySet) != null ? es : (entrySet = new EntrySetView<K,V>(this)); |
1090 |
} |
1091 |
|
1092 |
/** |
1093 |
* Returns the hash code value for this {@link Map}, i.e., |
1094 |
* the sum of, for each key-value pair in the map, |
1095 |
* {@code key.hashCode() ^ value.hashCode()}. |
1096 |
* |
1097 |
* @return the hash code value for this map |
1098 |
*/ |
1099 |
public int hashCode() { |
1100 |
int h = 0; |
1101 |
Node<K,V>[] t; |
1102 |
if ((t = table) != null) { |
1103 |
Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length); |
1104 |
for (Node<K,V> p; (p = it.advance()) != null; ) |
1105 |
h += p.key.hashCode() ^ p.val.hashCode(); |
1106 |
} |
1107 |
return h; |
1108 |
} |
1109 |
|
1110 |
/** |
1111 |
* Returns a string representation of this map. The string |
1112 |
* representation consists of a list of key-value mappings (in no |
1113 |
* particular order) enclosed in braces ("{@code {}}"). Adjacent |
1114 |
* mappings are separated by the characters {@code ", "} (comma |
1115 |
* and space). Each key-value mapping is rendered as the key |
1116 |
* followed by an equals sign ("{@code =}") followed by the |
1117 |
* associated value. |
1118 |
* |
1119 |
* @return a string representation of this map |
1120 |
*/ |
1121 |
public String toString() { |
1122 |
Node<K,V>[] t; |
1123 |
int f = (t = table) == null ? 0 : t.length; |
1124 |
Traverser<K,V> it = new Traverser<K,V>(t, f, 0, f); |
1125 |
StringBuilder sb = new StringBuilder(); |
1126 |
sb.append('{'); |
1127 |
Node<K,V> p; |
1128 |
if ((p = it.advance()) != null) { |
1129 |
for (;;) { |
1130 |
K k = p.key; |
1131 |
V v = p.val; |
1132 |
sb.append(k == this ? "(this Map)" : k); |
1133 |
sb.append('='); |
1134 |
sb.append(v == this ? "(this Map)" : v); |
1135 |
if ((p = it.advance()) == null) |
1136 |
break; |
1137 |
sb.append(',').append(' '); |
1138 |
} |
1139 |
} |
1140 |
return sb.append('}').toString(); |
1141 |
} |
1142 |
|
1143 |
/** |
1144 |
* Compares the specified object with this map for equality. |
1145 |
* Returns {@code true} if the given object is a map with the same |
1146 |
* mappings as this map. This operation may return misleading |
1147 |
* results if either map is concurrently modified during execution |
1148 |
* of this method. |
1149 |
* |
1150 |
* @param o object to be compared for equality with this map |
1151 |
* @return {@code true} if the specified object is equal to this map |
1152 |
*/ |
1153 |
public boolean equals(Object o) { |
1154 |
if (o != this) { |
1155 |
if (!(o instanceof Map)) |
1156 |
return false; |
1157 |
Map<?,?> m = (Map<?,?>) o; |
1158 |
Node<K,V>[] t; |
1159 |
int f = (t = table) == null ? 0 : t.length; |
1160 |
Traverser<K,V> it = new Traverser<K,V>(t, f, 0, f); |
1161 |
for (Node<K,V> p; (p = it.advance()) != null; ) { |
1162 |
V val = p.val; |
1163 |
Object v = m.get(p.key); |
1164 |
if (v == null || (v != val && !v.equals(val))) |
1165 |
return false; |
1166 |
} |
1167 |
for (Map.Entry<?,?> e : m.entrySet()) { |
1168 |
Object mk, mv, v; |
1169 |
if ((mk = e.getKey()) == null || |
1170 |
(mv = e.getValue()) == null || |
1171 |
(v = get(mk)) == null || |
1172 |
(mv != v && !mv.equals(v))) |
1173 |
return false; |
1174 |
} |
1175 |
} |
1176 |
return true; |
1177 |
} |
1178 |
|
1179 |
/** |
1180 |
* Stripped-down version of helper class used in previous version, |
1181 |
* declared for the sake of serialization compatibility |
1182 |
*/ |
1183 |
static class Segment<K,V> extends ReentrantLock implements Serializable { |
1184 |
private static final long serialVersionUID = 2249069246763182397L; |
1185 |
final float loadFactor; |
1186 |
Segment(float lf) { this.loadFactor = lf; } |
1187 |
} |
1188 |
|
1189 |
/** |
1190 |
* Saves the state of the {@code ConcurrentHashMap} instance to a |
1191 |
* stream (i.e., serializes it). |
1192 |
* @param s the stream |
1193 |
* @serialData |
1194 |
* the key (Object) and value (Object) |
1195 |
* for each key-value mapping, followed by a null pair. |
1196 |
* The key-value mappings are emitted in no particular order. |
1197 |
*/ |
1198 |
private void writeObject(java.io.ObjectOutputStream s) |
1199 |
throws java.io.IOException { |
1200 |
// For serialization compatibility |
1201 |
// Emulate segment calculation from previous version of this class |
1202 |
int sshift = 0; |
1203 |
int ssize = 1; |
1204 |
while (ssize < DEFAULT_CONCURRENCY_LEVEL) { |
1205 |
++sshift; |
1206 |
ssize <<= 1; |
1207 |
} |
1208 |
int segmentShift = 32 - sshift; |
1209 |
int segmentMask = ssize - 1; |
1210 |
@SuppressWarnings("unchecked") Segment<K,V>[] segments = (Segment<K,V>[]) |
1211 |
new Segment<?,?>[DEFAULT_CONCURRENCY_LEVEL]; |
1212 |
for (int i = 0; i < segments.length; ++i) |
1213 |
segments[i] = new Segment<K,V>(LOAD_FACTOR); |
1214 |
s.putFields().put("segments", segments); |
1215 |
s.putFields().put("segmentShift", segmentShift); |
1216 |
s.putFields().put("segmentMask", segmentMask); |
1217 |
s.writeFields(); |
1218 |
|
1219 |
Node<K,V>[] t; |
1220 |
if ((t = table) != null) { |
1221 |
Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length); |
1222 |
for (Node<K,V> p; (p = it.advance()) != null; ) { |
1223 |
s.writeObject(p.key); |
1224 |
s.writeObject(p.val); |
1225 |
} |
1226 |
} |
1227 |
s.writeObject(null); |
1228 |
s.writeObject(null); |
1229 |
segments = null; // throw away |
1230 |
} |
1231 |
|
1232 |
/** |
1233 |
* Reconstitutes the instance from a stream (that is, deserializes it). |
1234 |
* @param s the stream |
1235 |
*/ |
1236 |
private void readObject(java.io.ObjectInputStream s) |
1237 |
throws java.io.IOException, ClassNotFoundException { |
1238 |
/* |
1239 |
* To improve performance in typical cases, we create nodes |
1240 |
* while reading, then place in table once size is known. |
1241 |
* However, we must also validate uniqueness and deal with |
1242 |
* overpopulated bins while doing so, which requires |
1243 |
* specialized versions of putVal mechanics. |
1244 |
*/ |
1245 |
sizeCtl = -1; // force exclusion for table construction |
1246 |
s.defaultReadObject(); |
1247 |
long size = 0L; |
1248 |
Node<K,V> p = null; |
1249 |
for (;;) { |
1250 |
@SuppressWarnings("unchecked") K k = (K) s.readObject(); |
1251 |
@SuppressWarnings("unchecked") V v = (V) s.readObject(); |
1252 |
if (k != null && v != null) { |
1253 |
p = new Node<K,V>(spread(k.hashCode()), k, v, p); |
1254 |
++size; |
1255 |
} |
1256 |
else |
1257 |
break; |
1258 |
} |
1259 |
if (size == 0L) |
1260 |
sizeCtl = 0; |
1261 |
else { |
1262 |
int n; |
1263 |
if (size >= (long)(MAXIMUM_CAPACITY >>> 1)) |
1264 |
n = MAXIMUM_CAPACITY; |
1265 |
else { |
1266 |
int sz = (int)size; |
1267 |
n = tableSizeFor(sz + (sz >>> 1) + 1); |
1268 |
} |
1269 |
@SuppressWarnings({"rawtypes","unchecked"}) |
1270 |
Node<K,V>[] tab = (Node<K,V>[])new Node[n]; |
1271 |
int mask = n - 1; |
1272 |
long added = 0L; |
1273 |
while (p != null) { |
1274 |
boolean insertAtFront; |
1275 |
Node<K,V> next = p.next, first; |
1276 |
int h = p.hash, j = h & mask; |
1277 |
if ((first = tabAt(tab, j)) == null) |
1278 |
insertAtFront = true; |
1279 |
else { |
1280 |
K k = p.key; |
1281 |
if (first.hash < 0) { |
1282 |
TreeBin<K,V> t = (TreeBin<K,V>)first; |
1283 |
if (t.putTreeVal(h, k, p.val) == null) |
1284 |
++added; |
1285 |
insertAtFront = false; |
1286 |
} |
1287 |
else { |
1288 |
int binCount = 0; |
1289 |
insertAtFront = true; |
1290 |
Node<K,V> q; K qk; |
1291 |
for (q = first; q != null; q = q.next) { |
1292 |
if (q.hash == h && |
1293 |
((qk = q.key) == k || |
1294 |
(qk != null && k.equals(qk)))) { |
1295 |
insertAtFront = false; |
1296 |
break; |
1297 |
} |
1298 |
++binCount; |
1299 |
} |
1300 |
if (insertAtFront && binCount >= TREEIFY_THRESHOLD) { |
1301 |
insertAtFront = false; |
1302 |
++added; |
1303 |
p.next = first; |
1304 |
TreeNode<K,V> hd = null, tl = null; |
1305 |
for (q = p; q != null; q = q.next) { |
1306 |
TreeNode<K,V> t = new TreeNode<K,V> |
1307 |
(q.hash, q.key, q.val, null, null); |
1308 |
if ((t.prev = tl) == null) |
1309 |
hd = t; |
1310 |
else |
1311 |
tl.next = t; |
1312 |
tl = t; |
1313 |
} |
1314 |
setTabAt(tab, j, new TreeBin<K,V>(hd)); |
1315 |
} |
1316 |
} |
1317 |
} |
1318 |
if (insertAtFront) { |
1319 |
++added; |
1320 |
p.next = first; |
1321 |
setTabAt(tab, j, p); |
1322 |
} |
1323 |
p = next; |
1324 |
} |
1325 |
table = tab; |
1326 |
sizeCtl = n - (n >>> 2); |
1327 |
baseCount = added; |
1328 |
} |
1329 |
} |
1330 |
|
1331 |
// ConcurrentMap methods |
1332 |
|
1333 |
/** |
1334 |
* {@inheritDoc} |
1335 |
* |
1336 |
* @return the previous value associated with the specified key, |
1337 |
* or {@code null} if there was no mapping for the key |
1338 |
* @throws NullPointerException if the specified key or value is null |
1339 |
*/ |
1340 |
public V putIfAbsent(K key, V value) { |
1341 |
return putVal(key, value, true); |
1342 |
} |
1343 |
|
1344 |
/** |
1345 |
* {@inheritDoc} |
1346 |
* |
1347 |
* @throws NullPointerException if the specified key is null |
1348 |
*/ |
1349 |
public boolean remove(Object key, Object value) { |
1350 |
if (key == null) |
1351 |
throw new NullPointerException(); |
1352 |
return value != null && replaceNode(key, null, value) != null; |
1353 |
} |
1354 |
|
1355 |
/** |
1356 |
* {@inheritDoc} |
1357 |
* |
1358 |
* @throws NullPointerException if any of the arguments are null |
1359 |
*/ |
1360 |
public boolean replace(K key, V oldValue, V newValue) { |
1361 |
if (key == null || oldValue == null || newValue == null) |
1362 |
throw new NullPointerException(); |
1363 |
return replaceNode(key, newValue, oldValue) != null; |
1364 |
} |
1365 |
|
1366 |
/** |
1367 |
* {@inheritDoc} |
1368 |
* |
1369 |
* @return the previous value associated with the specified key, |
1370 |
* or {@code null} if there was no mapping for the key |
1371 |
* @throws NullPointerException if the specified key or value is null |
1372 |
*/ |
1373 |
public V replace(K key, V value) { |
1374 |
if (key == null || value == null) |
1375 |
throw new NullPointerException(); |
1376 |
return replaceNode(key, value, null); |
1377 |
} |
1378 |
// Hashtable legacy methods |
1379 |
|
1380 |
/** |
1381 |
* Legacy method testing if some key maps into the specified value |
1382 |
* in this table. This method is identical in functionality to |
1383 |
* {@link #containsValue(Object)}, and exists solely to ensure |
1384 |
* full compatibility with class {@link java.util.Hashtable}, |
1385 |
* which supported this method prior to introduction of the |
1386 |
* Java Collections framework. |
1387 |
* |
1388 |
* @param value a value to search for |
1389 |
* @return {@code true} if and only if some key maps to the |
1390 |
* {@code value} argument in this table as |
1391 |
* determined by the {@code equals} method; |
1392 |
* {@code false} otherwise |
1393 |
* @throws NullPointerException if the specified value is null |
1394 |
*/ |
1395 |
@Deprecated public boolean contains(Object value) { |
1396 |
return containsValue(value); |
1397 |
} |
1398 |
|
1399 |
/** |
1400 |
* Returns an enumeration of the keys in this table. |
1401 |
* |
1402 |
* @return an enumeration of the keys in this table |
1403 |
* @see #keySet() |
1404 |
*/ |
1405 |
public Enumeration<K> keys() { |
1406 |
Node<K,V>[] t; |
1407 |
int f = (t = table) == null ? 0 : t.length; |
1408 |
return new KeyIterator<K,V>(t, f, 0, f, this); |
1409 |
} |
1410 |
|
1411 |
/** |
1412 |
* Returns an enumeration of the values in this table. |
1413 |
* |
1414 |
* @return an enumeration of the values in this table |
1415 |
* @see #values() |
1416 |
*/ |
1417 |
public Enumeration<V> elements() { |
1418 |
Node<K,V>[] t; |
1419 |
int f = (t = table) == null ? 0 : t.length; |
1420 |
return new ValueIterator<K,V>(t, f, 0, f, this); |
1421 |
} |
1422 |
|
1423 |
// ConcurrentHashMap-only methods |
1424 |
|
1425 |
/** |
1426 |
* Returns the number of mappings. This method should be used |
1427 |
* instead of {@link #size} because a ConcurrentHashMap may |
1428 |
* contain more mappings than can be represented as an int. The |
1429 |
* value returned is an estimate; the actual count may differ if |
1430 |
* there are concurrent insertions or removals. |
1431 |
* |
1432 |
* @return the number of mappings |
1433 |
* @since 1.8 |
1434 |
*/ |
1435 |
public long mappingCount() { |
1436 |
long n = sumCount(); |
1437 |
return (n < 0L) ? 0L : n; // ignore transient negative values |
1438 |
} |
1439 |
|
1440 |
/** |
1441 |
* Creates a new {@link Set} backed by a ConcurrentHashMap |
1442 |
* from the given type to {@code Boolean.TRUE}. |
1443 |
* |
1444 |
* @return the new set |
1445 |
* @since 1.8 |
1446 |
*/ |
1447 |
public static <K> KeySetView<K,Boolean> newKeySet() { |
1448 |
return new KeySetView<K,Boolean> |
1449 |
(new ConcurrentHashMap<K,Boolean>(), Boolean.TRUE); |
1450 |
} |
1451 |
|
1452 |
/** |
1453 |
* Creates a new {@link Set} backed by a ConcurrentHashMap |
1454 |
* from the given type to {@code Boolean.TRUE}. |
1455 |
* |
1456 |
* @param initialCapacity The implementation performs internal |
1457 |
* sizing to accommodate this many elements. |
1458 |
* @throws IllegalArgumentException if the initial capacity of |
1459 |
* elements is negative |
1460 |
* @return the new set |
1461 |
* @since 1.8 |
1462 |
*/ |
1463 |
public static <K> KeySetView<K,Boolean> newKeySet(int initialCapacity) { |
1464 |
return new KeySetView<K,Boolean> |
1465 |
(new ConcurrentHashMap<K,Boolean>(initialCapacity), Boolean.TRUE); |
1466 |
} |
1467 |
|
1468 |
/** |
1469 |
* Returns a {@link Set} view of the keys in this map, using the |
1470 |
* given common mapped value for any additions (i.e., {@link |
1471 |
* Collection#add} and {@link Collection#addAll(Collection)}). |
1472 |
* This is of course only appropriate if it is acceptable to use |
1473 |
* the same value for all additions from this view. |
1474 |
* |
1475 |
* @param mappedValue the mapped value to use for any additions |
1476 |
* @return the set view |
1477 |
* @throws NullPointerException if the mappedValue is null |
1478 |
*/ |
1479 |
public KeySetView<K,V> keySet(V mappedValue) { |
1480 |
if (mappedValue == null) |
1481 |
throw new NullPointerException(); |
1482 |
return new KeySetView<K,V>(this, mappedValue); |
1483 |
} |
1484 |
|
1485 |
/* ---------------- Special Nodes -------------- */ |
1486 |
|
1487 |
/** |
1488 |
* A node inserted at head of bins during transfer operations. |
1489 |
*/ |
1490 |
static final class ForwardingNode<K,V> extends Node<K,V> { |
1491 |
final Node<K,V>[] nextTable; |
1492 |
ForwardingNode(Node<K,V>[] tab) { |
1493 |
super(MOVED, null, null, null); |
1494 |
this.nextTable = tab; |
1495 |
} |
1496 |
|
1497 |
Node<K,V> find(int h, Object k) { |
1498 |
Node<K,V> e; int n; |
1499 |
Node<K,V>[] tab = nextTable; |
1500 |
if (k != null && tab != null && (n = tab.length) > 0 && |
1501 |
(e = tabAt(tab, (n - 1) & h)) != null) { |
1502 |
do { |
1503 |
int eh; K ek; |
1504 |
if ((eh = e.hash) == h && |
1505 |
((ek = e.key) == k || (ek != null && k.equals(ek)))) |
1506 |
return e; |
1507 |
if (eh < 0) |
1508 |
return e.find(h, k); |
1509 |
} while ((e = e.next) != null); |
1510 |
} |
1511 |
return null; |
1512 |
} |
1513 |
} |
1514 |
|
1515 |
/** |
1516 |
* A place-holder node used in computeIfAbsent and compute |
1517 |
*/ |
1518 |
static final class ReservationNode<K,V> extends Node<K,V> { |
1519 |
ReservationNode() { |
1520 |
super(RESERVED, null, null, null); |
1521 |
} |
1522 |
|
1523 |
Node<K,V> find(int h, Object k) { |
1524 |
return null; |
1525 |
} |
1526 |
} |
1527 |
|
1528 |
/* ---------------- Table Initialization and Resizing -------------- */ |
1529 |
|
1530 |
/** |
1531 |
* Initializes table, using the size recorded in sizeCtl. |
1532 |
*/ |
1533 |
private final Node<K,V>[] initTable() { |
1534 |
Node<K,V>[] tab; int sc; |
1535 |
while ((tab = table) == null || tab.length == 0) { |
1536 |
if ((sc = sizeCtl) < 0) |
1537 |
Thread.yield(); // lost initialization race; just spin |
1538 |
else if (U.compareAndSwapInt(this, SIZECTL, sc, -1)) { |
1539 |
try { |
1540 |
if ((tab = table) == null || tab.length == 0) { |
1541 |
int n = (sc > 0) ? sc : DEFAULT_CAPACITY; |
1542 |
@SuppressWarnings({"rawtypes","unchecked"}) |
1543 |
Node<K,V>[] nt = (Node<K,V>[])new Node[n]; |
1544 |
table = tab = nt; |
1545 |
sc = n - (n >>> 2); |
1546 |
} |
1547 |
} finally { |
1548 |
sizeCtl = sc; |
1549 |
} |
1550 |
break; |
1551 |
} |
1552 |
} |
1553 |
return tab; |
1554 |
} |
1555 |
|
1556 |
/** |
1557 |
* Adds to count, and if table is too small and not already |
1558 |
* resizing, initiates transfer. If already resizing, helps |
1559 |
* perform transfer if work is available. Rechecks occupancy |
1560 |
* after a transfer to see if another resize is already needed |
1561 |
* because resizings are lagging additions. |
1562 |
* |
1563 |
* @param x the count to add |
1564 |
* @param check if <0, don't check resize, if <= 1 only check if uncontended |
1565 |
*/ |
1566 |
private final void addCount(long x, int check) { |
1567 |
CounterCell[] as; long b, s; |
1568 |
if ((as = counterCells) != null || |
1569 |
!U.compareAndSwapLong(this, BASECOUNT, b = baseCount, s = b + x)) { |
1570 |
CounterHashCode hc; CounterCell a; long v; int m; |
1571 |
boolean uncontended = true; |
1572 |
if ((hc = threadCounterHashCode.get()) == null || |
1573 |
as == null || (m = as.length - 1) < 0 || |
1574 |
(a = as[m & hc.code]) == null || |
1575 |
!(uncontended = |
1576 |
U.compareAndSwapLong(a, CELLVALUE, v = a.value, v + x))) { |
1577 |
fullAddCount(x, hc, uncontended); |
1578 |
return; |
1579 |
} |
1580 |
if (check <= 1) |
1581 |
return; |
1582 |
s = sumCount(); |
1583 |
} |
1584 |
if (check >= 0) { |
1585 |
Node<K,V>[] tab, nt; int sc; |
1586 |
while (s >= (long)(sc = sizeCtl) && (tab = table) != null && |
1587 |
tab.length < MAXIMUM_CAPACITY) { |
1588 |
if (sc < 0) { |
1589 |
if (sc == -1 || transferIndex <= transferOrigin || |
1590 |
(nt = nextTable) == null) |
1591 |
break; |
1592 |
if (U.compareAndSwapInt(this, SIZECTL, sc, sc - 1)) |
1593 |
transfer(tab, nt); |
1594 |
} |
1595 |
else if (U.compareAndSwapInt(this, SIZECTL, sc, -2)) |
1596 |
transfer(tab, null); |
1597 |
s = sumCount(); |
1598 |
} |
1599 |
} |
1600 |
} |
1601 |
|
1602 |
/** |
1603 |
* Helps transfer if a resize is in progress. |
1604 |
*/ |
1605 |
final Node<K,V>[] helpTransfer(Node<K,V>[] tab, Node<K,V> f) { |
1606 |
Node<K,V>[] nextTab; int sc; |
1607 |
if ((f instanceof ForwardingNode) && |
1608 |
(nextTab = ((ForwardingNode<K,V>)f).nextTable) != null) { |
1609 |
if (nextTab == nextTable && tab == table && |
1610 |
transferIndex > transferOrigin && (sc = sizeCtl) < -1 && |
1611 |
U.compareAndSwapInt(this, SIZECTL, sc, sc - 1)) |
1612 |
transfer(tab, nextTab); |
1613 |
return nextTab; |
1614 |
} |
1615 |
return table; |
1616 |
} |
1617 |
|
1618 |
/** |
1619 |
* Tries to presize table to accommodate the given number of elements. |
1620 |
* |
1621 |
* @param size number of elements (doesn't need to be perfectly accurate) |
1622 |
*/ |
1623 |
private final void tryPresize(int size) { |
1624 |
int c = (size >= (MAXIMUM_CAPACITY >>> 1)) ? MAXIMUM_CAPACITY : |
1625 |
tableSizeFor(size + (size >>> 1) + 1); |
1626 |
int sc; |
1627 |
while ((sc = sizeCtl) >= 0) { |
1628 |
Node<K,V>[] tab = table; int n; |
1629 |
if (tab == null || (n = tab.length) == 0) { |
1630 |
n = (sc > c) ? sc : c; |
1631 |
if (U.compareAndSwapInt(this, SIZECTL, sc, -1)) { |
1632 |
try { |
1633 |
if (table == tab) { |
1634 |
@SuppressWarnings({"rawtypes","unchecked"}) |
1635 |
Node<K,V>[] nt = (Node<K,V>[])new Node[n]; |
1636 |
table = nt; |
1637 |
sc = n - (n >>> 2); |
1638 |
} |
1639 |
} finally { |
1640 |
sizeCtl = sc; |
1641 |
} |
1642 |
} |
1643 |
} |
1644 |
else if (c <= sc || n >= MAXIMUM_CAPACITY) |
1645 |
break; |
1646 |
else if (tab == table && |
1647 |
U.compareAndSwapInt(this, SIZECTL, sc, -2)) |
1648 |
transfer(tab, null); |
1649 |
} |
1650 |
} |
1651 |
|
1652 |
/** |
1653 |
* Moves and/or copies the nodes in each bin to new table. See |
1654 |
* above for explanation. |
1655 |
*/ |
1656 |
private final void transfer(Node<K,V>[] tab, Node<K,V>[] nextTab) { |
1657 |
int n = tab.length, stride; |
1658 |
if ((stride = (NCPU > 1) ? (n >>> 3) / NCPU : n) < MIN_TRANSFER_STRIDE) |
1659 |
stride = MIN_TRANSFER_STRIDE; // subdivide range |
1660 |
if (nextTab == null) { // initiating |
1661 |
try { |
1662 |
@SuppressWarnings({"rawtypes","unchecked"}) |
1663 |
Node<K,V>[] nt = (Node<K,V>[])new Node[n << 1]; |
1664 |
nextTab = nt; |
1665 |
} catch (Throwable ex) { // try to cope with OOME |
1666 |
sizeCtl = Integer.MAX_VALUE; |
1667 |
return; |
1668 |
} |
1669 |
nextTable = nextTab; |
1670 |
transferOrigin = n; |
1671 |
transferIndex = n; |
1672 |
ForwardingNode<K,V> rev = new ForwardingNode<K,V>(tab); |
1673 |
for (int k = n; k > 0;) { // progressively reveal ready slots |
1674 |
int nextk = (k > stride) ? k - stride : 0; |
1675 |
for (int m = nextk; m < k; ++m) |
1676 |
nextTab[m] = rev; |
1677 |
for (int m = n + nextk; m < n + k; ++m) |
1678 |
nextTab[m] = rev; |
1679 |
U.putOrderedInt(this, TRANSFERORIGIN, k = nextk); |
1680 |
} |
1681 |
} |
1682 |
int nextn = nextTab.length; |
1683 |
ForwardingNode<K,V> fwd = new ForwardingNode<K,V>(nextTab); |
1684 |
boolean advance = true; |
1685 |
for (int i = 0, bound = 0;;) { |
1686 |
int nextIndex, nextBound, fh; Node<K,V> f; |
1687 |
while (advance) { |
1688 |
if (--i >= bound) |
1689 |
advance = false; |
1690 |
else if ((nextIndex = transferIndex) <= transferOrigin) { |
1691 |
i = -1; |
1692 |
advance = false; |
1693 |
} |
1694 |
else if (U.compareAndSwapInt |
1695 |
(this, TRANSFERINDEX, nextIndex, |
1696 |
nextBound = (nextIndex > stride ? |
1697 |
nextIndex - stride : 0))) { |
1698 |
bound = nextBound; |
1699 |
i = nextIndex - 1; |
1700 |
advance = false; |
1701 |
} |
1702 |
} |
1703 |
if (i < 0 || i >= n || i + n >= nextn) { |
1704 |
for (int sc;;) { |
1705 |
if (U.compareAndSwapInt(this, SIZECTL, sc = sizeCtl, ++sc)) { |
1706 |
if (sc == -1) { |
1707 |
nextTable = null; |
1708 |
table = nextTab; |
1709 |
sizeCtl = (n << 1) - (n >>> 1); |
1710 |
} |
1711 |
return; |
1712 |
} |
1713 |
} |
1714 |
} |
1715 |
else if ((f = tabAt(tab, i)) == null) { |
1716 |
if (casTabAt(tab, i, null, fwd)) { |
1717 |
setTabAt(nextTab, i, null); |
1718 |
setTabAt(nextTab, i + n, null); |
1719 |
advance = true; |
1720 |
} |
1721 |
} |
1722 |
else if ((fh = f.hash) == MOVED) |
1723 |
advance = true; // already processed |
1724 |
else { |
1725 |
synchronized (f) { |
1726 |
if (tabAt(tab, i) == f) { |
1727 |
Node<K,V> ln, hn; |
1728 |
if (fh >= 0) { |
1729 |
int runBit = fh & n; |
1730 |
Node<K,V> lastRun = f; |
1731 |
for (Node<K,V> p = f.next; p != null; p = p.next) { |
1732 |
int b = p.hash & n; |
1733 |
if (b != runBit) { |
1734 |
runBit = b; |
1735 |
lastRun = p; |
1736 |
} |
1737 |
} |
1738 |
if (runBit == 0) { |
1739 |
ln = lastRun; |
1740 |
hn = null; |
1741 |
} |
1742 |
else { |
1743 |
hn = lastRun; |
1744 |
ln = null; |
1745 |
} |
1746 |
for (Node<K,V> p = f; p != lastRun; p = p.next) { |
1747 |
int ph = p.hash; K pk = p.key; V pv = p.val; |
1748 |
if ((ph & n) == 0) |
1749 |
ln = new Node<K,V>(ph, pk, pv, ln); |
1750 |
else |
1751 |
hn = new Node<K,V>(ph, pk, pv, hn); |
1752 |
} |
1753 |
} |
1754 |
else if (f instanceof TreeBin) { |
1755 |
TreeBin<K,V> t = (TreeBin<K,V>)f; |
1756 |
TreeNode<K,V> lo = null, loTail = null; |
1757 |
TreeNode<K,V> hi = null, hiTail = null; |
1758 |
int lc = 0, hc = 0; |
1759 |
for (Node<K,V> e = t.first; e != null; e = e.next) { |
1760 |
int h = e.hash; |
1761 |
TreeNode<K,V> p = new TreeNode<K,V> |
1762 |
(h, e.key, e.val, null, null); |
1763 |
if ((h & n) == 0) { |
1764 |
if ((p.prev = loTail) == null) |
1765 |
lo = p; |
1766 |
else |
1767 |
loTail.next = p; |
1768 |
loTail = p; |
1769 |
++lc; |
1770 |
} |
1771 |
else { |
1772 |
if ((p.prev = hiTail) == null) |
1773 |
hi = p; |
1774 |
else |
1775 |
hiTail.next = p; |
1776 |
hiTail = p; |
1777 |
++hc; |
1778 |
} |
1779 |
} |
1780 |
ln = (lc <= UNTREEIFY_THRESHOLD ? untreeify(lo) : |
1781 |
(hc != 0) ? new TreeBin<K,V>(lo) : t); |
1782 |
hn = (hc <= UNTREEIFY_THRESHOLD ? untreeify(hi) : |
1783 |
(lc != 0) ? new TreeBin<K,V>(hi) : t); |
1784 |
} |
1785 |
else |
1786 |
ln = hn = null; |
1787 |
setTabAt(nextTab, i, ln); |
1788 |
setTabAt(nextTab, i + n, hn); |
1789 |
setTabAt(tab, i, fwd); |
1790 |
advance = true; |
1791 |
} |
1792 |
} |
1793 |
} |
1794 |
} |
1795 |
} |
1796 |
|
1797 |
/* ---------------- Conversion from/to TreeBins -------------- */ |
1798 |
|
1799 |
/** |
1800 |
* Replaces all linked nodes in bin at given index unless table is |
1801 |
* too small, in which case resizes instead. |
1802 |
*/ |
1803 |
private final void treeifyBin(Node<K,V>[] tab, int index) { |
1804 |
Node<K,V> b; int n, sc; |
1805 |
if (tab != null) { |
1806 |
if ((n = tab.length) < MIN_TREEIFY_CAPACITY) { |
1807 |
if (tab == table && (sc = sizeCtl) >= 0 && |
1808 |
U.compareAndSwapInt(this, SIZECTL, sc, -2)) |
1809 |
transfer(tab, null); |
1810 |
} |
1811 |
else if ((b = tabAt(tab, index)) != null) { |
1812 |
synchronized (b) { |
1813 |
if (tabAt(tab, index) == b) { |
1814 |
TreeNode<K,V> hd = null, tl = null; |
1815 |
for (Node<K,V> e = b; e != null; e = e.next) { |
1816 |
TreeNode<K,V> p = |
1817 |
new TreeNode<K,V>(e.hash, e.key, e.val, |
1818 |
null, null); |
1819 |
if ((p.prev = tl) == null) |
1820 |
hd = p; |
1821 |
else |
1822 |
tl.next = p; |
1823 |
tl = p; |
1824 |
} |
1825 |
setTabAt(tab, index, new TreeBin<K,V>(hd)); |
1826 |
} |
1827 |
} |
1828 |
} |
1829 |
} |
1830 |
} |
1831 |
|
1832 |
/** |
1833 |
* Returns a list on non-TreeNodes replacing those in given list |
1834 |
*/ |
1835 |
static <K,V> Node<K,V> untreeify(Node<K,V> b) { |
1836 |
Node<K,V> hd = null, tl = null; |
1837 |
for (Node<K,V> q = b; q != null; q = q.next) { |
1838 |
Node<K,V> p = new Node<K,V>(q.hash, q.key, q.val, null); |
1839 |
if (tl == null) |
1840 |
hd = p; |
1841 |
else |
1842 |
tl.next = p; |
1843 |
tl = p; |
1844 |
} |
1845 |
return hd; |
1846 |
} |
1847 |
|
1848 |
/* ---------------- TreeNodes -------------- */ |
1849 |
|
1850 |
/** |
1851 |
* Nodes for use in TreeBins |
1852 |
*/ |
1853 |
static final class TreeNode<K,V> extends Node<K,V> { |
1854 |
TreeNode<K,V> parent; // red-black tree links |
1855 |
TreeNode<K,V> left; |
1856 |
TreeNode<K,V> right; |
1857 |
TreeNode<K,V> prev; // needed to unlink next upon deletion |
1858 |
boolean red; |
1859 |
|
1860 |
TreeNode(int hash, K key, V val, Node<K,V> next, |
1861 |
TreeNode<K,V> parent) { |
1862 |
super(hash, key, val, next); |
1863 |
this.parent = parent; |
1864 |
} |
1865 |
|
1866 |
Node<K,V> find(int h, Object k) { |
1867 |
return findTreeNode(h, k, null); |
1868 |
} |
1869 |
|
1870 |
/** |
1871 |
* Returns the TreeNode (or null if not found) for the given key |
1872 |
* starting at given root. |
1873 |
*/ |
1874 |
final TreeNode<K,V> findTreeNode(int h, Object k, Class<?> kc) { |
1875 |
if (k != null) { |
1876 |
TreeNode<K,V> p = this; |
1877 |
do { |
1878 |
int ph, dir; K pk; TreeNode<K,V> q; |
1879 |
TreeNode<K,V> pl = p.left, pr = p.right; |
1880 |
if ((ph = p.hash) > h) |
1881 |
p = pl; |
1882 |
else if (ph < h) |
1883 |
p = pr; |
1884 |
else if ((pk = p.key) == k || (pk != null && k.equals(pk))) |
1885 |
return p; |
1886 |
else if (pl == null && pr == null) |
1887 |
break; |
1888 |
else if ((kc != null || |
1889 |
(kc = comparableClassFor(k)) != null) && |
1890 |
(dir = compareComparables(kc, k, pk)) != 0) |
1891 |
p = (dir < 0) ? pl : pr; |
1892 |
else if (pl == null) |
1893 |
p = pr; |
1894 |
else if (pr == null || |
1895 |
(q = pr.findTreeNode(h, k, kc)) == null) |
1896 |
p = pl; |
1897 |
else |
1898 |
return q; |
1899 |
} while (p != null); |
1900 |
} |
1901 |
return null; |
1902 |
} |
1903 |
} |
1904 |
|
1905 |
/* ---------------- TreeBins -------------- */ |
1906 |
|
1907 |
/** |
1908 |
* TreeNodes used at the heads of bins. TreeBins do not hold user |
1909 |
* keys or values, but instead point to list of TreeNodes and |
1910 |
* their root. They also maintain a parasitic read-write lock |
1911 |
* forcing writers (who hold bin lock) to wait for readers (who do |
1912 |
* not) to complete before tree restructuring operations. |
1913 |
*/ |
1914 |
static final class TreeBin<K,V> extends Node<K,V> { |
1915 |
TreeNode<K,V> root; |
1916 |
volatile TreeNode<K,V> first; |
1917 |
volatile Thread waiter; |
1918 |
volatile int lockState; |
1919 |
// values for lockState |
1920 |
static final int WRITER = 1; // set while holding write lock |
1921 |
static final int WAITER = 2; // set when waiting for write lock |
1922 |
static final int READER = 4; // increment value for setting read lock |
1923 |
|
1924 |
/** |
1925 |
* Creates bin with initial set of nodes headed by b. |
1926 |
*/ |
1927 |
TreeBin(TreeNode<K,V> b) { |
1928 |
super(TREEBIN, null, null, null); |
1929 |
this.first = b; |
1930 |
TreeNode<K,V> r = null; |
1931 |
for (TreeNode<K,V> x = b, next; x != null; x = next) { |
1932 |
next = (TreeNode<K,V>)x.next; |
1933 |
x.left = x.right = null; |
1934 |
if (r == null) { |
1935 |
x.parent = null; |
1936 |
x.red = false; |
1937 |
r = x; |
1938 |
} |
1939 |
else { |
1940 |
Object key = x.key; |
1941 |
int hash = x.hash; |
1942 |
Class<?> kc = null; |
1943 |
for (TreeNode<K,V> p = r;;) { |
1944 |
int dir, ph; |
1945 |
if ((ph = p.hash) > hash) |
1946 |
dir = -1; |
1947 |
else if (ph < hash) |
1948 |
dir = 1; |
1949 |
else if ((kc != null || |
1950 |
(kc = comparableClassFor(key)) != null)) |
1951 |
dir = compareComparables(kc, key, p.key); |
1952 |
else |
1953 |
dir = 0; |
1954 |
TreeNode<K,V> xp = p; |
1955 |
if ((p = (dir <= 0) ? p.left : p.right) == null) { |
1956 |
x.parent = xp; |
1957 |
if (dir <= 0) |
1958 |
xp.left = x; |
1959 |
else |
1960 |
xp.right = x; |
1961 |
r = balanceInsertion(r, x); |
1962 |
break; |
1963 |
} |
1964 |
} |
1965 |
} |
1966 |
} |
1967 |
this.root = r; |
1968 |
} |
1969 |
|
1970 |
/** |
1971 |
* Acquires write lock for tree restructuring |
1972 |
*/ |
1973 |
private final void lockRoot() { |
1974 |
if (!U.compareAndSwapInt(this, LOCKSTATE, 0, WRITER)) |
1975 |
contendedLock(); // offload to separate method |
1976 |
} |
1977 |
|
1978 |
/** |
1979 |
* Releases write lock for tree restructuring |
1980 |
*/ |
1981 |
private final void unlockRoot() { |
1982 |
lockState = 0; |
1983 |
} |
1984 |
|
1985 |
/** |
1986 |
* Possibly blocks awaiting root lock |
1987 |
*/ |
1988 |
private final void contendedLock() { |
1989 |
boolean waiting = false; |
1990 |
for (int s;;) { |
1991 |
if (((s = lockState) & WRITER) == 0) { |
1992 |
if (U.compareAndSwapInt(this, LOCKSTATE, s, WRITER)) { |
1993 |
if (waiting) |
1994 |
waiter = null; |
1995 |
return; |
1996 |
} |
1997 |
} |
1998 |
else if ((s | WAITER) == 0) { |
1999 |
if (U.compareAndSwapInt(this, LOCKSTATE, s, s | WAITER)) { |
2000 |
waiting = true; |
2001 |
waiter = Thread.currentThread(); |
2002 |
} |
2003 |
} |
2004 |
else if (waiting) |
2005 |
LockSupport.park(this); |
2006 |
} |
2007 |
} |
2008 |
|
2009 |
/** |
2010 |
* Returns matching node or null if none. Tries to search |
2011 |
* using tree compareisons from root, but continues linear |
2012 |
* search when lock not available. |
2013 |
*/ |
2014 |
final Node<K,V> find(int h, Object k) { |
2015 |
if (k != null) { |
2016 |
for (Node<K,V> e = first; e != null; e = e.next) { |
2017 |
int s; K ek; |
2018 |
if (((s = lockState) & (WAITER|WRITER)) != 0) { |
2019 |
if (e.hash == h && |
2020 |
((ek = e.key) == k || (ek != null && k.equals(ek)))) |
2021 |
return e; |
2022 |
} |
2023 |
else if (U.compareAndSwapInt(this, LOCKSTATE, s, |
2024 |
s + READER)) { |
2025 |
TreeNode<K,V> r, p; |
2026 |
try { |
2027 |
p = ((r = root) == null ? null : |
2028 |
r.findTreeNode(h, k, null)); |
2029 |
} finally { |
2030 |
|
2031 |
Thread w; |
2032 |
int ls; |
2033 |
do {} while (!U.compareAndSwapInt |
2034 |
(this, LOCKSTATE, |
2035 |
ls = lockState, ls - READER)); |
2036 |
if (ls == (READER|WAITER) && (w = waiter) != null) |
2037 |
LockSupport.unpark(w); |
2038 |
} |
2039 |
return p; |
2040 |
} |
2041 |
} |
2042 |
} |
2043 |
return null; |
2044 |
} |
2045 |
|
2046 |
/** |
2047 |
* Finds or adds a node. |
2048 |
* @return null if added |
2049 |
*/ |
2050 |
final TreeNode<K,V> putTreeVal(int h, K k, V v) { |
2051 |
Class<?> kc = null; |
2052 |
for (TreeNode<K,V> p = root;;) { |
2053 |
int dir, ph; K pk; TreeNode<K,V> q, pr; |
2054 |
if (p == null) { |
2055 |
first = root = new TreeNode<K,V>(h, k, v, null, null); |
2056 |
break; |
2057 |
} |
2058 |
else if ((ph = p.hash) > h) |
2059 |
dir = -1; |
2060 |
else if (ph < h) |
2061 |
dir = 1; |
2062 |
else if ((pk = p.key) == k || (pk != null && k.equals(pk))) |
2063 |
return p; |
2064 |
else if ((kc == null && |
2065 |
(kc = comparableClassFor(k)) == null) || |
2066 |
(dir = compareComparables(kc, k, pk)) == 0) { |
2067 |
if (p.left == null) |
2068 |
dir = 1; |
2069 |
else if ((pr = p.right) == null || |
2070 |
(q = pr.findTreeNode(h, k, kc)) == null) |
2071 |
dir = -1; |
2072 |
else |
2073 |
return q; |
2074 |
} |
2075 |
TreeNode<K,V> xp = p; |
2076 |
if ((p = (dir < 0) ? p.left : p.right) == null) { |
2077 |
TreeNode<K,V> x, f = first; |
2078 |
first = x = new TreeNode<K,V>(h, k, v, f, xp); |
2079 |
if (f != null) |
2080 |
f.prev = x; |
2081 |
if (dir < 0) |
2082 |
xp.left = x; |
2083 |
else |
2084 |
xp.right = x; |
2085 |
if (!xp.red) |
2086 |
x.red = true; |
2087 |
else { |
2088 |
lockRoot(); |
2089 |
try { |
2090 |
root = balanceInsertion(root, x); |
2091 |
} finally { |
2092 |
unlockRoot(); |
2093 |
} |
2094 |
} |
2095 |
break; |
2096 |
} |
2097 |
} |
2098 |
assert checkInvariants(root); |
2099 |
return null; |
2100 |
} |
2101 |
|
2102 |
/** |
2103 |
* Removes the given node, that must be present before this |
2104 |
* call. This is messier than typical red-black deletion code |
2105 |
* because we cannot swap the contents of an interior node |
2106 |
* with a leaf successor that is pinned by "next" pointers |
2107 |
* that are accessible independently of lock. So instead we |
2108 |
* swap the tree linkages. |
2109 |
* |
2110 |
* @return true if now too small so should be untreeified. |
2111 |
*/ |
2112 |
final boolean removeTreeNode(TreeNode<K,V> p) { |
2113 |
TreeNode<K,V> next = (TreeNode<K,V>)p.next; |
2114 |
TreeNode<K,V> pred = p.prev; // unlink traversal pointers |
2115 |
TreeNode<K,V> r, rl; |
2116 |
if (pred == null) |
2117 |
first = next; |
2118 |
else |
2119 |
pred.next = next; |
2120 |
if (next != null) |
2121 |
next.prev = pred; |
2122 |
if (first == null) { |
2123 |
root = null; |
2124 |
return true; |
2125 |
} |
2126 |
if ((r = root) == null || r.right == null || // too small |
2127 |
(rl = r.left) == null || rl.left == null) |
2128 |
return true; |
2129 |
lockRoot(); |
2130 |
try { |
2131 |
TreeNode<K,V> replacement; |
2132 |
TreeNode<K,V> pl = p.left; |
2133 |
TreeNode<K,V> pr = p.right; |
2134 |
if (pl != null && pr != null) { |
2135 |
TreeNode<K,V> s = pr, sl; |
2136 |
while ((sl = s.left) != null) // find successor |
2137 |
s = sl; |
2138 |
boolean c = s.red; s.red = p.red; p.red = c; // swap colors |
2139 |
TreeNode<K,V> sr = s.right; |
2140 |
TreeNode<K,V> pp = p.parent; |
2141 |
if (s == pr) { // p was s's direct parent |
2142 |
p.parent = s; |
2143 |
s.right = p; |
2144 |
} |
2145 |
else { |
2146 |
TreeNode<K,V> sp = s.parent; |
2147 |
if ((p.parent = sp) != null) { |
2148 |
if (s == sp.left) |
2149 |
sp.left = p; |
2150 |
else |
2151 |
sp.right = p; |
2152 |
} |
2153 |
if ((s.right = pr) != null) |
2154 |
pr.parent = s; |
2155 |
} |
2156 |
p.left = null; |
2157 |
if ((p.right = sr) != null) |
2158 |
sr.parent = p; |
2159 |
if ((s.left = pl) != null) |
2160 |
pl.parent = s; |
2161 |
if ((s.parent = pp) == null) |
2162 |
r = s; |
2163 |
else if (p == pp.left) |
2164 |
pp.left = s; |
2165 |
else |
2166 |
pp.right = s; |
2167 |
if (sr != null) |
2168 |
replacement = sr; |
2169 |
else |
2170 |
replacement = p; |
2171 |
} |
2172 |
else if (pl != null) |
2173 |
replacement = pl; |
2174 |
else if (pr != null) |
2175 |
replacement = pr; |
2176 |
else |
2177 |
replacement = p; |
2178 |
if (replacement != p) { |
2179 |
TreeNode<K,V> pp = replacement.parent = p.parent; |
2180 |
if (pp == null) |
2181 |
r = replacement; |
2182 |
else if (p == pp.left) |
2183 |
pp.left = replacement; |
2184 |
else |
2185 |
pp.right = replacement; |
2186 |
p.left = p.right = p.parent = null; |
2187 |
} |
2188 |
|
2189 |
root = (p.red) ? r : balanceDeletion(r, replacement); |
2190 |
|
2191 |
if (p == replacement) { // detach pointers |
2192 |
TreeNode<K,V> pp; |
2193 |
if ((pp = p.parent) != null) { |
2194 |
if (p == pp.left) |
2195 |
pp.left = null; |
2196 |
else if (p == pp.right) |
2197 |
pp.right = null; |
2198 |
p.parent = null; |
2199 |
} |
2200 |
} |
2201 |
} finally { |
2202 |
unlockRoot(); |
2203 |
} |
2204 |
assert checkInvariants(root); |
2205 |
return false; |
2206 |
} |
2207 |
|
2208 |
/* ------------------------------------------------------------ */ |
2209 |
// Red-black tree methods, all adapted from CLR |
2210 |
|
2211 |
static <K,V> TreeNode<K,V> rotateLeft(TreeNode<K,V> root, |
2212 |
TreeNode<K,V> p) { |
2213 |
TreeNode<K,V> r, pp, rl; |
2214 |
if (p != null && (r = p.right) != null) { |
2215 |
if ((rl = p.right = r.left) != null) |
2216 |
rl.parent = p; |
2217 |
if ((pp = r.parent = p.parent) == null) |
2218 |
(root = r).red = false; |
2219 |
else if (pp.left == p) |
2220 |
pp.left = r; |
2221 |
else |
2222 |
pp.right = r; |
2223 |
r.left = p; |
2224 |
p.parent = r; |
2225 |
} |
2226 |
return root; |
2227 |
} |
2228 |
|
2229 |
static <K,V> TreeNode<K,V> rotateRight(TreeNode<K,V> root, |
2230 |
TreeNode<K,V> p) { |
2231 |
TreeNode<K,V> l, pp, lr; |
2232 |
if (p != null && (l = p.left) != null) { |
2233 |
if ((lr = p.left = l.right) != null) |
2234 |
lr.parent = p; |
2235 |
if ((pp = l.parent = p.parent) == null) |
2236 |
(root = l).red = false; |
2237 |
else if (pp.right == p) |
2238 |
pp.right = l; |
2239 |
else |
2240 |
pp.left = l; |
2241 |
l.right = p; |
2242 |
p.parent = l; |
2243 |
} |
2244 |
return root; |
2245 |
} |
2246 |
|
2247 |
static <K,V> TreeNode<K,V> balanceInsertion(TreeNode<K,V> root, |
2248 |
TreeNode<K,V> x) { |
2249 |
x.red = true; |
2250 |
for (TreeNode<K,V> xp, xpp, xppl, xppr;;) { |
2251 |
if ((xp = x.parent) == null) { |
2252 |
x.red = false; |
2253 |
return x; |
2254 |
} |
2255 |
else if (!xp.red || (xpp = xp.parent) == null) |
2256 |
return root; |
2257 |
if (xp == (xppl = xpp.left)) { |
2258 |
if ((xppr = xpp.right) != null && xppr.red) { |
2259 |
xppr.red = false; |
2260 |
xp.red = false; |
2261 |
xpp.red = true; |
2262 |
x = xpp; |
2263 |
} |
2264 |
else { |
2265 |
if (x == xp.right) { |
2266 |
root = rotateLeft(root, x = xp); |
2267 |
xpp = (xp = x.parent) == null ? null : xp.parent; |
2268 |
} |
2269 |
if (xp != null) { |
2270 |
xp.red = false; |
2271 |
if (xpp != null) { |
2272 |
xpp.red = true; |
2273 |
root = rotateRight(root, xpp); |
2274 |
} |
2275 |
} |
2276 |
} |
2277 |
} |
2278 |
else { |
2279 |
if (xppl != null && xppl.red) { |
2280 |
xppl.red = false; |
2281 |
xp.red = false; |
2282 |
xpp.red = true; |
2283 |
x = xpp; |
2284 |
} |
2285 |
else { |
2286 |
if (x == xp.left) { |
2287 |
root = rotateRight(root, x = xp); |
2288 |
xpp = (xp = x.parent) == null ? null : xp.parent; |
2289 |
} |
2290 |
if (xp != null) { |
2291 |
xp.red = false; |
2292 |
if (xpp != null) { |
2293 |
xpp.red = true; |
2294 |
root = rotateLeft(root, xpp); |
2295 |
} |
2296 |
} |
2297 |
} |
2298 |
} |
2299 |
} |
2300 |
} |
2301 |
|
2302 |
static <K,V> TreeNode<K,V> balanceDeletion(TreeNode<K,V> root, |
2303 |
TreeNode<K,V> x) { |
2304 |
for (TreeNode<K,V> xp, xpl, xpr;;) { |
2305 |
if (x == null || x == root) |
2306 |
return root; |
2307 |
else if ((xp = x.parent) == null) { |
2308 |
x.red = false; |
2309 |
return x; |
2310 |
} |
2311 |
else if (x.red) { |
2312 |
x.red = false; |
2313 |
return root; |
2314 |
} |
2315 |
else if ((xpl = xp.left) == x) { |
2316 |
if ((xpr = xp.right) != null && xpr.red) { |
2317 |
xpr.red = false; |
2318 |
xp.red = true; |
2319 |
root = rotateLeft(root, xp); |
2320 |
xpr = (xp = x.parent) == null ? null : xp.right; |
2321 |
} |
2322 |
if (xpr == null) |
2323 |
x = xp; |
2324 |
else { |
2325 |
TreeNode<K,V> sl = xpr.left, sr = xpr.right; |
2326 |
if ((sr == null || !sr.red) && |
2327 |
(sl == null || !sl.red)) { |
2328 |
xpr.red = true; |
2329 |
x = xp; |
2330 |
} |
2331 |
else { |
2332 |
if (sr == null || !sr.red) { |
2333 |
if (sl != null) |
2334 |
sl.red = false; |
2335 |
xpr.red = true; |
2336 |
root = rotateRight(root, xpr); |
2337 |
xpr = (xp = x.parent) == null ? |
2338 |
null : xp.right; |
2339 |
} |
2340 |
if (xpr != null) { |
2341 |
xpr.red = (xp == null) ? false : xp.red; |
2342 |
if ((sr = xpr.right) != null) |
2343 |
sr.red = false; |
2344 |
} |
2345 |
if (xp != null) { |
2346 |
xp.red = false; |
2347 |
root = rotateLeft(root, xp); |
2348 |
} |
2349 |
x = root; |
2350 |
} |
2351 |
} |
2352 |
} |
2353 |
else { // symmetric |
2354 |
if (xpl != null && xpl.red) { |
2355 |
xpl.red = false; |
2356 |
xp.red = true; |
2357 |
root = rotateRight(root, xp); |
2358 |
xpl = (xp = x.parent) == null ? null : xp.left; |
2359 |
} |
2360 |
if (xpl == null) |
2361 |
x = xp; |
2362 |
else { |
2363 |
TreeNode<K,V> sl = xpl.left, sr = xpl.right; |
2364 |
if ((sl == null || !sl.red) && |
2365 |
(sr == null || !sr.red)) { |
2366 |
xpl.red = true; |
2367 |
x = xp; |
2368 |
} |
2369 |
else { |
2370 |
if (sl == null || !sl.red) { |
2371 |
if (sr != null) |
2372 |
sr.red = false; |
2373 |
xpl.red = true; |
2374 |
root = rotateLeft(root, xpl); |
2375 |
xpl = (xp = x.parent) == null ? |
2376 |
null : xp.left; |
2377 |
} |
2378 |
if (xpl != null) { |
2379 |
xpl.red = (xp == null) ? false : xp.red; |
2380 |
if ((sl = xpl.left) != null) |
2381 |
sl.red = false; |
2382 |
} |
2383 |
if (xp != null) { |
2384 |
xp.red = false; |
2385 |
root = rotateRight(root, xp); |
2386 |
} |
2387 |
x = root; |
2388 |
} |
2389 |
} |
2390 |
} |
2391 |
} |
2392 |
} |
2393 |
|
2394 |
/** |
2395 |
* Recursive invariant check |
2396 |
*/ |
2397 |
static <K,V> boolean checkInvariants(TreeNode<K,V> t) { |
2398 |
TreeNode<K,V> tp = t.parent, tl = t.left, tr = t.right, |
2399 |
tb = t.prev, tn = (TreeNode<K,V>)t.next; |
2400 |
if (tb != null && tb.next != t) |
2401 |
return false; |
2402 |
if (tn != null && tn.prev != t) |
2403 |
return false; |
2404 |
if (tp != null && t != tp.left && t != tp.right) |
2405 |
return false; |
2406 |
if (tl != null && (tl.parent != t || tl.hash > t.hash)) |
2407 |
return false; |
2408 |
if (tr != null && (tr.parent != t || tr.hash < t.hash)) |
2409 |
return false; |
2410 |
if (t.red && tl != null && tl.red && tr != null && tr.red) |
2411 |
return false; |
2412 |
if (tl != null && !checkInvariants(tl)) |
2413 |
return false; |
2414 |
if (tr != null && !checkInvariants(tr)) |
2415 |
return false; |
2416 |
return true; |
2417 |
} |
2418 |
|
2419 |
private static final sun.misc.Unsafe U; |
2420 |
private static final long LOCKSTATE; |
2421 |
static { |
2422 |
try { |
2423 |
U = sun.misc.Unsafe.getUnsafe(); |
2424 |
Class<?> k = TreeBin.class; |
2425 |
LOCKSTATE = U.objectFieldOffset |
2426 |
(k.getDeclaredField("lockState")); |
2427 |
} catch (Exception e) { |
2428 |
throw new Error(e); |
2429 |
} |
2430 |
} |
2431 |
} |
2432 |
|
2433 |
/* ----------------Table Traversal -------------- */ |
2434 |
|
2435 |
/** |
2436 |
* Encapsulates traversal for methods such as containsValue; also |
2437 |
* serves as a base class for other iterators. |
2438 |
* |
2439 |
* Method advance visits once each still-valid node that was |
2440 |
* reachable upon iterator construction. It might miss some that |
2441 |
* were added to a bin after the bin was visited, which is OK wrt |
2442 |
* consistency guarantees. Maintaining this property in the face |
2443 |
* of possible ongoing resizes requires a fair amount of |
2444 |
* bookkeeping state that is difficult to optimize away amidst |
2445 |
* volatile accesses. Even so, traversal maintains reasonable |
2446 |
* throughput. |
2447 |
* |
2448 |
* Normally, iteration proceeds bin-by-bin traversing lists. |
2449 |
* However, if the table has been resized, then all future steps |
2450 |
* must traverse both the bin at the current index as well as at |
2451 |
* (index + baseSize); and so on for further resizings. To |
2452 |
* paranoically cope with potential sharing by users of iterators |
2453 |
* across threads, iteration terminates if a bounds checks fails |
2454 |
* for a table read. |
2455 |
*/ |
2456 |
static class Traverser<K,V> { |
2457 |
Node<K,V>[] tab; // current table; updated if resized |
2458 |
Node<K,V> next; // the next entry to use |
2459 |
int index; // index of bin to use next |
2460 |
int baseIndex; // current index of initial table |
2461 |
int baseLimit; // index bound for initial table |
2462 |
final int baseSize; // initial table size |
2463 |
|
2464 |
Traverser(Node<K,V>[] tab, int size, int index, int limit) { |
2465 |
this.tab = tab; |
2466 |
this.baseSize = size; |
2467 |
this.baseIndex = this.index = index; |
2468 |
this.baseLimit = limit; |
2469 |
this.next = null; |
2470 |
} |
2471 |
|
2472 |
/** |
2473 |
* Advances if possible, returning next valid node, or null if none. |
2474 |
*/ |
2475 |
final Node<K,V> advance() { |
2476 |
Node<K,V> e; |
2477 |
if ((e = next) != null) |
2478 |
e = e.next; |
2479 |
for (;;) { |
2480 |
Node<K,V>[] t; int i, n; K ek; // must use locals in checks |
2481 |
if (e != null) |
2482 |
return next = e; |
2483 |
if (baseIndex >= baseLimit || (t = tab) == null || |
2484 |
(n = t.length) <= (i = index) || i < 0) |
2485 |
return next = null; |
2486 |
if ((e = tabAt(t, index)) != null && e.hash < 0) { |
2487 |
if (e instanceof ForwardingNode) { |
2488 |
tab = ((ForwardingNode<K,V>)e).nextTable; |
2489 |
e = null; |
2490 |
continue; |
2491 |
} |
2492 |
else if (e instanceof TreeBin) |
2493 |
e = ((TreeBin<K,V>)e).first; |
2494 |
else |
2495 |
e = null; |
2496 |
} |
2497 |
if ((index += baseSize) >= n) |
2498 |
index = ++baseIndex; // visit upper slots if present |
2499 |
} |
2500 |
} |
2501 |
} |
2502 |
|
2503 |
/** |
2504 |
* Base of key, value, and entry Iterators. Adds fields to |
2505 |
* Traverser to support iterator.remove |
2506 |
*/ |
2507 |
static class BaseIterator<K,V> extends Traverser<K,V> { |
2508 |
final ConcurrentHashMap<K,V> map; |
2509 |
Node<K,V> lastReturned; |
2510 |
BaseIterator(Node<K,V>[] tab, int size, int index, int limit, |
2511 |
ConcurrentHashMap<K,V> map) { |
2512 |
super(tab, size, index, limit); |
2513 |
this.map = map; |
2514 |
advance(); |
2515 |
} |
2516 |
|
2517 |
public final boolean hasNext() { return next != null; } |
2518 |
public final boolean hasMoreElements() { return next != null; } |
2519 |
|
2520 |
public final void remove() { |
2521 |
Node<K,V> p; |
2522 |
if ((p = lastReturned) == null) |
2523 |
throw new IllegalStateException(); |
2524 |
lastReturned = null; |
2525 |
map.replaceNode(p.key, null, null); |
2526 |
} |
2527 |
} |
2528 |
|
2529 |
static final class KeyIterator<K,V> extends BaseIterator<K,V> |
2530 |
implements Iterator<K>, Enumeration<K> { |
2531 |
KeyIterator(Node<K,V>[] tab, int index, int size, int limit, |
2532 |
ConcurrentHashMap<K,V> map) { |
2533 |
super(tab, index, size, limit, map); |
2534 |
} |
2535 |
|
2536 |
public final K next() { |
2537 |
Node<K,V> p; |
2538 |
if ((p = next) == null) |
2539 |
throw new NoSuchElementException(); |
2540 |
K k = p.key; |
2541 |
lastReturned = p; |
2542 |
advance(); |
2543 |
return k; |
2544 |
} |
2545 |
|
2546 |
public final K nextElement() { return next(); } |
2547 |
} |
2548 |
|
2549 |
static final class ValueIterator<K,V> extends BaseIterator<K,V> |
2550 |
implements Iterator<V>, Enumeration<V> { |
2551 |
ValueIterator(Node<K,V>[] tab, int index, int size, int limit, |
2552 |
ConcurrentHashMap<K,V> map) { |
2553 |
super(tab, index, size, limit, map); |
2554 |
} |
2555 |
|
2556 |
public final V next() { |
2557 |
Node<K,V> p; |
2558 |
if ((p = next) == null) |
2559 |
throw new NoSuchElementException(); |
2560 |
V v = p.val; |
2561 |
lastReturned = p; |
2562 |
advance(); |
2563 |
return v; |
2564 |
} |
2565 |
|
2566 |
public final V nextElement() { return next(); } |
2567 |
} |
2568 |
|
2569 |
static final class EntryIterator<K,V> extends BaseIterator<K,V> |
2570 |
implements Iterator<Map.Entry<K,V>> { |
2571 |
EntryIterator(Node<K,V>[] tab, int index, int size, int limit, |
2572 |
ConcurrentHashMap<K,V> map) { |
2573 |
super(tab, index, size, limit, map); |
2574 |
} |
2575 |
|
2576 |
public final Map.Entry<K,V> next() { |
2577 |
Node<K,V> p; |
2578 |
if ((p = next) == null) |
2579 |
throw new NoSuchElementException(); |
2580 |
K k = p.key; |
2581 |
V v = p.val; |
2582 |
lastReturned = p; |
2583 |
advance(); |
2584 |
return new MapEntry<K,V>(k, v, map); |
2585 |
} |
2586 |
} |
2587 |
|
2588 |
/** |
2589 |
* Exported Entry for EntryIterator |
2590 |
*/ |
2591 |
static final class MapEntry<K,V> implements Map.Entry<K,V> { |
2592 |
final K key; // non-null |
2593 |
V val; // non-null |
2594 |
final ConcurrentHashMap<K,V> map; |
2595 |
MapEntry(K key, V val, ConcurrentHashMap<K,V> map) { |
2596 |
this.key = key; |
2597 |
this.val = val; |
2598 |
this.map = map; |
2599 |
} |
2600 |
public K getKey() { return key; } |
2601 |
public V getValue() { return val; } |
2602 |
public int hashCode() { return key.hashCode() ^ val.hashCode(); } |
2603 |
public String toString() { return key + "=" + val; } |
2604 |
|
2605 |
public boolean equals(Object o) { |
2606 |
Object k, v; Map.Entry<?,?> e; |
2607 |
return ((o instanceof Map.Entry) && |
2608 |
(k = (e = (Map.Entry<?,?>)o).getKey()) != null && |
2609 |
(v = e.getValue()) != null && |
2610 |
(k == key || k.equals(key)) && |
2611 |
(v == val || v.equals(val))); |
2612 |
} |
2613 |
|
2614 |
/** |
2615 |
* Sets our entry's value and writes through to the map. The |
2616 |
* value to return is somewhat arbitrary here. Since we do not |
2617 |
* necessarily track asynchronous changes, the most recent |
2618 |
* "previous" value could be different from what we return (or |
2619 |
* could even have been removed, in which case the put will |
2620 |
* re-establish). We do not and cannot guarantee more. |
2621 |
*/ |
2622 |
public V setValue(V value) { |
2623 |
if (value == null) throw new NullPointerException(); |
2624 |
V v = val; |
2625 |
val = value; |
2626 |
map.put(key, value); |
2627 |
return v; |
2628 |
} |
2629 |
} |
2630 |
|
2631 |
/* ----------------Views -------------- */ |
2632 |
|
2633 |
/** |
2634 |
* Base class for views. |
2635 |
*/ |
2636 |
abstract static class CollectionView<K,V,E> |
2637 |
implements Collection<E>, java.io.Serializable { |
2638 |
private static final long serialVersionUID = 7249069246763182397L; |
2639 |
final ConcurrentHashMap<K,V> map; |
2640 |
CollectionView(ConcurrentHashMap<K,V> map) { this.map = map; } |
2641 |
|
2642 |
/** |
2643 |
* Returns the map backing this view. |
2644 |
* |
2645 |
* @return the map backing this view |
2646 |
*/ |
2647 |
public ConcurrentHashMap<K,V> getMap() { return map; } |
2648 |
|
2649 |
/** |
2650 |
* Removes all of the elements from this view, by removing all |
2651 |
* the mappings from the map backing this view. |
2652 |
*/ |
2653 |
public final void clear() { map.clear(); } |
2654 |
public final int size() { return map.size(); } |
2655 |
public final boolean isEmpty() { return map.isEmpty(); } |
2656 |
|
2657 |
// implementations below rely on concrete classes supplying these |
2658 |
// abstract methods |
2659 |
/** |
2660 |
* Returns a "weakly consistent" iterator that will never |
2661 |
* throw {@link ConcurrentModificationException}, and |
2662 |
* guarantees to traverse elements as they existed upon |
2663 |
* construction of the iterator, and may (but is not |
2664 |
* guaranteed to) reflect any modifications subsequent to |
2665 |
* construction. |
2666 |
*/ |
2667 |
public abstract Iterator<E> iterator(); |
2668 |
public abstract boolean contains(Object o); |
2669 |
public abstract boolean remove(Object o); |
2670 |
|
2671 |
private static final String oomeMsg = "Required array size too large"; |
2672 |
|
2673 |
public final Object[] toArray() { |
2674 |
long sz = map.mappingCount(); |
2675 |
if (sz > MAX_ARRAY_SIZE) |
2676 |
throw new OutOfMemoryError(oomeMsg); |
2677 |
int n = (int)sz; |
2678 |
Object[] r = new Object[n]; |
2679 |
int i = 0; |
2680 |
for (E e : this) { |
2681 |
if (i == n) { |
2682 |
if (n >= MAX_ARRAY_SIZE) |
2683 |
throw new OutOfMemoryError(oomeMsg); |
2684 |
if (n >= MAX_ARRAY_SIZE - (MAX_ARRAY_SIZE >>> 1) - 1) |
2685 |
n = MAX_ARRAY_SIZE; |
2686 |
else |
2687 |
n += (n >>> 1) + 1; |
2688 |
r = Arrays.copyOf(r, n); |
2689 |
} |
2690 |
r[i++] = e; |
2691 |
} |
2692 |
return (i == n) ? r : Arrays.copyOf(r, i); |
2693 |
} |
2694 |
|
2695 |
@SuppressWarnings("unchecked") |
2696 |
public final <T> T[] toArray(T[] a) { |
2697 |
long sz = map.mappingCount(); |
2698 |
if (sz > MAX_ARRAY_SIZE) |
2699 |
throw new OutOfMemoryError(oomeMsg); |
2700 |
int m = (int)sz; |
2701 |
T[] r = (a.length >= m) ? a : |
2702 |
(T[])java.lang.reflect.Array |
2703 |
.newInstance(a.getClass().getComponentType(), m); |
2704 |
int n = r.length; |
2705 |
int i = 0; |
2706 |
for (E e : this) { |
2707 |
if (i == n) { |
2708 |
if (n >= MAX_ARRAY_SIZE) |
2709 |
throw new OutOfMemoryError(oomeMsg); |
2710 |
if (n >= MAX_ARRAY_SIZE - (MAX_ARRAY_SIZE >>> 1) - 1) |
2711 |
n = MAX_ARRAY_SIZE; |
2712 |
else |
2713 |
n += (n >>> 1) + 1; |
2714 |
r = Arrays.copyOf(r, n); |
2715 |
} |
2716 |
r[i++] = (T)e; |
2717 |
} |
2718 |
if (a == r && i < n) { |
2719 |
r[i] = null; // null-terminate |
2720 |
return r; |
2721 |
} |
2722 |
return (i == n) ? r : Arrays.copyOf(r, i); |
2723 |
} |
2724 |
|
2725 |
/** |
2726 |
* Returns a string representation of this collection. |
2727 |
* The string representation consists of the string representations |
2728 |
* of the collection's elements in the order they are returned by |
2729 |
* its iterator, enclosed in square brackets ({@code "[]"}). |
2730 |
* Adjacent elements are separated by the characters {@code ", "} |
2731 |
* (comma and space). Elements are converted to strings as by |
2732 |
* {@link String#valueOf(Object)}. |
2733 |
* |
2734 |
* @return a string representation of this collection |
2735 |
*/ |
2736 |
public final String toString() { |
2737 |
StringBuilder sb = new StringBuilder(); |
2738 |
sb.append('['); |
2739 |
Iterator<E> it = iterator(); |
2740 |
if (it.hasNext()) { |
2741 |
for (;;) { |
2742 |
Object e = it.next(); |
2743 |
sb.append(e == this ? "(this Collection)" : e); |
2744 |
if (!it.hasNext()) |
2745 |
break; |
2746 |
sb.append(',').append(' '); |
2747 |
} |
2748 |
} |
2749 |
return sb.append(']').toString(); |
2750 |
} |
2751 |
|
2752 |
public final boolean containsAll(Collection<?> c) { |
2753 |
if (c != this) { |
2754 |
for (Object e : c) { |
2755 |
if (e == null || !contains(e)) |
2756 |
return false; |
2757 |
} |
2758 |
} |
2759 |
return true; |
2760 |
} |
2761 |
|
2762 |
public final boolean removeAll(Collection<?> c) { |
2763 |
boolean modified = false; |
2764 |
for (Iterator<E> it = iterator(); it.hasNext();) { |
2765 |
if (c.contains(it.next())) { |
2766 |
it.remove(); |
2767 |
modified = true; |
2768 |
} |
2769 |
} |
2770 |
return modified; |
2771 |
} |
2772 |
|
2773 |
public final boolean retainAll(Collection<?> c) { |
2774 |
boolean modified = false; |
2775 |
for (Iterator<E> it = iterator(); it.hasNext();) { |
2776 |
if (!c.contains(it.next())) { |
2777 |
it.remove(); |
2778 |
modified = true; |
2779 |
} |
2780 |
} |
2781 |
return modified; |
2782 |
} |
2783 |
|
2784 |
} |
2785 |
|
2786 |
/** |
2787 |
* A view of a ConcurrentHashMap as a {@link Set} of keys, in |
2788 |
* which additions may optionally be enabled by mapping to a |
2789 |
* common value. This class cannot be directly instantiated. |
2790 |
* See {@link #keySet() keySet()}, |
2791 |
* {@link #keySet(Object) keySet(V)}, |
2792 |
* {@link #newKeySet() newKeySet()}, |
2793 |
* {@link #newKeySet(int) newKeySet(int)}. |
2794 |
* |
2795 |
* @since 1.8 |
2796 |
*/ |
2797 |
public static class KeySetView<K,V> extends CollectionView<K,V,K> |
2798 |
implements Set<K>, java.io.Serializable { |
2799 |
private static final long serialVersionUID = 7249069246763182397L; |
2800 |
private final V value; |
2801 |
KeySetView(ConcurrentHashMap<K,V> map, V value) { // non-public |
2802 |
super(map); |
2803 |
this.value = value; |
2804 |
} |
2805 |
|
2806 |
/** |
2807 |
* Returns the default mapped value for additions, |
2808 |
* or {@code null} if additions are not supported. |
2809 |
* |
2810 |
* @return the default mapped value for additions, or {@code null} |
2811 |
* if not supported |
2812 |
*/ |
2813 |
public V getMappedValue() { return value; } |
2814 |
|
2815 |
/** |
2816 |
* {@inheritDoc} |
2817 |
* @throws NullPointerException if the specified key is null |
2818 |
*/ |
2819 |
public boolean contains(Object o) { return map.containsKey(o); } |
2820 |
|
2821 |
/** |
2822 |
* Removes the key from this map view, by removing the key (and its |
2823 |
* corresponding value) from the backing map. This method does |
2824 |
* nothing if the key is not in the map. |
2825 |
* |
2826 |
* @param o the key to be removed from the backing map |
2827 |
* @return {@code true} if the backing map contained the specified key |
2828 |
* @throws NullPointerException if the specified key is null |
2829 |
*/ |
2830 |
public boolean remove(Object o) { return map.remove(o) != null; } |
2831 |
|
2832 |
/** |
2833 |
* @return an iterator over the keys of the backing map |
2834 |
*/ |
2835 |
public Iterator<K> iterator() { |
2836 |
Node<K,V>[] t; |
2837 |
ConcurrentHashMap<K,V> m = map; |
2838 |
int f = (t = m.table) == null ? 0 : t.length; |
2839 |
return new KeyIterator<K,V>(t, f, 0, f, m); |
2840 |
} |
2841 |
|
2842 |
/** |
2843 |
* Adds the specified key to this set view by mapping the key to |
2844 |
* the default mapped value in the backing map, if defined. |
2845 |
* |
2846 |
* @param e key to be added |
2847 |
* @return {@code true} if this set changed as a result of the call |
2848 |
* @throws NullPointerException if the specified key is null |
2849 |
* @throws UnsupportedOperationException if no default mapped value |
2850 |
* for additions was provided |
2851 |
*/ |
2852 |
public boolean add(K e) { |
2853 |
V v; |
2854 |
if ((v = value) == null) |
2855 |
throw new UnsupportedOperationException(); |
2856 |
return map.putVal(e, v, true) == null; |
2857 |
} |
2858 |
|
2859 |
/** |
2860 |
* Adds all of the elements in the specified collection to this set, |
2861 |
* as if by calling {@link #add} on each one. |
2862 |
* |
2863 |
* @param c the elements to be inserted into this set |
2864 |
* @return {@code true} if this set changed as a result of the call |
2865 |
* @throws NullPointerException if the collection or any of its |
2866 |
* elements are {@code null} |
2867 |
* @throws UnsupportedOperationException if no default mapped value |
2868 |
* for additions was provided |
2869 |
*/ |
2870 |
public boolean addAll(Collection<? extends K> c) { |
2871 |
boolean added = false; |
2872 |
V v; |
2873 |
if ((v = value) == null) |
2874 |
throw new UnsupportedOperationException(); |
2875 |
for (K e : c) { |
2876 |
if (map.putVal(e, v, true) == null) |
2877 |
added = true; |
2878 |
} |
2879 |
return added; |
2880 |
} |
2881 |
|
2882 |
public int hashCode() { |
2883 |
int h = 0; |
2884 |
for (K e : this) |
2885 |
h += e.hashCode(); |
2886 |
return h; |
2887 |
} |
2888 |
|
2889 |
public boolean equals(Object o) { |
2890 |
Set<?> c; |
2891 |
return ((o instanceof Set) && |
2892 |
((c = (Set<?>)o) == this || |
2893 |
(containsAll(c) && c.containsAll(this)))); |
2894 |
} |
2895 |
|
2896 |
} |
2897 |
|
2898 |
/** |
2899 |
* A view of a ConcurrentHashMap as a {@link Collection} of |
2900 |
* values, in which additions are disabled. This class cannot be |
2901 |
* directly instantiated. See {@link #values()}. |
2902 |
*/ |
2903 |
static final class ValuesView<K,V> extends CollectionView<K,V,V> |
2904 |
implements Collection<V>, java.io.Serializable { |
2905 |
private static final long serialVersionUID = 2249069246763182397L; |
2906 |
ValuesView(ConcurrentHashMap<K,V> map) { super(map); } |
2907 |
public final boolean contains(Object o) { |
2908 |
return map.containsValue(o); |
2909 |
} |
2910 |
|
2911 |
public final boolean remove(Object o) { |
2912 |
if (o != null) { |
2913 |
for (Iterator<V> it = iterator(); it.hasNext();) { |
2914 |
if (o.equals(it.next())) { |
2915 |
it.remove(); |
2916 |
return true; |
2917 |
} |
2918 |
} |
2919 |
} |
2920 |
return false; |
2921 |
} |
2922 |
|
2923 |
public final Iterator<V> iterator() { |
2924 |
ConcurrentHashMap<K,V> m = map; |
2925 |
Node<K,V>[] t; |
2926 |
int f = (t = m.table) == null ? 0 : t.length; |
2927 |
return new ValueIterator<K,V>(t, f, 0, f, m); |
2928 |
} |
2929 |
|
2930 |
public final boolean add(V e) { |
2931 |
throw new UnsupportedOperationException(); |
2932 |
} |
2933 |
public final boolean addAll(Collection<? extends V> c) { |
2934 |
throw new UnsupportedOperationException(); |
2935 |
} |
2936 |
|
2937 |
} |
2938 |
|
2939 |
/** |
2940 |
* A view of a ConcurrentHashMap as a {@link Set} of (key, value) |
2941 |
* entries. This class cannot be directly instantiated. See |
2942 |
* {@link #entrySet()}. |
2943 |
*/ |
2944 |
static final class EntrySetView<K,V> extends CollectionView<K,V,Map.Entry<K,V>> |
2945 |
implements Set<Map.Entry<K,V>>, java.io.Serializable { |
2946 |
private static final long serialVersionUID = 2249069246763182397L; |
2947 |
EntrySetView(ConcurrentHashMap<K,V> map) { super(map); } |
2948 |
|
2949 |
public boolean contains(Object o) { |
2950 |
Object k, v, r; Map.Entry<?,?> e; |
2951 |
return ((o instanceof Map.Entry) && |
2952 |
(k = (e = (Map.Entry<?,?>)o).getKey()) != null && |
2953 |
(r = map.get(k)) != null && |
2954 |
(v = e.getValue()) != null && |
2955 |
(v == r || v.equals(r))); |
2956 |
} |
2957 |
|
2958 |
public boolean remove(Object o) { |
2959 |
Object k, v; Map.Entry<?,?> e; |
2960 |
return ((o instanceof Map.Entry) && |
2961 |
(k = (e = (Map.Entry<?,?>)o).getKey()) != null && |
2962 |
(v = e.getValue()) != null && |
2963 |
map.remove(k, v)); |
2964 |
} |
2965 |
|
2966 |
/** |
2967 |
* @return an iterator over the entries of the backing map |
2968 |
*/ |
2969 |
public Iterator<Map.Entry<K,V>> iterator() { |
2970 |
ConcurrentHashMap<K,V> m = map; |
2971 |
Node<K,V>[] t; |
2972 |
int f = (t = m.table) == null ? 0 : t.length; |
2973 |
return new EntryIterator<K,V>(t, f, 0, f, m); |
2974 |
} |
2975 |
|
2976 |
public boolean add(Entry<K,V> e) { |
2977 |
return map.putVal(e.getKey(), e.getValue(), false) == null; |
2978 |
} |
2979 |
|
2980 |
public boolean addAll(Collection<? extends Entry<K,V>> c) { |
2981 |
boolean added = false; |
2982 |
for (Entry<K,V> e : c) { |
2983 |
if (add(e)) |
2984 |
added = true; |
2985 |
} |
2986 |
return added; |
2987 |
} |
2988 |
|
2989 |
public final int hashCode() { |
2990 |
int h = 0; |
2991 |
Node<K,V>[] t; |
2992 |
if ((t = map.table) != null) { |
2993 |
Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length); |
2994 |
for (Node<K,V> p; (p = it.advance()) != null; ) { |
2995 |
h += p.hashCode(); |
2996 |
} |
2997 |
} |
2998 |
return h; |
2999 |
} |
3000 |
|
3001 |
public final boolean equals(Object o) { |
3002 |
Set<?> c; |
3003 |
return ((o instanceof Set) && |
3004 |
((c = (Set<?>)o) == this || |
3005 |
(containsAll(c) && c.containsAll(this)))); |
3006 |
} |
3007 |
|
3008 |
} |
3009 |
|
3010 |
|
3011 |
/* ---------------- Counters -------------- */ |
3012 |
|
3013 |
// Adapted from LongAdder and Striped64. |
3014 |
// See their internal docs for explanation. |
3015 |
|
3016 |
// A padded cell for distributing counts |
3017 |
static final class CounterCell { |
3018 |
volatile long p0, p1, p2, p3, p4, p5, p6; |
3019 |
volatile long value; |
3020 |
volatile long q0, q1, q2, q3, q4, q5, q6; |
3021 |
CounterCell(long x) { value = x; } |
3022 |
} |
3023 |
|
3024 |
/** |
3025 |
* Holder for the thread-local hash code determining which |
3026 |
* CounterCell to use. The code is initialized via the |
3027 |
* counterHashCodeGenerator, but may be moved upon collisions. |
3028 |
*/ |
3029 |
static final class CounterHashCode { |
3030 |
int code; |
3031 |
} |
3032 |
|
3033 |
/** |
3034 |
* Generates initial value for per-thread CounterHashCodes |
3035 |
*/ |
3036 |
static final AtomicInteger counterHashCodeGenerator = new AtomicInteger(); |
3037 |
|
3038 |
/** |
3039 |
* Increment for counterHashCodeGenerator. See class ThreadLocal |
3040 |
* for explanation. |
3041 |
*/ |
3042 |
static final int SEED_INCREMENT = 0x61c88647; |
3043 |
|
3044 |
/** |
3045 |
* Per-thread counter hash codes. Shared across all instances. |
3046 |
*/ |
3047 |
static final ThreadLocal<CounterHashCode> threadCounterHashCode = |
3048 |
new ThreadLocal<CounterHashCode>(); |
3049 |
|
3050 |
|
3051 |
final long sumCount() { |
3052 |
CounterCell[] as = counterCells; CounterCell a; |
3053 |
long sum = baseCount; |
3054 |
if (as != null) { |
3055 |
for (int i = 0; i < as.length; ++i) { |
3056 |
if ((a = as[i]) != null) |
3057 |
sum += a.value; |
3058 |
} |
3059 |
} |
3060 |
return sum; |
3061 |
} |
3062 |
|
3063 |
// See LongAdder version for explanation |
3064 |
private final void fullAddCount(long x, CounterHashCode hc, |
3065 |
boolean wasUncontended) { |
3066 |
int h; |
3067 |
if (hc == null) { |
3068 |
hc = new CounterHashCode(); |
3069 |
int s = counterHashCodeGenerator.addAndGet(SEED_INCREMENT); |
3070 |
h = hc.code = (s == 0) ? 1 : s; // Avoid zero |
3071 |
threadCounterHashCode.set(hc); |
3072 |
} |
3073 |
else |
3074 |
h = hc.code; |
3075 |
boolean collide = false; // True if last slot nonempty |
3076 |
for (;;) { |
3077 |
CounterCell[] as; CounterCell a; int n; long v; |
3078 |
if ((as = counterCells) != null && (n = as.length) > 0) { |
3079 |
if ((a = as[(n - 1) & h]) == null) { |
3080 |
if (cellsBusy == 0) { // Try to attach new Cell |
3081 |
CounterCell r = new CounterCell(x); // Optimistic create |
3082 |
if (cellsBusy == 0 && |
3083 |
U.compareAndSwapInt(this, CELLSBUSY, 0, 1)) { |
3084 |
boolean created = false; |
3085 |
try { // Recheck under lock |
3086 |
CounterCell[] rs; int m, j; |
3087 |
if ((rs = counterCells) != null && |
3088 |
(m = rs.length) > 0 && |
3089 |
rs[j = (m - 1) & h] == null) { |
3090 |
rs[j] = r; |
3091 |
created = true; |
3092 |
} |
3093 |
} finally { |
3094 |
cellsBusy = 0; |
3095 |
} |
3096 |
if (created) |
3097 |
break; |
3098 |
continue; // Slot is now non-empty |
3099 |
} |
3100 |
} |
3101 |
collide = false; |
3102 |
} |
3103 |
else if (!wasUncontended) // CAS already known to fail |
3104 |
wasUncontended = true; // Continue after rehash |
3105 |
else if (U.compareAndSwapLong(a, CELLVALUE, v = a.value, v + x)) |
3106 |
break; |
3107 |
else if (counterCells != as || n >= NCPU) |
3108 |
collide = false; // At max size or stale |
3109 |
else if (!collide) |
3110 |
collide = true; |
3111 |
else if (cellsBusy == 0 && |
3112 |
U.compareAndSwapInt(this, CELLSBUSY, 0, 1)) { |
3113 |
try { |
3114 |
if (counterCells == as) {// Expand table unless stale |
3115 |
CounterCell[] rs = new CounterCell[n << 1]; |
3116 |
for (int i = 0; i < n; ++i) |
3117 |
rs[i] = as[i]; |
3118 |
counterCells = rs; |
3119 |
} |
3120 |
} finally { |
3121 |
cellsBusy = 0; |
3122 |
} |
3123 |
collide = false; |
3124 |
continue; // Retry with expanded table |
3125 |
} |
3126 |
h ^= h << 13; // Rehash |
3127 |
h ^= h >>> 17; |
3128 |
h ^= h << 5; |
3129 |
} |
3130 |
else if (cellsBusy == 0 && counterCells == as && |
3131 |
U.compareAndSwapInt(this, CELLSBUSY, 0, 1)) { |
3132 |
boolean init = false; |
3133 |
try { // Initialize table |
3134 |
if (counterCells == as) { |
3135 |
CounterCell[] rs = new CounterCell[2]; |
3136 |
rs[h & 1] = new CounterCell(x); |
3137 |
counterCells = rs; |
3138 |
init = true; |
3139 |
} |
3140 |
} finally { |
3141 |
cellsBusy = 0; |
3142 |
} |
3143 |
if (init) |
3144 |
break; |
3145 |
} |
3146 |
else if (U.compareAndSwapLong(this, BASECOUNT, v = baseCount, v + x)) |
3147 |
break; // Fall back on using base |
3148 |
} |
3149 |
hc.code = h; // Record index for next time |
3150 |
} |
3151 |
|
3152 |
// Unsafe mechanics |
3153 |
private static final sun.misc.Unsafe U; |
3154 |
private static final long SIZECTL; |
3155 |
private static final long TRANSFERINDEX; |
3156 |
private static final long TRANSFERORIGIN; |
3157 |
private static final long BASECOUNT; |
3158 |
private static final long CELLSBUSY; |
3159 |
private static final long CELLVALUE; |
3160 |
private static final long ABASE; |
3161 |
private static final int ASHIFT; |
3162 |
|
3163 |
static { |
3164 |
try { |
3165 |
U = sun.misc.Unsafe.getUnsafe(); |
3166 |
Class<?> k = ConcurrentHashMap.class; |
3167 |
SIZECTL = U.objectFieldOffset |
3168 |
(k.getDeclaredField("sizeCtl")); |
3169 |
TRANSFERINDEX = U.objectFieldOffset |
3170 |
(k.getDeclaredField("transferIndex")); |
3171 |
TRANSFERORIGIN = U.objectFieldOffset |
3172 |
(k.getDeclaredField("transferOrigin")); |
3173 |
BASECOUNT = U.objectFieldOffset |
3174 |
(k.getDeclaredField("baseCount")); |
3175 |
CELLSBUSY = U.objectFieldOffset |
3176 |
(k.getDeclaredField("cellsBusy")); |
3177 |
Class<?> ck = CounterCell.class; |
3178 |
CELLVALUE = U.objectFieldOffset |
3179 |
(ck.getDeclaredField("value")); |
3180 |
Class<?> sc = Node[].class; |
3181 |
ABASE = U.arrayBaseOffset(sc); |
3182 |
int scale = U.arrayIndexScale(sc); |
3183 |
if ((scale & (scale - 1)) != 0) |
3184 |
throw new Error("data type scale not a power of two"); |
3185 |
ASHIFT = 31 - Integer.numberOfLeadingZeros(scale); |
3186 |
} catch (Exception e) { |
3187 |
throw new Error(e); |
3188 |
} |
3189 |
} |
3190 |
|
3191 |
} |