1 /*
2 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
3 *
4 * This code is free software; you can redistribute it and/or modify it
5 * under the terms of the GNU General Public License version 2 only, as
6 * published by the Free Software Foundation. Oracle designates this
7 * particular file as subject to the "Classpath" exception as provided
8 * by Oracle in the LICENSE file that accompanied this code.
9 *
10 * This code is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
13 * version 2 for more details (a copy is included in the LICENSE file that
14 * accompanied this code).
15 *
16 * You should have received a copy of the GNU General Public License version
17 * 2 along with this work; if not, write to the Free Software Foundation,
18 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
19 *
20 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
21 * or visit www.oracle.com if you need additional information or have any
22 * questions.
23 */
24
25 /*
26 * This file is available under and governed by the GNU General Public
27 * License version 2 only, as published by the Free Software Foundation.
28 * However, the following notice accompanied the original version of this
29 * file:
30 *
31 * Written by Doug Lea with assistance from members of JCP JSR-166
32 * Expert Group and released to the public domain, as explained at
33 * http://creativecommons.org/publicdomain/zero/1.0/
34 */
35
36 package java.util.concurrent;
37
38 import java.io.ObjectStreamField;
39 import java.io.Serializable;
40 import java.lang.reflect.ParameterizedType;
41 import java.lang.reflect.Type;
42 import java.util.AbstractMap;
43 import java.util.Arrays;
44 import java.util.Collection;
45 import java.util.Comparator;
46 import java.util.Enumeration;
47 import java.util.HashMap;
48 import java.util.Hashtable;
49 import java.util.Iterator;
50 import java.util.Map;
51 import java.util.NoSuchElementException;
52 import java.util.Set;
53 import java.util.Spliterator;
54 import java.util.concurrent.ConcurrentMap;
55 import java.util.concurrent.ForkJoinPool;
56 import java.util.concurrent.atomic.AtomicReference;
57 import java.util.concurrent.locks.LockSupport;
58 import java.util.concurrent.locks.ReentrantLock;
59 import java.util.function.BiConsumer;
60 import java.util.function.BiFunction;
61 import java.util.function.BinaryOperator;
62 import java.util.function.Consumer;
63 import java.util.function.DoubleBinaryOperator;
64 import java.util.function.Function;
65 import java.util.function.IntBinaryOperator;
66 import java.util.function.LongBinaryOperator;
67 import java.util.function.Predicate;
68 import java.util.function.ToDoubleBiFunction;
69 import java.util.function.ToDoubleFunction;
70 import java.util.function.ToIntBiFunction;
71 import java.util.function.ToIntFunction;
72 import java.util.function.ToLongBiFunction;
73 import java.util.function.ToLongFunction;
74 import java.util.stream.Stream;
75
76 /**
77 * A hash table supporting full concurrency of retrievals and
78 * high expected concurrency for updates. This class obeys the
79 * same functional specification as {@link java.util.Hashtable}, and
80 * includes versions of methods corresponding to each method of
81 * {@code Hashtable}. However, even though all operations are
82 * thread-safe, retrieval operations do <em>not</em> entail locking,
83 * and there is <em>not</em> any support for locking the entire table
84 * in a way that prevents all access. This class is fully
85 * interoperable with {@code Hashtable} in programs that rely on its
86 * thread safety but not on its synchronization details.
87 *
88 * <p>Retrieval operations (including {@code get}) generally do not
89 * block, so may overlap with update operations (including {@code put}
90 * and {@code remove}). Retrievals reflect the results of the most
91 * recently <em>completed</em> update operations holding upon their
92 * onset. (More formally, an update operation for a given key bears a
93 * <em>happens-before</em> relation with any (non-null) retrieval for
94 * that key reporting the updated value.) For aggregate operations
95 * such as {@code putAll} and {@code clear}, concurrent retrievals may
96 * reflect insertion or removal of only some entries. Similarly,
97 * Iterators, Spliterators and Enumerations return elements reflecting the
98 * state of the hash table at some point at or since the creation of the
99 * iterator/enumeration. They do <em>not</em> throw {@link
100 * java.util.ConcurrentModificationException ConcurrentModificationException}.
101 * However, iterators are designed to be used by only one thread at a time.
102 * Bear in mind that the results of aggregate status methods including
103 * {@code size}, {@code isEmpty}, and {@code containsValue} are typically
104 * useful only when a map is not undergoing concurrent updates in other threads.
105 * Otherwise the results of these methods reflect transient states
106 * that may be adequate for monitoring or estimation purposes, but not
107 * for program control.
108 *
109 * <p>The table is dynamically expanded when there are too many
110 * collisions (i.e., keys that have distinct hash codes but fall into
111 * the same slot modulo the table size), with the expected average
112 * effect of maintaining roughly two bins per mapping (corresponding
113 * to a 0.75 load factor threshold for resizing). There may be much
114 * variance around this average as mappings are added and removed, but
115 * overall, this maintains a commonly accepted time/space tradeoff for
116 * hash tables. However, resizing this or any other kind of hash
117 * table may be a relatively slow operation. When possible, it is a
118 * good idea to provide a size estimate as an optional {@code
119 * initialCapacity} constructor argument. An additional optional
120 * {@code loadFactor} constructor argument provides a further means of
121 * customizing initial table capacity by specifying the table density
122 * to be used in calculating the amount of space to allocate for the
123 * given number of elements. Also, for compatibility with previous
124 * versions of this class, constructors may optionally specify an
125 * expected {@code concurrencyLevel} as an additional hint for
126 * internal sizing. Note that using many keys with exactly the same
127 * {@code hashCode()} is a sure way to slow down performance of any
128 * hash table. To ameliorate impact, when keys are {@link Comparable},
129 * this class may use comparison order among keys to help break ties.
130 *
131 * <p>A {@link Set} projection of a ConcurrentHashMap may be created
132 * (using {@link #newKeySet()} or {@link #newKeySet(int)}), or viewed
133 * (using {@link #keySet(Object)} when only keys are of interest, and the
134 * mapped values are (perhaps transiently) not used or all take the
135 * same mapping value.
136 *
137 * <p>A ConcurrentHashMap can be used as a scalable frequency map (a
138 * form of histogram or multiset) by using {@link
139 * java.util.concurrent.atomic.LongAdder} values and initializing via
140 * {@link #computeIfAbsent computeIfAbsent}. For example, to add a count
141 * to a {@code ConcurrentHashMap<String,LongAdder> freqs}, you can use
142 * {@code freqs.computeIfAbsent(key, k -> new LongAdder()).increment();}
143 *
144 * <p>This class and its views and iterators implement all of the
145 * <em>optional</em> methods of the {@link Map} and {@link Iterator}
146 * interfaces.
147 *
148 * <p>Like {@link Hashtable} but unlike {@link HashMap}, this class
149 * does <em>not</em> allow {@code null} to be used as a key or value.
150 *
151 * <p>ConcurrentHashMaps support a set of sequential and parallel bulk
152 * operations that, unlike most {@link Stream} methods, are designed
153 * to be safely, and often sensibly, applied even with maps that are
154 * being concurrently updated by other threads; for example, when
155 * computing a snapshot summary of the values in a shared registry.
156 * There are three kinds of operation, each with four forms, accepting
157 * functions with Keys, Values, Entries, and (Key, Value) arguments
158 * and/or return values. Because the elements of a ConcurrentHashMap
159 * are not ordered in any particular way, and may be processed in
160 * different orders in different parallel executions, the correctness
161 * of supplied functions should not depend on any ordering, or on any
162 * other objects or values that may transiently change while
163 * computation is in progress; and except for forEach actions, should
164 * ideally be side-effect-free. Bulk operations on {@link java.util.Map.Entry}
165 * objects do not support method {@code setValue}.
166 *
167 * <ul>
168 * <li> forEach: Perform a given action on each element.
169 * A variant form applies a given transformation on each element
170 * before performing the action.</li>
171 *
172 * <li> search: Return the first available non-null result of
173 * applying a given function on each element; skipping further
174 * search when a result is found.</li>
175 *
176 * <li> reduce: Accumulate each element. The supplied reduction
177 * function cannot rely on ordering (more formally, it should be
178 * both associative and commutative). There are five variants:
179 *
180 * <ul>
181 *
182 * <li> Plain reductions. (There is not a form of this method for
183 * (key, value) function arguments since there is no corresponding
184 * return type.)</li>
185 *
186 * <li> Mapped reductions that accumulate the results of a given
187 * function applied to each element.</li>
188 *
189 * <li> Reductions to scalar doubles, longs, and ints, using a
190 * given basis value.</li>
191 *
192 * </ul>
193 * </li>
194 * </ul>
195 *
196 * <p>These bulk operations accept a {@code parallelismThreshold}
197 * argument. Methods proceed sequentially if the current map size is
198 * estimated to be less than the given threshold. Using a value of
199 * {@code Long.MAX_VALUE} suppresses all parallelism. Using a value
200 * of {@code 1} results in maximal parallelism by partitioning into
201 * enough subtasks to fully utilize the {@link
202 * ForkJoinPool#commonPool()} that is used for all parallel
203 * computations. Normally, you would initially choose one of these
204 * extreme values, and then measure performance of using in-between
205 * values that trade off overhead versus throughput.
206 *
207 * <p>The concurrency properties of bulk operations follow
208 * from those of ConcurrentHashMap: Any non-null result returned
209 * from {@code get(key)} and related access methods bears a
210 * happens-before relation with the associated insertion or
211 * update. The result of any bulk operation reflects the
212 * composition of these per-element relations (but is not
213 * necessarily atomic with respect to the map as a whole unless it
214 * is somehow known to be quiescent). Conversely, because keys
215 * and values in the map are never null, null serves as a reliable
216 * atomic indicator of the current lack of any result. To
217 * maintain this property, null serves as an implicit basis for
218 * all non-scalar reduction operations. For the double, long, and
219 * int versions, the basis should be one that, when combined with
220 * any other value, returns that other value (more formally, it
221 * should be the identity element for the reduction). Most common
222 * reductions have these properties; for example, computing a sum
223 * with basis 0 or a minimum with basis MAX_VALUE.
224 *
225 * <p>Search and transformation functions provided as arguments
226 * should similarly return null to indicate the lack of any result
227 * (in which case it is not used). In the case of mapped
228 * reductions, this also enables transformations to serve as
229 * filters, returning null (or, in the case of primitive
230 * specializations, the identity basis) if the element should not
231 * be combined. You can create compound transformations and
232 * filterings by composing them yourself under this "null means
233 * there is nothing there now" rule before using them in search or
234 * reduce operations.
235 *
236 * <p>Methods accepting and/or returning Entry arguments maintain
237 * key-value associations. They may be useful for example when
238 * finding the key for the greatest value. Note that "plain" Entry
239 * arguments can be supplied using {@code new
240 * AbstractMap.SimpleEntry(k,v)}.
241 *
242 * <p>Bulk operations may complete abruptly, throwing an
243 * exception encountered in the application of a supplied
244 * function. Bear in mind when handling such exceptions that other
245 * concurrently executing functions could also have thrown
246 * exceptions, or would have done so if the first exception had
247 * not occurred.
248 *
249 * <p>Speedups for parallel compared to sequential forms are common
250 * but not guaranteed. Parallel operations involving brief functions
251 * on small maps may execute more slowly than sequential forms if the
252 * underlying work to parallelize the computation is more expensive
253 * than the computation itself. Similarly, parallelization may not
254 * lead to much actual parallelism if all processors are busy
255 * performing unrelated tasks.
256 *
257 * <p>All arguments to all task methods must be non-null.
258 *
259 * <p>This class is a member of the
260 * <a href="{@docRoot}/../technotes/guides/collections/index.html">
261 * Java Collections Framework</a>.
262 *
263 * @since 1.5
264 * @author Doug Lea
265 * @param <K> the type of keys maintained by this map
266 * @param <V> the type of mapped values
267 */
268 public class ConcurrentHashMap<K,V> extends AbstractMap<K,V>
269 implements ConcurrentMap<K,V>, Serializable {
270 private static final long serialVersionUID = 7249069246763182397L;
271
272 /*
273 * Overview:
274 *
275 * The primary design goal of this hash table is to maintain
276 * concurrent readability (typically method get(), but also
277 * iterators and related methods) while minimizing update
278 * contention. Secondary goals are to keep space consumption about
279 * the same or better than java.util.HashMap, and to support high
280 * initial insertion rates on an empty table by many threads.
281 *
282 * This map usually acts as a binned (bucketed) hash table. Each
283 * key-value mapping is held in a Node. Most nodes are instances
284 * of the basic Node class with hash, key, value, and next
285 * fields. However, various subclasses exist: TreeNodes are
286 * arranged in balanced trees, not lists. TreeBins hold the roots
287 * of sets of TreeNodes. ForwardingNodes are placed at the heads
288 * of bins during resizing. ReservationNodes are used as
289 * placeholders while establishing values in computeIfAbsent and
290 * related methods. The types TreeBin, ForwardingNode, and
291 * ReservationNode do not hold normal user keys, values, or
292 * hashes, and are readily distinguishable during search etc
293 * because they have negative hash fields and null key and value
294 * fields. (These special nodes are either uncommon or transient,
295 * so the impact of carrying around some unused fields is
296 * insignificant.)
297 *
298 * The table is lazily initialized to a power-of-two size upon the
299 * first insertion. Each bin in the table normally contains a
300 * list of Nodes (most often, the list has only zero or one Node).
301 * Table accesses require volatile/atomic reads, writes, and
302 * CASes. Because there is no other way to arrange this without
303 * adding further indirections, we use intrinsics
304 * (sun.misc.Unsafe) operations.
305 *
306 * We use the top (sign) bit of Node hash fields for control
307 * purposes -- it is available anyway because of addressing
308 * constraints. Nodes with negative hash fields are specially
309 * handled or ignored in map methods.
310 *
311 * Insertion (via put or its variants) of the first node in an
312 * empty bin is performed by just CASing it to the bin. This is
313 * by far the most common case for put operations under most
314 * key/hash distributions. Other update operations (insert,
315 * delete, and replace) require locks. We do not want to waste
316 * the space required to associate a distinct lock object with
317 * each bin, so instead use the first node of a bin list itself as
318 * a lock. Locking support for these locks relies on builtin
319 * "synchronized" monitors.
320 *
321 * Using the first node of a list as a lock does not by itself
322 * suffice though: When a node is locked, any update must first
323 * validate that it is still the first node after locking it, and
324 * retry if not. Because new nodes are always appended to lists,
325 * once a node is first in a bin, it remains first until deleted
326 * or the bin becomes invalidated (upon resizing).
327 *
328 * The main disadvantage of per-bin locks is that other update
329 * operations on other nodes in a bin list protected by the same
330 * lock can stall, for example when user equals() or mapping
331 * functions take a long time. However, statistically, under
332 * random hash codes, this is not a common problem. Ideally, the
333 * frequency of nodes in bins follows a Poisson distribution
334 * (http://en.wikipedia.org/wiki/Poisson_distribution) with a
335 * parameter of about 0.5 on average, given the resizing threshold
336 * of 0.75, although with a large variance because of resizing
337 * granularity. Ignoring variance, the expected occurrences of
338 * list size k are (exp(-0.5) * pow(0.5, k) / factorial(k)). The
339 * first values are:
340 *
341 * 0: 0.60653066
342 * 1: 0.30326533
343 * 2: 0.07581633
344 * 3: 0.01263606
345 * 4: 0.00157952
346 * 5: 0.00015795
347 * 6: 0.00001316
348 * 7: 0.00000094
349 * 8: 0.00000006
350 * more: less than 1 in ten million
351 *
352 * Lock contention probability for two threads accessing distinct
353 * elements is roughly 1 / (8 * #elements) under random hashes.
354 *
355 * Actual hash code distributions encountered in practice
356 * sometimes deviate significantly from uniform randomness. This
357 * includes the case when N > (1<<30), so some keys MUST collide.
358 * Similarly for dumb or hostile usages in which multiple keys are
359 * designed to have identical hash codes or ones that differs only
360 * in masked-out high bits. So we use a secondary strategy that
361 * applies when the number of nodes in a bin exceeds a
362 * threshold. These TreeBins use a balanced tree to hold nodes (a
363 * specialized form of red-black trees), bounding search time to
364 * O(log N). Each search step in a TreeBin is at least twice as
365 * slow as in a regular list, but given that N cannot exceed
366 * (1<<64) (before running out of addresses) this bounds search
367 * steps, lock hold times, etc, to reasonable constants (roughly
368 * 100 nodes inspected per operation worst case) so long as keys
369 * are Comparable (which is very common -- String, Long, etc).
370 * TreeBin nodes (TreeNodes) also maintain the same "next"
371 * traversal pointers as regular nodes, so can be traversed in
372 * iterators in the same way.
373 *
374 * The table is resized when occupancy exceeds a percentage
375 * threshold (nominally, 0.75, but see below). Any thread
376 * noticing an overfull bin may assist in resizing after the
377 * initiating thread allocates and sets up the replacement array.
378 * However, rather than stalling, these other threads may proceed
379 * with insertions etc. The use of TreeBins shields us from the
380 * worst case effects of overfilling while resizes are in
381 * progress. Resizing proceeds by transferring bins, one by one,
382 * from the table to the next table. However, threads claim small
383 * blocks of indices to transfer (via field transferIndex) before
384 * doing so, reducing contention. A generation stamp in field
385 * sizeCtl ensures that resizings do not overlap. Because we are
386 * using power-of-two expansion, the elements from each bin must
387 * either stay at same index, or move with a power of two
388 * offset. We eliminate unnecessary node creation by catching
389 * cases where old nodes can be reused because their next fields
390 * won't change. On average, only about one-sixth of them need
391 * cloning when a table doubles. The nodes they replace will be
392 * garbage collectable as soon as they are no longer referenced by
393 * any reader thread that may be in the midst of concurrently
394 * traversing table. Upon transfer, the old table bin contains
395 * only a special forwarding node (with hash field "MOVED") that
396 * contains the next table as its key. On encountering a
397 * forwarding node, access and update operations restart, using
398 * the new table.
399 *
400 * Each bin transfer requires its bin lock, which can stall
401 * waiting for locks while resizing. However, because other
402 * threads can join in and help resize rather than contend for
403 * locks, average aggregate waits become shorter as resizing
404 * progresses. The transfer operation must also ensure that all
405 * accessible bins in both the old and new table are usable by any
406 * traversal. This is arranged in part by proceeding from the
407 * last bin (table.length - 1) up towards the first. Upon seeing
408 * a forwarding node, traversals (see class Traverser) arrange to
409 * move to the new table without revisiting nodes. To ensure that
410 * no intervening nodes are skipped even when moved out of order,
411 * a stack (see class TableStack) is created on first encounter of
412 * a forwarding node during a traversal, to maintain its place if
413 * later processing the current table. The need for these
414 * save/restore mechanics is relatively rare, but when one
415 * forwarding node is encountered, typically many more will be.
416 * So Traversers use a simple caching scheme to avoid creating so
417 * many new TableStack nodes. (Thanks to Peter Levart for
418 * suggesting use of a stack here.)
419 *
420 * The traversal scheme also applies to partial traversals of
421 * ranges of bins (via an alternate Traverser constructor)
422 * to support partitioned aggregate operations. Also, read-only
423 * operations give up if ever forwarded to a null table, which
424 * provides support for shutdown-style clearing, which is also not
425 * currently implemented.
426 *
427 * Lazy table initialization minimizes footprint until first use,
428 * and also avoids resizings when the first operation is from a
429 * putAll, constructor with map argument, or deserialization.
430 * These cases attempt to override the initial capacity settings,
431 * but harmlessly fail to take effect in cases of races.
432 *
433 * The element count is maintained using a specialization of
434 * LongAdder. We need to incorporate a specialization rather than
435 * just use a LongAdder in order to access implicit
436 * contention-sensing that leads to creation of multiple
437 * CounterCells. The counter mechanics avoid contention on
438 * updates but can encounter cache thrashing if read too
439 * frequently during concurrent access. To avoid reading so often,
440 * resizing under contention is attempted only upon adding to a
441 * bin already holding two or more nodes. Under uniform hash
442 * distributions, the probability of this occurring at threshold
443 * is around 13%, meaning that only about 1 in 8 puts check
444 * threshold (and after resizing, many fewer do so).
445 *
446 * TreeBins use a special form of comparison for search and
447 * related operations (which is the main reason we cannot use
448 * existing collections such as TreeMaps). TreeBins contain
449 * Comparable elements, but may contain others, as well as
450 * elements that are Comparable but not necessarily Comparable for
451 * the same T, so we cannot invoke compareTo among them. To handle
452 * this, the tree is ordered primarily by hash value, then by
453 * Comparable.compareTo order if applicable. On lookup at a node,
454 * if elements are not comparable or compare as 0 then both left
455 * and right children may need to be searched in the case of tied
456 * hash values. (This corresponds to the full list search that
457 * would be necessary if all elements were non-Comparable and had
458 * tied hashes.) On insertion, to keep a total ordering (or as
459 * close as is required here) across rebalancings, we compare
460 * classes and identityHashCodes as tie-breakers. The red-black
461 * balancing code is updated from pre-jdk-collections
462 * (http://gee.cs.oswego.edu/dl/classes/collections/RBCell.java)
463 * based in turn on Cormen, Leiserson, and Rivest "Introduction to
464 * Algorithms" (CLR).
465 *
466 * TreeBins also require an additional locking mechanism. While
467 * list traversal is always possible by readers even during
468 * updates, tree traversal is not, mainly because of tree-rotations
469 * that may change the root node and/or its linkages. TreeBins
470 * include a simple read-write lock mechanism parasitic on the
471 * main bin-synchronization strategy: Structural adjustments
472 * associated with an insertion or removal are already bin-locked
473 * (and so cannot conflict with other writers) but must wait for
474 * ongoing readers to finish. Since there can be only one such
475 * waiter, we use a simple scheme using a single "waiter" field to
476 * block writers. However, readers need never block. If the root
477 * lock is held, they proceed along the slow traversal path (via
478 * next-pointers) until the lock becomes available or the list is
479 * exhausted, whichever comes first. These cases are not fast, but
480 * maximize aggregate expected throughput.
481 *
482 * Maintaining API and serialization compatibility with previous
483 * versions of this class introduces several oddities. Mainly: We
484 * leave untouched but unused constructor arguments refering to
485 * concurrencyLevel. We accept a loadFactor constructor argument,
486 * but apply it only to initial table capacity (which is the only
487 * time that we can guarantee to honor it.) We also declare an
488 * unused "Segment" class that is instantiated in minimal form
489 * only when serializing.
490 *
491 * Also, solely for compatibility with previous versions of this
492 * class, it extends AbstractMap, even though all of its methods
493 * are overridden, so it is just useless baggage.
494 *
495 * This file is organized to make things a little easier to follow
496 * while reading than they might otherwise: First the main static
497 * declarations and utilities, then fields, then main public
498 * methods (with a few factorings of multiple public methods into
499 * internal ones), then sizing methods, trees, traversers, and
500 * bulk operations.
501 */
502
503 /* ---------------- Constants -------------- */
504
505 /**
506 * The largest possible table capacity. This value must be
507 * exactly 1<<30 to stay within Java array allocation and indexing
508 * bounds for power of two table sizes, and is further required
509 * because the top two bits of 32bit hash fields are used for
510 * control purposes.
511 */
512 private static final int MAXIMUM_CAPACITY = 1 << 30;
513
514 /**
515 * The default initial table capacity. Must be a power of 2
516 * (i.e., at least 1) and at most MAXIMUM_CAPACITY.
517 */
518 private static final int DEFAULT_CAPACITY = 16;
519
520 /**
521 * The largest possible (non-power of two) array size.
522 * Needed by toArray and related methods.
523 */
524 static final int MAX_ARRAY_SIZE = Integer.MAX_VALUE - 8;
525
526 /**
527 * The default concurrency level for this table. Unused but
528 * defined for compatibility with previous versions of this class.
529 */
530 private static final int DEFAULT_CONCURRENCY_LEVEL = 16;
531
532 /**
533 * The load factor for this table. Overrides of this value in
534 * constructors affect only the initial table capacity. The
535 * actual floating point value isn't normally used -- it is
536 * simpler to use expressions such as {@code n - (n >>> 2)} for
537 * the associated resizing threshold.
538 */
539 private static final float LOAD_FACTOR = 0.75f;
540
541 /**
542 * The bin count threshold for using a tree rather than list for a
543 * bin. Bins are converted to trees when adding an element to a
544 * bin with at least this many nodes. The value must be greater
545 * than 2, and should be at least 8 to mesh with assumptions in
546 * tree removal about conversion back to plain bins upon
547 * shrinkage.
548 */
549 static final int TREEIFY_THRESHOLD = 8;
550
551 /**
552 * The bin count threshold for untreeifying a (split) bin during a
553 * resize operation. Should be less than TREEIFY_THRESHOLD, and at
554 * most 6 to mesh with shrinkage detection under removal.
555 */
556 static final int UNTREEIFY_THRESHOLD = 6;
557
558 /**
559 * The smallest table capacity for which bins may be treeified.
560 * (Otherwise the table is resized if too many nodes in a bin.)
561 * The value should be at least 4 * TREEIFY_THRESHOLD to avoid
562 * conflicts between resizing and treeification thresholds.
563 */
564 static final int MIN_TREEIFY_CAPACITY = 64;
565
566 /**
567 * Minimum number of rebinnings per transfer step. Ranges are
568 * subdivided to allow multiple resizer threads. This value
569 * serves as a lower bound to avoid resizers encountering
570 * excessive memory contention. The value should be at least
571 * DEFAULT_CAPACITY.
572 */
573 private static final int MIN_TRANSFER_STRIDE = 16;
574
575 /**
576 * The number of bits used for generation stamp in sizeCtl.
577 * Must be at least 6 for 32bit arrays.
578 */
579 private static int RESIZE_STAMP_BITS = 16;
580
581 /**
582 * The maximum number of threads that can help resize.
583 * Must fit in 32 - RESIZE_STAMP_BITS bits.
584 */
585 private static final int MAX_RESIZERS = (1 << (32 - RESIZE_STAMP_BITS)) - 1;
586
587 /**
588 * The bit shift for recording size stamp in sizeCtl.
589 */
590 private static final int RESIZE_STAMP_SHIFT = 32 - RESIZE_STAMP_BITS;
591
592 /*
593 * Encodings for Node hash fields. See above for explanation.
594 */
595 static final int MOVED = -1; // hash for forwarding nodes
596 static final int TREEBIN = -2; // hash for roots of trees
597 static final int RESERVED = -3; // hash for transient reservations
598 static final int HASH_BITS = 0x7fffffff; // usable bits of normal node hash
599
600 /** Number of CPUS, to place bounds on some sizings */
601 static final int NCPU = Runtime.getRuntime().availableProcessors();
602
603 /** For serialization compatibility. */
604 private static final ObjectStreamField[] serialPersistentFields = {
605 new ObjectStreamField("segments", Segment[].class),
606 new ObjectStreamField("segmentMask", Integer.TYPE),
607 new ObjectStreamField("segmentShift", Integer.TYPE)
608 };
609
610 /* ---------------- Nodes -------------- */
611
612 /**
613 * Key-value entry. This class is never exported out as a
614 * user-mutable Map.Entry (i.e., one supporting setValue; see
615 * MapEntry below), but can be used for read-only traversals used
616 * in bulk tasks. Subclasses of Node with a negative hash field
617 * are special, and contain null keys and values (but are never
618 * exported). Otherwise, keys and vals are never null.
619 */
620 static class Node<K,V> implements Map.Entry<K,V> {
621 final int hash;
622 final K key;
623 volatile V val;
624 volatile Node<K,V> next;
625
626 Node(int hash, K key, V val, Node<K,V> next) {
627 this.hash = hash;
628 this.key = key;
629 this.val = val;
630 this.next = next;
631 }
632
633 public final K getKey() { return key; }
634 public final V getValue() { return val; }
635 public final int hashCode() { return key.hashCode() ^ val.hashCode(); }
636 public final String toString(){ return key + "=" + val; }
637 public final V setValue(V value) {
638 throw new UnsupportedOperationException();
639 }
640
641 public final boolean equals(Object o) {
642 Object k, v, u; Map.Entry<?,?> e;
643 return ((o instanceof Map.Entry) &&
644 (k = (e = (Map.Entry<?,?>)o).getKey()) != null &&
645 (v = e.getValue()) != null &&
646 (k == key || k.equals(key)) &&
647 (v == (u = val) || v.equals(u)));
648 }
649
650 /**
651 * Virtualized support for map.get(); overridden in subclasses.
652 */
653 Node<K,V> find(int h, Object k) {
654 Node<K,V> e = this;
655 if (k != null) {
656 do {
657 K ek;
658 if (e.hash == h &&
659 ((ek = e.key) == k || (ek != null && k.equals(ek))))
660 return e;
661 } while ((e = e.next) != null);
662 }
663 return null;
664 }
665 }
666
667 /* ---------------- Static utilities -------------- */
668
669 /**
670 * Spreads (XORs) higher bits of hash to lower and also forces top
671 * bit to 0. Because the table uses power-of-two masking, sets of
672 * hashes that vary only in bits above the current mask will
673 * always collide. (Among known examples are sets of Float keys
674 * holding consecutive whole numbers in small tables.) So we
675 * apply a transform that spreads the impact of higher bits
676 * downward. There is a tradeoff between speed, utility, and
677 * quality of bit-spreading. Because many common sets of hashes
678 * are already reasonably distributed (so don't benefit from
679 * spreading), and because we use trees to handle large sets of
680 * collisions in bins, we just XOR some shifted bits in the
681 * cheapest possible way to reduce systematic lossage, as well as
682 * to incorporate impact of the highest bits that would otherwise
683 * never be used in index calculations because of table bounds.
684 */
685 static final int spread(int h) {
686 return (h ^ (h >>> 16)) & HASH_BITS;
687 }
688
689 /**
690 * Returns a power of two table size for the given desired capacity.
691 * See Hackers Delight, sec 3.2
692 */
693 private static final int tableSizeFor(int c) {
694 int n = c - 1;
695 n |= n >>> 1;
696 n |= n >>> 2;
697 n |= n >>> 4;
698 n |= n >>> 8;
699 n |= n >>> 16;
700 return (n < 0) ? 1 : (n >= MAXIMUM_CAPACITY) ? MAXIMUM_CAPACITY : n + 1;
701 }
702
703 /**
704 * Returns x's Class if it is of the form "class C implements
705 * Comparable<C>", else null.
706 */
707 static Class<?> comparableClassFor(Object x) {
708 if (x instanceof Comparable) {
709 Class<?> c; Type[] ts, as; Type t; ParameterizedType p;
710 if ((c = x.getClass()) == String.class) // bypass checks
711 return c;
712 if ((ts = c.getGenericInterfaces()) != null) {
713 for (int i = 0; i < ts.length; ++i) {
714 if (((t = ts[i]) instanceof ParameterizedType) &&
715 ((p = (ParameterizedType)t).getRawType() ==
716 Comparable.class) &&
717 (as = p.getActualTypeArguments()) != null &&
718 as.length == 1 && as[0] == c) // type arg is c
719 return c;
720 }
721 }
722 }
723 return null;
724 }
725
726 /**
727 * Returns k.compareTo(x) if x matches kc (k's screened comparable
728 * class), else 0.
729 */
730 @SuppressWarnings({"rawtypes","unchecked"}) // for cast to Comparable
731 static int compareComparables(Class<?> kc, Object k, Object x) {
732 return (x == null || x.getClass() != kc ? 0 :
733 ((Comparable)k).compareTo(x));
734 }
735
736 /* ---------------- Table element access -------------- */
737
738 /*
739 * Volatile access methods are used for table elements as well as
740 * elements of in-progress next table while resizing. All uses of
741 * the tab arguments must be null checked by callers. All callers
742 * also paranoically precheck that tab's length is not zero (or an
743 * equivalent check), thus ensuring that any index argument taking
744 * the form of a hash value anded with (length - 1) is a valid
745 * index. Note that, to be correct wrt arbitrary concurrency
746 * errors by users, these checks must operate on local variables,
747 * which accounts for some odd-looking inline assignments below.
748 * Note that calls to setTabAt always occur within locked regions,
749 * and so in principle require only release ordering, not
750 * full volatile semantics, but are currently coded as volatile
751 * writes to be conservative.
752 */
753
754 @SuppressWarnings("unchecked")
755 static final <K,V> Node<K,V> tabAt(Node<K,V>[] tab, int i) {
756 return (Node<K,V>)U.getObjectVolatile(tab, ((long)i << ASHIFT) + ABASE);
757 }
758
759 static final <K,V> boolean casTabAt(Node<K,V>[] tab, int i,
760 Node<K,V> c, Node<K,V> v) {
761 return U.compareAndSwapObject(tab, ((long)i << ASHIFT) + ABASE, c, v);
762 }
763
764 static final <K,V> void setTabAt(Node<K,V>[] tab, int i, Node<K,V> v) {
765 U.putObjectVolatile(tab, ((long)i << ASHIFT) + ABASE, v);
766 }
767
768 /* ---------------- Fields -------------- */
769
770 /**
771 * The array of bins. Lazily initialized upon first insertion.
772 * Size is always a power of two. Accessed directly by iterators.
773 */
774 transient volatile Node<K,V>[] table;
775
776 /**
777 * The next table to use; non-null only while resizing.
778 */
779 private transient volatile Node<K,V>[] nextTable;
780
781 /**
782 * Base counter value, used mainly when there is no contention,
783 * but also as a fallback during table initialization
784 * races. Updated via CAS.
785 */
786 private transient volatile long baseCount;
787
788 /**
789 * Table initialization and resizing control. When negative, the
790 * table is being initialized or resized: -1 for initialization,
791 * else -(1 + the number of active resizing threads). Otherwise,
792 * when table is null, holds the initial table size to use upon
793 * creation, or 0 for default. After initialization, holds the
794 * next element count value upon which to resize the table.
795 */
796 private transient volatile int sizeCtl;
797
798 /**
799 * The next table index (plus one) to split while resizing.
800 */
801 private transient volatile int transferIndex;
802
803 /**
804 * Spinlock (locked via CAS) used when resizing and/or creating CounterCells.
805 */
806 private transient volatile int cellsBusy;
807
808 /**
809 * Table of counter cells. When non-null, size is a power of 2.
810 */
811 private transient volatile CounterCell[] counterCells;
812
813 // views
814 private transient KeySetView<K,V> keySet;
815 private transient ValuesView<K,V> values;
816 private transient EntrySetView<K,V> entrySet;
817
818
819 /* ---------------- Public operations -------------- */
820
821 /**
822 * Creates a new, empty map with the default initial table size (16).
823 */
824 public ConcurrentHashMap() {
825 }
826
827 /**
828 * Creates a new, empty map with an initial table size
829 * accommodating the specified number of elements without the need
830 * to dynamically resize.
831 *
832 * @param initialCapacity The implementation performs internal
833 * sizing to accommodate this many elements.
834 * @throws IllegalArgumentException if the initial capacity of
835 * elements is negative
836 */
837 public ConcurrentHashMap(int initialCapacity) {
838 if (initialCapacity < 0)
839 throw new IllegalArgumentException();
840 int cap = ((initialCapacity >= (MAXIMUM_CAPACITY >>> 1)) ?
841 MAXIMUM_CAPACITY :
842 tableSizeFor(initialCapacity + (initialCapacity >>> 1) + 1));
843 this.sizeCtl = cap;
844 }
845
846 /**
847 * Creates a new map with the same mappings as the given map.
848 *
849 * @param m the map
850 */
851 public ConcurrentHashMap(Map<? extends K, ? extends V> m) {
852 this.sizeCtl = DEFAULT_CAPACITY;
853 putAll(m);
854 }
855
856 /**
857 * Creates a new, empty map with an initial table size based on
858 * the given number of elements ({@code initialCapacity}) and
859 * initial table density ({@code loadFactor}).
860 *
861 * @param initialCapacity the initial capacity. The implementation
862 * performs internal sizing to accommodate this many elements,
863 * given the specified load factor.
864 * @param loadFactor the load factor (table density) for
865 * establishing the initial table size
866 * @throws IllegalArgumentException if the initial capacity of
867 * elements is negative or the load factor is nonpositive
868 *
869 * @since 1.6
870 */
871 public ConcurrentHashMap(int initialCapacity, float loadFactor) {
872 this(initialCapacity, loadFactor, 1);
873 }
874
875 /**
876 * Creates a new, empty map with an initial table size based on
877 * the given number of elements ({@code initialCapacity}), table
878 * density ({@code loadFactor}), and number of concurrently
879 * updating threads ({@code concurrencyLevel}).
880 *
881 * @param initialCapacity the initial capacity. The implementation
882 * performs internal sizing to accommodate this many elements,
883 * given the specified load factor.
884 * @param loadFactor the load factor (table density) for
885 * establishing the initial table size
886 * @param concurrencyLevel the estimated number of concurrently
887 * updating threads. The implementation may use this value as
888 * a sizing hint.
889 * @throws IllegalArgumentException if the initial capacity is
890 * negative or the load factor or concurrencyLevel are
891 * nonpositive
892 */
893 public ConcurrentHashMap(int initialCapacity,
894 float loadFactor, int concurrencyLevel) {
895 if (!(loadFactor > 0.0f) || initialCapacity < 0 || concurrencyLevel <= 0)
896 throw new IllegalArgumentException();
897 if (initialCapacity < concurrencyLevel) // Use at least as many bins
898 initialCapacity = concurrencyLevel; // as estimated threads
899 long size = (long)(1.0 + (long)initialCapacity / loadFactor);
900 int cap = (size >= (long)MAXIMUM_CAPACITY) ?
901 MAXIMUM_CAPACITY : tableSizeFor((int)size);
902 this.sizeCtl = cap;
903 }
904
905 // Original (since JDK1.2) Map methods
906
907 /**
908 * {@inheritDoc}
909 */
910 public int size() {
911 long n = sumCount();
912 return ((n < 0L) ? 0 :
913 (n > (long)Integer.MAX_VALUE) ? Integer.MAX_VALUE :
914 (int)n);
915 }
916
917 /**
918 * {@inheritDoc}
919 */
920 public boolean isEmpty() {
921 return sumCount() <= 0L; // ignore transient negative values
922 }
923
924 /**
925 * Returns the value to which the specified key is mapped,
926 * or {@code null} if this map contains no mapping for the key.
927 *
928 * <p>More formally, if this map contains a mapping from a key
929 * {@code k} to a value {@code v} such that {@code key.equals(k)},
930 * then this method returns {@code v}; otherwise it returns
931 * {@code null}. (There can be at most one such mapping.)
932 *
933 * @throws NullPointerException if the specified key is null
934 */
935 public V get(Object key) {
936 Node<K,V>[] tab; Node<K,V> e, p; int n, eh; K ek;
937 int h = spread(key.hashCode());
938 if ((tab = table) != null && (n = tab.length) > 0 &&
939 (e = tabAt(tab, (n - 1) & h)) != null) {
940 if ((eh = e.hash) == h) {
941 if ((ek = e.key) == key || (ek != null && key.equals(ek)))
942 return e.val;
943 }
944 else if (eh < 0)
945 return (p = e.find(h, key)) != null ? p.val : null;
946 while ((e = e.next) != null) {
947 if (e.hash == h &&
948 ((ek = e.key) == key || (ek != null && key.equals(ek))))
949 return e.val;
950 }
951 }
952 return null;
953 }
954
955 /**
956 * Tests if the specified object is a key in this table.
957 *
958 * @param key possible key
959 * @return {@code true} if and only if the specified object
960 * is a key in this table, as determined by the
961 * {@code equals} method; {@code false} otherwise
962 * @throws NullPointerException if the specified key is null
963 */
964 public boolean containsKey(Object key) {
965 return get(key) != null;
966 }
967
968 /**
969 * Returns {@code true} if this map maps one or more keys to the
970 * specified value. Note: This method may require a full traversal
971 * of the map, and is much slower than method {@code containsKey}.
972 *
973 * @param value value whose presence in this map is to be tested
974 * @return {@code true} if this map maps one or more keys to the
975 * specified value
976 * @throws NullPointerException if the specified value is null
977 */
978 public boolean containsValue(Object value) {
979 if (value == null)
980 throw new NullPointerException();
981 Node<K,V>[] t;
982 if ((t = table) != null) {
983 Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length);
984 for (Node<K,V> p; (p = it.advance()) != null; ) {
985 V v;
986 if ((v = p.val) == value || (v != null && value.equals(v)))
987 return true;
988 }
989 }
990 return false;
991 }
992
993 /**
994 * Maps the specified key to the specified value in this table.
995 * Neither the key nor the value can be null.
996 *
997 * <p>The value can be retrieved by calling the {@code get} method
998 * with a key that is equal to the original key.
999 *
1000 * @param key key with which the specified value is to be associated
1001 * @param value value to be associated with the specified key
1002 * @return the previous value associated with {@code key}, or
1003 * {@code null} if there was no mapping for {@code key}
1004 * @throws NullPointerException if the specified key or value is null
1005 */
1006 public V put(K key, V value) {
1007 return putVal(key, value, false);
1008 }
1009
1010 /** Implementation for put and putIfAbsent */
1011 final V putVal(K key, V value, boolean onlyIfAbsent) {
1012 if (key == null || value == null) throw new NullPointerException();
1013 int hash = spread(key.hashCode());
1014 int binCount = 0;
1015 for (Node<K,V>[] tab = table;;) {
1016 Node<K,V> f; int n, i, fh;
1017 if (tab == null || (n = tab.length) == 0)
1018 tab = initTable();
1019 else if ((f = tabAt(tab, i = (n - 1) & hash)) == null) {
1020 if (casTabAt(tab, i, null,
1021 new Node<K,V>(hash, key, value, null)))
1022 break; // no lock when adding to empty bin
1023 }
1024 else if ((fh = f.hash) == MOVED)
1025 tab = helpTransfer(tab, f);
1026 else {
1027 V oldVal = null;
1028 synchronized (f) {
1029 if (tabAt(tab, i) == f) {
1030 if (fh >= 0) {
1031 binCount = 1;
1032 for (Node<K,V> e = f;; ++binCount) {
1033 K ek;
1034 if (e.hash == hash &&
1035 ((ek = e.key) == key ||
1036 (ek != null && key.equals(ek)))) {
1037 oldVal = e.val;
1038 if (!onlyIfAbsent)
1039 e.val = value;
1040 break;
1041 }
1042 Node<K,V> pred = e;
1043 if ((e = e.next) == null) {
1044 pred.next = new Node<K,V>(hash, key,
1045 value, null);
1046 break;
1047 }
1048 }
1049 }
1050 else if (f instanceof TreeBin) {
1051 Node<K,V> p;
1052 binCount = 2;
1053 if ((p = ((TreeBin<K,V>)f).putTreeVal(hash, key,
1054 value)) != null) {
1055 oldVal = p.val;
1056 if (!onlyIfAbsent)
1057 p.val = value;
1058 }
1059 }
1060 }
1061 }
1062 if (binCount != 0) {
1063 if (binCount >= TREEIFY_THRESHOLD)
1064 treeifyBin(tab, i);
1065 if (oldVal != null)
1066 return oldVal;
1067 break;
1068 }
1069 }
1070 }
1071 addCount(1L, binCount);
1072 return null;
1073 }
1074
1075 /**
1076 * Copies all of the mappings from the specified map to this one.
1077 * These mappings replace any mappings that this map had for any of the
1078 * keys currently in the specified map.
1079 *
1080 * @param m mappings to be stored in this map
1081 */
1082 public void putAll(Map<? extends K, ? extends V> m) {
1083 tryPresize(m.size());
1084 for (Map.Entry<? extends K, ? extends V> e : m.entrySet())
1085 putVal(e.getKey(), e.getValue(), false);
1086 }
1087
1088 /**
1089 * Removes the key (and its corresponding value) from this map.
1090 * This method does nothing if the key is not in the map.
1091 *
1092 * @param key the key that needs to be removed
1093 * @return the previous value associated with {@code key}, or
1094 * {@code null} if there was no mapping for {@code key}
1095 * @throws NullPointerException if the specified key is null
1096 */
1097 public V remove(Object key) {
1098 return replaceNode(key, null, null);
1099 }
1100
1101 /**
1102 * Implementation for the four public remove/replace methods:
1103 * Replaces node value with v, conditional upon match of cv if
1104 * non-null. If resulting value is null, delete.
1105 */
1106 final V replaceNode(Object key, V value, Object cv) {
1107 int hash = spread(key.hashCode());
1108 for (Node<K,V>[] tab = table;;) {
1109 Node<K,V> f; int n, i, fh;
1110 if (tab == null || (n = tab.length) == 0 ||
1111 (f = tabAt(tab, i = (n - 1) & hash)) == null)
1112 break;
1113 else if ((fh = f.hash) == MOVED)
1114 tab = helpTransfer(tab, f);
1115 else {
1116 V oldVal = null;
1117 boolean validated = false;
1118 synchronized (f) {
1119 if (tabAt(tab, i) == f) {
1120 if (fh >= 0) {
1121 validated = true;
1122 for (Node<K,V> e = f, pred = null;;) {
1123 K ek;
1124 if (e.hash == hash &&
1125 ((ek = e.key) == key ||
1126 (ek != null && key.equals(ek)))) {
1127 V ev = e.val;
1128 if (cv == null || cv == ev ||
1129 (ev != null && cv.equals(ev))) {
1130 oldVal = ev;
1131 if (value != null)
1132 e.val = value;
1133 else if (pred != null)
1134 pred.next = e.next;
1135 else
1136 setTabAt(tab, i, e.next);
1137 }
1138 break;
1139 }
1140 pred = e;
1141 if ((e = e.next) == null)
1142 break;
1143 }
1144 }
1145 else if (f instanceof TreeBin) {
1146 validated = true;
1147 TreeBin<K,V> t = (TreeBin<K,V>)f;
1148 TreeNode<K,V> r, p;
1149 if ((r = t.root) != null &&
1150 (p = r.findTreeNode(hash, key, null)) != null) {
1151 V pv = p.val;
1152 if (cv == null || cv == pv ||
1153 (pv != null && cv.equals(pv))) {
1154 oldVal = pv;
1155 if (value != null)
1156 p.val = value;
1157 else if (t.removeTreeNode(p))
1158 setTabAt(tab, i, untreeify(t.first));
1159 }
1160 }
1161 }
1162 }
1163 }
1164 if (validated) {
1165 if (oldVal != null) {
1166 if (value == null)
1167 addCount(-1L, -1);
1168 return oldVal;
1169 }
1170 break;
1171 }
1172 }
1173 }
1174 return null;
1175 }
1176
1177 /**
1178 * Removes all of the mappings from this map.
1179 */
1180 public void clear() {
1181 long delta = 0L; // negative number of deletions
1182 int i = 0;
1183 Node<K,V>[] tab = table;
1184 while (tab != null && i < tab.length) {
1185 int fh;
1186 Node<K,V> f = tabAt(tab, i);
1187 if (f == null)
1188 ++i;
1189 else if ((fh = f.hash) == MOVED) {
1190 tab = helpTransfer(tab, f);
1191 i = 0; // restart
1192 }
1193 else {
1194 synchronized (f) {
1195 if (tabAt(tab, i) == f) {
1196 Node<K,V> p = (fh >= 0 ? f :
1197 (f instanceof TreeBin) ?
1198 ((TreeBin<K,V>)f).first : null);
1199 while (p != null) {
1200 --delta;
1201 p = p.next;
1202 }
1203 setTabAt(tab, i++, null);
1204 }
1205 }
1206 }
1207 }
1208 if (delta != 0L)
1209 addCount(delta, -1);
1210 }
1211
1212 /**
1213 * Returns a {@link Set} view of the keys contained in this map.
1214 * The set is backed by the map, so changes to the map are
1215 * reflected in the set, and vice-versa. The set supports element
1216 * removal, which removes the corresponding mapping from this map,
1217 * via the {@code Iterator.remove}, {@code Set.remove},
1218 * {@code removeAll}, {@code retainAll}, and {@code clear}
1219 * operations. It does not support the {@code add} or
1220 * {@code addAll} operations.
1221 *
1222 * <p>The view's iterators and spliterators are
1223 * <a href="package-summary.html#Weakly"><i>weakly consistent</i></a>.
1224 *
1225 * <p>The view's {@code spliterator} reports {@link Spliterator#CONCURRENT},
1226 * {@link Spliterator#DISTINCT}, and {@link Spliterator#NONNULL}.
1227 *
1228 * @return the set view
1229 */
1230 public KeySetView<K,V> keySet() {
1231 KeySetView<K,V> ks;
1232 return (ks = keySet) != null ? ks : (keySet = new KeySetView<K,V>(this, null));
1233 }
1234
1235 /**
1236 * Returns a {@link Collection} view of the values contained in this map.
1237 * The collection is backed by the map, so changes to the map are
1238 * reflected in the collection, and vice-versa. The collection
1239 * supports element removal, which removes the corresponding
1240 * mapping from this map, via the {@code Iterator.remove},
1241 * {@code Collection.remove}, {@code removeAll},
1242 * {@code retainAll}, and {@code clear} operations. It does not
1243 * support the {@code add} or {@code addAll} operations.
1244 *
1245 * <p>The view's iterators and spliterators are
1246 * <a href="package-summary.html#Weakly"><i>weakly consistent</i></a>.
1247 *
1248 * <p>The view's {@code spliterator} reports {@link Spliterator#CONCURRENT}
1249 * and {@link Spliterator#NONNULL}.
1250 *
1251 * @return the collection view
1252 */
1253 public Collection<V> values() {
1254 ValuesView<K,V> vs;
1255 return (vs = values) != null ? vs : (values = new ValuesView<K,V>(this));
1256 }
1257
1258 /**
1259 * Returns a {@link Set} view of the mappings contained in this map.
1260 * The set is backed by the map, so changes to the map are
1261 * reflected in the set, and vice-versa. The set supports element
1262 * removal, which removes the corresponding mapping from the map,
1263 * via the {@code Iterator.remove}, {@code Set.remove},
1264 * {@code removeAll}, {@code retainAll}, and {@code clear}
1265 * operations.
1266 *
1267 * <p>The view's iterators and spliterators are
1268 * <a href="package-summary.html#Weakly"><i>weakly consistent</i></a>.
1269 *
1270 * <p>The view's {@code spliterator} reports {@link Spliterator#CONCURRENT},
1271 * {@link Spliterator#DISTINCT}, and {@link Spliterator#NONNULL}.
1272 *
1273 * @return the set view
1274 */
1275 public Set<Map.Entry<K,V>> entrySet() {
1276 EntrySetView<K,V> es;
1277 return (es = entrySet) != null ? es : (entrySet = new EntrySetView<K,V>(this));
1278 }
1279
1280 /**
1281 * Returns the hash code value for this {@link Map}, i.e.,
1282 * the sum of, for each key-value pair in the map,
1283 * {@code key.hashCode() ^ value.hashCode()}.
1284 *
1285 * @return the hash code value for this map
1286 */
1287 public int hashCode() {
1288 int h = 0;
1289 Node<K,V>[] t;
1290 if ((t = table) != null) {
1291 Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length);
1292 for (Node<K,V> p; (p = it.advance()) != null; )
1293 h += p.key.hashCode() ^ p.val.hashCode();
1294 }
1295 return h;
1296 }
1297
1298 /**
1299 * Returns a string representation of this map. The string
1300 * representation consists of a list of key-value mappings (in no
1301 * particular order) enclosed in braces ("{@code {}}"). Adjacent
1302 * mappings are separated by the characters {@code ", "} (comma
1303 * and space). Each key-value mapping is rendered as the key
1304 * followed by an equals sign ("{@code =}") followed by the
1305 * associated value.
1306 *
1307 * @return a string representation of this map
1308 */
1309 public String toString() {
1310 Node<K,V>[] t;
1311 int f = (t = table) == null ? 0 : t.length;
1312 Traverser<K,V> it = new Traverser<K,V>(t, f, 0, f);
1313 StringBuilder sb = new StringBuilder();
1314 sb.append('{');
1315 Node<K,V> p;
1316 if ((p = it.advance()) != null) {
1317 for (;;) {
1318 K k = p.key;
1319 V v = p.val;
1320 sb.append(k == this ? "(this Map)" : k);
1321 sb.append('=');
1322 sb.append(v == this ? "(this Map)" : v);
1323 if ((p = it.advance()) == null)
1324 break;
1325 sb.append(',').append(' ');
1326 }
1327 }
1328 return sb.append('}').toString();
1329 }
1330
1331 /**
1332 * Compares the specified object with this map for equality.
1333 * Returns {@code true} if the given object is a map with the same
1334 * mappings as this map. This operation may return misleading
1335 * results if either map is concurrently modified during execution
1336 * of this method.
1337 *
1338 * @param o object to be compared for equality with this map
1339 * @return {@code true} if the specified object is equal to this map
1340 */
1341 public boolean equals(Object o) {
1342 if (o != this) {
1343 if (!(o instanceof Map))
1344 return false;
1345 Map<?,?> m = (Map<?,?>) o;
1346 Node<K,V>[] t;
1347 int f = (t = table) == null ? 0 : t.length;
1348 Traverser<K,V> it = new Traverser<K,V>(t, f, 0, f);
1349 for (Node<K,V> p; (p = it.advance()) != null; ) {
1350 V val = p.val;
1351 Object v = m.get(p.key);
1352 if (v == null || (v != val && !v.equals(val)))
1353 return false;
1354 }
1355 for (Map.Entry<?,?> e : m.entrySet()) {
1356 Object mk, mv, v;
1357 if ((mk = e.getKey()) == null ||
1358 (mv = e.getValue()) == null ||
1359 (v = get(mk)) == null ||
1360 (mv != v && !mv.equals(v)))
1361 return false;
1362 }
1363 }
1364 return true;
1365 }
1366
1367 /**
1368 * Stripped-down version of helper class used in previous version,
1369 * declared for the sake of serialization compatibility
1370 */
1371 static class Segment<K,V> extends ReentrantLock implements Serializable {
1372 private static final long serialVersionUID = 2249069246763182397L;
1373 final float loadFactor;
1374 Segment(float lf) { this.loadFactor = lf; }
1375 }
1376
1377 /**
1378 * Saves the state of the {@code ConcurrentHashMap} instance to a
1379 * stream (i.e., serializes it).
1380 * @param s the stream
1381 * @throws java.io.IOException if an I/O error occurs
1382 * @serialData
1383 * the key (Object) and value (Object)
1384 * for each key-value mapping, followed by a null pair.
1385 * The key-value mappings are emitted in no particular order.
1386 */
1387 private void writeObject(java.io.ObjectOutputStream s)
1388 throws java.io.IOException {
1389 // For serialization compatibility
1390 // Emulate segment calculation from previous version of this class
1391 int sshift = 0;
1392 int ssize = 1;
1393 while (ssize < DEFAULT_CONCURRENCY_LEVEL) {
1394 ++sshift;
1395 ssize <<= 1;
1396 }
1397 int segmentShift = 32 - sshift;
1398 int segmentMask = ssize - 1;
1399 @SuppressWarnings("unchecked")
1400 Segment<K,V>[] segments = (Segment<K,V>[])
1401 new Segment<?,?>[DEFAULT_CONCURRENCY_LEVEL];
1402 for (int i = 0; i < segments.length; ++i)
1403 segments[i] = new Segment<K,V>(LOAD_FACTOR);
1404 s.putFields().put("segments", segments);
1405 s.putFields().put("segmentShift", segmentShift);
1406 s.putFields().put("segmentMask", segmentMask);
1407 s.writeFields();
1408
1409 Node<K,V>[] t;
1410 if ((t = table) != null) {
1411 Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length);
1412 for (Node<K,V> p; (p = it.advance()) != null; ) {
1413 s.writeObject(p.key);
1414 s.writeObject(p.val);
1415 }
1416 }
1417 s.writeObject(null);
1418 s.writeObject(null);
1419 segments = null; // throw away
1420 }
1421
1422 /**
1423 * Reconstitutes the instance from a stream (that is, deserializes it).
1424 * @param s the stream
1425 * @throws ClassNotFoundException if the class of a serialized object
1426 * could not be found
1427 * @throws java.io.IOException if an I/O error occurs
1428 */
1429 private void readObject(java.io.ObjectInputStream s)
1430 throws java.io.IOException, ClassNotFoundException {
1431 /*
1432 * To improve performance in typical cases, we create nodes
1433 * while reading, then place in table once size is known.
1434 * However, we must also validate uniqueness and deal with
1435 * overpopulated bins while doing so, which requires
1436 * specialized versions of putVal mechanics.
1437 */
1438 sizeCtl = -1; // force exclusion for table construction
1439 s.defaultReadObject();
1440 long size = 0L;
1441 Node<K,V> p = null;
1442 for (;;) {
1443 @SuppressWarnings("unchecked")
1444 K k = (K) s.readObject();
1445 @SuppressWarnings("unchecked")
1446 V v = (V) s.readObject();
1447 if (k != null && v != null) {
1448 p = new Node<K,V>(spread(k.hashCode()), k, v, p);
1449 ++size;
1450 }
1451 else
1452 break;
1453 }
1454 if (size == 0L)
1455 sizeCtl = 0;
1456 else {
1457 int n;
1458 if (size >= (long)(MAXIMUM_CAPACITY >>> 1))
1459 n = MAXIMUM_CAPACITY;
1460 else {
1461 int sz = (int)size;
1462 n = tableSizeFor(sz + (sz >>> 1) + 1);
1463 }
1464 @SuppressWarnings("unchecked")
1465 Node<K,V>[] tab = (Node<K,V>[])new Node<?,?>[n];
1466 int mask = n - 1;
1467 long added = 0L;
1468 while (p != null) {
1469 boolean insertAtFront;
1470 Node<K,V> next = p.next, first;
1471 int h = p.hash, j = h & mask;
1472 if ((first = tabAt(tab, j)) == null)
1473 insertAtFront = true;
1474 else {
1475 K k = p.key;
1476 if (first.hash < 0) {
1477 TreeBin<K,V> t = (TreeBin<K,V>)first;
1478 if (t.putTreeVal(h, k, p.val) == null)
1479 ++added;
1480 insertAtFront = false;
1481 }
1482 else {
1483 int binCount = 0;
1484 insertAtFront = true;
1485 Node<K,V> q; K qk;
1486 for (q = first; q != null; q = q.next) {
1487 if (q.hash == h &&
1488 ((qk = q.key) == k ||
1489 (qk != null && k.equals(qk)))) {
1490 insertAtFront = false;
1491 break;
1492 }
1493 ++binCount;
1494 }
1495 if (insertAtFront && binCount >= TREEIFY_THRESHOLD) {
1496 insertAtFront = false;
1497 ++added;
1498 p.next = first;
1499 TreeNode<K,V> hd = null, tl = null;
1500 for (q = p; q != null; q = q.next) {
1501 TreeNode<K,V> t = new TreeNode<K,V>
1502 (q.hash, q.key, q.val, null, null);
1503 if ((t.prev = tl) == null)
1504 hd = t;
1505 else
1506 tl.next = t;
1507 tl = t;
1508 }
1509 setTabAt(tab, j, new TreeBin<K,V>(hd));
1510 }
1511 }
1512 }
1513 if (insertAtFront) {
1514 ++added;
1515 p.next = first;
1516 setTabAt(tab, j, p);
1517 }
1518 p = next;
1519 }
1520 table = tab;
1521 sizeCtl = n - (n >>> 2);
1522 baseCount = added;
1523 }
1524 }
1525
1526 // ConcurrentMap methods
1527
1528 /**
1529 * {@inheritDoc}
1530 *
1531 * @return the previous value associated with the specified key,
1532 * or {@code null} if there was no mapping for the key
1533 * @throws NullPointerException if the specified key or value is null
1534 */
1535 public V putIfAbsent(K key, V value) {
1536 return putVal(key, value, true);
1537 }
1538
1539 /**
1540 * {@inheritDoc}
1541 *
1542 * @throws NullPointerException if the specified key is null
1543 */
1544 public boolean remove(Object key, Object value) {
1545 if (key == null)
1546 throw new NullPointerException();
1547 return value != null && replaceNode(key, null, value) != null;
1548 }
1549
1550 /**
1551 * {@inheritDoc}
1552 *
1553 * @throws NullPointerException if any of the arguments are null
1554 */
1555 public boolean replace(K key, V oldValue, V newValue) {
1556 if (key == null || oldValue == null || newValue == null)
1557 throw new NullPointerException();
1558 return replaceNode(key, newValue, oldValue) != null;
1559 }
1560
1561 /**
1562 * {@inheritDoc}
1563 *
1564 * @return the previous value associated with the specified key,
1565 * or {@code null} if there was no mapping for the key
1566 * @throws NullPointerException if the specified key or value is null
1567 */
1568 public V replace(K key, V value) {
1569 if (key == null || value == null)
1570 throw new NullPointerException();
1571 return replaceNode(key, value, null);
1572 }
1573
1574 // Overrides of JDK8+ Map extension method defaults
1575
1576 /**
1577 * Returns the value to which the specified key is mapped, or the
1578 * given default value if this map contains no mapping for the
1579 * key.
1580 *
1581 * @param key the key whose associated value is to be returned
1582 * @param defaultValue the value to return if this map contains
1583 * no mapping for the given key
1584 * @return the mapping for the key, if present; else the default value
1585 * @throws NullPointerException if the specified key is null
1586 */
1587 public V getOrDefault(Object key, V defaultValue) {
1588 V v;
1589 return (v = get(key)) == null ? defaultValue : v;
1590 }
1591
1592 public void forEach(BiConsumer<? super K, ? super V> action) {
1593 if (action == null) throw new NullPointerException();
1594 Node<K,V>[] t;
1595 if ((t = table) != null) {
1596 Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length);
1597 for (Node<K,V> p; (p = it.advance()) != null; ) {
1598 action.accept(p.key, p.val);
1599 }
1600 }
1601 }
1602
1603 public void replaceAll(BiFunction<? super K, ? super V, ? extends V> function) {
1604 if (function == null) throw new NullPointerException();
1605 Node<K,V>[] t;
1606 if ((t = table) != null) {
1607 Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length);
1608 for (Node<K,V> p; (p = it.advance()) != null; ) {
1609 V oldValue = p.val;
1610 for (K key = p.key;;) {
1611 V newValue = function.apply(key, oldValue);
1612 if (newValue == null)
1613 throw new NullPointerException();
1614 if (replaceNode(key, newValue, oldValue) != null ||
1615 (oldValue = get(key)) == null)
1616 break;
1617 }
1618 }
1619 }
1620 }
1621
1622 /**
1623 * Helper method for EntrySet.removeIf
1624 */
1625 boolean removeEntryIf(Predicate<? super Entry<K, V>> function) {
1626 if (function == null) throw new NullPointerException();
1627 Node<K,V>[] t;
1628 boolean removed = false;
1629 if ((t = table) != null) {
1630 Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length);
1631 for (Node<K,V> p; (p = it.advance()) != null; ) {
1632 K k = p.key;
1633 V v = p.val;
1634 Map.Entry<K,V> e = new AbstractMap.SimpleImmutableEntry<>(k, v);
1635 if (function.test(e) && replaceNode(k, null, v) != null)
1636 removed = true;
1637 }
1638 }
1639 return removed;
1640 }
1641
1642 /**
1643 * Helper method for Values.removeIf
1644 */
1645 boolean removeValueIf(Predicate<? super V> function) {
1646 if (function == null) throw new NullPointerException();
1647 Node<K,V>[] t;
1648 boolean removed = false;
1649 if ((t = table) != null) {
1650 Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length);
1651 for (Node<K,V> p; (p = it.advance()) != null; ) {
1652 K k = p.key;
1653 V v = p.val;
1654 if (function.test(v) && replaceNode(k, null, v) != null)
1655 removed = true;
1656 }
1657 }
1658 return removed;
1659 }
1660
1661 /**
1662 * If the specified key is not already associated with a value,
1663 * attempts to compute its value using the given mapping function
1664 * and enters it into this map unless {@code null}. The entire
1665 * method invocation is performed atomically, so the function is
1666 * applied at most once per key. Some attempted update operations
1667 * on this map by other threads may be blocked while computation
1668 * is in progress, so the computation should be short and simple,
1669 * and must not attempt to update any other mappings of this map.
1670 *
1671 * @param key key with which the specified value is to be associated
1672 * @param mappingFunction the function to compute a value
1673 * @return the current (existing or computed) value associated with
1674 * the specified key, or null if the computed value is null
1675 * @throws NullPointerException if the specified key or mappingFunction
1676 * is null
1677 * @throws IllegalStateException if the computation detectably
1678 * attempts a recursive update to this map that would
1679 * otherwise never complete
1680 * @throws RuntimeException or Error if the mappingFunction does so,
1681 * in which case the mapping is left unestablished
1682 */
1683 public V computeIfAbsent(K key, Function<? super K, ? extends V> mappingFunction) {
1684 if (key == null || mappingFunction == null)
1685 throw new NullPointerException();
1686 int h = spread(key.hashCode());
1687 V val = null;
1688 int binCount = 0;
1689 for (Node<K,V>[] tab = table;;) {
1690 Node<K,V> f; int n, i, fh;
1691 if (tab == null || (n = tab.length) == 0)
1692 tab = initTable();
1693 else if ((f = tabAt(tab, i = (n - 1) & h)) == null) {
1694 Node<K,V> r = new ReservationNode<K,V>();
1695 synchronized (r) {
1696 if (casTabAt(tab, i, null, r)) {
1697 binCount = 1;
1698 Node<K,V> node = null;
1699 try {
1700 if ((val = mappingFunction.apply(key)) != null)
1701 node = new Node<K,V>(h, key, val, null);
1702 } finally {
1703 setTabAt(tab, i, node);
1704 }
1705 }
1706 }
1707 if (binCount != 0)
1708 break;
1709 }
1710 else if ((fh = f.hash) == MOVED)
1711 tab = helpTransfer(tab, f);
1712 else {
1713 boolean added = false;
1714 synchronized (f) {
1715 if (tabAt(tab, i) == f) {
1716 if (fh >= 0) {
1717 binCount = 1;
1718 for (Node<K,V> e = f;; ++binCount) {
1719 K ek; V ev;
1720 if (e.hash == h &&
1721 ((ek = e.key) == key ||
1722 (ek != null && key.equals(ek)))) {
1723 val = e.val;
1724 break;
1725 }
1726 Node<K,V> pred = e;
1727 if ((e = e.next) == null) {
1728 if ((val = mappingFunction.apply(key)) != null) {
1729 added = true;
1730 pred.next = new Node<K,V>(h, key, val, null);
1731 }
1732 break;
1733 }
1734 }
1735 }
1736 else if (f instanceof TreeBin) {
1737 binCount = 2;
1738 TreeBin<K,V> t = (TreeBin<K,V>)f;
1739 TreeNode<K,V> r, p;
1740 if ((r = t.root) != null &&
1741 (p = r.findTreeNode(h, key, null)) != null)
1742 val = p.val;
1743 else if ((val = mappingFunction.apply(key)) != null) {
1744 added = true;
1745 t.putTreeVal(h, key, val);
1746 }
1747 }
1748 }
1749 }
1750 if (binCount != 0) {
1751 if (binCount >= TREEIFY_THRESHOLD)
1752 treeifyBin(tab, i);
1753 if (!added)
1754 return val;
1755 break;
1756 }
1757 }
1758 }
1759 if (val != null)
1760 addCount(1L, binCount);
1761 return val;
1762 }
1763
1764 /**
1765 * If the value for the specified key is present, attempts to
1766 * compute a new mapping given the key and its current mapped
1767 * value. The entire method invocation is performed atomically.
1768 * Some attempted update operations on this map by other threads
1769 * may be blocked while computation is in progress, so the
1770 * computation should be short and simple, and must not attempt to
1771 * update any other mappings of this map.
1772 *
1773 * @param key key with which a value may be associated
1774 * @param remappingFunction the function to compute a value
1775 * @return the new value associated with the specified key, or null if none
1776 * @throws NullPointerException if the specified key or remappingFunction
1777 * is null
1778 * @throws IllegalStateException if the computation detectably
1779 * attempts a recursive update to this map that would
1780 * otherwise never complete
1781 * @throws RuntimeException or Error if the remappingFunction does so,
1782 * in which case the mapping is unchanged
1783 */
1784 public V computeIfPresent(K key, BiFunction<? super K, ? super V, ? extends V> remappingFunction) {
1785 if (key == null || remappingFunction == null)
1786 throw new NullPointerException();
1787 int h = spread(key.hashCode());
1788 V val = null;
1789 int delta = 0;
1790 int binCount = 0;
1791 for (Node<K,V>[] tab = table;;) {
1792 Node<K,V> f; int n, i, fh;
1793 if (tab == null || (n = tab.length) == 0)
1794 tab = initTable();
1795 else if ((f = tabAt(tab, i = (n - 1) & h)) == null)
1796 break;
1797 else if ((fh = f.hash) == MOVED)
1798 tab = helpTransfer(tab, f);
1799 else {
1800 synchronized (f) {
1801 if (tabAt(tab, i) == f) {
1802 if (fh >= 0) {
1803 binCount = 1;
1804 for (Node<K,V> e = f, pred = null;; ++binCount) {
1805 K ek;
1806 if (e.hash == h &&
1807 ((ek = e.key) == key ||
1808 (ek != null && key.equals(ek)))) {
1809 val = remappingFunction.apply(key, e.val);
1810 if (val != null)
1811 e.val = val;
1812 else {
1813 delta = -1;
1814 Node<K,V> en = e.next;
1815 if (pred != null)
1816 pred.next = en;
1817 else
1818 setTabAt(tab, i, en);
1819 }
1820 break;
1821 }
1822 pred = e;
1823 if ((e = e.next) == null)
1824 break;
1825 }
1826 }
1827 else if (f instanceof TreeBin) {
1828 binCount = 2;
1829 TreeBin<K,V> t = (TreeBin<K,V>)f;
1830 TreeNode<K,V> r, p;
1831 if ((r = t.root) != null &&
1832 (p = r.findTreeNode(h, key, null)) != null) {
1833 val = remappingFunction.apply(key, p.val);
1834 if (val != null)
1835 p.val = val;
1836 else {
1837 delta = -1;
1838 if (t.removeTreeNode(p))
1839 setTabAt(tab, i, untreeify(t.first));
1840 }
1841 }
1842 }
1843 }
1844 }
1845 if (binCount != 0)
1846 break;
1847 }
1848 }
1849 if (delta != 0)
1850 addCount((long)delta, binCount);
1851 return val;
1852 }
1853
1854 /**
1855 * Attempts to compute a mapping for the specified key and its
1856 * current mapped value (or {@code null} if there is no current
1857 * mapping). The entire method invocation is performed atomically.
1858 * Some attempted update operations on this map by other threads
1859 * may be blocked while computation is in progress, so the
1860 * computation should be short and simple, and must not attempt to
1861 * update any other mappings of this Map.
1862 *
1863 * @param key key with which the specified value is to be associated
1864 * @param remappingFunction the function to compute a value
1865 * @return the new value associated with the specified key, or null if none
1866 * @throws NullPointerException if the specified key or remappingFunction
1867 * is null
1868 * @throws IllegalStateException if the computation detectably
1869 * attempts a recursive update to this map that would
1870 * otherwise never complete
1871 * @throws RuntimeException or Error if the remappingFunction does so,
1872 * in which case the mapping is unchanged
1873 */
1874 public V compute(K key,
1875 BiFunction<? super K, ? super V, ? extends V> remappingFunction) {
1876 if (key == null || remappingFunction == null)
1877 throw new NullPointerException();
1878 int h = spread(key.hashCode());
1879 V val = null;
1880 int delta = 0;
1881 int binCount = 0;
1882 for (Node<K,V>[] tab = table;;) {
1883 Node<K,V> f; int n, i, fh;
1884 if (tab == null || (n = tab.length) == 0)
1885 tab = initTable();
1886 else if ((f = tabAt(tab, i = (n - 1) & h)) == null) {
1887 Node<K,V> r = new ReservationNode<K,V>();
1888 synchronized (r) {
1889 if (casTabAt(tab, i, null, r)) {
1890 binCount = 1;
1891 Node<K,V> node = null;
1892 try {
1893 if ((val = remappingFunction.apply(key, null)) != null) {
1894 delta = 1;
1895 node = new Node<K,V>(h, key, val, null);
1896 }
1897 } finally {
1898 setTabAt(tab, i, node);
1899 }
1900 }
1901 }
1902 if (binCount != 0)
1903 break;
1904 }
1905 else if ((fh = f.hash) == MOVED)
1906 tab = helpTransfer(tab, f);
1907 else {
1908 synchronized (f) {
1909 if (tabAt(tab, i) == f) {
1910 if (fh >= 0) {
1911 binCount = 1;
1912 for (Node<K,V> e = f, pred = null;; ++binCount) {
1913 K ek;
1914 if (e.hash == h &&
1915 ((ek = e.key) == key ||
1916 (ek != null && key.equals(ek)))) {
1917 val = remappingFunction.apply(key, e.val);
1918 if (val != null)
1919 e.val = val;
1920 else {
1921 delta = -1;
1922 Node<K,V> en = e.next;
1923 if (pred != null)
1924 pred.next = en;
1925 else
1926 setTabAt(tab, i, en);
1927 }
1928 break;
1929 }
1930 pred = e;
1931 if ((e = e.next) == null) {
1932 val = remappingFunction.apply(key, null);
1933 if (val != null) {
1934 delta = 1;
1935 pred.next =
1936 new Node<K,V>(h, key, val, null);
1937 }
1938 break;
1939 }
1940 }
1941 }
1942 else if (f instanceof TreeBin) {
1943 binCount = 1;
1944 TreeBin<K,V> t = (TreeBin<K,V>)f;
1945 TreeNode<K,V> r, p;
1946 if ((r = t.root) != null)
1947 p = r.findTreeNode(h, key, null);
1948 else
1949 p = null;
1950 V pv = (p == null) ? null : p.val;
1951 val = remappingFunction.apply(key, pv);
1952 if (val != null) {
1953 if (p != null)
1954 p.val = val;
1955 else {
1956 delta = 1;
1957 t.putTreeVal(h, key, val);
1958 }
1959 }
1960 else if (p != null) {
1961 delta = -1;
1962 if (t.removeTreeNode(p))
1963 setTabAt(tab, i, untreeify(t.first));
1964 }
1965 }
1966 }
1967 }
1968 if (binCount != 0) {
1969 if (binCount >= TREEIFY_THRESHOLD)
1970 treeifyBin(tab, i);
1971 break;
1972 }
1973 }
1974 }
1975 if (delta != 0)
1976 addCount((long)delta, binCount);
1977 return val;
1978 }
1979
1980 /**
1981 * If the specified key is not already associated with a
1982 * (non-null) value, associates it with the given value.
1983 * Otherwise, replaces the value with the results of the given
1984 * remapping function, or removes if {@code null}. The entire
1985 * method invocation is performed atomically. Some attempted
1986 * update operations on this map by other threads may be blocked
1987 * while computation is in progress, so the computation should be
1988 * short and simple, and must not attempt to update any other
1989 * mappings of this Map.
1990 *
1991 * @param key key with which the specified value is to be associated
1992 * @param value the value to use if absent
1993 * @param remappingFunction the function to recompute a value if present
1994 * @return the new value associated with the specified key, or null if none
1995 * @throws NullPointerException if the specified key or the
1996 * remappingFunction is null
1997 * @throws RuntimeException or Error if the remappingFunction does so,
1998 * in which case the mapping is unchanged
1999 */
2000 public V merge(K key, V value, BiFunction<? super V, ? super V, ? extends V> remappingFunction) {
2001 if (key == null || value == null || remappingFunction == null)
2002 throw new NullPointerException();
2003 int h = spread(key.hashCode());
2004 V val = null;
2005 int delta = 0;
2006 int binCount = 0;
2007 for (Node<K,V>[] tab = table;;) {
2008 Node<K,V> f; int n, i, fh;
2009 if (tab == null || (n = tab.length) == 0)
2010 tab = initTable();
2011 else if ((f = tabAt(tab, i = (n - 1) & h)) == null) {
2012 if (casTabAt(tab, i, null, new Node<K,V>(h, key, value, null))) {
2013 delta = 1;
2014 val = value;
2015 break;
2016 }
2017 }
2018 else if ((fh = f.hash) == MOVED)
2019 tab = helpTransfer(tab, f);
2020 else {
2021 synchronized (f) {
2022 if (tabAt(tab, i) == f) {
2023 if (fh >= 0) {
2024 binCount = 1;
2025 for (Node<K,V> e = f, pred = null;; ++binCount) {
2026 K ek;
2027 if (e.hash == h &&
2028 ((ek = e.key) == key ||
2029 (ek != null && key.equals(ek)))) {
2030 val = remappingFunction.apply(e.val, value);
2031 if (val != null)
2032 e.val = val;
2033 else {
2034 delta = -1;
2035 Node<K,V> en = e.next;
2036 if (pred != null)
2037 pred.next = en;
2038 else
2039 setTabAt(tab, i, en);
2040 }
2041 break;
2042 }
2043 pred = e;
2044 if ((e = e.next) == null) {
2045 delta = 1;
2046 val = value;
2047 pred.next =
2048 new Node<K,V>(h, key, val, null);
2049 break;
2050 }
2051 }
2052 }
2053 else if (f instanceof TreeBin) {
2054 binCount = 2;
2055 TreeBin<K,V> t = (TreeBin<K,V>)f;
2056 TreeNode<K,V> r = t.root;
2057 TreeNode<K,V> p = (r == null) ? null :
2058 r.findTreeNode(h, key, null);
2059 val = (p == null) ? value :
2060 remappingFunction.apply(p.val, value);
2061 if (val != null) {
2062 if (p != null)
2063 p.val = val;
2064 else {
2065 delta = 1;
2066 t.putTreeVal(h, key, val);
2067 }
2068 }
2069 else if (p != null) {
2070 delta = -1;
2071 if (t.removeTreeNode(p))
2072 setTabAt(tab, i, untreeify(t.first));
2073 }
2074 }
2075 }
2076 }
2077 if (binCount != 0) {
2078 if (binCount >= TREEIFY_THRESHOLD)
2079 treeifyBin(tab, i);
2080 break;
2081 }
2082 }
2083 }
2084 if (delta != 0)
2085 addCount((long)delta, binCount);
2086 return val;
2087 }
2088
2089 // Hashtable legacy methods
2090
2091 /**
2092 * Legacy method testing if some key maps into the specified value
2093 * in this table. This method is identical in functionality to
2094 * {@link #containsValue(Object)}, and exists solely to ensure
2095 * full compatibility with class {@link java.util.Hashtable},
2096 * which supported this method prior to introduction of the
2097 * Java Collections framework.
2098 *
2099 * @param value a value to search for
2100 * @return {@code true} if and only if some key maps to the
2101 * {@code value} argument in this table as
2102 * determined by the {@code equals} method;
2103 * {@code false} otherwise
2104 * @throws NullPointerException if the specified value is null
2105 */
2106 public boolean contains(Object value) {
2107 return containsValue(value);
2108 }
2109
2110 /**
2111 * Returns an enumeration of the keys in this table.
2112 *
2113 * @return an enumeration of the keys in this table
2114 * @see #keySet()
2115 */
2116 public Enumeration<K> keys() {
2117 Node<K,V>[] t;
2118 int f = (t = table) == null ? 0 : t.length;
2119 return new KeyIterator<K,V>(t, f, 0, f, this);
2120 }
2121
2122 /**
2123 * Returns an enumeration of the values in this table.
2124 *
2125 * @return an enumeration of the values in this table
2126 * @see #values()
2127 */
2128 public Enumeration<V> elements() {
2129 Node<K,V>[] t;
2130 int f = (t = table) == null ? 0 : t.length;
2131 return new ValueIterator<K,V>(t, f, 0, f, this);
2132 }
2133
2134 // ConcurrentHashMap-only methods
2135
2136 /**
2137 * Returns the number of mappings. This method should be used
2138 * instead of {@link #size} because a ConcurrentHashMap may
2139 * contain more mappings than can be represented as an int. The
2140 * value returned is an estimate; the actual count may differ if
2141 * there are concurrent insertions or removals.
2142 *
2143 * @return the number of mappings
2144 * @since 1.8
2145 */
2146 public long mappingCount() {
2147 long n = sumCount();
2148 return (n < 0L) ? 0L : n; // ignore transient negative values
2149 }
2150
2151 /**
2152 * Creates a new {@link Set} backed by a ConcurrentHashMap
2153 * from the given type to {@code Boolean.TRUE}.
2154 *
2155 * @param <K> the element type of the returned set
2156 * @return the new set
2157 * @since 1.8
2158 */
2159 public static <K> KeySetView<K,Boolean> newKeySet() {
2160 return new KeySetView<K,Boolean>
2161 (new ConcurrentHashMap<K,Boolean>(), Boolean.TRUE);
2162 }
2163
2164 /**
2165 * Creates a new {@link Set} backed by a ConcurrentHashMap
2166 * from the given type to {@code Boolean.TRUE}.
2167 *
2168 * @param initialCapacity The implementation performs internal
2169 * sizing to accommodate this many elements.
2170 * @param <K> the element type of the returned set
2171 * @return the new set
2172 * @throws IllegalArgumentException if the initial capacity of
2173 * elements is negative
2174 * @since 1.8
2175 */
2176 public static <K> KeySetView<K,Boolean> newKeySet(int initialCapacity) {
2177 return new KeySetView<K,Boolean>
2178 (new ConcurrentHashMap<K,Boolean>(initialCapacity), Boolean.TRUE);
2179 }
2180
2181 /**
2182 * Returns a {@link Set} view of the keys in this map, using the
2183 * given common mapped value for any additions (i.e., {@link
2184 * Collection#add} and {@link Collection#addAll(Collection)}).
2185 * This is of course only appropriate if it is acceptable to use
2186 * the same value for all additions from this view.
2187 *
2188 * @param mappedValue the mapped value to use for any additions
2189 * @return the set view
2190 * @throws NullPointerException if the mappedValue is null
2191 */
2192 public KeySetView<K,V> keySet(V mappedValue) {
2193 if (mappedValue == null)
2194 throw new NullPointerException();
2195 return new KeySetView<K,V>(this, mappedValue);
2196 }
2197
2198 /* ---------------- Special Nodes -------------- */
2199
2200 /**
2201 * A node inserted at head of bins during transfer operations.
2202 */
2203 static final class ForwardingNode<K,V> extends Node<K,V> {
2204 final Node<K,V>[] nextTable;
2205 ForwardingNode(Node<K,V>[] tab) {
2206 super(MOVED, null, null, null);
2207 this.nextTable = tab;
2208 }
2209
2210 Node<K,V> find(int h, Object k) {
2211 // loop to avoid arbitrarily deep recursion on forwarding nodes
2212 outer: for (Node<K,V>[] tab = nextTable;;) {
2213 Node<K,V> e; int n;
2214 if (k == null || tab == null || (n = tab.length) == 0 ||
2215 (e = tabAt(tab, (n - 1) & h)) == null)
2216 return null;
2217 for (;;) {
2218 int eh; K ek;
2219 if ((eh = e.hash) == h &&
2220 ((ek = e.key) == k || (ek != null && k.equals(ek))))
2221 return e;
2222 if (eh < 0) {
2223 if (e instanceof ForwardingNode) {
2224 tab = ((ForwardingNode<K,V>)e).nextTable;
2225 continue outer;
2226 }
2227 else
2228 return e.find(h, k);
2229 }
2230 if ((e = e.next) == null)
2231 return null;
2232 }
2233 }
2234 }
2235 }
2236
2237 /**
2238 * A place-holder node used in computeIfAbsent and compute
2239 */
2240 static final class ReservationNode<K,V> extends Node<K,V> {
2241 ReservationNode() {
2242 super(RESERVED, null, null, null);
2243 }
2244
2245 Node<K,V> find(int h, Object k) {
2246 return null;
2247 }
2248 }
2249
2250 /* ---------------- Table Initialization and Resizing -------------- */
2251
2252 /**
2253 * Returns the stamp bits for resizing a table of size n.
2254 * Must be negative when shifted left by RESIZE_STAMP_SHIFT.
2255 */
2256 static final int resizeStamp(int n) {
2257 return Integer.numberOfLeadingZeros(n) | (1 << (RESIZE_STAMP_BITS - 1));
2258 }
2259
2260 /**
2261 * Initializes table, using the size recorded in sizeCtl.
2262 */
2263 private final Node<K,V>[] initTable() {
2264 Node<K,V>[] tab; int sc;
2265 while ((tab = table) == null || tab.length == 0) {
2266 if ((sc = sizeCtl) < 0)
2267 Thread.yield(); // lost initialization race; just spin
2268 else if (U.compareAndSwapInt(this, SIZECTL, sc, -1)) {
2269 try {
2270 if ((tab = table) == null || tab.length == 0) {
2271 int n = (sc > 0) ? sc : DEFAULT_CAPACITY;
2272 @SuppressWarnings("unchecked")
2273 Node<K,V>[] nt = (Node<K,V>[])new Node<?,?>[n];
2274 table = tab = nt;
2275 sc = n - (n >>> 2);
2276 }
2277 } finally {
2278 sizeCtl = sc;
2279 }
2280 break;
2281 }
2282 }
2283 return tab;
2284 }
2285
2286 /**
2287 * Adds to count, and if table is too small and not already
2288 * resizing, initiates transfer. If already resizing, helps
2289 * perform transfer if work is available. Rechecks occupancy
2290 * after a transfer to see if another resize is already needed
2291 * because resizings are lagging additions.
2292 *
2293 * @param x the count to add
2294 * @param check if <0, don't check resize, if <= 1 only check if uncontended
2295 */
2296 private final void addCount(long x, int check) {
2297 CounterCell[] as; long b, s;
2298 if ((as = counterCells) != null ||
2299 !U.compareAndSwapLong(this, BASECOUNT, b = baseCount, s = b + x)) {
2300 CounterCell a; long v; int m;
2301 boolean uncontended = true;
2302 if (as == null || (m = as.length - 1) < 0 ||
2303 (a = as[ThreadLocalRandom.getProbe() & m]) == null ||
2304 !(uncontended =
2305 U.compareAndSwapLong(a, CELLVALUE, v = a.value, v + x))) {
2306 fullAddCount(x, uncontended);
2307 return;
2308 }
2309 if (check <= 1)
2310 return;
2311 s = sumCount();
2312 }
2313 if (check >= 0) {
2314 Node<K,V>[] tab, nt; int n, sc;
2315 while (s >= (long)(sc = sizeCtl) && (tab = table) != null &&
2316 (n = tab.length) < MAXIMUM_CAPACITY) {
2317 int rs = resizeStamp(n);
2318 if (sc < 0) {
2319 if ((sc >>> RESIZE_STAMP_SHIFT) != rs || sc == rs + 1 ||
2320 sc == rs + MAX_RESIZERS || (nt = nextTable) == null ||
2321 transferIndex <= 0)
2322 break;
2323 if (U.compareAndSwapInt(this, SIZECTL, sc, sc + 1))
2324 transfer(tab, nt);
2325 }
2326 else if (U.compareAndSwapInt(this, SIZECTL, sc,
2327 (rs << RESIZE_STAMP_SHIFT) + 2))
2328 transfer(tab, null);
2329 s = sumCount();
2330 }
2331 }
2332 }
2333
2334 /**
2335 * Helps transfer if a resize is in progress.
2336 */
2337 final Node<K,V>[] helpTransfer(Node<K,V>[] tab, Node<K,V> f) {
2338 Node<K,V>[] nextTab; int sc;
2339 if (tab != null && (f instanceof ForwardingNode) &&
2340 (nextTab = ((ForwardingNode<K,V>)f).nextTable) != null) {
2341 int rs = resizeStamp(tab.length);
2342 while (nextTab == nextTable && table == tab &&
2343 (sc = sizeCtl) < 0) {
2344 if ((sc >>> RESIZE_STAMP_SHIFT) != rs || sc == rs + 1 ||
2345 sc == rs + MAX_RESIZERS || transferIndex <= 0)
2346 break;
2347 if (U.compareAndSwapInt(this, SIZECTL, sc, sc + 1)) {
2348 transfer(tab, nextTab);
2349 break;
2350 }
2351 }
2352 return nextTab;
2353 }
2354 return table;
2355 }
2356
2357 /**
2358 * Tries to presize table to accommodate the given number of elements.
2359 *
2360 * @param size number of elements (doesn't need to be perfectly accurate)
2361 */
2362 private final void tryPresize(int size) {
2363 int c = (size >= (MAXIMUM_CAPACITY >>> 1)) ? MAXIMUM_CAPACITY :
2364 tableSizeFor(size + (size >>> 1) + 1);
2365 int sc;
2366 while ((sc = sizeCtl) >= 0) {
2367 Node<K,V>[] tab = table; int n;
2368 if (tab == null || (n = tab.length) == 0) {
2369 n = (sc > c) ? sc : c;
2370 if (U.compareAndSwapInt(this, SIZECTL, sc, -1)) {
2371 try {
2372 if (table == tab) {
2373 @SuppressWarnings("unchecked")
2374 Node<K,V>[] nt = (Node<K,V>[])new Node<?,?>[n];
2375 table = nt;
2376 sc = n - (n >>> 2);
2377 }
2378 } finally {
2379 sizeCtl = sc;
2380 }
2381 }
2382 }
2383 else if (c <= sc || n >= MAXIMUM_CAPACITY)
2384 break;
2385 else if (tab == table) {
2386 int rs = resizeStamp(n);
2387 if (sc < 0) {
2388 Node<K,V>[] nt;
2389 if ((sc >>> RESIZE_STAMP_SHIFT) != rs || sc == rs + 1 ||
2390 sc == rs + MAX_RESIZERS || (nt = nextTable) == null ||
2391 transferIndex <= 0)
2392 break;
2393 if (U.compareAndSwapInt(this, SIZECTL, sc, sc + 1))
2394 transfer(tab, nt);
2395 }
2396 else if (U.compareAndSwapInt(this, SIZECTL, sc,
2397 (rs << RESIZE_STAMP_SHIFT) + 2))
2398 transfer(tab, null);
2399 }
2400 }
2401 }
2402
2403 /**
2404 * Moves and/or copies the nodes in each bin to new table. See
2405 * above for explanation.
2406 */
2407 private final void transfer(Node<K,V>[] tab, Node<K,V>[] nextTab) {
2408 int n = tab.length, stride;
2409 if ((stride = (NCPU > 1) ? (n >>> 3) / NCPU : n) < MIN_TRANSFER_STRIDE)
2410 stride = MIN_TRANSFER_STRIDE; // subdivide range
2411 if (nextTab == null) { // initiating
2412 try {
2413 @SuppressWarnings("unchecked")
2414 Node<K,V>[] nt = (Node<K,V>[])new Node<?,?>[n << 1];
2415 nextTab = nt;
2416 } catch (Throwable ex) { // try to cope with OOME
2417 sizeCtl = Integer.MAX_VALUE;
2418 return;
2419 }
2420 nextTable = nextTab;
2421 transferIndex = n;
2422 }
2423 int nextn = nextTab.length;
2424 ForwardingNode<K,V> fwd = new ForwardingNode<K,V>(nextTab);
2425 boolean advance = true;
2426 boolean finishing = false; // to ensure sweep before committing nextTab
2427 for (int i = 0, bound = 0;;) {
2428 Node<K,V> f; int fh;
2429 while (advance) {
2430 int nextIndex, nextBound;
2431 if (--i >= bound || finishing)
2432 advance = false;
2433 else if ((nextIndex = transferIndex) <= 0) {
2434 i = -1;
2435 advance = false;
2436 }
2437 else if (U.compareAndSwapInt
2438 (this, TRANSFERINDEX, nextIndex,
2439 nextBound = (nextIndex > stride ?
2440 nextIndex - stride : 0))) {
2441 bound = nextBound;
2442 i = nextIndex - 1;
2443 advance = false;
2444 }
2445 }
2446 if (i < 0 || i >= n || i + n >= nextn) {
2447 int sc;
2448 if (finishing) {
2449 nextTable = null;
2450 table = nextTab;
2451 sizeCtl = (n << 1) - (n >>> 1);
2452 return;
2453 }
2454 if (U.compareAndSwapInt(this, SIZECTL, sc = sizeCtl, sc - 1)) {
2455 if ((sc - 2) != resizeStamp(n) << RESIZE_STAMP_SHIFT)
2456 return;
2457 finishing = advance = true;
2458 i = n; // recheck before commit
2459 }
2460 }
2461 else if ((f = tabAt(tab, i)) == null)
2462 advance = casTabAt(tab, i, null, fwd);
2463 else if ((fh = f.hash) == MOVED)
2464 advance = true; // already processed
2465 else {
2466 synchronized (f) {
2467 if (tabAt(tab, i) == f) {
2468 Node<K,V> ln, hn;
2469 if (fh >= 0) {
2470 int runBit = fh & n;
2471 Node<K,V> lastRun = f;
2472 for (Node<K,V> p = f.next; p != null; p = p.next) {
2473 int b = p.hash & n;
2474 if (b != runBit) {
2475 runBit = b;
2476 lastRun = p;
2477 }
2478 }
2479 if (runBit == 0) {
2480 ln = lastRun;
2481 hn = null;
2482 }
2483 else {
2484 hn = lastRun;
2485 ln = null;
2486 }
2487 for (Node<K,V> p = f; p != lastRun; p = p.next) {
2488 int ph = p.hash; K pk = p.key; V pv = p.val;
2489 if ((ph & n) == 0)
2490 ln = new Node<K,V>(ph, pk, pv, ln);
2491 else
2492 hn = new Node<K,V>(ph, pk, pv, hn);
2493 }
2494 setTabAt(nextTab, i, ln);
2495 setTabAt(nextTab, i + n, hn);
2496 setTabAt(tab, i, fwd);
2497 advance = true;
2498 }
2499 else if (f instanceof TreeBin) {
2500 TreeBin<K,V> t = (TreeBin<K,V>)f;
2501 TreeNode<K,V> lo = null, loTail = null;
2502 TreeNode<K,V> hi = null, hiTail = null;
2503 int lc = 0, hc = 0;
2504 for (Node<K,V> e = t.first; e != null; e = e.next) {
2505 int h = e.hash;
2506 TreeNode<K,V> p = new TreeNode<K,V>
2507 (h, e.key, e.val, null, null);
2508 if ((h & n) == 0) {
2509 if ((p.prev = loTail) == null)
2510 lo = p;
2511 else
2512 loTail.next = p;
2513 loTail = p;
2514 ++lc;
2515 }
2516 else {
2517 if ((p.prev = hiTail) == null)
2518 hi = p;
2519 else
2520 hiTail.next = p;
2521 hiTail = p;
2522 ++hc;
2523 }
2524 }
2525 ln = (lc <= UNTREEIFY_THRESHOLD) ? untreeify(lo) :
2526 (hc != 0) ? new TreeBin<K,V>(lo) : t;
2527 hn = (hc <= UNTREEIFY_THRESHOLD) ? untreeify(hi) :
2528 (lc != 0) ? new TreeBin<K,V>(hi) : t;
2529 setTabAt(nextTab, i, ln);
2530 setTabAt(nextTab, i + n, hn);
2531 setTabAt(tab, i, fwd);
2532 advance = true;
2533 }
2534 }
2535 }
2536 }
2537 }
2538 }
2539
2540 /* ---------------- Counter support -------------- */
2541
2542 /**
2543 * A padded cell for distributing counts. Adapted from LongAdder
2544 * and Striped64. See their internal docs for explanation.
2545 */
2546 @sun.misc.Contended static final class CounterCell {
2547 volatile long value;
2548 CounterCell(long x) { value = x; }
2549 }
2550
2551 final long sumCount() {
2552 CounterCell[] as = counterCells; CounterCell a;
2553 long sum = baseCount;
2554 if (as != null) {
2555 for (int i = 0; i < as.length; ++i) {
2556 if ((a = as[i]) != null)
2557 sum += a.value;
2558 }
2559 }
2560 return sum;
2561 }
2562
2563 // See LongAdder version for explanation
2564 private final void fullAddCount(long x, boolean wasUncontended) {
2565 int h;
2566 if ((h = ThreadLocalRandom.getProbe()) == 0) {
2567 ThreadLocalRandom.localInit(); // force initialization
2568 h = ThreadLocalRandom.getProbe();
2569 wasUncontended = true;
2570 }
2571 boolean collide = false; // True if last slot nonempty
2572 for (;;) {
2573 CounterCell[] as; CounterCell a; int n; long v;
2574 if ((as = counterCells) != null && (n = as.length) > 0) {
2575 if ((a = as[(n - 1) & h]) == null) {
2576 if (cellsBusy == 0) { // Try to attach new Cell
2577 CounterCell r = new CounterCell(x); // Optimistic create
2578 if (cellsBusy == 0 &&
2579 U.compareAndSwapInt(this, CELLSBUSY, 0, 1)) {
2580 boolean created = false;
2581 try { // Recheck under lock
2582 CounterCell[] rs; int m, j;
2583 if ((rs = counterCells) != null &&
2584 (m = rs.length) > 0 &&
2585 rs[j = (m - 1) & h] == null) {
2586 rs[j] = r;
2587 created = true;
2588 }
2589 } finally {
2590 cellsBusy = 0;
2591 }
2592 if (created)
2593 break;
2594 continue; // Slot is now non-empty
2595 }
2596 }
2597 collide = false;
2598 }
2599 else if (!wasUncontended) // CAS already known to fail
2600 wasUncontended = true; // Continue after rehash
2601 else if (U.compareAndSwapLong(a, CELLVALUE, v = a.value, v + x))
2602 break;
2603 else if (counterCells != as || n >= NCPU)
2604 collide = false; // At max size or stale
2605 else if (!collide)
2606 collide = true;
2607 else if (cellsBusy == 0 &&
2608 U.compareAndSwapInt(this, CELLSBUSY, 0, 1)) {
2609 try {
2610 if (counterCells == as) {// Expand table unless stale
2611 CounterCell[] rs = new CounterCell[n << 1];
2612 for (int i = 0; i < n; ++i)
2613 rs[i] = as[i];
2614 counterCells = rs;
2615 }
2616 } finally {
2617 cellsBusy = 0;
2618 }
2619 collide = false;
2620 continue; // Retry with expanded table
2621 }
2622 h = ThreadLocalRandom.advanceProbe(h);
2623 }
2624 else if (cellsBusy == 0 && counterCells == as &&
2625 U.compareAndSwapInt(this, CELLSBUSY, 0, 1)) {
2626 boolean init = false;
2627 try { // Initialize table
2628 if (counterCells == as) {
2629 CounterCell[] rs = new CounterCell[2];
2630 rs[h & 1] = new CounterCell(x);
2631 counterCells = rs;
2632 init = true;
2633 }
2634 } finally {
2635 cellsBusy = 0;
2636 }
2637 if (init)
2638 break;
2639 }
2640 else if (U.compareAndSwapLong(this, BASECOUNT, v = baseCount, v + x))
2641 break; // Fall back on using base
2642 }
2643 }
2644
2645 /* ---------------- Conversion from/to TreeBins -------------- */
2646
2647 /**
2648 * Replaces all linked nodes in bin at given index unless table is
2649 * too small, in which case resizes instead.
2650 */
2651 private final void treeifyBin(Node<K,V>[] tab, int index) {
2652 Node<K,V> b; int n, sc;
2653 if (tab != null) {
2654 if ((n = tab.length) < MIN_TREEIFY_CAPACITY)
2655 tryPresize(n << 1);
2656 else if ((b = tabAt(tab, index)) != null && b.hash >= 0) {
2657 synchronized (b) {
2658 if (tabAt(tab, index) == b) {
2659 TreeNode<K,V> hd = null, tl = null;
2660 for (Node<K,V> e = b; e != null; e = e.next) {
2661 TreeNode<K,V> p =
2662 new TreeNode<K,V>(e.hash, e.key, e.val,
2663 null, null);
2664 if ((p.prev = tl) == null)
2665 hd = p;
2666 else
2667 tl.next = p;
2668 tl = p;
2669 }
2670 setTabAt(tab, index, new TreeBin<K,V>(hd));
2671 }
2672 }
2673 }
2674 }
2675 }
2676
2677 /**
2678 * Returns a list on non-TreeNodes replacing those in given list.
2679 */
2680 static <K,V> Node<K,V> untreeify(Node<K,V> b) {
2681 Node<K,V> hd = null, tl = null;
2682 for (Node<K,V> q = b; q != null; q = q.next) {
2683 Node<K,V> p = new Node<K,V>(q.hash, q.key, q.val, null);
2684 if (tl == null)
2685 hd = p;
2686 else
2687 tl.next = p;
2688 tl = p;
2689 }
2690 return hd;
2691 }
2692
2693 /* ---------------- TreeNodes -------------- */
2694
2695 /**
2696 * Nodes for use in TreeBins
2697 */
2698 static final class TreeNode<K,V> extends Node<K,V> {
2699 TreeNode<K,V> parent; // red-black tree links
2700 TreeNode<K,V> left;
2701 TreeNode<K,V> right;
2702 TreeNode<K,V> prev; // needed to unlink next upon deletion
2703 boolean red;
2704
2705 TreeNode(int hash, K key, V val, Node<K,V> next,
2706 TreeNode<K,V> parent) {
2707 super(hash, key, val, next);
2708 this.parent = parent;
2709 }
2710
2711 Node<K,V> find(int h, Object k) {
2712 return findTreeNode(h, k, null);
2713 }
2714
2715 /**
2716 * Returns the TreeNode (or null if not found) for the given key
2717 * starting at given root.
2718 */
2719 final TreeNode<K,V> findTreeNode(int h, Object k, Class<?> kc) {
2720 if (k != null) {
2721 TreeNode<K,V> p = this;
2722 do {
2723 int ph, dir; K pk; TreeNode<K,V> q;
2724 TreeNode<K,V> pl = p.left, pr = p.right;
2725 if ((ph = p.hash) > h)
2726 p = pl;
2727 else if (ph < h)
2728 p = pr;
2729 else if ((pk = p.key) == k || (pk != null && k.equals(pk)))
2730 return p;
2731 else if (pl == null)
2732 p = pr;
2733 else if (pr == null)
2734 p = pl;
2735 else if ((kc != null ||
2736 (kc = comparableClassFor(k)) != null) &&
2737 (dir = compareComparables(kc, k, pk)) != 0)
2738 p = (dir < 0) ? pl : pr;
2739 else if ((q = pr.findTreeNode(h, k, kc)) != null)
2740 return q;
2741 else
2742 p = pl;
2743 } while (p != null);
2744 }
2745 return null;
2746 }
2747 }
2748
2749 /* ---------------- TreeBins -------------- */
2750
2751 /**
2752 * TreeNodes used at the heads of bins. TreeBins do not hold user
2753 * keys or values, but instead point to list of TreeNodes and
2754 * their root. They also maintain a parasitic read-write lock
2755 * forcing writers (who hold bin lock) to wait for readers (who do
2756 * not) to complete before tree restructuring operations.
2757 */
2758 static final class TreeBin<K,V> extends Node<K,V> {
2759 TreeNode<K,V> root;
2760 volatile TreeNode<K,V> first;
2761 volatile Thread waiter;
2762 volatile int lockState;
2763 // values for lockState
2764 static final int WRITER = 1; // set while holding write lock
2765 static final int WAITER = 2; // set when waiting for write lock
2766 static final int READER = 4; // increment value for setting read lock
2767
2768 /**
2769 * Tie-breaking utility for ordering insertions when equal
2770 * hashCodes and non-comparable. We don't require a total
2771 * order, just a consistent insertion rule to maintain
2772 * equivalence across rebalancings. Tie-breaking further than
2773 * necessary simplifies testing a bit.
2774 */
2775 static int tieBreakOrder(Object a, Object b) {
2776 int d;
2777 if (a == null || b == null ||
2778 (d = a.getClass().getName().
2779 compareTo(b.getClass().getName())) == 0)
2780 d = (System.identityHashCode(a) <= System.identityHashCode(b) ?
2781 -1 : 1);
2782 return d;
2783 }
2784
2785 /**
2786 * Creates bin with initial set of nodes headed by b.
2787 */
2788 TreeBin(TreeNode<K,V> b) {
2789 super(TREEBIN, null, null, null);
2790 this.first = b;
2791 TreeNode<K,V> r = null;
2792 for (TreeNode<K,V> x = b, next; x != null; x = next) {
2793 next = (TreeNode<K,V>)x.next;
2794 x.left = x.right = null;
2795 if (r == null) {
2796 x.parent = null;
2797 x.red = false;
2798 r = x;
2799 }
2800 else {
2801 K k = x.key;
2802 int h = x.hash;
2803 Class<?> kc = null;
2804 for (TreeNode<K,V> p = r;;) {
2805 int dir, ph;
2806 K pk = p.key;
2807 if ((ph = p.hash) > h)
2808 dir = -1;
2809 else if (ph < h)
2810 dir = 1;
2811 else if ((kc == null &&
2812 (kc = comparableClassFor(k)) == null) ||
2813 (dir = compareComparables(kc, k, pk)) == 0)
2814 dir = tieBreakOrder(k, pk);
2815 TreeNode<K,V> xp = p;
2816 if ((p = (dir <= 0) ? p.left : p.right) == null) {
2817 x.parent = xp;
2818 if (dir <= 0)
2819 xp.left = x;
2820 else
2821 xp.right = x;
2822 r = balanceInsertion(r, x);
2823 break;
2824 }
2825 }
2826 }
2827 }
2828 this.root = r;
2829 assert checkInvariants(root);
2830 }
2831
2832 /**
2833 * Acquires write lock for tree restructuring.
2834 */
2835 private final void lockRoot() {
2836 if (!U.compareAndSwapInt(this, LOCKSTATE, 0, WRITER))
2837 contendedLock(); // offload to separate method
2838 }
2839
2840 /**
2841 * Releases write lock for tree restructuring.
2842 */
2843 private final void unlockRoot() {
2844 lockState = 0;
2845 }
2846
2847 /**
2848 * Possibly blocks awaiting root lock.
2849 */
2850 private final void contendedLock() {
2851 boolean waiting = false;
2852 for (int s;;) {
2853 if (((s = lockState) & ~WAITER) == 0) {
2854 if (U.compareAndSwapInt(this, LOCKSTATE, s, WRITER)) {
2855 if (waiting)
2856 waiter = null;
2857 return;
2858 }
2859 }
2860 else if ((s & WAITER) == 0) {
2861 if (U.compareAndSwapInt(this, LOCKSTATE, s, s | WAITER)) {
2862 waiting = true;
2863 waiter = Thread.currentThread();
2864 }
2865 }
2866 else if (waiting)
2867 LockSupport.park(this);
2868 }
2869 }
2870
2871 /**
2872 * Returns matching node or null if none. Tries to search
2873 * using tree comparisons from root, but continues linear
2874 * search when lock not available.
2875 */
2876 final Node<K,V> find(int h, Object k) {
2877 if (k != null) {
2878 for (Node<K,V> e = first; e != null; ) {
2879 int s; K ek;
2880 if (((s = lockState) & (WAITER|WRITER)) != 0) {
2881 if (e.hash == h &&
2882 ((ek = e.key) == k || (ek != null && k.equals(ek))))
2883 return e;
2884 e = e.next;
2885 }
2886 else if (U.compareAndSwapInt(this, LOCKSTATE, s,
2887 s + READER)) {
2888 TreeNode<K,V> r, p;
2889 try {
2890 p = ((r = root) == null ? null :
2891 r.findTreeNode(h, k, null));
2892 } finally {
2893 Thread w;
2894 if (U.getAndAddInt(this, LOCKSTATE, -READER) ==
2895 (READER|WAITER) && (w = waiter) != null)
2896 LockSupport.unpark(w);
2897 }
2898 return p;
2899 }
2900 }
2901 }
2902 return null;
2903 }
2904
2905 /**
2906 * Finds or adds a node.
2907 * @return null if added
2908 */
2909 final TreeNode<K,V> putTreeVal(int h, K k, V v) {
2910 Class<?> kc = null;
2911 boolean searched = false;
2912 for (TreeNode<K,V> p = root;;) {
2913 int dir, ph; K pk;
2914 if (p == null) {
2915 first = root = new TreeNode<K,V>(h, k, v, null, null);
2916 break;
2917 }
2918 else if ((ph = p.hash) > h)
2919 dir = -1;
2920 else if (ph < h)
2921 dir = 1;
2922 else if ((pk = p.key) == k || (pk != null && k.equals(pk)))
2923 return p;
2924 else if ((kc == null &&
2925 (kc = comparableClassFor(k)) == null) ||
2926 (dir = compareComparables(kc, k, pk)) == 0) {
2927 if (!searched) {
2928 TreeNode<K,V> q, ch;
2929 searched = true;
2930 if (((ch = p.left) != null &&
2931 (q = ch.findTreeNode(h, k, kc)) != null) ||
2932 ((ch = p.right) != null &&
2933 (q = ch.findTreeNode(h, k, kc)) != null))
2934 return q;
2935 }
2936 dir = tieBreakOrder(k, pk);
2937 }
2938
2939 TreeNode<K,V> xp = p;
2940 if ((p = (dir <= 0) ? p.left : p.right) == null) {
2941 TreeNode<K,V> x, f = first;
2942 first = x = new TreeNode<K,V>(h, k, v, f, xp);
2943 if (f != null)
2944 f.prev = x;
2945 if (dir <= 0)
2946 xp.left = x;
2947 else
2948 xp.right = x;
2949 if (!xp.red)
2950 x.red = true;
2951 else {
2952 lockRoot();
2953 try {
2954 root = balanceInsertion(root, x);
2955 } finally {
2956 unlockRoot();
2957 }
2958 }
2959 break;
2960 }
2961 }
2962 assert checkInvariants(root);
2963 return null;
2964 }
2965
2966 /**
2967 * Removes the given node, that must be present before this
2968 * call. This is messier than typical red-black deletion code
2969 * because we cannot swap the contents of an interior node
2970 * with a leaf successor that is pinned by "next" pointers
2971 * that are accessible independently of lock. So instead we
2972 * swap the tree linkages.
2973 *
2974 * @return true if now too small, so should be untreeified
2975 */
2976 final boolean removeTreeNode(TreeNode<K,V> p) {
2977 TreeNode<K,V> next = (TreeNode<K,V>)p.next;
2978 TreeNode<K,V> pred = p.prev; // unlink traversal pointers
2979 TreeNode<K,V> r, rl;
2980 if (pred == null)
2981 first = next;
2982 else
2983 pred.next = next;
2984 if (next != null)
2985 next.prev = pred;
2986 if (first == null) {
2987 root = null;
2988 return true;
2989 }
2990 if ((r = root) == null || r.right == null || // too small
2991 (rl = r.left) == null || rl.left == null)
2992 return true;
2993 lockRoot();
2994 try {
2995 TreeNode<K,V> replacement;
2996 TreeNode<K,V> pl = p.left;
2997 TreeNode<K,V> pr = p.right;
2998 if (pl != null && pr != null) {
2999 TreeNode<K,V> s = pr, sl;
3000 while ((sl = s.left) != null) // find successor
3001 s = sl;
3002 boolean c = s.red; s.red = p.red; p.red = c; // swap colors
3003 TreeNode<K,V> sr = s.right;
3004 TreeNode<K,V> pp = p.parent;
3005 if (s == pr) { // p was s's direct parent
3006 p.parent = s;
3007 s.right = p;
3008 }
3009 else {
3010 TreeNode<K,V> sp = s.parent;
3011 if ((p.parent = sp) != null) {
3012 if (s == sp.left)
3013 sp.left = p;
3014 else
3015 sp.right = p;
3016 }
3017 if ((s.right = pr) != null)
3018 pr.parent = s;
3019 }
3020 p.left = null;
3021 if ((p.right = sr) != null)
3022 sr.parent = p;
3023 if ((s.left = pl) != null)
3024 pl.parent = s;
3025 if ((s.parent = pp) == null)
3026 r = s;
3027 else if (p == pp.left)
3028 pp.left = s;
3029 else
3030 pp.right = s;
3031 if (sr != null)
3032 replacement = sr;
3033 else
3034 replacement = p;
3035 }
3036 else if (pl != null)
3037 replacement = pl;
3038 else if (pr != null)
3039 replacement = pr;
3040 else
3041 replacement = p;
3042 if (replacement != p) {
3043 TreeNode<K,V> pp = replacement.parent = p.parent;
3044 if (pp == null)
3045 r = replacement;
3046 else if (p == pp.left)
3047 pp.left = replacement;
3048 else
3049 pp.right = replacement;
3050 p.left = p.right = p.parent = null;
3051 }
3052
3053 root = (p.red) ? r : balanceDeletion(r, replacement);
3054
3055 if (p == replacement) { // detach pointers
3056 TreeNode<K,V> pp;
3057 if ((pp = p.parent) != null) {
3058 if (p == pp.left)
3059 pp.left = null;
3060 else if (p == pp.right)
3061 pp.right = null;
3062 p.parent = null;
3063 }
3064 }
3065 } finally {
3066 unlockRoot();
3067 }
3068 assert checkInvariants(root);
3069 return false;
3070 }
3071
3072 /* ------------------------------------------------------------ */
3073 // Red-black tree methods, all adapted from CLR
3074
3075 static <K,V> TreeNode<K,V> rotateLeft(TreeNode<K,V> root,
3076 TreeNode<K,V> p) {
3077 TreeNode<K,V> r, pp, rl;
3078 if (p != null && (r = p.right) != null) {
3079 if ((rl = p.right = r.left) != null)
3080 rl.parent = p;
3081 if ((pp = r.parent = p.parent) == null)
3082 (root = r).red = false;
3083 else if (pp.left == p)
3084 pp.left = r;
3085 else
3086 pp.right = r;
3087 r.left = p;
3088 p.parent = r;
3089 }
3090 return root;
3091 }
3092
3093 static <K,V> TreeNode<K,V> rotateRight(TreeNode<K,V> root,
3094 TreeNode<K,V> p) {
3095 TreeNode<K,V> l, pp, lr;
3096 if (p != null && (l = p.left) != null) {
3097 if ((lr = p.left = l.right) != null)
3098 lr.parent = p;
3099 if ((pp = l.parent = p.parent) == null)
3100 (root = l).red = false;
3101 else if (pp.right == p)
3102 pp.right = l;
3103 else
3104 pp.left = l;
3105 l.right = p;
3106 p.parent = l;
3107 }
3108 return root;
3109 }
3110
3111 static <K,V> TreeNode<K,V> balanceInsertion(TreeNode<K,V> root,
3112 TreeNode<K,V> x) {
3113 x.red = true;
3114 for (TreeNode<K,V> xp, xpp, xppl, xppr;;) {
3115 if ((xp = x.parent) == null) {
3116 x.red = false;
3117 return x;
3118 }
3119 else if (!xp.red || (xpp = xp.parent) == null)
3120 return root;
3121 if (xp == (xppl = xpp.left)) {
3122 if ((xppr = xpp.right) != null && xppr.red) {
3123 xppr.red = false;
3124 xp.red = false;
3125 xpp.red = true;
3126 x = xpp;
3127 }
3128 else {
3129 if (x == xp.right) {
3130 root = rotateLeft(root, x = xp);
3131 xpp = (xp = x.parent) == null ? null : xp.parent;
3132 }
3133 if (xp != null) {
3134 xp.red = false;
3135 if (xpp != null) {
3136 xpp.red = true;
3137 root = rotateRight(root, xpp);
3138 }
3139 }
3140 }
3141 }
3142 else {
3143 if (xppl != null && xppl.red) {
3144 xppl.red = false;
3145 xp.red = false;
3146 xpp.red = true;
3147 x = xpp;
3148 }
3149 else {
3150 if (x == xp.left) {
3151 root = rotateRight(root, x = xp);
3152 xpp = (xp = x.parent) == null ? null : xp.parent;
3153 }
3154 if (xp != null) {
3155 xp.red = false;
3156 if (xpp != null) {
3157 xpp.red = true;
3158 root = rotateLeft(root, xpp);
3159 }
3160 }
3161 }
3162 }
3163 }
3164 }
3165
3166 static <K,V> TreeNode<K,V> balanceDeletion(TreeNode<K,V> root,
3167 TreeNode<K,V> x) {
3168 for (TreeNode<K,V> xp, xpl, xpr;;) {
3169 if (x == null || x == root)
3170 return root;
3171 else if ((xp = x.parent) == null) {
3172 x.red = false;
3173 return x;
3174 }
3175 else if (x.red) {
3176 x.red = false;
3177 return root;
3178 }
3179 else if ((xpl = xp.left) == x) {
3180 if ((xpr = xp.right) != null && xpr.red) {
3181 xpr.red = false;
3182 xp.red = true;
3183 root = rotateLeft(root, xp);
3184 xpr = (xp = x.parent) == null ? null : xp.right;
3185 }
3186 if (xpr == null)
3187 x = xp;
3188 else {
3189 TreeNode<K,V> sl = xpr.left, sr = xpr.right;
3190 if ((sr == null || !sr.red) &&
3191 (sl == null || !sl.red)) {
3192 xpr.red = true;
3193 x = xp;
3194 }
3195 else {
3196 if (sr == null || !sr.red) {
3197 if (sl != null)
3198 sl.red = false;
3199 xpr.red = true;
3200 root = rotateRight(root, xpr);
3201 xpr = (xp = x.parent) == null ?
3202 null : xp.right;
3203 }
3204 if (xpr != null) {
3205 xpr.red = (xp == null) ? false : xp.red;
3206 if ((sr = xpr.right) != null)
3207 sr.red = false;
3208 }
3209 if (xp != null) {
3210 xp.red = false;
3211 root = rotateLeft(root, xp);
3212 }
3213 x = root;
3214 }
3215 }
3216 }
3217 else { // symmetric
3218 if (xpl != null && xpl.red) {
3219 xpl.red = false;
3220 xp.red = true;
3221 root = rotateRight(root, xp);
3222 xpl = (xp = x.parent) == null ? null : xp.left;
3223 }
3224 if (xpl == null)
3225 x = xp;
3226 else {
3227 TreeNode<K,V> sl = xpl.left, sr = xpl.right;
3228 if ((sl == null || !sl.red) &&
3229 (sr == null || !sr.red)) {
3230 xpl.red = true;
3231 x = xp;
3232 }
3233 else {
3234 if (sl == null || !sl.red) {
3235 if (sr != null)
3236 sr.red = false;
3237 xpl.red = true;
3238 root = rotateLeft(root, xpl);
3239 xpl = (xp = x.parent) == null ?
3240 null : xp.left;
3241 }
3242 if (xpl != null) {
3243 xpl.red = (xp == null) ? false : xp.red;
3244 if ((sl = xpl.left) != null)
3245 sl.red = false;
3246 }
3247 if (xp != null) {
3248 xp.red = false;
3249 root = rotateRight(root, xp);
3250 }
3251 x = root;
3252 }
3253 }
3254 }
3255 }
3256 }
3257
3258 /**
3259 * Recursive invariant check
3260 */
3261 static <K,V> boolean checkInvariants(TreeNode<K,V> t) {
3262 TreeNode<K,V> tp = t.parent, tl = t.left, tr = t.right,
3263 tb = t.prev, tn = (TreeNode<K,V>)t.next;
3264 if (tb != null && tb.next != t)
3265 return false;
3266 if (tn != null && tn.prev != t)
3267 return false;
3268 if (tp != null && t != tp.left && t != tp.right)
3269 return false;
3270 if (tl != null && (tl.parent != t || tl.hash > t.hash))
3271 return false;
3272 if (tr != null && (tr.parent != t || tr.hash < t.hash))
3273 return false;
3274 if (t.red && tl != null && tl.red && tr != null && tr.red)
3275 return false;
3276 if (tl != null && !checkInvariants(tl))
3277 return false;
3278 if (tr != null && !checkInvariants(tr))
3279 return false;
3280 return true;
3281 }
3282
3283 private static final sun.misc.Unsafe U;
3284 private static final long LOCKSTATE;
3285 static {
3286 try {
3287 U = sun.misc.Unsafe.getUnsafe();
3288 Class<?> k = TreeBin.class;
3289 LOCKSTATE = U.objectFieldOffset
3290 (k.getDeclaredField("lockState"));
3291 } catch (Exception e) {
3292 throw new Error(e);
3293 }
3294 }
3295 }
3296
3297 /* ----------------Table Traversal -------------- */
3298
3299 /**
3300 * Records the table, its length, and current traversal index for a
3301 * traverser that must process a region of a forwarded table before
3302 * proceeding with current table.
3303 */
3304 static final class TableStack<K,V> {
3305 int length;
3306 int index;
3307 Node<K,V>[] tab;
3308 TableStack<K,V> next;
3309 }
3310
3311 /**
3312 * Encapsulates traversal for methods such as containsValue; also
3313 * serves as a base class for other iterators and spliterators.
3314 *
3315 * Method advance visits once each still-valid node that was
3316 * reachable upon iterator construction. It might miss some that
3317 * were added to a bin after the bin was visited, which is OK wrt
3318 * consistency guarantees. Maintaining this property in the face
3319 * of possible ongoing resizes requires a fair amount of
3320 * bookkeeping state that is difficult to optimize away amidst
3321 * volatile accesses. Even so, traversal maintains reasonable
3322 * throughput.
3323 *
3324 * Normally, iteration proceeds bin-by-bin traversing lists.
3325 * However, if the table has been resized, then all future steps
3326 * must traverse both the bin at the current index as well as at
3327 * (index + baseSize); and so on for further resizings. To
3328 * paranoically cope with potential sharing by users of iterators
3329 * across threads, iteration terminates if a bounds checks fails
3330 * for a table read.
3331 */
3332 static class Traverser<K,V> {
3333 Node<K,V>[] tab; // current table; updated if resized
3334 Node<K,V> next; // the next entry to use
3335 TableStack<K,V> stack, spare; // to save/restore on ForwardingNodes
3336 int index; // index of bin to use next
3337 int baseIndex; // current index of initial table
3338 int baseLimit; // index bound for initial table
3339 final int baseSize; // initial table size
3340
3341 Traverser(Node<K,V>[] tab, int size, int index, int limit) {
3342 this.tab = tab;
3343 this.baseSize = size;
3344 this.baseIndex = this.index = index;
3345 this.baseLimit = limit;
3346 this.next = null;
3347 }
3348
3349 /**
3350 * Advances if possible, returning next valid node, or null if none.
3351 */
3352 final Node<K,V> advance() {
3353 Node<K,V> e;
3354 if ((e = next) != null)
3355 e = e.next;
3356 for (;;) {
3357 Node<K,V>[] t; int i, n; // must use locals in checks
3358 if (e != null)
3359 return next = e;
3360 if (baseIndex >= baseLimit || (t = tab) == null ||
3361 (n = t.length) <= (i = index) || i < 0)
3362 return next = null;
3363 if ((e = tabAt(t, i)) != null && e.hash < 0) {
3364 if (e instanceof ForwardingNode) {
3365 tab = ((ForwardingNode<K,V>)e).nextTable;
3366 e = null;
3367 pushState(t, i, n);
3368 continue;
3369 }
3370 else if (e instanceof TreeBin)
3371 e = ((TreeBin<K,V>)e).first;
3372 else
3373 e = null;
3374 }
3375 if (stack != null)
3376 recoverState(n);
3377 else if ((index = i + baseSize) >= n)
3378 index = ++baseIndex; // visit upper slots if present
3379 }
3380 }
3381
3382 /**
3383 * Saves traversal state upon encountering a forwarding node.
3384 */
3385 private void pushState(Node<K,V>[] t, int i, int n) {
3386 TableStack<K,V> s = spare; // reuse if possible
3387 if (s != null)
3388 spare = s.next;
3389 else
3390 s = new TableStack<K,V>();
3391 s.tab = t;
3392 s.length = n;
3393 s.index = i;
3394 s.next = stack;
3395 stack = s;
3396 }
3397
3398 /**
3399 * Possibly pops traversal state.
3400 *
3401 * @param n length of current table
3402 */
3403 private void recoverState(int n) {
3404 TableStack<K,V> s; int len;
3405 while ((s = stack) != null && (index += (len = s.length)) >= n) {
3406 n = len;
3407 index = s.index;
3408 tab = s.tab;
3409 s.tab = null;
3410 TableStack<K,V> next = s.next;
3411 s.next = spare; // save for reuse
3412 stack = next;
3413 spare = s;
3414 }
3415 if (s == null && (index += baseSize) >= n)
3416 index = ++baseIndex;
3417 }
3418 }
3419
3420 /**
3421 * Base of key, value, and entry Iterators. Adds fields to
3422 * Traverser to support iterator.remove.
3423 */
3424 static class BaseIterator<K,V> extends Traverser<K,V> {
3425 final ConcurrentHashMap<K,V> map;
3426 Node<K,V> lastReturned;
3427 BaseIterator(Node<K,V>[] tab, int size, int index, int limit,
3428 ConcurrentHashMap<K,V> map) {
3429 super(tab, size, index, limit);
3430 this.map = map;
3431 advance();
3432 }
3433
3434 public final boolean hasNext() { return next != null; }
3435 public final boolean hasMoreElements() { return next != null; }
3436
3437 public final void remove() {
3438 Node<K,V> p;
3439 if ((p = lastReturned) == null)
3440 throw new IllegalStateException();
3441 lastReturned = null;
3442 map.replaceNode(p.key, null, null);
3443 }
3444 }
3445
3446 static final class KeyIterator<K,V> extends BaseIterator<K,V>
3447 implements Iterator<K>, Enumeration<K> {
3448 KeyIterator(Node<K,V>[] tab, int index, int size, int limit,
3449 ConcurrentHashMap<K,V> map) {
3450 super(tab, index, size, limit, map);
3451 }
3452
3453 public final K next() {
3454 Node<K,V> p;
3455 if ((p = next) == null)
3456 throw new NoSuchElementException();
3457 K k = p.key;
3458 lastReturned = p;
3459 advance();
3460 return k;
3461 }
3462
3463 public final K nextElement() { return next(); }
3464 }
3465
3466 static final class ValueIterator<K,V> extends BaseIterator<K,V>
3467 implements Iterator<V>, Enumeration<V> {
3468 ValueIterator(Node<K,V>[] tab, int index, int size, int limit,
3469 ConcurrentHashMap<K,V> map) {
3470 super(tab, index, size, limit, map);
3471 }
3472
3473 public final V next() {
3474 Node<K,V> p;
3475 if ((p = next) == null)
3476 throw new NoSuchElementException();
3477 V v = p.val;
3478 lastReturned = p;
3479 advance();
3480 return v;
3481 }
3482
3483 public final V nextElement() { return next(); }
3484 }
3485
3486 static final class EntryIterator<K,V> extends BaseIterator<K,V>
3487 implements Iterator<Map.Entry<K,V>> {
3488 EntryIterator(Node<K,V>[] tab, int index, int size, int limit,
3489 ConcurrentHashMap<K,V> map) {
3490 super(tab, index, size, limit, map);
3491 }
3492
3493 public final Map.Entry<K,V> next() {
3494 Node<K,V> p;
3495 if ((p = next) == null)
3496 throw new NoSuchElementException();
3497 K k = p.key;
3498 V v = p.val;
3499 lastReturned = p;
3500 advance();
3501 return new MapEntry<K,V>(k, v, map);
3502 }
3503 }
3504
3505 /**
3506 * Exported Entry for EntryIterator
3507 */
3508 static final class MapEntry<K,V> implements Map.Entry<K,V> {
3509 final K key; // non-null
3510 V val; // non-null
3511 final ConcurrentHashMap<K,V> map;
3512 MapEntry(K key, V val, ConcurrentHashMap<K,V> map) {
3513 this.key = key;
3514 this.val = val;
3515 this.map = map;
3516 }
3517 public K getKey() { return key; }
3518 public V getValue() { return val; }
3519 public int hashCode() { return key.hashCode() ^ val.hashCode(); }
3520 public String toString() { return key + "=" + val; }
3521
3522 public boolean equals(Object o) {
3523 Object k, v; Map.Entry<?,?> e;
3524 return ((o instanceof Map.Entry) &&
3525 (k = (e = (Map.Entry<?,?>)o).getKey()) != null &&
3526 (v = e.getValue()) != null &&
3527 (k == key || k.equals(key)) &&
3528 (v == val || v.equals(val)));
3529 }
3530
3531 /**
3532 * Sets our entry's value and writes through to the map. The
3533 * value to return is somewhat arbitrary here. Since we do not
3534 * necessarily track asynchronous changes, the most recent
3535 * "previous" value could be different from what we return (or
3536 * could even have been removed, in which case the put will
3537 * re-establish). We do not and cannot guarantee more.
3538 */
3539 public V setValue(V value) {
3540 if (value == null) throw new NullPointerException();
3541 V v = val;
3542 val = value;
3543 map.put(key, value);
3544 return v;
3545 }
3546 }
3547
3548 static final class KeySpliterator<K,V> extends Traverser<K,V>
3549 implements Spliterator<K> {
3550 long est; // size estimate
3551 KeySpliterator(Node<K,V>[] tab, int size, int index, int limit,
3552 long est) {
3553 super(tab, size, index, limit);
3554 this.est = est;
3555 }
3556
3557 public Spliterator<K> trySplit() {
3558 int i, f, h;
3559 return (h = ((i = baseIndex) + (f = baseLimit)) >>> 1) <= i ? null :
3560 new KeySpliterator<K,V>(tab, baseSize, baseLimit = h,
3561 f, est >>>= 1);
3562 }
3563
3564 public void forEachRemaining(Consumer<? super K> action) {
3565 if (action == null) throw new NullPointerException();
3566 for (Node<K,V> p; (p = advance()) != null;)
3567 action.accept(p.key);
3568 }
3569
3570 public boolean tryAdvance(Consumer<? super K> action) {
3571 if (action == null) throw new NullPointerException();
3572 Node<K,V> p;
3573 if ((p = advance()) == null)
3574 return false;
3575 action.accept(p.key);
3576 return true;
3577 }
3578
3579 public long estimateSize() { return est; }
3580
3581 public int characteristics() {
3582 return Spliterator.DISTINCT | Spliterator.CONCURRENT |
3583 Spliterator.NONNULL;
3584 }
3585 }
3586
3587 static final class ValueSpliterator<K,V> extends Traverser<K,V>
3588 implements Spliterator<V> {
3589 long est; // size estimate
3590 ValueSpliterator(Node<K,V>[] tab, int size, int index, int limit,
3591 long est) {
3592 super(tab, size, index, limit);
3593 this.est = est;
3594 }
3595
3596 public Spliterator<V> trySplit() {
3597 int i, f, h;
3598 return (h = ((i = baseIndex) + (f = baseLimit)) >>> 1) <= i ? null :
3599 new ValueSpliterator<K,V>(tab, baseSize, baseLimit = h,
3600 f, est >>>= 1);
3601 }
3602
3603 public void forEachRemaining(Consumer<? super V> action) {
3604 if (action == null) throw new NullPointerException();
3605 for (Node<K,V> p; (p = advance()) != null;)
3606 action.accept(p.val);
3607 }
3608
3609 public boolean tryAdvance(Consumer<? super V> action) {
3610 if (action == null) throw new NullPointerException();
3611 Node<K,V> p;
3612 if ((p = advance()) == null)
3613 return false;
3614 action.accept(p.val);
3615 return true;
3616 }
3617
3618 public long estimateSize() { return est; }
3619
3620 public int characteristics() {
3621 return Spliterator.CONCURRENT | Spliterator.NONNULL;
3622 }
3623 }
3624
3625 static final class EntrySpliterator<K,V> extends Traverser<K,V>
3626 implements Spliterator<Map.Entry<K,V>> {
3627 final ConcurrentHashMap<K,V> map; // To export MapEntry
3628 long est; // size estimate
3629 EntrySpliterator(Node<K,V>[] tab, int size, int index, int limit,
3630 long est, ConcurrentHashMap<K,V> map) {
3631 super(tab, size, index, limit);
3632 this.map = map;
3633 this.est = est;
3634 }
3635
3636 public Spliterator<Map.Entry<K,V>> trySplit() {
3637 int i, f, h;
3638 return (h = ((i = baseIndex) + (f = baseLimit)) >>> 1) <= i ? null :
3639 new EntrySpliterator<K,V>(tab, baseSize, baseLimit = h,
3640 f, est >>>= 1, map);
3641 }
3642
3643 public void forEachRemaining(Consumer<? super Map.Entry<K,V>> action) {
3644 if (action == null) throw new NullPointerException();
3645 for (Node<K,V> p; (p = advance()) != null; )
3646 action.accept(new MapEntry<K,V>(p.key, p.val, map));
3647 }
3648
3649 public boolean tryAdvance(Consumer<? super Map.Entry<K,V>> action) {
3650 if (action == null) throw new NullPointerException();
3651 Node<K,V> p;
3652 if ((p = advance()) == null)
3653 return false;
3654 action.accept(new MapEntry<K,V>(p.key, p.val, map));
3655 return true;
3656 }
3657
3658 public long estimateSize() { return est; }
3659
3660 public int characteristics() {
3661 return Spliterator.DISTINCT | Spliterator.CONCURRENT |
3662 Spliterator.NONNULL;
3663 }
3664 }
3665
3666 // Parallel bulk operations
3667
3668 /**
3669 * Computes initial batch value for bulk tasks. The returned value
3670 * is approximately exp2 of the number of times (minus one) to
3671 * split task by two before executing leaf action. This value is
3672 * faster to compute and more convenient to use as a guide to
3673 * splitting than is the depth, since it is used while dividing by
3674 * two anyway.
3675 */
3676 final int batchFor(long b) {
3677 long n;
3678 if (b == Long.MAX_VALUE || (n = sumCount()) <= 1L || n < b)
3679 return 0;
3680 int sp = ForkJoinPool.getCommonPoolParallelism() << 2; // slack of 4
3681 return (b <= 0L || (n /= b) >= sp) ? sp : (int)n;
3682 }
3683
3684 /**
3685 * Performs the given action for each (key, value).
3686 *
3687 * @param parallelismThreshold the (estimated) number of elements
3688 * needed for this operation to be executed in parallel
3689 * @param action the action
3690 * @since 1.8
3691 */
3692 public void forEach(long parallelismThreshold,
3693 BiConsumer<? super K,? super V> action) {
3694 if (action == null) throw new NullPointerException();
3695 new ForEachMappingTask<K,V>
3696 (null, batchFor(parallelismThreshold), 0, 0, table,
3697 action).invoke();
3698 }
3699
3700 /**
3701 * Performs the given action for each non-null transformation
3702 * of each (key, value).
3703 *
3704 * @param parallelismThreshold the (estimated) number of elements
3705 * needed for this operation to be executed in parallel
3706 * @param transformer a function returning the transformation
3707 * for an element, or null if there is no transformation (in
3708 * which case the action is not applied)
3709 * @param action the action
3710 * @param <U> the return type of the transformer
3711 * @since 1.8
3712 */
3713 public <U> void forEach(long parallelismThreshold,
3714 BiFunction<? super K, ? super V, ? extends U> transformer,
3715 Consumer<? super U> action) {
3716 if (transformer == null || action == null)
3717 throw new NullPointerException();
3718 new ForEachTransformedMappingTask<K,V,U>
3719 (null, batchFor(parallelismThreshold), 0, 0, table,
3720 transformer, action).invoke();
3721 }
3722
3723 /**
3724 * Returns a non-null result from applying the given search
3725 * function on each (key, value), or null if none. Upon
3726 * success, further element processing is suppressed and the
3727 * results of any other parallel invocations of the search
3728 * function are ignored.
3729 *
3730 * @param parallelismThreshold the (estimated) number of elements
3731 * needed for this operation to be executed in parallel
3732 * @param searchFunction a function returning a non-null
3733 * result on success, else null
3734 * @param <U> the return type of the search function
3735 * @return a non-null result from applying the given search
3736 * function on each (key, value), or null if none
3737 * @since 1.8
3738 */
3739 public <U> U search(long parallelismThreshold,
3740 BiFunction<? super K, ? super V, ? extends U> searchFunction) {
3741 if (searchFunction == null) throw new NullPointerException();
3742 return new SearchMappingsTask<K,V,U>
3743 (null, batchFor(parallelismThreshold), 0, 0, table,
3744 searchFunction, new AtomicReference<U>()).invoke();
3745 }
3746
3747 /**
3748 * Returns the result of accumulating the given transformation
3749 * of all (key, value) pairs using the given reducer to
3750 * combine values, or null if none.
3751 *
3752 * @param parallelismThreshold the (estimated) number of elements
3753 * needed for this operation to be executed in parallel
3754 * @param transformer a function returning the transformation
3755 * for an element, or null if there is no transformation (in
3756 * which case it is not combined)
3757 * @param reducer a commutative associative combining function
3758 * @param <U> the return type of the transformer
3759 * @return the result of accumulating the given transformation
3760 * of all (key, value) pairs
3761 * @since 1.8
3762 */
3763 public <U> U reduce(long parallelismThreshold,
3764 BiFunction<? super K, ? super V, ? extends U> transformer,
3765 BiFunction<? super U, ? super U, ? extends U> reducer) {
3766 if (transformer == null || reducer == null)
3767 throw new NullPointerException();
3768 return new MapReduceMappingsTask<K,V,U>
3769 (null, batchFor(parallelismThreshold), 0, 0, table,
3770 null, transformer, reducer).invoke();
3771 }
3772
3773 /**
3774 * Returns the result of accumulating the given transformation
3775 * of all (key, value) pairs using the given reducer to
3776 * combine values, and the given basis as an identity value.
3777 *
3778 * @param parallelismThreshold the (estimated) number of elements
3779 * needed for this operation to be executed in parallel
3780 * @param transformer a function returning the transformation
3781 * for an element
3782 * @param basis the identity (initial default value) for the reduction
3783 * @param reducer a commutative associative combining function
3784 * @return the result of accumulating the given transformation
3785 * of all (key, value) pairs
3786 * @since 1.8
3787 */
3788 public double reduceToDouble(long parallelismThreshold,
3789 ToDoubleBiFunction<? super K, ? super V> transformer,
3790 double basis,
3791 DoubleBinaryOperator reducer) {
3792 if (transformer == null || reducer == null)
3793 throw new NullPointerException();
3794 return new MapReduceMappingsToDoubleTask<K,V>
3795 (null, batchFor(parallelismThreshold), 0, 0, table,
3796 null, transformer, basis, reducer).invoke();
3797 }
3798
3799 /**
3800 * Returns the result of accumulating the given transformation
3801 * of all (key, value) pairs using the given reducer to
3802 * combine values, and the given basis as an identity value.
3803 *
3804 * @param parallelismThreshold the (estimated) number of elements
3805 * needed for this operation to be executed in parallel
3806 * @param transformer a function returning the transformation
3807 * for an element
3808 * @param basis the identity (initial default value) for the reduction
3809 * @param reducer a commutative associative combining function
3810 * @return the result of accumulating the given transformation
3811 * of all (key, value) pairs
3812 * @since 1.8
3813 */
3814 public long reduceToLong(long parallelismThreshold,
3815 ToLongBiFunction<? super K, ? super V> transformer,
3816 long basis,
3817 LongBinaryOperator reducer) {
3818 if (transformer == null || reducer == null)
3819 throw new NullPointerException();
3820 return new MapReduceMappingsToLongTask<K,V>
3821 (null, batchFor(parallelismThreshold), 0, 0, table,
3822 null, transformer, basis, reducer).invoke();
3823 }
3824
3825 /**
3826 * Returns the result of accumulating the given transformation
3827 * of all (key, value) pairs using the given reducer to
3828 * combine values, and the given basis as an identity value.
3829 *
3830 * @param parallelismThreshold the (estimated) number of elements
3831 * needed for this operation to be executed in parallel
3832 * @param transformer a function returning the transformation
3833 * for an element
3834 * @param basis the identity (initial default value) for the reduction
3835 * @param reducer a commutative associative combining function
3836 * @return the result of accumulating the given transformation
3837 * of all (key, value) pairs
3838 * @since 1.8
3839 */
3840 public int reduceToInt(long parallelismThreshold,
3841 ToIntBiFunction<? super K, ? super V> transformer,
3842 int basis,
3843 IntBinaryOperator reducer) {
3844 if (transformer == null || reducer == null)
3845 throw new NullPointerException();
3846 return new MapReduceMappingsToIntTask<K,V>
3847 (null, batchFor(parallelismThreshold), 0, 0, table,
3848 null, transformer, basis, reducer).invoke();
3849 }
3850
3851 /**
3852 * Performs the given action for each key.
3853 *
3854 * @param parallelismThreshold the (estimated) number of elements
3855 * needed for this operation to be executed in parallel
3856 * @param action the action
3857 * @since 1.8
3858 */
3859 public void forEachKey(long parallelismThreshold,
3860 Consumer<? super K> action) {
3861 if (action == null) throw new NullPointerException();
3862 new ForEachKeyTask<K,V>
3863 (null, batchFor(parallelismThreshold), 0, 0, table,
3864 action).invoke();
3865 }
3866
3867 /**
3868 * Performs the given action for each non-null transformation
3869 * of each key.
3870 *
3871 * @param parallelismThreshold the (estimated) number of elements
3872 * needed for this operation to be executed in parallel
3873 * @param transformer a function returning the transformation
3874 * for an element, or null if there is no transformation (in
3875 * which case the action is not applied)
3876 * @param action the action
3877 * @param <U> the return type of the transformer
3878 * @since 1.8
3879 */
3880 public <U> void forEachKey(long parallelismThreshold,
3881 Function<? super K, ? extends U> transformer,
3882 Consumer<? super U> action) {
3883 if (transformer == null || action == null)
3884 throw new NullPointerException();
3885 new ForEachTransformedKeyTask<K,V,U>
3886 (null, batchFor(parallelismThreshold), 0, 0, table,
3887 transformer, action).invoke();
3888 }
3889
3890 /**
3891 * Returns a non-null result from applying the given search
3892 * function on each key, or null if none. Upon success,
3893 * further element processing is suppressed and the results of
3894 * any other parallel invocations of the search function are
3895 * ignored.
3896 *
3897 * @param parallelismThreshold the (estimated) number of elements
3898 * needed for this operation to be executed in parallel
3899 * @param searchFunction a function returning a non-null
3900 * result on success, else null
3901 * @param <U> the return type of the search function
3902 * @return a non-null result from applying the given search
3903 * function on each key, or null if none
3904 * @since 1.8
3905 */
3906 public <U> U searchKeys(long parallelismThreshold,
3907 Function<? super K, ? extends U> searchFunction) {
3908 if (searchFunction == null) throw new NullPointerException();
3909 return new SearchKeysTask<K,V,U>
3910 (null, batchFor(parallelismThreshold), 0, 0, table,
3911 searchFunction, new AtomicReference<U>()).invoke();
3912 }
3913
3914 /**
3915 * Returns the result of accumulating all keys using the given
3916 * reducer to combine values, or null if none.
3917 *
3918 * @param parallelismThreshold the (estimated) number of elements
3919 * needed for this operation to be executed in parallel
3920 * @param reducer a commutative associative combining function
3921 * @return the result of accumulating all keys using the given
3922 * reducer to combine values, or null if none
3923 * @since 1.8
3924 */
3925 public K reduceKeys(long parallelismThreshold,
3926 BiFunction<? super K, ? super K, ? extends K> reducer) {
3927 if (reducer == null) throw new NullPointerException();
3928 return new ReduceKeysTask<K,V>
3929 (null, batchFor(parallelismThreshold), 0, 0, table,
3930 null, reducer).invoke();
3931 }
3932
3933 /**
3934 * Returns the result of accumulating the given transformation
3935 * of all keys using the given reducer to combine values, or
3936 * null if none.
3937 *
3938 * @param parallelismThreshold the (estimated) number of elements
3939 * needed for this operation to be executed in parallel
3940 * @param transformer a function returning the transformation
3941 * for an element, or null if there is no transformation (in
3942 * which case it is not combined)
3943 * @param reducer a commutative associative combining function
3944 * @param <U> the return type of the transformer
3945 * @return the result of accumulating the given transformation
3946 * of all keys
3947 * @since 1.8
3948 */
3949 public <U> U reduceKeys(long parallelismThreshold,
3950 Function<? super K, ? extends U> transformer,
3951 BiFunction<? super U, ? super U, ? extends U> reducer) {
3952 if (transformer == null || reducer == null)
3953 throw new NullPointerException();
3954 return new MapReduceKeysTask<K,V,U>
3955 (null, batchFor(parallelismThreshold), 0, 0, table,
3956 null, transformer, reducer).invoke();
3957 }
3958
3959 /**
3960 * Returns the result of accumulating the given transformation
3961 * of all keys using the given reducer to combine values, and
3962 * the given basis as an identity value.
3963 *
3964 * @param parallelismThreshold the (estimated) number of elements
3965 * needed for this operation to be executed in parallel
3966 * @param transformer a function returning the transformation
3967 * for an element
3968 * @param basis the identity (initial default value) for the reduction
3969 * @param reducer a commutative associative combining function
3970 * @return the result of accumulating the given transformation
3971 * of all keys
3972 * @since 1.8
3973 */
3974 public double reduceKeysToDouble(long parallelismThreshold,
3975 ToDoubleFunction<? super K> transformer,
3976 double basis,
3977 DoubleBinaryOperator reducer) {
3978 if (transformer == null || reducer == null)
3979 throw new NullPointerException();
3980 return new MapReduceKeysToDoubleTask<K,V>
3981 (null, batchFor(parallelismThreshold), 0, 0, table,
3982 null, transformer, basis, reducer).invoke();
3983 }
3984
3985 /**
3986 * Returns the result of accumulating the given transformation
3987 * of all keys using the given reducer to combine values, and
3988 * the given basis as an identity value.
3989 *
3990 * @param parallelismThreshold the (estimated) number of elements
3991 * needed for this operation to be executed in parallel
3992 * @param transformer a function returning the transformation
3993 * for an element
3994 * @param basis the identity (initial default value) for the reduction
3995 * @param reducer a commutative associative combining function
3996 * @return the result of accumulating the given transformation
3997 * of all keys
3998 * @since 1.8
3999 */
4000 public long reduceKeysToLong(long parallelismThreshold,
4001 ToLongFunction<? super K> transformer,
4002 long basis,
4003 LongBinaryOperator reducer) {
4004 if (transformer == null || reducer == null)
4005 throw new NullPointerException();
4006 return new MapReduceKeysToLongTask<K,V>
4007 (null, batchFor(parallelismThreshold), 0, 0, table,
4008 null, transformer, basis, reducer).invoke();
4009 }
4010
4011 /**
4012 * Returns the result of accumulating the given transformation
4013 * of all keys using the given reducer to combine values, and
4014 * the given basis as an identity value.
4015 *
4016 * @param parallelismThreshold the (estimated) number of elements
4017 * needed for this operation to be executed in parallel
4018 * @param transformer a function returning the transformation
4019 * for an element
4020 * @param basis the identity (initial default value) for the reduction
4021 * @param reducer a commutative associative combining function
4022 * @return the result of accumulating the given transformation
4023 * of all keys
4024 * @since 1.8
4025 */
4026 public int reduceKeysToInt(long parallelismThreshold,
4027 ToIntFunction<? super K> transformer,
4028 int basis,
4029 IntBinaryOperator reducer) {
4030 if (transformer == null || reducer == null)
4031 throw new NullPointerException();
4032 return new MapReduceKeysToIntTask<K,V>
4033 (null, batchFor(parallelismThreshold), 0, 0, table,
4034 null, transformer, basis, reducer).invoke();
4035 }
4036
4037 /**
4038 * Performs the given action for each value.
4039 *
4040 * @param parallelismThreshold the (estimated) number of elements
4041 * needed for this operation to be executed in parallel
4042 * @param action the action
4043 * @since 1.8
4044 */
4045 public void forEachValue(long parallelismThreshold,
4046 Consumer<? super V> action) {
4047 if (action == null)
4048 throw new NullPointerException();
4049 new ForEachValueTask<K,V>
4050 (null, batchFor(parallelismThreshold), 0, 0, table,
4051 action).invoke();
4052 }
4053
4054 /**
4055 * Performs the given action for each non-null transformation
4056 * of each value.
4057 *
4058 * @param parallelismThreshold the (estimated) number of elements
4059 * needed for this operation to be executed in parallel
4060 * @param transformer a function returning the transformation
4061 * for an element, or null if there is no transformation (in
4062 * which case the action is not applied)
4063 * @param action the action
4064 * @param <U> the return type of the transformer
4065 * @since 1.8
4066 */
4067 public <U> void forEachValue(long parallelismThreshold,
4068 Function<? super V, ? extends U> transformer,
4069 Consumer<? super U> action) {
4070 if (transformer == null || action == null)
4071 throw new NullPointerException();
4072 new ForEachTransformedValueTask<K,V,U>
4073 (null, batchFor(parallelismThreshold), 0, 0, table,
4074 transformer, action).invoke();
4075 }
4076
4077 /**
4078 * Returns a non-null result from applying the given search
4079 * function on each value, or null if none. Upon success,
4080 * further element processing is suppressed and the results of
4081 * any other parallel invocations of the search function are
4082 * ignored.
4083 *
4084 * @param parallelismThreshold the (estimated) number of elements
4085 * needed for this operation to be executed in parallel
4086 * @param searchFunction a function returning a non-null
4087 * result on success, else null
4088 * @param <U> the return type of the search function
4089 * @return a non-null result from applying the given search
4090 * function on each value, or null if none
4091 * @since 1.8
4092 */
4093 public <U> U searchValues(long parallelismThreshold,
4094 Function<? super V, ? extends U> searchFunction) {
4095 if (searchFunction == null) throw new NullPointerException();
4096 return new SearchValuesTask<K,V,U>
4097 (null, batchFor(parallelismThreshold), 0, 0, table,
4098 searchFunction, new AtomicReference<U>()).invoke();
4099 }
4100
4101 /**
4102 * Returns the result of accumulating all values using the
4103 * given reducer to combine values, or null if none.
4104 *
4105 * @param parallelismThreshold the (estimated) number of elements
4106 * needed for this operation to be executed in parallel
4107 * @param reducer a commutative associative combining function
4108 * @return the result of accumulating all values
4109 * @since 1.8
4110 */
4111 public V reduceValues(long parallelismThreshold,
4112 BiFunction<? super V, ? super V, ? extends V> reducer) {
4113 if (reducer == null) throw new NullPointerException();
4114 return new ReduceValuesTask<K,V>
4115 (null, batchFor(parallelismThreshold), 0, 0, table,
4116 null, reducer).invoke();
4117 }
4118
4119 /**
4120 * Returns the result of accumulating the given transformation
4121 * of all values using the given reducer to combine values, or
4122 * null if none.
4123 *
4124 * @param parallelismThreshold the (estimated) number of elements
4125 * needed for this operation to be executed in parallel
4126 * @param transformer a function returning the transformation
4127 * for an element, or null if there is no transformation (in
4128 * which case it is not combined)
4129 * @param reducer a commutative associative combining function
4130 * @param <U> the return type of the transformer
4131 * @return the result of accumulating the given transformation
4132 * of all values
4133 * @since 1.8
4134 */
4135 public <U> U reduceValues(long parallelismThreshold,
4136 Function<? super V, ? extends U> transformer,
4137 BiFunction<? super U, ? super U, ? extends U> reducer) {
4138 if (transformer == null || reducer == null)
4139 throw new NullPointerException();
4140 return new MapReduceValuesTask<K,V,U>
4141 (null, batchFor(parallelismThreshold), 0, 0, table,
4142 null, transformer, reducer).invoke();
4143 }
4144
4145 /**
4146 * Returns the result of accumulating the given transformation
4147 * of all values using the given reducer to combine values,
4148 * and the given basis as an identity value.
4149 *
4150 * @param parallelismThreshold the (estimated) number of elements
4151 * needed for this operation to be executed in parallel
4152 * @param transformer a function returning the transformation
4153 * for an element
4154 * @param basis the identity (initial default value) for the reduction
4155 * @param reducer a commutative associative combining function
4156 * @return the result of accumulating the given transformation
4157 * of all values
4158 * @since 1.8
4159 */
4160 public double reduceValuesToDouble(long parallelismThreshold,
4161 ToDoubleFunction<? super V> transformer,
4162 double basis,
4163 DoubleBinaryOperator reducer) {
4164 if (transformer == null || reducer == null)
4165 throw new NullPointerException();
4166 return new MapReduceValuesToDoubleTask<K,V>
4167 (null, batchFor(parallelismThreshold), 0, 0, table,
4168 null, transformer, basis, reducer).invoke();
4169 }
4170
4171 /**
4172 * Returns the result of accumulating the given transformation
4173 * of all values using the given reducer to combine values,
4174 * and the given basis as an identity value.
4175 *
4176 * @param parallelismThreshold the (estimated) number of elements
4177 * needed for this operation to be executed in parallel
4178 * @param transformer a function returning the transformation
4179 * for an element
4180 * @param basis the identity (initial default value) for the reduction
4181 * @param reducer a commutative associative combining function
4182 * @return the result of accumulating the given transformation
4183 * of all values
4184 * @since 1.8
4185 */
4186 public long reduceValuesToLong(long parallelismThreshold,
4187 ToLongFunction<? super V> transformer,
4188 long basis,
4189 LongBinaryOperator reducer) {
4190 if (transformer == null || reducer == null)
4191 throw new NullPointerException();
4192 return new MapReduceValuesToLongTask<K,V>
4193 (null, batchFor(parallelismThreshold), 0, 0, table,
4194 null, transformer, basis, reducer).invoke();
4195 }
4196
4197 /**
4198 * Returns the result of accumulating the given transformation
4199 * of all values using the given reducer to combine values,
4200 * and the given basis as an identity value.
4201 *
4202 * @param parallelismThreshold the (estimated) number of elements
4203 * needed for this operation to be executed in parallel
4204 * @param transformer a function returning the transformation
4205 * for an element
4206 * @param basis the identity (initial default value) for the reduction
4207 * @param reducer a commutative associative combining function
4208 * @return the result of accumulating the given transformation
4209 * of all values
4210 * @since 1.8
4211 */
4212 public int reduceValuesToInt(long parallelismThreshold,
4213 ToIntFunction<? super V> transformer,
4214 int basis,
4215 IntBinaryOperator reducer) {
4216 if (transformer == null || reducer == null)
4217 throw new NullPointerException();
4218 return new MapReduceValuesToIntTask<K,V>
4219 (null, batchFor(parallelismThreshold), 0, 0, table,
4220 null, transformer, basis, reducer).invoke();
4221 }
4222
4223 /**
4224 * Performs the given action for each entry.
4225 *
4226 * @param parallelismThreshold the (estimated) number of elements
4227 * needed for this operation to be executed in parallel
4228 * @param action the action
4229 * @since 1.8
4230 */
4231 public void forEachEntry(long parallelismThreshold,
4232 Consumer<? super Map.Entry<K,V>> action) {
4233 if (action == null) throw new NullPointerException();
4234 new ForEachEntryTask<K,V>(null, batchFor(parallelismThreshold), 0, 0, table,
4235 action).invoke();
4236 }
4237
4238 /**
4239 * Performs the given action for each non-null transformation
4240 * of each entry.
4241 *
4242 * @param parallelismThreshold the (estimated) number of elements
4243 * needed for this operation to be executed in parallel
4244 * @param transformer a function returning the transformation
4245 * for an element, or null if there is no transformation (in
4246 * which case the action is not applied)
4247 * @param action the action
4248 * @param <U> the return type of the transformer
4249 * @since 1.8
4250 */
4251 public <U> void forEachEntry(long parallelismThreshold,
4252 Function<Map.Entry<K,V>, ? extends U> transformer,
4253 Consumer<? super U> action) {
4254 if (transformer == null || action == null)
4255 throw new NullPointerException();
4256 new ForEachTransformedEntryTask<K,V,U>
4257 (null, batchFor(parallelismThreshold), 0, 0, table,
4258 transformer, action).invoke();
4259 }
4260
4261 /**
4262 * Returns a non-null result from applying the given search
4263 * function on each entry, or null if none. Upon success,
4264 * further element processing is suppressed and the results of
4265 * any other parallel invocations of the search function are
4266 * ignored.
4267 *
4268 * @param parallelismThreshold the (estimated) number of elements
4269 * needed for this operation to be executed in parallel
4270 * @param searchFunction a function returning a non-null
4271 * result on success, else null
4272 * @param <U> the return type of the search function
4273 * @return a non-null result from applying the given search
4274 * function on each entry, or null if none
4275 * @since 1.8
4276 */
4277 public <U> U searchEntries(long parallelismThreshold,
4278 Function<Map.Entry<K,V>, ? extends U> searchFunction) {
4279 if (searchFunction == null) throw new NullPointerException();
4280 return new SearchEntriesTask<K,V,U>
4281 (null, batchFor(parallelismThreshold), 0, 0, table,
4282 searchFunction, new AtomicReference<U>()).invoke();
4283 }
4284
4285 /**
4286 * Returns the result of accumulating all entries using the
4287 * given reducer to combine values, or null if none.
4288 *
4289 * @param parallelismThreshold the (estimated) number of elements
4290 * needed for this operation to be executed in parallel
4291 * @param reducer a commutative associative combining function
4292 * @return the result of accumulating all entries
4293 * @since 1.8
4294 */
4295 public Map.Entry<K,V> reduceEntries(long parallelismThreshold,
4296 BiFunction<Map.Entry<K,V>, Map.Entry<K,V>, ? extends Map.Entry<K,V>> reducer) {
4297 if (reducer == null) throw new NullPointerException();
4298 return new ReduceEntriesTask<K,V>
4299 (null, batchFor(parallelismThreshold), 0, 0, table,
4300 null, reducer).invoke();
4301 }
4302
4303 /**
4304 * Returns the result of accumulating the given transformation
4305 * of all entries using the given reducer to combine values,
4306 * or null if none.
4307 *
4308 * @param parallelismThreshold the (estimated) number of elements
4309 * needed for this operation to be executed in parallel
4310 * @param transformer a function returning the transformation
4311 * for an element, or null if there is no transformation (in
4312 * which case it is not combined)
4313 * @param reducer a commutative associative combining function
4314 * @param <U> the return type of the transformer
4315 * @return the result of accumulating the given transformation
4316 * of all entries
4317 * @since 1.8
4318 */
4319 public <U> U reduceEntries(long parallelismThreshold,
4320 Function<Map.Entry<K,V>, ? extends U> transformer,
4321 BiFunction<? super U, ? super U, ? extends U> reducer) {
4322 if (transformer == null || reducer == null)
4323 throw new NullPointerException();
4324 return new MapReduceEntriesTask<K,V,U>
4325 (null, batchFor(parallelismThreshold), 0, 0, table,
4326 null, transformer, reducer).invoke();
4327 }
4328
4329 /**
4330 * Returns the result of accumulating the given transformation
4331 * of all entries using the given reducer to combine values,
4332 * and the given basis as an identity value.
4333 *
4334 * @param parallelismThreshold the (estimated) number of elements
4335 * needed for this operation to be executed in parallel
4336 * @param transformer a function returning the transformation
4337 * for an element
4338 * @param basis the identity (initial default value) for the reduction
4339 * @param reducer a commutative associative combining function
4340 * @return the result of accumulating the given transformation
4341 * of all entries
4342 * @since 1.8
4343 */
4344 public double reduceEntriesToDouble(long parallelismThreshold,
4345 ToDoubleFunction<Map.Entry<K,V>> transformer,
4346 double basis,
4347 DoubleBinaryOperator reducer) {
4348 if (transformer == null || reducer == null)
4349 throw new NullPointerException();
4350 return new MapReduceEntriesToDoubleTask<K,V>
4351 (null, batchFor(parallelismThreshold), 0, 0, table,
4352 null, transformer, basis, reducer).invoke();
4353 }
4354
4355 /**
4356 * Returns the result of accumulating the given transformation
4357 * of all entries using the given reducer to combine values,
4358 * and the given basis as an identity value.
4359 *
4360 * @param parallelismThreshold the (estimated) number of elements
4361 * needed for this operation to be executed in parallel
4362 * @param transformer a function returning the transformation
4363 * for an element
4364 * @param basis the identity (initial default value) for the reduction
4365 * @param reducer a commutative associative combining function
4366 * @return the result of accumulating the given transformation
4367 * of all entries
4368 * @since 1.8
4369 */
4370 public long reduceEntriesToLong(long parallelismThreshold,
4371 ToLongFunction<Map.Entry<K,V>> transformer,
4372 long basis,
4373 LongBinaryOperator reducer) {
4374 if (transformer == null || reducer == null)
4375 throw new NullPointerException();
4376 return new MapReduceEntriesToLongTask<K,V>
4377 (null, batchFor(parallelismThreshold), 0, 0, table,
4378 null, transformer, basis, reducer).invoke();
4379 }
4380
4381 /**
4382 * Returns the result of accumulating the given transformation
4383 * of all entries using the given reducer to combine values,
4384 * and the given basis as an identity value.
4385 *
4386 * @param parallelismThreshold the (estimated) number of elements
4387 * needed for this operation to be executed in parallel
4388 * @param transformer a function returning the transformation
4389 * for an element
4390 * @param basis the identity (initial default value) for the reduction
4391 * @param reducer a commutative associative combining function
4392 * @return the result of accumulating the given transformation
4393 * of all entries
4394 * @since 1.8
4395 */
4396 public int reduceEntriesToInt(long parallelismThreshold,
4397 ToIntFunction<Map.Entry<K,V>> transformer,
4398 int basis,
4399 IntBinaryOperator reducer) {
4400 if (transformer == null || reducer == null)
4401 throw new NullPointerException();
4402 return new MapReduceEntriesToIntTask<K,V>
4403 (null, batchFor(parallelismThreshold), 0, 0, table,
4404 null, transformer, basis, reducer).invoke();
4405 }
4406
4407
4408 /* ----------------Views -------------- */
4409
4410 /**
4411 * Base class for views.
4412 */
4413 abstract static class CollectionView<K,V,E>
4414 implements Collection<E>, java.io.Serializable {
4415 private static final long serialVersionUID = 7249069246763182397L;
4416 final ConcurrentHashMap<K,V> map;
4417 CollectionView(ConcurrentHashMap<K,V> map) { this.map = map; }
4418
4419 /**
4420 * Returns the map backing this view.
4421 *
4422 * @return the map backing this view
4423 */
4424 public ConcurrentHashMap<K,V> getMap() { return map; }
4425
4426 /**
4427 * Removes all of the elements from this view, by removing all
4428 * the mappings from the map backing this view.
4429 */
4430 public final void clear() { map.clear(); }
4431 public final int size() { return map.size(); }
4432 public final boolean isEmpty() { return map.isEmpty(); }
4433
4434 // implementations below rely on concrete classes supplying these
4435 // abstract methods
4436 /**
4437 * Returns an iterator over the elements in this collection.
4438 *
4439 * <p>The returned iterator is
4440 * <a href="package-summary.html#Weakly"><i>weakly consistent</i></a>.
4441 *
4442 * @return an iterator over the elements in this collection
4443 */
4444 public abstract Iterator<E> iterator();
4445 public abstract boolean contains(Object o);
4446 public abstract boolean remove(Object o);
4447
4448 private static final String oomeMsg = "Required array size too large";
4449
4450 public final Object[] toArray() {
4451 long sz = map.mappingCount();
4452 if (sz > MAX_ARRAY_SIZE)
4453 throw new OutOfMemoryError(oomeMsg);
4454 int n = (int)sz;
4455 Object[] r = new Object[n];
4456 int i = 0;
4457 for (E e : this) {
4458 if (i == n) {
4459 if (n >= MAX_ARRAY_SIZE)
4460 throw new OutOfMemoryError(oomeMsg);
4461 if (n >= MAX_ARRAY_SIZE - (MAX_ARRAY_SIZE >>> 1) - 1)
4462 n = MAX_ARRAY_SIZE;
4463 else
4464 n += (n >>> 1) + 1;
4465 r = Arrays.copyOf(r, n);
4466 }
4467 r[i++] = e;
4468 }
4469 return (i == n) ? r : Arrays.copyOf(r, i);
4470 }
4471
4472 @SuppressWarnings("unchecked")
4473 public final <T> T[] toArray(T[] a) {
4474 long sz = map.mappingCount();
4475 if (sz > MAX_ARRAY_SIZE)
4476 throw new OutOfMemoryError(oomeMsg);
4477 int m = (int)sz;
4478 T[] r = (a.length >= m) ? a :
4479 (T[])java.lang.reflect.Array
4480 .newInstance(a.getClass().getComponentType(), m);
4481 int n = r.length;
4482 int i = 0;
4483 for (E e : this) {
4484 if (i == n) {
4485 if (n >= MAX_ARRAY_SIZE)
4486 throw new OutOfMemoryError(oomeMsg);
4487 if (n >= MAX_ARRAY_SIZE - (MAX_ARRAY_SIZE >>> 1) - 1)
4488 n = MAX_ARRAY_SIZE;
4489 else
4490 n += (n >>> 1) + 1;
4491 r = Arrays.copyOf(r, n);
4492 }
4493 r[i++] = (T)e;
4494 }
4495 if (a == r && i < n) {
4496 r[i] = null; // null-terminate
4497 return r;
4498 }
4499 return (i == n) ? r : Arrays.copyOf(r, i);
4500 }
4501
4502 /**
4503 * Returns a string representation of this collection.
4504 * The string representation consists of the string representations
4505 * of the collection's elements in the order they are returned by
4506 * its iterator, enclosed in square brackets ({@code "[]"}).
4507 * Adjacent elements are separated by the characters {@code ", "}
4508 * (comma and space). Elements are converted to strings as by
4509 * {@link String#valueOf(Object)}.
4510 *
4511 * @return a string representation of this collection
4512 */
4513 public final String toString() {
4514 StringBuilder sb = new StringBuilder();
4515 sb.append('[');
4516 Iterator<E> it = iterator();
4517 if (it.hasNext()) {
4518 for (;;) {
4519 Object e = it.next();
4520 sb.append(e == this ? "(this Collection)" : e);
4521 if (!it.hasNext())
4522 break;
4523 sb.append(',').append(' ');
4524 }
4525 }
4526 return sb.append(']').toString();
4527 }
4528
4529 public final boolean containsAll(Collection<?> c) {
4530 if (c != this) {
4531 for (Object e : c) {
4532 if (e == null || !contains(e))
4533 return false;
4534 }
4535 }
4536 return true;
4537 }
4538
4539 public final boolean removeAll(Collection<?> c) {
4540 if (c == null) throw new NullPointerException();
4541 boolean modified = false;
4542 for (Iterator<E> it = iterator(); it.hasNext();) {
4543 if (c.contains(it.next())) {
4544 it.remove();
4545 modified = true;
4546 }
4547 }
4548 return modified;
4549 }
4550
4551 public final boolean retainAll(Collection<?> c) {
4552 if (c == null) throw new NullPointerException();
4553 boolean modified = false;
4554 for (Iterator<E> it = iterator(); it.hasNext();) {
4555 if (!c.contains(it.next())) {
4556 it.remove();
4557 modified = true;
4558 }
4559 }
4560 return modified;
4561 }
4562
4563 }
4564
4565 /**
4566 * A view of a ConcurrentHashMap as a {@link Set} of keys, in
4567 * which additions may optionally be enabled by mapping to a
4568 * common value. This class cannot be directly instantiated.
4569 * See {@link #keySet() keySet()},
4570 * {@link #keySet(Object) keySet(V)},
4571 * {@link #newKeySet() newKeySet()},
4572 * {@link #newKeySet(int) newKeySet(int)}.
4573 *
4574 * @since 1.8
4575 */
4576 public static class KeySetView<K,V> extends CollectionView<K,V,K>
4577 implements Set<K>, java.io.Serializable {
4578 private static final long serialVersionUID = 7249069246763182397L;
4579 private final V value;
4580 KeySetView(ConcurrentHashMap<K,V> map, V value) { // non-public
4581 super(map);
4582 this.value = value;
4583 }
4584
4585 /**
4586 * Returns the default mapped value for additions,
4587 * or {@code null} if additions are not supported.
4588 *
4589 * @return the default mapped value for additions, or {@code null}
4590 * if not supported
4591 */
4592 public V getMappedValue() { return value; }
4593
4594 /**
4595 * {@inheritDoc}
4596 * @throws NullPointerException if the specified key is null
4597 */
4598 public boolean contains(Object o) { return map.containsKey(o); }
4599
4600 /**
4601 * Removes the key from this map view, by removing the key (and its
4602 * corresponding value) from the backing map. This method does
4603 * nothing if the key is not in the map.
4604 *
4605 * @param o the key to be removed from the backing map
4606 * @return {@code true} if the backing map contained the specified key
4607 * @throws NullPointerException if the specified key is null
4608 */
4609 public boolean remove(Object o) { return map.remove(o) != null; }
4610
4611 /**
4612 * @return an iterator over the keys of the backing map
4613 */
4614 public Iterator<K> iterator() {
4615 Node<K,V>[] t;
4616 ConcurrentHashMap<K,V> m = map;
4617 int f = (t = m.table) == null ? 0 : t.length;
4618 return new KeyIterator<K,V>(t, f, 0, f, m);
4619 }
4620
4621 /**
4622 * Adds the specified key to this set view by mapping the key to
4623 * the default mapped value in the backing map, if defined.
4624 *
4625 * @param e key to be added
4626 * @return {@code true} if this set changed as a result of the call
4627 * @throws NullPointerException if the specified key is null
4628 * @throws UnsupportedOperationException if no default mapped value
4629 * for additions was provided
4630 */
4631 public boolean add(K e) {
4632 V v;
4633 if ((v = value) == null)
4634 throw new UnsupportedOperationException();
4635 return map.putVal(e, v, true) == null;
4636 }
4637
4638 /**
4639 * Adds all of the elements in the specified collection to this set,
4640 * as if by calling {@link #add} on each one.
4641 *
4642 * @param c the elements to be inserted into this set
4643 * @return {@code true} if this set changed as a result of the call
4644 * @throws NullPointerException if the collection or any of its
4645 * elements are {@code null}
4646 * @throws UnsupportedOperationException if no default mapped value
4647 * for additions was provided
4648 */
4649 public boolean addAll(Collection<? extends K> c) {
4650 boolean added = false;
4651 V v;
4652 if ((v = value) == null)
4653 throw new UnsupportedOperationException();
4654 for (K e : c) {
4655 if (map.putVal(e, v, true) == null)
4656 added = true;
4657 }
4658 return added;
4659 }
4660
4661 public int hashCode() {
4662 int h = 0;
4663 for (K e : this)
4664 h += e.hashCode();
4665 return h;
4666 }
4667
4668 public boolean equals(Object o) {
4669 Set<?> c;
4670 return ((o instanceof Set) &&
4671 ((c = (Set<?>)o) == this ||
4672 (containsAll(c) && c.containsAll(this))));
4673 }
4674
4675 public Spliterator<K> spliterator() {
4676 Node<K,V>[] t;
4677 ConcurrentHashMap<K,V> m = map;
4678 long n = m.sumCount();
4679 int f = (t = m.table) == null ? 0 : t.length;
4680 return new KeySpliterator<K,V>(t, f, 0, f, n < 0L ? 0L : n);
4681 }
4682
4683 public void forEach(Consumer<? super K> action) {
4684 if (action == null) throw new NullPointerException();
4685 Node<K,V>[] t;
4686 if ((t = map.table) != null) {
4687 Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length);
4688 for (Node<K,V> p; (p = it.advance()) != null; )
4689 action.accept(p.key);
4690 }
4691 }
4692 }
4693
4694 /**
4695 * A view of a ConcurrentHashMap as a {@link Collection} of
4696 * values, in which additions are disabled. This class cannot be
4697 * directly instantiated. See {@link #values()}.
4698 */
4699 static final class ValuesView<K,V> extends CollectionView<K,V,V>
4700 implements Collection<V>, java.io.Serializable {
4701 private static final long serialVersionUID = 2249069246763182397L;
4702 ValuesView(ConcurrentHashMap<K,V> map) { super(map); }
4703 public final boolean contains(Object o) {
4704 return map.containsValue(o);
4705 }
4706
4707 public final boolean remove(Object o) {
4708 if (o != null) {
4709 for (Iterator<V> it = iterator(); it.hasNext();) {
4710 if (o.equals(it.next())) {
4711 it.remove();
4712 return true;
4713 }
4714 }
4715 }
4716 return false;
4717 }
4718
4719 public final Iterator<V> iterator() {
4720 ConcurrentHashMap<K,V> m = map;
4721 Node<K,V>[] t;
4722 int f = (t = m.table) == null ? 0 : t.length;
4723 return new ValueIterator<K,V>(t, f, 0, f, m);
4724 }
4725
4726 public final boolean add(V e) {
4727 throw new UnsupportedOperationException();
4728 }
4729 public final boolean addAll(Collection<? extends V> c) {
4730 throw new UnsupportedOperationException();
4731 }
4732
4733 public boolean removeIf(Predicate<? super V> filter) {
4734 return map.removeValueIf(filter);
4735 }
4736
4737 public Spliterator<V> spliterator() {
4738 Node<K,V>[] t;
4739 ConcurrentHashMap<K,V> m = map;
4740 long n = m.sumCount();
4741 int f = (t = m.table) == null ? 0 : t.length;
4742 return new ValueSpliterator<K,V>(t, f, 0, f, n < 0L ? 0L : n);
4743 }
4744
4745 public void forEach(Consumer<? super V> action) {
4746 if (action == null) throw new NullPointerException();
4747 Node<K,V>[] t;
4748 if ((t = map.table) != null) {
4749 Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length);
4750 for (Node<K,V> p; (p = it.advance()) != null; )
4751 action.accept(p.val);
4752 }
4753 }
4754 }
4755
4756 /**
4757 * A view of a ConcurrentHashMap as a {@link Set} of (key, value)
4758 * entries. This class cannot be directly instantiated. See
4759 * {@link #entrySet()}.
4760 */
4761 static final class EntrySetView<K,V> extends CollectionView<K,V,Map.Entry<K,V>>
4762 implements Set<Map.Entry<K,V>>, java.io.Serializable {
4763 private static final long serialVersionUID = 2249069246763182397L;
4764 EntrySetView(ConcurrentHashMap<K,V> map) { super(map); }
4765
4766 public boolean contains(Object o) {
4767 Object k, v, r; Map.Entry<?,?> e;
4768 return ((o instanceof Map.Entry) &&
4769 (k = (e = (Map.Entry<?,?>)o).getKey()) != null &&
4770 (r = map.get(k)) != null &&
4771 (v = e.getValue()) != null &&
4772 (v == r || v.equals(r)));
4773 }
4774
4775 public boolean remove(Object o) {
4776 Object k, v; Map.Entry<?,?> e;
4777 return ((o instanceof Map.Entry) &&
4778 (k = (e = (Map.Entry<?,?>)o).getKey()) != null &&
4779 (v = e.getValue()) != null &&
4780 map.remove(k, v));
4781 }
4782
4783 /**
4784 * @return an iterator over the entries of the backing map
4785 */
4786 public Iterator<Map.Entry<K,V>> iterator() {
4787 ConcurrentHashMap<K,V> m = map;
4788 Node<K,V>[] t;
4789 int f = (t = m.table) == null ? 0 : t.length;
4790 return new EntryIterator<K,V>(t, f, 0, f, m);
4791 }
4792
4793 public boolean add(Entry<K,V> e) {
4794 return map.putVal(e.getKey(), e.getValue(), false) == null;
4795 }
4796
4797 public boolean addAll(Collection<? extends Entry<K,V>> c) {
4798 boolean added = false;
4799 for (Entry<K,V> e : c) {
4800 if (add(e))
4801 added = true;
4802 }
4803 return added;
4804 }
4805
4806 public boolean removeIf(Predicate<? super Entry<K, V>> filter) {
4807 return map.removeEntryIf(filter);
4808 }
4809
4810 public final int hashCode() {
4811 int h = 0;
4812 Node<K,V>[] t;
4813 if ((t = map.table) != null) {
4814 Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length);
4815 for (Node<K,V> p; (p = it.advance()) != null; ) {
4816 h += p.hashCode();
4817 }
4818 }
4819 return h;
4820 }
4821
4822 public final boolean equals(Object o) {
4823 Set<?> c;
4824 return ((o instanceof Set) &&
4825 ((c = (Set<?>)o) == this ||
4826 (containsAll(c) && c.containsAll(this))));
4827 }
4828
4829 public Spliterator<Map.Entry<K,V>> spliterator() {
4830 Node<K,V>[] t;
4831 ConcurrentHashMap<K,V> m = map;
4832 long n = m.sumCount();
4833 int f = (t = m.table) == null ? 0 : t.length;
4834 return new EntrySpliterator<K,V>(t, f, 0, f, n < 0L ? 0L : n, m);
4835 }
4836
4837 public void forEach(Consumer<? super Map.Entry<K,V>> action) {
4838 if (action == null) throw new NullPointerException();
4839 Node<K,V>[] t;
4840 if ((t = map.table) != null) {
4841 Traverser<K,V> it = new Traverser<K,V>(t, t.length, 0, t.length);
4842 for (Node<K,V> p; (p = it.advance()) != null; )
4843 action.accept(new MapEntry<K,V>(p.key, p.val, map));
4844 }
4845 }
4846
4847 }
4848
4849 // -------------------------------------------------------
4850
4851 /**
4852 * Base class for bulk tasks. Repeats some fields and code from
4853 * class Traverser, because we need to subclass CountedCompleter.
4854 */
4855 @SuppressWarnings("serial")
4856 abstract static class BulkTask<K,V,R> extends CountedCompleter<R> {
4857 Node<K,V>[] tab; // same as Traverser
4858 Node<K,V> next;
4859 TableStack<K,V> stack, spare;
4860 int index;
4861 int baseIndex;
4862 int baseLimit;
4863 final int baseSize;
4864 int batch; // split control
4865
4866 BulkTask(BulkTask<K,V,?> par, int b, int i, int f, Node<K,V>[] t) {
4867 super(par);
4868 this.batch = b;
4869 this.index = this.baseIndex = i;
4870 if ((this.tab = t) == null)
4871 this.baseSize = this.baseLimit = 0;
4872 else if (par == null)
4873 this.baseSize = this.baseLimit = t.length;
4874 else {
4875 this.baseLimit = f;
4876 this.baseSize = par.baseSize;
4877 }
4878 }
4879
4880 /**
4881 * Same as Traverser version
4882 */
4883 final Node<K,V> advance() {
4884 Node<K,V> e;
4885 if ((e = next) != null)
4886 e = e.next;
4887 for (;;) {
4888 Node<K,V>[] t; int i, n;
4889 if (e != null)
4890 return next = e;
4891 if (baseIndex >= baseLimit || (t = tab) == null ||
4892 (n = t.length) <= (i = index) || i < 0)
4893 return next = null;
4894 if ((e = tabAt(t, i)) != null && e.hash < 0) {
4895 if (e instanceof ForwardingNode) {
4896 tab = ((ForwardingNode<K,V>)e).nextTable;
4897 e = null;
4898 pushState(t, i, n);
4899 continue;
4900 }
4901 else if (e instanceof TreeBin)
4902 e = ((TreeBin<K,V>)e).first;
4903 else
4904 e = null;
4905 }
4906 if (stack != null)
4907 recoverState(n);
4908 else if ((index = i + baseSize) >= n)
4909 index = ++baseIndex;
4910 }
4911 }
4912
4913 private void pushState(Node<K,V>[] t, int i, int n) {
4914 TableStack<K,V> s = spare;
4915 if (s != null)
4916 spare = s.next;
4917 else
4918 s = new TableStack<K,V>();
4919 s.tab = t;
4920 s.length = n;
4921 s.index = i;
4922 s.next = stack;
4923 stack = s;
4924 }
4925
4926 private void recoverState(int n) {
4927 TableStack<K,V> s; int len;
4928 while ((s = stack) != null && (index += (len = s.length)) >= n) {
4929 n = len;
4930 index = s.index;
4931 tab = s.tab;
4932 s.tab = null;
4933 TableStack<K,V> next = s.next;
4934 s.next = spare; // save for reuse
4935 stack = next;
4936 spare = s;
4937 }
4938 if (s == null && (index += baseSize) >= n)
4939 index = ++baseIndex;
4940 }
4941 }
4942
4943 /*
4944 * Task classes. Coded in a regular but ugly format/style to
4945 * simplify checks that each variant differs in the right way from
4946 * others. The null screenings exist because compilers cannot tell
4947 * that we've already null-checked task arguments, so we force
4948 * simplest hoisted bypass to help avoid convoluted traps.
4949 */
4950 @SuppressWarnings("serial")
4951 static final class ForEachKeyTask<K,V>
4952 extends BulkTask<K,V,Void> {
4953 final Consumer<? super K> action;
4954 ForEachKeyTask
4955 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
4956 Consumer<? super K> action) {
4957 super(p, b, i, f, t);
4958 this.action = action;
4959 }
4960 public final void compute() {
4961 final Consumer<? super K> action;
4962 if ((action = this.action) != null) {
4963 for (int i = baseIndex, f, h; batch > 0 &&
4964 (h = ((f = baseLimit) + i) >>> 1) > i;) {
4965 addToPendingCount(1);
4966 new ForEachKeyTask<K,V>
4967 (this, batch >>>= 1, baseLimit = h, f, tab,
4968 action).fork();
4969 }
4970 for (Node<K,V> p; (p = advance()) != null;)
4971 action.accept(p.key);
4972 propagateCompletion();
4973 }
4974 }
4975 }
4976
4977 @SuppressWarnings("serial")
4978 static final class ForEachValueTask<K,V>
4979 extends BulkTask<K,V,Void> {
4980 final Consumer<? super V> action;
4981 ForEachValueTask
4982 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
4983 Consumer<? super V> action) {
4984 super(p, b, i, f, t);
4985 this.action = action;
4986 }
4987 public final void compute() {
4988 final Consumer<? super V> action;
4989 if ((action = this.action) != null) {
4990 for (int i = baseIndex, f, h; batch > 0 &&
4991 (h = ((f = baseLimit) + i) >>> 1) > i;) {
4992 addToPendingCount(1);
4993 new ForEachValueTask<K,V>
4994 (this, batch >>>= 1, baseLimit = h, f, tab,
4995 action).fork();
4996 }
4997 for (Node<K,V> p; (p = advance()) != null;)
4998 action.accept(p.val);
4999 propagateCompletion();
5000 }
5001 }
5002 }
5003
5004 @SuppressWarnings("serial")
5005 static final class ForEachEntryTask<K,V>
5006 extends BulkTask<K,V,Void> {
5007 final Consumer<? super Entry<K,V>> action;
5008 ForEachEntryTask
5009 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5010 Consumer<? super Entry<K,V>> action) {
5011 super(p, b, i, f, t);
5012 this.action = action;
5013 }
5014 public final void compute() {
5015 final Consumer<? super Entry<K,V>> action;
5016 if ((action = this.action) != null) {
5017 for (int i = baseIndex, f, h; batch > 0 &&
5018 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5019 addToPendingCount(1);
5020 new ForEachEntryTask<K,V>
5021 (this, batch >>>= 1, baseLimit = h, f, tab,
5022 action).fork();
5023 }
5024 for (Node<K,V> p; (p = advance()) != null; )
5025 action.accept(p);
5026 propagateCompletion();
5027 }
5028 }
5029 }
5030
5031 @SuppressWarnings("serial")
5032 static final class ForEachMappingTask<K,V>
5033 extends BulkTask<K,V,Void> {
5034 final BiConsumer<? super K, ? super V> action;
5035 ForEachMappingTask
5036 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5037 BiConsumer<? super K,? super V> action) {
5038 super(p, b, i, f, t);
5039 this.action = action;
5040 }
5041 public final void compute() {
5042 final BiConsumer<? super K, ? super V> action;
5043 if ((action = this.action) != null) {
5044 for (int i = baseIndex, f, h; batch > 0 &&
5045 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5046 addToPendingCount(1);
5047 new ForEachMappingTask<K,V>
5048 (this, batch >>>= 1, baseLimit = h, f, tab,
5049 action).fork();
5050 }
5051 for (Node<K,V> p; (p = advance()) != null; )
5052 action.accept(p.key, p.val);
5053 propagateCompletion();
5054 }
5055 }
5056 }
5057
5058 @SuppressWarnings("serial")
5059 static final class ForEachTransformedKeyTask<K,V,U>
5060 extends BulkTask<K,V,Void> {
5061 final Function<? super K, ? extends U> transformer;
5062 final Consumer<? super U> action;
5063 ForEachTransformedKeyTask
5064 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5065 Function<? super K, ? extends U> transformer, Consumer<? super U> action) {
5066 super(p, b, i, f, t);
5067 this.transformer = transformer; this.action = action;
5068 }
5069 public final void compute() {
5070 final Function<? super K, ? extends U> transformer;
5071 final Consumer<? super U> action;
5072 if ((transformer = this.transformer) != null &&
5073 (action = this.action) != null) {
5074 for (int i = baseIndex, f, h; batch > 0 &&
5075 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5076 addToPendingCount(1);
5077 new ForEachTransformedKeyTask<K,V,U>
5078 (this, batch >>>= 1, baseLimit = h, f, tab,
5079 transformer, action).fork();
5080 }
5081 for (Node<K,V> p; (p = advance()) != null; ) {
5082 U u;
5083 if ((u = transformer.apply(p.key)) != null)
5084 action.accept(u);
5085 }
5086 propagateCompletion();
5087 }
5088 }
5089 }
5090
5091 @SuppressWarnings("serial")
5092 static final class ForEachTransformedValueTask<K,V,U>
5093 extends BulkTask<K,V,Void> {
5094 final Function<? super V, ? extends U> transformer;
5095 final Consumer<? super U> action;
5096 ForEachTransformedValueTask
5097 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5098 Function<? super V, ? extends U> transformer, Consumer<? super U> action) {
5099 super(p, b, i, f, t);
5100 this.transformer = transformer; this.action = action;
5101 }
5102 public final void compute() {
5103 final Function<? super V, ? extends U> transformer;
5104 final Consumer<? super U> action;
5105 if ((transformer = this.transformer) != null &&
5106 (action = this.action) != null) {
5107 for (int i = baseIndex, f, h; batch > 0 &&
5108 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5109 addToPendingCount(1);
5110 new ForEachTransformedValueTask<K,V,U>
5111 (this, batch >>>= 1, baseLimit = h, f, tab,
5112 transformer, action).fork();
5113 }
5114 for (Node<K,V> p; (p = advance()) != null; ) {
5115 U u;
5116 if ((u = transformer.apply(p.val)) != null)
5117 action.accept(u);
5118 }
5119 propagateCompletion();
5120 }
5121 }
5122 }
5123
5124 @SuppressWarnings("serial")
5125 static final class ForEachTransformedEntryTask<K,V,U>
5126 extends BulkTask<K,V,Void> {
5127 final Function<Map.Entry<K,V>, ? extends U> transformer;
5128 final Consumer<? super U> action;
5129 ForEachTransformedEntryTask
5130 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5131 Function<Map.Entry<K,V>, ? extends U> transformer, Consumer<? super U> action) {
5132 super(p, b, i, f, t);
5133 this.transformer = transformer; this.action = action;
5134 }
5135 public final void compute() {
5136 final Function<Map.Entry<K,V>, ? extends U> transformer;
5137 final Consumer<? super U> action;
5138 if ((transformer = this.transformer) != null &&
5139 (action = this.action) != null) {
5140 for (int i = baseIndex, f, h; batch > 0 &&
5141 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5142 addToPendingCount(1);
5143 new ForEachTransformedEntryTask<K,V,U>
5144 (this, batch >>>= 1, baseLimit = h, f, tab,
5145 transformer, action).fork();
5146 }
5147 for (Node<K,V> p; (p = advance()) != null; ) {
5148 U u;
5149 if ((u = transformer.apply(p)) != null)
5150 action.accept(u);
5151 }
5152 propagateCompletion();
5153 }
5154 }
5155 }
5156
5157 @SuppressWarnings("serial")
5158 static final class ForEachTransformedMappingTask<K,V,U>
5159 extends BulkTask<K,V,Void> {
5160 final BiFunction<? super K, ? super V, ? extends U> transformer;
5161 final Consumer<? super U> action;
5162 ForEachTransformedMappingTask
5163 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5164 BiFunction<? super K, ? super V, ? extends U> transformer,
5165 Consumer<? super U> action) {
5166 super(p, b, i, f, t);
5167 this.transformer = transformer; this.action = action;
5168 }
5169 public final void compute() {
5170 final BiFunction<? super K, ? super V, ? extends U> transformer;
5171 final Consumer<? super U> action;
5172 if ((transformer = this.transformer) != null &&
5173 (action = this.action) != null) {
5174 for (int i = baseIndex, f, h; batch > 0 &&
5175 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5176 addToPendingCount(1);
5177 new ForEachTransformedMappingTask<K,V,U>
5178 (this, batch >>>= 1, baseLimit = h, f, tab,
5179 transformer, action).fork();
5180 }
5181 for (Node<K,V> p; (p = advance()) != null; ) {
5182 U u;
5183 if ((u = transformer.apply(p.key, p.val)) != null)
5184 action.accept(u);
5185 }
5186 propagateCompletion();
5187 }
5188 }
5189 }
5190
5191 @SuppressWarnings("serial")
5192 static final class SearchKeysTask<K,V,U>
5193 extends BulkTask<K,V,U> {
5194 final Function<? super K, ? extends U> searchFunction;
5195 final AtomicReference<U> result;
5196 SearchKeysTask
5197 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5198 Function<? super K, ? extends U> searchFunction,
5199 AtomicReference<U> result) {
5200 super(p, b, i, f, t);
5201 this.searchFunction = searchFunction; this.result = result;
5202 }
5203 public final U getRawResult() { return result.get(); }
5204 public final void compute() {
5205 final Function<? super K, ? extends U> searchFunction;
5206 final AtomicReference<U> result;
5207 if ((searchFunction = this.searchFunction) != null &&
5208 (result = this.result) != null) {
5209 for (int i = baseIndex, f, h; batch > 0 &&
5210 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5211 if (result.get() != null)
5212 return;
5213 addToPendingCount(1);
5214 new SearchKeysTask<K,V,U>
5215 (this, batch >>>= 1, baseLimit = h, f, tab,
5216 searchFunction, result).fork();
5217 }
5218 while (result.get() == null) {
5219 U u;
5220 Node<K,V> p;
5221 if ((p = advance()) == null) {
5222 propagateCompletion();
5223 break;
5224 }
5225 if ((u = searchFunction.apply(p.key)) != null) {
5226 if (result.compareAndSet(null, u))
5227 quietlyCompleteRoot();
5228 break;
5229 }
5230 }
5231 }
5232 }
5233 }
5234
5235 @SuppressWarnings("serial")
5236 static final class SearchValuesTask<K,V,U>
5237 extends BulkTask<K,V,U> {
5238 final Function<? super V, ? extends U> searchFunction;
5239 final AtomicReference<U> result;
5240 SearchValuesTask
5241 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5242 Function<? super V, ? extends U> searchFunction,
5243 AtomicReference<U> result) {
5244 super(p, b, i, f, t);
5245 this.searchFunction = searchFunction; this.result = result;
5246 }
5247 public final U getRawResult() { return result.get(); }
5248 public final void compute() {
5249 final Function<? super V, ? extends U> searchFunction;
5250 final AtomicReference<U> result;
5251 if ((searchFunction = this.searchFunction) != null &&
5252 (result = this.result) != null) {
5253 for (int i = baseIndex, f, h; batch > 0 &&
5254 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5255 if (result.get() != null)
5256 return;
5257 addToPendingCount(1);
5258 new SearchValuesTask<K,V,U>
5259 (this, batch >>>= 1, baseLimit = h, f, tab,
5260 searchFunction, result).fork();
5261 }
5262 while (result.get() == null) {
5263 U u;
5264 Node<K,V> p;
5265 if ((p = advance()) == null) {
5266 propagateCompletion();
5267 break;
5268 }
5269 if ((u = searchFunction.apply(p.val)) != null) {
5270 if (result.compareAndSet(null, u))
5271 quietlyCompleteRoot();
5272 break;
5273 }
5274 }
5275 }
5276 }
5277 }
5278
5279 @SuppressWarnings("serial")
5280 static final class SearchEntriesTask<K,V,U>
5281 extends BulkTask<K,V,U> {
5282 final Function<Entry<K,V>, ? extends U> searchFunction;
5283 final AtomicReference<U> result;
5284 SearchEntriesTask
5285 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5286 Function<Entry<K,V>, ? extends U> searchFunction,
5287 AtomicReference<U> result) {
5288 super(p, b, i, f, t);
5289 this.searchFunction = searchFunction; this.result = result;
5290 }
5291 public final U getRawResult() { return result.get(); }
5292 public final void compute() {
5293 final Function<Entry<K,V>, ? extends U> searchFunction;
5294 final AtomicReference<U> result;
5295 if ((searchFunction = this.searchFunction) != null &&
5296 (result = this.result) != null) {
5297 for (int i = baseIndex, f, h; batch > 0 &&
5298 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5299 if (result.get() != null)
5300 return;
5301 addToPendingCount(1);
5302 new SearchEntriesTask<K,V,U>
5303 (this, batch >>>= 1, baseLimit = h, f, tab,
5304 searchFunction, result).fork();
5305 }
5306 while (result.get() == null) {
5307 U u;
5308 Node<K,V> p;
5309 if ((p = advance()) == null) {
5310 propagateCompletion();
5311 break;
5312 }
5313 if ((u = searchFunction.apply(p)) != null) {
5314 if (result.compareAndSet(null, u))
5315 quietlyCompleteRoot();
5316 return;
5317 }
5318 }
5319 }
5320 }
5321 }
5322
5323 @SuppressWarnings("serial")
5324 static final class SearchMappingsTask<K,V,U>
5325 extends BulkTask<K,V,U> {
5326 final BiFunction<? super K, ? super V, ? extends U> searchFunction;
5327 final AtomicReference<U> result;
5328 SearchMappingsTask
5329 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5330 BiFunction<? super K, ? super V, ? extends U> searchFunction,
5331 AtomicReference<U> result) {
5332 super(p, b, i, f, t);
5333 this.searchFunction = searchFunction; this.result = result;
5334 }
5335 public final U getRawResult() { return result.get(); }
5336 public final void compute() {
5337 final BiFunction<? super K, ? super V, ? extends U> searchFunction;
5338 final AtomicReference<U> result;
5339 if ((searchFunction = this.searchFunction) != null &&
5340 (result = this.result) != null) {
5341 for (int i = baseIndex, f, h; batch > 0 &&
5342 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5343 if (result.get() != null)
5344 return;
5345 addToPendingCount(1);
5346 new SearchMappingsTask<K,V,U>
5347 (this, batch >>>= 1, baseLimit = h, f, tab,
5348 searchFunction, result).fork();
5349 }
5350 while (result.get() == null) {
5351 U u;
5352 Node<K,V> p;
5353 if ((p = advance()) == null) {
5354 propagateCompletion();
5355 break;
5356 }
5357 if ((u = searchFunction.apply(p.key, p.val)) != null) {
5358 if (result.compareAndSet(null, u))
5359 quietlyCompleteRoot();
5360 break;
5361 }
5362 }
5363 }
5364 }
5365 }
5366
5367 @SuppressWarnings("serial")
5368 static final class ReduceKeysTask<K,V>
5369 extends BulkTask<K,V,K> {
5370 final BiFunction<? super K, ? super K, ? extends K> reducer;
5371 K result;
5372 ReduceKeysTask<K,V> rights, nextRight;
5373 ReduceKeysTask
5374 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5375 ReduceKeysTask<K,V> nextRight,
5376 BiFunction<? super K, ? super K, ? extends K> reducer) {
5377 super(p, b, i, f, t); this.nextRight = nextRight;
5378 this.reducer = reducer;
5379 }
5380 public final K getRawResult() { return result; }
5381 public final void compute() {
5382 final BiFunction<? super K, ? super K, ? extends K> reducer;
5383 if ((reducer = this.reducer) != null) {
5384 for (int i = baseIndex, f, h; batch > 0 &&
5385 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5386 addToPendingCount(1);
5387 (rights = new ReduceKeysTask<K,V>
5388 (this, batch >>>= 1, baseLimit = h, f, tab,
5389 rights, reducer)).fork();
5390 }
5391 K r = null;
5392 for (Node<K,V> p; (p = advance()) != null; ) {
5393 K u = p.key;
5394 r = (r == null) ? u : u == null ? r : reducer.apply(r, u);
5395 }
5396 result = r;
5397 CountedCompleter<?> c;
5398 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5399 @SuppressWarnings("unchecked")
5400 ReduceKeysTask<K,V>
5401 t = (ReduceKeysTask<K,V>)c,
5402 s = t.rights;
5403 while (s != null) {
5404 K tr, sr;
5405 if ((sr = s.result) != null)
5406 t.result = (((tr = t.result) == null) ? sr :
5407 reducer.apply(tr, sr));
5408 s = t.rights = s.nextRight;
5409 }
5410 }
5411 }
5412 }
5413 }
5414
5415 @SuppressWarnings("serial")
5416 static final class ReduceValuesTask<K,V>
5417 extends BulkTask<K,V,V> {
5418 final BiFunction<? super V, ? super V, ? extends V> reducer;
5419 V result;
5420 ReduceValuesTask<K,V> rights, nextRight;
5421 ReduceValuesTask
5422 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5423 ReduceValuesTask<K,V> nextRight,
5424 BiFunction<? super V, ? super V, ? extends V> reducer) {
5425 super(p, b, i, f, t); this.nextRight = nextRight;
5426 this.reducer = reducer;
5427 }
5428 public final V getRawResult() { return result; }
5429 public final void compute() {
5430 final BiFunction<? super V, ? super V, ? extends V> reducer;
5431 if ((reducer = this.reducer) != null) {
5432 for (int i = baseIndex, f, h; batch > 0 &&
5433 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5434 addToPendingCount(1);
5435 (rights = new ReduceValuesTask<K,V>
5436 (this, batch >>>= 1, baseLimit = h, f, tab,
5437 rights, reducer)).fork();
5438 }
5439 V r = null;
5440 for (Node<K,V> p; (p = advance()) != null; ) {
5441 V v = p.val;
5442 r = (r == null) ? v : reducer.apply(r, v);
5443 }
5444 result = r;
5445 CountedCompleter<?> c;
5446 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5447 @SuppressWarnings("unchecked")
5448 ReduceValuesTask<K,V>
5449 t = (ReduceValuesTask<K,V>)c,
5450 s = t.rights;
5451 while (s != null) {
5452 V tr, sr;
5453 if ((sr = s.result) != null)
5454 t.result = (((tr = t.result) == null) ? sr :
5455 reducer.apply(tr, sr));
5456 s = t.rights = s.nextRight;
5457 }
5458 }
5459 }
5460 }
5461 }
5462
5463 @SuppressWarnings("serial")
5464 static final class ReduceEntriesTask<K,V>
5465 extends BulkTask<K,V,Map.Entry<K,V>> {
5466 final BiFunction<Map.Entry<K,V>, Map.Entry<K,V>, ? extends Map.Entry<K,V>> reducer;
5467 Map.Entry<K,V> result;
5468 ReduceEntriesTask<K,V> rights, nextRight;
5469 ReduceEntriesTask
5470 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5471 ReduceEntriesTask<K,V> nextRight,
5472 BiFunction<Entry<K,V>, Map.Entry<K,V>, ? extends Map.Entry<K,V>> reducer) {
5473 super(p, b, i, f, t); this.nextRight = nextRight;
5474 this.reducer = reducer;
5475 }
5476 public final Map.Entry<K,V> getRawResult() { return result; }
5477 public final void compute() {
5478 final BiFunction<Map.Entry<K,V>, Map.Entry<K,V>, ? extends Map.Entry<K,V>> reducer;
5479 if ((reducer = this.reducer) != null) {
5480 for (int i = baseIndex, f, h; batch > 0 &&
5481 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5482 addToPendingCount(1);
5483 (rights = new ReduceEntriesTask<K,V>
5484 (this, batch >>>= 1, baseLimit = h, f, tab,
5485 rights, reducer)).fork();
5486 }
5487 Map.Entry<K,V> r = null;
5488 for (Node<K,V> p; (p = advance()) != null; )
5489 r = (r == null) ? p : reducer.apply(r, p);
5490 result = r;
5491 CountedCompleter<?> c;
5492 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5493 @SuppressWarnings("unchecked")
5494 ReduceEntriesTask<K,V>
5495 t = (ReduceEntriesTask<K,V>)c,
5496 s = t.rights;
5497 while (s != null) {
5498 Map.Entry<K,V> tr, sr;
5499 if ((sr = s.result) != null)
5500 t.result = (((tr = t.result) == null) ? sr :
5501 reducer.apply(tr, sr));
5502 s = t.rights = s.nextRight;
5503 }
5504 }
5505 }
5506 }
5507 }
5508
5509 @SuppressWarnings("serial")
5510 static final class MapReduceKeysTask<K,V,U>
5511 extends BulkTask<K,V,U> {
5512 final Function<? super K, ? extends U> transformer;
5513 final BiFunction<? super U, ? super U, ? extends U> reducer;
5514 U result;
5515 MapReduceKeysTask<K,V,U> rights, nextRight;
5516 MapReduceKeysTask
5517 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5518 MapReduceKeysTask<K,V,U> nextRight,
5519 Function<? super K, ? extends U> transformer,
5520 BiFunction<? super U, ? super U, ? extends U> reducer) {
5521 super(p, b, i, f, t); this.nextRight = nextRight;
5522 this.transformer = transformer;
5523 this.reducer = reducer;
5524 }
5525 public final U getRawResult() { return result; }
5526 public final void compute() {
5527 final Function<? super K, ? extends U> transformer;
5528 final BiFunction<? super U, ? super U, ? extends U> reducer;
5529 if ((transformer = this.transformer) != null &&
5530 (reducer = this.reducer) != null) {
5531 for (int i = baseIndex, f, h; batch > 0 &&
5532 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5533 addToPendingCount(1);
5534 (rights = new MapReduceKeysTask<K,V,U>
5535 (this, batch >>>= 1, baseLimit = h, f, tab,
5536 rights, transformer, reducer)).fork();
5537 }
5538 U r = null;
5539 for (Node<K,V> p; (p = advance()) != null; ) {
5540 U u;
5541 if ((u = transformer.apply(p.key)) != null)
5542 r = (r == null) ? u : reducer.apply(r, u);
5543 }
5544 result = r;
5545 CountedCompleter<?> c;
5546 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5547 @SuppressWarnings("unchecked")
5548 MapReduceKeysTask<K,V,U>
5549 t = (MapReduceKeysTask<K,V,U>)c,
5550 s = t.rights;
5551 while (s != null) {
5552 U tr, sr;
5553 if ((sr = s.result) != null)
5554 t.result = (((tr = t.result) == null) ? sr :
5555 reducer.apply(tr, sr));
5556 s = t.rights = s.nextRight;
5557 }
5558 }
5559 }
5560 }
5561 }
5562
5563 @SuppressWarnings("serial")
5564 static final class MapReduceValuesTask<K,V,U>
5565 extends BulkTask<K,V,U> {
5566 final Function<? super V, ? extends U> transformer;
5567 final BiFunction<? super U, ? super U, ? extends U> reducer;
5568 U result;
5569 MapReduceValuesTask<K,V,U> rights, nextRight;
5570 MapReduceValuesTask
5571 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5572 MapReduceValuesTask<K,V,U> nextRight,
5573 Function<? super V, ? extends U> transformer,
5574 BiFunction<? super U, ? super U, ? extends U> reducer) {
5575 super(p, b, i, f, t); this.nextRight = nextRight;
5576 this.transformer = transformer;
5577 this.reducer = reducer;
5578 }
5579 public final U getRawResult() { return result; }
5580 public final void compute() {
5581 final Function<? super V, ? extends U> transformer;
5582 final BiFunction<? super U, ? super U, ? extends U> reducer;
5583 if ((transformer = this.transformer) != null &&
5584 (reducer = this.reducer) != null) {
5585 for (int i = baseIndex, f, h; batch > 0 &&
5586 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5587 addToPendingCount(1);
5588 (rights = new MapReduceValuesTask<K,V,U>
5589 (this, batch >>>= 1, baseLimit = h, f, tab,
5590 rights, transformer, reducer)).fork();
5591 }
5592 U r = null;
5593 for (Node<K,V> p; (p = advance()) != null; ) {
5594 U u;
5595 if ((u = transformer.apply(p.val)) != null)
5596 r = (r == null) ? u : reducer.apply(r, u);
5597 }
5598 result = r;
5599 CountedCompleter<?> c;
5600 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5601 @SuppressWarnings("unchecked")
5602 MapReduceValuesTask<K,V,U>
5603 t = (MapReduceValuesTask<K,V,U>)c,
5604 s = t.rights;
5605 while (s != null) {
5606 U tr, sr;
5607 if ((sr = s.result) != null)
5608 t.result = (((tr = t.result) == null) ? sr :
5609 reducer.apply(tr, sr));
5610 s = t.rights = s.nextRight;
5611 }
5612 }
5613 }
5614 }
5615 }
5616
5617 @SuppressWarnings("serial")
5618 static final class MapReduceEntriesTask<K,V,U>
5619 extends BulkTask<K,V,U> {
5620 final Function<Map.Entry<K,V>, ? extends U> transformer;
5621 final BiFunction<? super U, ? super U, ? extends U> reducer;
5622 U result;
5623 MapReduceEntriesTask<K,V,U> rights, nextRight;
5624 MapReduceEntriesTask
5625 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5626 MapReduceEntriesTask<K,V,U> nextRight,
5627 Function<Map.Entry<K,V>, ? extends U> transformer,
5628 BiFunction<? super U, ? super U, ? extends U> reducer) {
5629 super(p, b, i, f, t); this.nextRight = nextRight;
5630 this.transformer = transformer;
5631 this.reducer = reducer;
5632 }
5633 public final U getRawResult() { return result; }
5634 public final void compute() {
5635 final Function<Map.Entry<K,V>, ? extends U> transformer;
5636 final BiFunction<? super U, ? super U, ? extends U> reducer;
5637 if ((transformer = this.transformer) != null &&
5638 (reducer = this.reducer) != null) {
5639 for (int i = baseIndex, f, h; batch > 0 &&
5640 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5641 addToPendingCount(1);
5642 (rights = new MapReduceEntriesTask<K,V,U>
5643 (this, batch >>>= 1, baseLimit = h, f, tab,
5644 rights, transformer, reducer)).fork();
5645 }
5646 U r = null;
5647 for (Node<K,V> p; (p = advance()) != null; ) {
5648 U u;
5649 if ((u = transformer.apply(p)) != null)
5650 r = (r == null) ? u : reducer.apply(r, u);
5651 }
5652 result = r;
5653 CountedCompleter<?> c;
5654 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5655 @SuppressWarnings("unchecked")
5656 MapReduceEntriesTask<K,V,U>
5657 t = (MapReduceEntriesTask<K,V,U>)c,
5658 s = t.rights;
5659 while (s != null) {
5660 U tr, sr;
5661 if ((sr = s.result) != null)
5662 t.result = (((tr = t.result) == null) ? sr :
5663 reducer.apply(tr, sr));
5664 s = t.rights = s.nextRight;
5665 }
5666 }
5667 }
5668 }
5669 }
5670
5671 @SuppressWarnings("serial")
5672 static final class MapReduceMappingsTask<K,V,U>
5673 extends BulkTask<K,V,U> {
5674 final BiFunction<? super K, ? super V, ? extends U> transformer;
5675 final BiFunction<? super U, ? super U, ? extends U> reducer;
5676 U result;
5677 MapReduceMappingsTask<K,V,U> rights, nextRight;
5678 MapReduceMappingsTask
5679 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5680 MapReduceMappingsTask<K,V,U> nextRight,
5681 BiFunction<? super K, ? super V, ? extends U> transformer,
5682 BiFunction<? super U, ? super U, ? extends U> reducer) {
5683 super(p, b, i, f, t); this.nextRight = nextRight;
5684 this.transformer = transformer;
5685 this.reducer = reducer;
5686 }
5687 public final U getRawResult() { return result; }
5688 public final void compute() {
5689 final BiFunction<? super K, ? super V, ? extends U> transformer;
5690 final BiFunction<? super U, ? super U, ? extends U> reducer;
5691 if ((transformer = this.transformer) != null &&
5692 (reducer = this.reducer) != null) {
5693 for (int i = baseIndex, f, h; batch > 0 &&
5694 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5695 addToPendingCount(1);
5696 (rights = new MapReduceMappingsTask<K,V,U>
5697 (this, batch >>>= 1, baseLimit = h, f, tab,
5698 rights, transformer, reducer)).fork();
5699 }
5700 U r = null;
5701 for (Node<K,V> p; (p = advance()) != null; ) {
5702 U u;
5703 if ((u = transformer.apply(p.key, p.val)) != null)
5704 r = (r == null) ? u : reducer.apply(r, u);
5705 }
5706 result = r;
5707 CountedCompleter<?> c;
5708 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5709 @SuppressWarnings("unchecked")
5710 MapReduceMappingsTask<K,V,U>
5711 t = (MapReduceMappingsTask<K,V,U>)c,
5712 s = t.rights;
5713 while (s != null) {
5714 U tr, sr;
5715 if ((sr = s.result) != null)
5716 t.result = (((tr = t.result) == null) ? sr :
5717 reducer.apply(tr, sr));
5718 s = t.rights = s.nextRight;
5719 }
5720 }
5721 }
5722 }
5723 }
5724
5725 @SuppressWarnings("serial")
5726 static final class MapReduceKeysToDoubleTask<K,V>
5727 extends BulkTask<K,V,Double> {
5728 final ToDoubleFunction<? super K> transformer;
5729 final DoubleBinaryOperator reducer;
5730 final double basis;
5731 double result;
5732 MapReduceKeysToDoubleTask<K,V> rights, nextRight;
5733 MapReduceKeysToDoubleTask
5734 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5735 MapReduceKeysToDoubleTask<K,V> nextRight,
5736 ToDoubleFunction<? super K> transformer,
5737 double basis,
5738 DoubleBinaryOperator reducer) {
5739 super(p, b, i, f, t); this.nextRight = nextRight;
5740 this.transformer = transformer;
5741 this.basis = basis; this.reducer = reducer;
5742 }
5743 public final Double getRawResult() { return result; }
5744 public final void compute() {
5745 final ToDoubleFunction<? super K> transformer;
5746 final DoubleBinaryOperator reducer;
5747 if ((transformer = this.transformer) != null &&
5748 (reducer = this.reducer) != null) {
5749 double r = this.basis;
5750 for (int i = baseIndex, f, h; batch > 0 &&
5751 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5752 addToPendingCount(1);
5753 (rights = new MapReduceKeysToDoubleTask<K,V>
5754 (this, batch >>>= 1, baseLimit = h, f, tab,
5755 rights, transformer, r, reducer)).fork();
5756 }
5757 for (Node<K,V> p; (p = advance()) != null; )
5758 r = reducer.applyAsDouble(r, transformer.applyAsDouble(p.key));
5759 result = r;
5760 CountedCompleter<?> c;
5761 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5762 @SuppressWarnings("unchecked")
5763 MapReduceKeysToDoubleTask<K,V>
5764 t = (MapReduceKeysToDoubleTask<K,V>)c,
5765 s = t.rights;
5766 while (s != null) {
5767 t.result = reducer.applyAsDouble(t.result, s.result);
5768 s = t.rights = s.nextRight;
5769 }
5770 }
5771 }
5772 }
5773 }
5774
5775 @SuppressWarnings("serial")
5776 static final class MapReduceValuesToDoubleTask<K,V>
5777 extends BulkTask<K,V,Double> {
5778 final ToDoubleFunction<? super V> transformer;
5779 final DoubleBinaryOperator reducer;
5780 final double basis;
5781 double result;
5782 MapReduceValuesToDoubleTask<K,V> rights, nextRight;
5783 MapReduceValuesToDoubleTask
5784 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5785 MapReduceValuesToDoubleTask<K,V> nextRight,
5786 ToDoubleFunction<? super V> transformer,
5787 double basis,
5788 DoubleBinaryOperator reducer) {
5789 super(p, b, i, f, t); this.nextRight = nextRight;
5790 this.transformer = transformer;
5791 this.basis = basis; this.reducer = reducer;
5792 }
5793 public final Double getRawResult() { return result; }
5794 public final void compute() {
5795 final ToDoubleFunction<? super V> transformer;
5796 final DoubleBinaryOperator reducer;
5797 if ((transformer = this.transformer) != null &&
5798 (reducer = this.reducer) != null) {
5799 double r = this.basis;
5800 for (int i = baseIndex, f, h; batch > 0 &&
5801 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5802 addToPendingCount(1);
5803 (rights = new MapReduceValuesToDoubleTask<K,V>
5804 (this, batch >>>= 1, baseLimit = h, f, tab,
5805 rights, transformer, r, reducer)).fork();
5806 }
5807 for (Node<K,V> p; (p = advance()) != null; )
5808 r = reducer.applyAsDouble(r, transformer.applyAsDouble(p.val));
5809 result = r;
5810 CountedCompleter<?> c;
5811 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5812 @SuppressWarnings("unchecked")
5813 MapReduceValuesToDoubleTask<K,V>
5814 t = (MapReduceValuesToDoubleTask<K,V>)c,
5815 s = t.rights;
5816 while (s != null) {
5817 t.result = reducer.applyAsDouble(t.result, s.result);
5818 s = t.rights = s.nextRight;
5819 }
5820 }
5821 }
5822 }
5823 }
5824
5825 @SuppressWarnings("serial")
5826 static final class MapReduceEntriesToDoubleTask<K,V>
5827 extends BulkTask<K,V,Double> {
5828 final ToDoubleFunction<Map.Entry<K,V>> transformer;
5829 final DoubleBinaryOperator reducer;
5830 final double basis;
5831 double result;
5832 MapReduceEntriesToDoubleTask<K,V> rights, nextRight;
5833 MapReduceEntriesToDoubleTask
5834 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5835 MapReduceEntriesToDoubleTask<K,V> nextRight,
5836 ToDoubleFunction<Map.Entry<K,V>> transformer,
5837 double basis,
5838 DoubleBinaryOperator reducer) {
5839 super(p, b, i, f, t); this.nextRight = nextRight;
5840 this.transformer = transformer;
5841 this.basis = basis; this.reducer = reducer;
5842 }
5843 public final Double getRawResult() { return result; }
5844 public final void compute() {
5845 final ToDoubleFunction<Map.Entry<K,V>> transformer;
5846 final DoubleBinaryOperator reducer;
5847 if ((transformer = this.transformer) != null &&
5848 (reducer = this.reducer) != null) {
5849 double r = this.basis;
5850 for (int i = baseIndex, f, h; batch > 0 &&
5851 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5852 addToPendingCount(1);
5853 (rights = new MapReduceEntriesToDoubleTask<K,V>
5854 (this, batch >>>= 1, baseLimit = h, f, tab,
5855 rights, transformer, r, reducer)).fork();
5856 }
5857 for (Node<K,V> p; (p = advance()) != null; )
5858 r = reducer.applyAsDouble(r, transformer.applyAsDouble(p));
5859 result = r;
5860 CountedCompleter<?> c;
5861 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5862 @SuppressWarnings("unchecked")
5863 MapReduceEntriesToDoubleTask<K,V>
5864 t = (MapReduceEntriesToDoubleTask<K,V>)c,
5865 s = t.rights;
5866 while (s != null) {
5867 t.result = reducer.applyAsDouble(t.result, s.result);
5868 s = t.rights = s.nextRight;
5869 }
5870 }
5871 }
5872 }
5873 }
5874
5875 @SuppressWarnings("serial")
5876 static final class MapReduceMappingsToDoubleTask<K,V>
5877 extends BulkTask<K,V,Double> {
5878 final ToDoubleBiFunction<? super K, ? super V> transformer;
5879 final DoubleBinaryOperator reducer;
5880 final double basis;
5881 double result;
5882 MapReduceMappingsToDoubleTask<K,V> rights, nextRight;
5883 MapReduceMappingsToDoubleTask
5884 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5885 MapReduceMappingsToDoubleTask<K,V> nextRight,
5886 ToDoubleBiFunction<? super K, ? super V> transformer,
5887 double basis,
5888 DoubleBinaryOperator reducer) {
5889 super(p, b, i, f, t); this.nextRight = nextRight;
5890 this.transformer = transformer;
5891 this.basis = basis; this.reducer = reducer;
5892 }
5893 public final Double getRawResult() { return result; }
5894 public final void compute() {
5895 final ToDoubleBiFunction<? super K, ? super V> transformer;
5896 final DoubleBinaryOperator reducer;
5897 if ((transformer = this.transformer) != null &&
5898 (reducer = this.reducer) != null) {
5899 double r = this.basis;
5900 for (int i = baseIndex, f, h; batch > 0 &&
5901 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5902 addToPendingCount(1);
5903 (rights = new MapReduceMappingsToDoubleTask<K,V>
5904 (this, batch >>>= 1, baseLimit = h, f, tab,
5905 rights, transformer, r, reducer)).fork();
5906 }
5907 for (Node<K,V> p; (p = advance()) != null; )
5908 r = reducer.applyAsDouble(r, transformer.applyAsDouble(p.key, p.val));
5909 result = r;
5910 CountedCompleter<?> c;
5911 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5912 @SuppressWarnings("unchecked")
5913 MapReduceMappingsToDoubleTask<K,V>
5914 t = (MapReduceMappingsToDoubleTask<K,V>)c,
5915 s = t.rights;
5916 while (s != null) {
5917 t.result = reducer.applyAsDouble(t.result, s.result);
5918 s = t.rights = s.nextRight;
5919 }
5920 }
5921 }
5922 }
5923 }
5924
5925 @SuppressWarnings("serial")
5926 static final class MapReduceKeysToLongTask<K,V>
5927 extends BulkTask<K,V,Long> {
5928 final ToLongFunction<? super K> transformer;
5929 final LongBinaryOperator reducer;
5930 final long basis;
5931 long result;
5932 MapReduceKeysToLongTask<K,V> rights, nextRight;
5933 MapReduceKeysToLongTask
5934 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5935 MapReduceKeysToLongTask<K,V> nextRight,
5936 ToLongFunction<? super K> transformer,
5937 long basis,
5938 LongBinaryOperator reducer) {
5939 super(p, b, i, f, t); this.nextRight = nextRight;
5940 this.transformer = transformer;
5941 this.basis = basis; this.reducer = reducer;
5942 }
5943 public final Long getRawResult() { return result; }
5944 public final void compute() {
5945 final ToLongFunction<? super K> transformer;
5946 final LongBinaryOperator reducer;
5947 if ((transformer = this.transformer) != null &&
5948 (reducer = this.reducer) != null) {
5949 long r = this.basis;
5950 for (int i = baseIndex, f, h; batch > 0 &&
5951 (h = ((f = baseLimit) + i) >>> 1) > i;) {
5952 addToPendingCount(1);
5953 (rights = new MapReduceKeysToLongTask<K,V>
5954 (this, batch >>>= 1, baseLimit = h, f, tab,
5955 rights, transformer, r, reducer)).fork();
5956 }
5957 for (Node<K,V> p; (p = advance()) != null; )
5958 r = reducer.applyAsLong(r, transformer.applyAsLong(p.key));
5959 result = r;
5960 CountedCompleter<?> c;
5961 for (c = firstComplete(); c != null; c = c.nextComplete()) {
5962 @SuppressWarnings("unchecked")
5963 MapReduceKeysToLongTask<K,V>
5964 t = (MapReduceKeysToLongTask<K,V>)c,
5965 s = t.rights;
5966 while (s != null) {
5967 t.result = reducer.applyAsLong(t.result, s.result);
5968 s = t.rights = s.nextRight;
5969 }
5970 }
5971 }
5972 }
5973 }
5974
5975 @SuppressWarnings("serial")
5976 static final class MapReduceValuesToLongTask<K,V>
5977 extends BulkTask<K,V,Long> {
5978 final ToLongFunction<? super V> transformer;
5979 final LongBinaryOperator reducer;
5980 final long basis;
5981 long result;
5982 MapReduceValuesToLongTask<K,V> rights, nextRight;
5983 MapReduceValuesToLongTask
5984 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
5985 MapReduceValuesToLongTask<K,V> nextRight,
5986 ToLongFunction<? super V> transformer,
5987 long basis,
5988 LongBinaryOperator reducer) {
5989 super(p, b, i, f, t); this.nextRight = nextRight;
5990 this.transformer = transformer;
5991 this.basis = basis; this.reducer = reducer;
5992 }
5993 public final Long getRawResult() { return result; }
5994 public final void compute() {
5995 final ToLongFunction<? super V> transformer;
5996 final LongBinaryOperator reducer;
5997 if ((transformer = this.transformer) != null &&
5998 (reducer = this.reducer) != null) {
5999 long r = this.basis;
6000 for (int i = baseIndex, f, h; batch > 0 &&
6001 (h = ((f = baseLimit) + i) >>> 1) > i;) {
6002 addToPendingCount(1);
6003 (rights = new MapReduceValuesToLongTask<K,V>
6004 (this, batch >>>= 1, baseLimit = h, f, tab,
6005 rights, transformer, r, reducer)).fork();
6006 }
6007 for (Node<K,V> p; (p = advance()) != null; )
6008 r = reducer.applyAsLong(r, transformer.applyAsLong(p.val));
6009 result = r;
6010 CountedCompleter<?> c;
6011 for (c = firstComplete(); c != null; c = c.nextComplete()) {
6012 @SuppressWarnings("unchecked")
6013 MapReduceValuesToLongTask<K,V>
6014 t = (MapReduceValuesToLongTask<K,V>)c,
6015 s = t.rights;
6016 while (s != null) {
6017 t.result = reducer.applyAsLong(t.result, s.result);
6018 s = t.rights = s.nextRight;
6019 }
6020 }
6021 }
6022 }
6023 }
6024
6025 @SuppressWarnings("serial")
6026 static final class MapReduceEntriesToLongTask<K,V>
6027 extends BulkTask<K,V,Long> {
6028 final ToLongFunction<Map.Entry<K,V>> transformer;
6029 final LongBinaryOperator reducer;
6030 final long basis;
6031 long result;
6032 MapReduceEntriesToLongTask<K,V> rights, nextRight;
6033 MapReduceEntriesToLongTask
6034 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
6035 MapReduceEntriesToLongTask<K,V> nextRight,
6036 ToLongFunction<Map.Entry<K,V>> transformer,
6037 long basis,
6038 LongBinaryOperator reducer) {
6039 super(p, b, i, f, t); this.nextRight = nextRight;
6040 this.transformer = transformer;
6041 this.basis = basis; this.reducer = reducer;
6042 }
6043 public final Long getRawResult() { return result; }
6044 public final void compute() {
6045 final ToLongFunction<Map.Entry<K,V>> transformer;
6046 final LongBinaryOperator reducer;
6047 if ((transformer = this.transformer) != null &&
6048 (reducer = this.reducer) != null) {
6049 long r = this.basis;
6050 for (int i = baseIndex, f, h; batch > 0 &&
6051 (h = ((f = baseLimit) + i) >>> 1) > i;) {
6052 addToPendingCount(1);
6053 (rights = new MapReduceEntriesToLongTask<K,V>
6054 (this, batch >>>= 1, baseLimit = h, f, tab,
6055 rights, transformer, r, reducer)).fork();
6056 }
6057 for (Node<K,V> p; (p = advance()) != null; )
6058 r = reducer.applyAsLong(r, transformer.applyAsLong(p));
6059 result = r;
6060 CountedCompleter<?> c;
6061 for (c = firstComplete(); c != null; c = c.nextComplete()) {
6062 @SuppressWarnings("unchecked")
6063 MapReduceEntriesToLongTask<K,V>
6064 t = (MapReduceEntriesToLongTask<K,V>)c,
6065 s = t.rights;
6066 while (s != null) {
6067 t.result = reducer.applyAsLong(t.result, s.result);
6068 s = t.rights = s.nextRight;
6069 }
6070 }
6071 }
6072 }
6073 }
6074
6075 @SuppressWarnings("serial")
6076 static final class MapReduceMappingsToLongTask<K,V>
6077 extends BulkTask<K,V,Long> {
6078 final ToLongBiFunction<? super K, ? super V> transformer;
6079 final LongBinaryOperator reducer;
6080 final long basis;
6081 long result;
6082 MapReduceMappingsToLongTask<K,V> rights, nextRight;
6083 MapReduceMappingsToLongTask
6084 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
6085 MapReduceMappingsToLongTask<K,V> nextRight,
6086 ToLongBiFunction<? super K, ? super V> transformer,
6087 long basis,
6088 LongBinaryOperator reducer) {
6089 super(p, b, i, f, t); this.nextRight = nextRight;
6090 this.transformer = transformer;
6091 this.basis = basis; this.reducer = reducer;
6092 }
6093 public final Long getRawResult() { return result; }
6094 public final void compute() {
6095 final ToLongBiFunction<? super K, ? super V> transformer;
6096 final LongBinaryOperator reducer;
6097 if ((transformer = this.transformer) != null &&
6098 (reducer = this.reducer) != null) {
6099 long r = this.basis;
6100 for (int i = baseIndex, f, h; batch > 0 &&
6101 (h = ((f = baseLimit) + i) >>> 1) > i;) {
6102 addToPendingCount(1);
6103 (rights = new MapReduceMappingsToLongTask<K,V>
6104 (this, batch >>>= 1, baseLimit = h, f, tab,
6105 rights, transformer, r, reducer)).fork();
6106 }
6107 for (Node<K,V> p; (p = advance()) != null; )
6108 r = reducer.applyAsLong(r, transformer.applyAsLong(p.key, p.val));
6109 result = r;
6110 CountedCompleter<?> c;
6111 for (c = firstComplete(); c != null; c = c.nextComplete()) {
6112 @SuppressWarnings("unchecked")
6113 MapReduceMappingsToLongTask<K,V>
6114 t = (MapReduceMappingsToLongTask<K,V>)c,
6115 s = t.rights;
6116 while (s != null) {
6117 t.result = reducer.applyAsLong(t.result, s.result);
6118 s = t.rights = s.nextRight;
6119 }
6120 }
6121 }
6122 }
6123 }
6124
6125 @SuppressWarnings("serial")
6126 static final class MapReduceKeysToIntTask<K,V>
6127 extends BulkTask<K,V,Integer> {
6128 final ToIntFunction<? super K> transformer;
6129 final IntBinaryOperator reducer;
6130 final int basis;
6131 int result;
6132 MapReduceKeysToIntTask<K,V> rights, nextRight;
6133 MapReduceKeysToIntTask
6134 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
6135 MapReduceKeysToIntTask<K,V> nextRight,
6136 ToIntFunction<? super K> transformer,
6137 int basis,
6138 IntBinaryOperator reducer) {
6139 super(p, b, i, f, t); this.nextRight = nextRight;
6140 this.transformer = transformer;
6141 this.basis = basis; this.reducer = reducer;
6142 }
6143 public final Integer getRawResult() { return result; }
6144 public final void compute() {
6145 final ToIntFunction<? super K> transformer;
6146 final IntBinaryOperator reducer;
6147 if ((transformer = this.transformer) != null &&
6148 (reducer = this.reducer) != null) {
6149 int r = this.basis;
6150 for (int i = baseIndex, f, h; batch > 0 &&
6151 (h = ((f = baseLimit) + i) >>> 1) > i;) {
6152 addToPendingCount(1);
6153 (rights = new MapReduceKeysToIntTask<K,V>
6154 (this, batch >>>= 1, baseLimit = h, f, tab,
6155 rights, transformer, r, reducer)).fork();
6156 }
6157 for (Node<K,V> p; (p = advance()) != null; )
6158 r = reducer.applyAsInt(r, transformer.applyAsInt(p.key));
6159 result = r;
6160 CountedCompleter<?> c;
6161 for (c = firstComplete(); c != null; c = c.nextComplete()) {
6162 @SuppressWarnings("unchecked")
6163 MapReduceKeysToIntTask<K,V>
6164 t = (MapReduceKeysToIntTask<K,V>)c,
6165 s = t.rights;
6166 while (s != null) {
6167 t.result = reducer.applyAsInt(t.result, s.result);
6168 s = t.rights = s.nextRight;
6169 }
6170 }
6171 }
6172 }
6173 }
6174
6175 @SuppressWarnings("serial")
6176 static final class MapReduceValuesToIntTask<K,V>
6177 extends BulkTask<K,V,Integer> {
6178 final ToIntFunction<? super V> transformer;
6179 final IntBinaryOperator reducer;
6180 final int basis;
6181 int result;
6182 MapReduceValuesToIntTask<K,V> rights, nextRight;
6183 MapReduceValuesToIntTask
6184 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
6185 MapReduceValuesToIntTask<K,V> nextRight,
6186 ToIntFunction<? super V> transformer,
6187 int basis,
6188 IntBinaryOperator reducer) {
6189 super(p, b, i, f, t); this.nextRight = nextRight;
6190 this.transformer = transformer;
6191 this.basis = basis; this.reducer = reducer;
6192 }
6193 public final Integer getRawResult() { return result; }
6194 public final void compute() {
6195 final ToIntFunction<? super V> transformer;
6196 final IntBinaryOperator reducer;
6197 if ((transformer = this.transformer) != null &&
6198 (reducer = this.reducer) != null) {
6199 int r = this.basis;
6200 for (int i = baseIndex, f, h; batch > 0 &&
6201 (h = ((f = baseLimit) + i) >>> 1) > i;) {
6202 addToPendingCount(1);
6203 (rights = new MapReduceValuesToIntTask<K,V>
6204 (this, batch >>>= 1, baseLimit = h, f, tab,
6205 rights, transformer, r, reducer)).fork();
6206 }
6207 for (Node<K,V> p; (p = advance()) != null; )
6208 r = reducer.applyAsInt(r, transformer.applyAsInt(p.val));
6209 result = r;
6210 CountedCompleter<?> c;
6211 for (c = firstComplete(); c != null; c = c.nextComplete()) {
6212 @SuppressWarnings("unchecked")
6213 MapReduceValuesToIntTask<K,V>
6214 t = (MapReduceValuesToIntTask<K,V>)c,
6215 s = t.rights;
6216 while (s != null) {
6217 t.result = reducer.applyAsInt(t.result, s.result);
6218 s = t.rights = s.nextRight;
6219 }
6220 }
6221 }
6222 }
6223 }
6224
6225 @SuppressWarnings("serial")
6226 static final class MapReduceEntriesToIntTask<K,V>
6227 extends BulkTask<K,V,Integer> {
6228 final ToIntFunction<Map.Entry<K,V>> transformer;
6229 final IntBinaryOperator reducer;
6230 final int basis;
6231 int result;
6232 MapReduceEntriesToIntTask<K,V> rights, nextRight;
6233 MapReduceEntriesToIntTask
6234 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
6235 MapReduceEntriesToIntTask<K,V> nextRight,
6236 ToIntFunction<Map.Entry<K,V>> transformer,
6237 int basis,
6238 IntBinaryOperator reducer) {
6239 super(p, b, i, f, t); this.nextRight = nextRight;
6240 this.transformer = transformer;
6241 this.basis = basis; this.reducer = reducer;
6242 }
6243 public final Integer getRawResult() { return result; }
6244 public final void compute() {
6245 final ToIntFunction<Map.Entry<K,V>> transformer;
6246 final IntBinaryOperator reducer;
6247 if ((transformer = this.transformer) != null &&
6248 (reducer = this.reducer) != null) {
6249 int r = this.basis;
6250 for (int i = baseIndex, f, h; batch > 0 &&
6251 (h = ((f = baseLimit) + i) >>> 1) > i;) {
6252 addToPendingCount(1);
6253 (rights = new MapReduceEntriesToIntTask<K,V>
6254 (this, batch >>>= 1, baseLimit = h, f, tab,
6255 rights, transformer, r, reducer)).fork();
6256 }
6257 for (Node<K,V> p; (p = advance()) != null; )
6258 r = reducer.applyAsInt(r, transformer.applyAsInt(p));
6259 result = r;
6260 CountedCompleter<?> c;
6261 for (c = firstComplete(); c != null; c = c.nextComplete()) {
6262 @SuppressWarnings("unchecked")
6263 MapReduceEntriesToIntTask<K,V>
6264 t = (MapReduceEntriesToIntTask<K,V>)c,
6265 s = t.rights;
6266 while (s != null) {
6267 t.result = reducer.applyAsInt(t.result, s.result);
6268 s = t.rights = s.nextRight;
6269 }
6270 }
6271 }
6272 }
6273 }
6274
6275 @SuppressWarnings("serial")
6276 static final class MapReduceMappingsToIntTask<K,V>
6277 extends BulkTask<K,V,Integer> {
6278 final ToIntBiFunction<? super K, ? super V> transformer;
6279 final IntBinaryOperator reducer;
6280 final int basis;
6281 int result;
6282 MapReduceMappingsToIntTask<K,V> rights, nextRight;
6283 MapReduceMappingsToIntTask
6284 (BulkTask<K,V,?> p, int b, int i, int f, Node<K,V>[] t,
6285 MapReduceMappingsToIntTask<K,V> nextRight,
6286 ToIntBiFunction<? super K, ? super V> transformer,
6287 int basis,
6288 IntBinaryOperator reducer) {
6289 super(p, b, i, f, t); this.nextRight = nextRight;
6290 this.transformer = transformer;
6291 this.basis = basis; this.reducer = reducer;
6292 }
6293 public final Integer getRawResult() { return result; }
6294 public final void compute() {
6295 final ToIntBiFunction<? super K, ? super V> transformer;
6296 final IntBinaryOperator reducer;
6297 if ((transformer = this.transformer) != null &&
6298 (reducer = this.reducer) != null) {
6299 int r = this.basis;
6300 for (int i = baseIndex, f, h; batch > 0 &&
6301 (h = ((f = baseLimit) + i) >>> 1) > i;) {
6302 addToPendingCount(1);
6303 (rights = new MapReduceMappingsToIntTask<K,V>
6304 (this, batch >>>= 1, baseLimit = h, f, tab,
6305 rights, transformer, r, reducer)).fork();
6306 }
6307 for (Node<K,V> p; (p = advance()) != null; )
6308 r = reducer.applyAsInt(r, transformer.applyAsInt(p.key, p.val));
6309 result = r;
6310 CountedCompleter<?> c;
6311 for (c = firstComplete(); c != null; c = c.nextComplete()) {
6312 @SuppressWarnings("unchecked")
6313 MapReduceMappingsToIntTask<K,V>
6314 t = (MapReduceMappingsToIntTask<K,V>)c,
6315 s = t.rights;
6316 while (s != null) {
6317 t.result = reducer.applyAsInt(t.result, s.result);
6318 s = t.rights = s.nextRight;
6319 }
6320 }
6321 }
6322 }
6323 }
6324
6325 // Unsafe mechanics
6326 private static final sun.misc.Unsafe U;
6327 private static final long SIZECTL;
6328 private static final long TRANSFERINDEX;
6329 private static final long BASECOUNT;
6330 private static final long CELLSBUSY;
6331 private static final long CELLVALUE;
6332 private static final long ABASE;
6333 private static final int ASHIFT;
6334
6335 static {
6336 try {
6337 U = sun.misc.Unsafe.getUnsafe();
6338 Class<?> k = ConcurrentHashMap.class;
6339 SIZECTL = U.objectFieldOffset
6340 (k.getDeclaredField("sizeCtl"));
6341 TRANSFERINDEX = U.objectFieldOffset
6342 (k.getDeclaredField("transferIndex"));
6343 BASECOUNT = U.objectFieldOffset
6344 (k.getDeclaredField("baseCount"));
6345 CELLSBUSY = U.objectFieldOffset
6346 (k.getDeclaredField("cellsBusy"));
6347 Class<?> ck = CounterCell.class;
6348 CELLVALUE = U.objectFieldOffset
6349 (ck.getDeclaredField("value"));
6350 Class<?> ak = Node[].class;
6351 ABASE = U.arrayBaseOffset(ak);
6352 int scale = U.arrayIndexScale(ak);
6353 if ((scale & (scale - 1)) != 0)
6354 throw new Error("data type scale not a power of two");
6355 ASHIFT = 31 - Integer.numberOfLeadingZeros(scale);
6356 } catch (Exception e) {
6357 throw new Error(e);
6358 }
6359 }
6360 }