1/*
2 * Written by Doug Lea with assistance from members of JCP JSR-166
3 * Expert Group and released to the public domain, as explained at
4 * http://creativecommons.org/publicdomain/zero/1.0/
5 */
6
7package java.util.concurrent;
8
9import java.io.Serializable;
10import java.util.AbstractCollection;
11import java.util.AbstractMap;
12import java.util.AbstractSet;
13import java.util.ArrayList;
14import java.util.Collection;
15import java.util.Collections;
16import java.util.Comparator;
17import java.util.Iterator;
18import java.util.List;
19import java.util.Map;
20import java.util.NavigableMap;
21import java.util.NavigableSet;
22import java.util.NoSuchElementException;
23import java.util.Set;
24import java.util.SortedMap;
25import java.util.Spliterator;
26import java.util.function.BiConsumer;
27import java.util.function.BiFunction;
28import java.util.function.Consumer;
29import java.util.function.Function;
30import java.util.function.Predicate;
31
32// BEGIN android-note
33// removed link to collections framework docs
34// END android-note
35
36/**
37 * A scalable concurrent {@link ConcurrentNavigableMap} implementation.
38 * The map is sorted according to the {@linkplain Comparable natural
39 * ordering} of its keys, or by a {@link Comparator} provided at map
40 * creation time, depending on which constructor is used.
41 *
42 * <p>This class implements a concurrent variant of <a
43 * href="http://en.wikipedia.org/wiki/Skip_list" target="_top">SkipLists</a>
44 * providing expected average <i>log(n)</i> time cost for the
45 * {@code containsKey}, {@code get}, {@code put} and
46 * {@code remove} operations and their variants.  Insertion, removal,
47 * update, and access operations safely execute concurrently by
48 * multiple threads.
49 *
50 * <p>Iterators and spliterators are
51 * <a href="package-summary.html#Weakly"><i>weakly consistent</i></a>.
52 *
53 * <p>Ascending key ordered views and their iterators are faster than
54 * descending ones.
55 *
56 * <p>All {@code Map.Entry} pairs returned by methods in this class
57 * and its views represent snapshots of mappings at the time they were
58 * produced. They do <em>not</em> support the {@code Entry.setValue}
59 * method. (Note however that it is possible to change mappings in the
60 * associated map using {@code put}, {@code putIfAbsent}, or
61 * {@code replace}, depending on exactly which effect you need.)
62 *
63 * <p>Beware that, unlike in most collections, the {@code size}
64 * method is <em>not</em> a constant-time operation. Because of the
65 * asynchronous nature of these maps, determining the current number
66 * of elements requires a traversal of the elements, and so may report
67 * inaccurate results if this collection is modified during traversal.
68 * Additionally, the bulk operations {@code putAll}, {@code equals},
69 * {@code toArray}, {@code containsValue}, and {@code clear} are
70 * <em>not</em> guaranteed to be performed atomically. For example, an
71 * iterator operating concurrently with a {@code putAll} operation
72 * might view only some of the added elements.
73 *
74 * <p>This class and its views and iterators implement all of the
75 * <em>optional</em> methods of the {@link Map} and {@link Iterator}
76 * interfaces. Like most other concurrent collections, this class does
77 * <em>not</em> permit the use of {@code null} keys or values because some
78 * null return values cannot be reliably distinguished from the absence of
79 * elements.
80 *
81 * @author Doug Lea
82 * @param <K> the type of keys maintained by this map
83 * @param <V> the type of mapped values
84 * @since 1.6
85 */
86public class ConcurrentSkipListMap<K,V> extends AbstractMap<K,V>
87    implements ConcurrentNavigableMap<K,V>, Cloneable, Serializable {
88    /*
89     * This class implements a tree-like two-dimensionally linked skip
90     * list in which the index levels are represented in separate
91     * nodes from the base nodes holding data.  There are two reasons
92     * for taking this approach instead of the usual array-based
93     * structure: 1) Array based implementations seem to encounter
94     * more complexity and overhead 2) We can use cheaper algorithms
95     * for the heavily-traversed index lists than can be used for the
96     * base lists.  Here's a picture of some of the basics for a
97     * possible list with 2 levels of index:
98     *
99     * Head nodes          Index nodes
100     * +-+    right        +-+                      +-+
101     * |2|---------------->| |--------------------->| |->null
102     * +-+                 +-+                      +-+
103     *  | down              |                        |
104     *  v                   v                        v
105     * +-+            +-+  +-+       +-+            +-+       +-+
106     * |1|----------->| |->| |------>| |----------->| |------>| |->null
107     * +-+            +-+  +-+       +-+            +-+       +-+
108     *  v              |    |         |              |         |
109     * Nodes  next     v    v         v              v         v
110     * +-+  +-+  +-+  +-+  +-+  +-+  +-+  +-+  +-+  +-+  +-+  +-+
111     * | |->|A|->|B|->|C|->|D|->|E|->|F|->|G|->|H|->|I|->|J|->|K|->null
112     * +-+  +-+  +-+  +-+  +-+  +-+  +-+  +-+  +-+  +-+  +-+  +-+
113     *
114     * The base lists use a variant of the HM linked ordered set
115     * algorithm. See Tim Harris, "A pragmatic implementation of
116     * non-blocking linked lists"
117     * http://www.cl.cam.ac.uk/~tlh20/publications.html and Maged
118     * Michael "High Performance Dynamic Lock-Free Hash Tables and
119     * List-Based Sets"
120     * http://www.research.ibm.com/people/m/michael/pubs.htm.  The
121     * basic idea in these lists is to mark the "next" pointers of
122     * deleted nodes when deleting to avoid conflicts with concurrent
123     * insertions, and when traversing to keep track of triples
124     * (predecessor, node, successor) in order to detect when and how
125     * to unlink these deleted nodes.
126     *
127     * Rather than using mark-bits to mark list deletions (which can
128     * be slow and space-intensive using AtomicMarkedReference), nodes
129     * use direct CAS'able next pointers.  On deletion, instead of
130     * marking a pointer, they splice in another node that can be
131     * thought of as standing for a marked pointer (indicating this by
132     * using otherwise impossible field values).  Using plain nodes
133     * acts roughly like "boxed" implementations of marked pointers,
134     * but uses new nodes only when nodes are deleted, not for every
135     * link.  This requires less space and supports faster
136     * traversal. Even if marked references were better supported by
137     * JVMs, traversal using this technique might still be faster
138     * because any search need only read ahead one more node than
139     * otherwise required (to check for trailing marker) rather than
140     * unmasking mark bits or whatever on each read.
141     *
142     * This approach maintains the essential property needed in the HM
143     * algorithm of changing the next-pointer of a deleted node so
144     * that any other CAS of it will fail, but implements the idea by
145     * changing the pointer to point to a different node, not by
146     * marking it.  While it would be possible to further squeeze
147     * space by defining marker nodes not to have key/value fields, it
148     * isn't worth the extra type-testing overhead.  The deletion
149     * markers are rarely encountered during traversal and are
150     * normally quickly garbage collected. (Note that this technique
151     * would not work well in systems without garbage collection.)
152     *
153     * In addition to using deletion markers, the lists also use
154     * nullness of value fields to indicate deletion, in a style
155     * similar to typical lazy-deletion schemes.  If a node's value is
156     * null, then it is considered logically deleted and ignored even
157     * though it is still reachable. This maintains proper control of
158     * concurrent replace vs delete operations -- an attempted replace
159     * must fail if a delete beat it by nulling field, and a delete
160     * must return the last non-null value held in the field. (Note:
161     * Null, rather than some special marker, is used for value fields
162     * here because it just so happens to mesh with the Map API
163     * requirement that method get returns null if there is no
164     * mapping, which allows nodes to remain concurrently readable
165     * even when deleted. Using any other marker value here would be
166     * messy at best.)
167     *
168     * Here's the sequence of events for a deletion of node n with
169     * predecessor b and successor f, initially:
170     *
171     *        +------+       +------+      +------+
172     *   ...  |   b  |------>|   n  |----->|   f  | ...
173     *        +------+       +------+      +------+
174     *
175     * 1. CAS n's value field from non-null to null.
176     *    From this point on, no public operations encountering
177     *    the node consider this mapping to exist. However, other
178     *    ongoing insertions and deletions might still modify
179     *    n's next pointer.
180     *
181     * 2. CAS n's next pointer to point to a new marker node.
182     *    From this point on, no other nodes can be appended to n.
183     *    which avoids deletion errors in CAS-based linked lists.
184     *
185     *        +------+       +------+      +------+       +------+
186     *   ...  |   b  |------>|   n  |----->|marker|------>|   f  | ...
187     *        +------+       +------+      +------+       +------+
188     *
189     * 3. CAS b's next pointer over both n and its marker.
190     *    From this point on, no new traversals will encounter n,
191     *    and it can eventually be GCed.
192     *        +------+                                    +------+
193     *   ...  |   b  |----------------------------------->|   f  | ...
194     *        +------+                                    +------+
195     *
196     * A failure at step 1 leads to simple retry due to a lost race
197     * with another operation. Steps 2-3 can fail because some other
198     * thread noticed during a traversal a node with null value and
199     * helped out by marking and/or unlinking.  This helping-out
200     * ensures that no thread can become stuck waiting for progress of
201     * the deleting thread.  The use of marker nodes slightly
202     * complicates helping-out code because traversals must track
203     * consistent reads of up to four nodes (b, n, marker, f), not
204     * just (b, n, f), although the next field of a marker is
205     * immutable, and once a next field is CAS'ed to point to a
206     * marker, it never again changes, so this requires less care.
207     *
208     * Skip lists add indexing to this scheme, so that the base-level
209     * traversals start close to the locations being found, inserted
210     * or deleted -- usually base level traversals only traverse a few
211     * nodes. This doesn't change the basic algorithm except for the
212     * need to make sure base traversals start at predecessors (here,
213     * b) that are not (structurally) deleted, otherwise retrying
214     * after processing the deletion.
215     *
216     * Index levels are maintained as lists with volatile next fields,
217     * using CAS to link and unlink.  Races are allowed in index-list
218     * operations that can (rarely) fail to link in a new index node
219     * or delete one. (We can't do this of course for data nodes.)
220     * However, even when this happens, the index lists remain sorted,
221     * so correctly serve as indices.  This can impact performance,
222     * but since skip lists are probabilistic anyway, the net result
223     * is that under contention, the effective "p" value may be lower
224     * than its nominal value. And race windows are kept small enough
225     * that in practice these failures are rare, even under a lot of
226     * contention.
227     *
228     * The fact that retries (for both base and index lists) are
229     * relatively cheap due to indexing allows some minor
230     * simplifications of retry logic. Traversal restarts are
231     * performed after most "helping-out" CASes. This isn't always
232     * strictly necessary, but the implicit backoffs tend to help
233     * reduce other downstream failed CAS's enough to outweigh restart
234     * cost.  This worsens the worst case, but seems to improve even
235     * highly contended cases.
236     *
237     * Unlike most skip-list implementations, index insertion and
238     * deletion here require a separate traversal pass occurring after
239     * the base-level action, to add or remove index nodes.  This adds
240     * to single-threaded overhead, but improves contended
241     * multithreaded performance by narrowing interference windows,
242     * and allows deletion to ensure that all index nodes will be made
243     * unreachable upon return from a public remove operation, thus
244     * avoiding unwanted garbage retention. This is more important
245     * here than in some other data structures because we cannot null
246     * out node fields referencing user keys since they might still be
247     * read by other ongoing traversals.
248     *
249     * Indexing uses skip list parameters that maintain good search
250     * performance while using sparser-than-usual indices: The
251     * hardwired parameters k=1, p=0.5 (see method doPut) mean
252     * that about one-quarter of the nodes have indices. Of those that
253     * do, half have one level, a quarter have two, and so on (see
254     * Pugh's Skip List Cookbook, sec 3.4).  The expected total space
255     * requirement for a map is slightly less than for the current
256     * implementation of java.util.TreeMap.
257     *
258     * Changing the level of the index (i.e, the height of the
259     * tree-like structure) also uses CAS. The head index has initial
260     * level/height of one. Creation of an index with height greater
261     * than the current level adds a level to the head index by
262     * CAS'ing on a new top-most head. To maintain good performance
263     * after a lot of removals, deletion methods heuristically try to
264     * reduce the height if the topmost levels appear to be empty.
265     * This may encounter races in which it possible (but rare) to
266     * reduce and "lose" a level just as it is about to contain an
267     * index (that will then never be encountered). This does no
268     * structural harm, and in practice appears to be a better option
269     * than allowing unrestrained growth of levels.
270     *
271     * The code for all this is more verbose than you'd like. Most
272     * operations entail locating an element (or position to insert an
273     * element). The code to do this can't be nicely factored out
274     * because subsequent uses require a snapshot of predecessor
275     * and/or successor and/or value fields which can't be returned
276     * all at once, at least not without creating yet another object
277     * to hold them -- creating such little objects is an especially
278     * bad idea for basic internal search operations because it adds
279     * to GC overhead.  (This is one of the few times I've wished Java
280     * had macros.) Instead, some traversal code is interleaved within
281     * insertion and removal operations.  The control logic to handle
282     * all the retry conditions is sometimes twisty. Most search is
283     * broken into 2 parts. findPredecessor() searches index nodes
284     * only, returning a base-level predecessor of the key. findNode()
285     * finishes out the base-level search. Even with this factoring,
286     * there is a fair amount of near-duplication of code to handle
287     * variants.
288     *
289     * To produce random values without interference across threads,
290     * we use within-JDK thread local random support (via the
291     * "secondary seed", to avoid interference with user-level
292     * ThreadLocalRandom.)
293     *
294     * A previous version of this class wrapped non-comparable keys
295     * with their comparators to emulate Comparables when using
296     * comparators vs Comparables.  However, JVMs now appear to better
297     * handle infusing comparator-vs-comparable choice into search
298     * loops. Static method cpr(comparator, x, y) is used for all
299     * comparisons, which works well as long as the comparator
300     * argument is set up outside of loops (thus sometimes passed as
301     * an argument to internal methods) to avoid field re-reads.
302     *
303     * For explanation of algorithms sharing at least a couple of
304     * features with this one, see Mikhail Fomitchev's thesis
305     * (http://www.cs.yorku.ca/~mikhail/), Keir Fraser's thesis
306     * (http://www.cl.cam.ac.uk/users/kaf24/), and Hakan Sundell's
307     * thesis (http://www.cs.chalmers.se/~phs/).
308     *
309     * Given the use of tree-like index nodes, you might wonder why
310     * this doesn't use some kind of search tree instead, which would
311     * support somewhat faster search operations. The reason is that
312     * there are no known efficient lock-free insertion and deletion
313     * algorithms for search trees. The immutability of the "down"
314     * links of index nodes (as opposed to mutable "left" fields in
315     * true trees) makes this tractable using only CAS operations.
316     *
317     * Notation guide for local variables
318     * Node:         b, n, f    for  predecessor, node, successor
319     * Index:        q, r, d    for index node, right, down.
320     *               t          for another index node
321     * Head:         h
322     * Levels:       j
323     * Keys:         k, key
324     * Values:       v, value
325     * Comparisons:  c
326     */
327
328    private static final long serialVersionUID = -8627078645895051609L;
329
330    /**
331     * Special value used to identify base-level header.
332     */
333    static final Object BASE_HEADER = new Object();
334
335    /**
336     * The topmost head index of the skiplist.
337     */
338    private transient volatile HeadIndex<K,V> head;
339
340    /**
341     * The comparator used to maintain order in this map, or null if
342     * using natural ordering.  (Non-private to simplify access in
343     * nested classes.)
344     * @serial
345     */
346    final Comparator<? super K> comparator;
347
348    /** Lazily initialized key set */
349    private transient KeySet<K,V> keySet;
350    /** Lazily initialized entry set */
351    private transient EntrySet<K,V> entrySet;
352    /** Lazily initialized values collection */
353    private transient Values<K,V> values;
354    /** Lazily initialized descending key set */
355    private transient ConcurrentNavigableMap<K,V> descendingMap;
356
357    /**
358     * Initializes or resets state. Needed by constructors, clone,
359     * clear, readObject. and ConcurrentSkipListSet.clone.
360     * (Note that comparator must be separately initialized.)
361     */
362    private void initialize() {
363        keySet = null;
364        entrySet = null;
365        values = null;
366        descendingMap = null;
367        head = new HeadIndex<K,V>(new Node<K,V>(null, BASE_HEADER, null),
368                                  null, null, 1);
369    }
370
371    /**
372     * compareAndSet head node.
373     */
374    private boolean casHead(HeadIndex<K,V> cmp, HeadIndex<K,V> val) {
375        return U.compareAndSwapObject(this, HEAD, cmp, val);
376    }
377
378    /* ---------------- Nodes -------------- */
379
380    /**
381     * Nodes hold keys and values, and are singly linked in sorted
382     * order, possibly with some intervening marker nodes. The list is
383     * headed by a dummy node accessible as head.node. The value field
384     * is declared only as Object because it takes special non-V
385     * values for marker and header nodes.
386     */
387    static final class Node<K,V> {
388        final K key;
389        volatile Object value;
390        volatile Node<K,V> next;
391
392        /**
393         * Creates a new regular node.
394         */
395        Node(K key, Object value, Node<K,V> next) {
396            this.key = key;
397            this.value = value;
398            this.next = next;
399        }
400
401        /**
402         * Creates a new marker node. A marker is distinguished by
403         * having its value field point to itself.  Marker nodes also
404         * have null keys, a fact that is exploited in a few places,
405         * but this doesn't distinguish markers from the base-level
406         * header node (head.node), which also has a null key.
407         */
408        Node(Node<K,V> next) {
409            this.key = null;
410            this.value = this;
411            this.next = next;
412        }
413
414        /**
415         * compareAndSet value field.
416         */
417        boolean casValue(Object cmp, Object val) {
418            return U.compareAndSwapObject(this, VALUE, cmp, val);
419        }
420
421        /**
422         * compareAndSet next field.
423         */
424        boolean casNext(Node<K,V> cmp, Node<K,V> val) {
425            return U.compareAndSwapObject(this, NEXT, cmp, val);
426        }
427
428        /**
429         * Returns true if this node is a marker. This method isn't
430         * actually called in any current code checking for markers
431         * because callers will have already read value field and need
432         * to use that read (not another done here) and so directly
433         * test if value points to node.
434         *
435         * @return true if this node is a marker node
436         */
437        boolean isMarker() {
438            return value == this;
439        }
440
441        /**
442         * Returns true if this node is the header of base-level list.
443         * @return true if this node is header node
444         */
445        boolean isBaseHeader() {
446            return value == BASE_HEADER;
447        }
448
449        /**
450         * Tries to append a deletion marker to this node.
451         * @param f the assumed current successor of this node
452         * @return true if successful
453         */
454        boolean appendMarker(Node<K,V> f) {
455            return casNext(f, new Node<K,V>(f));
456        }
457
458        /**
459         * Helps out a deletion by appending marker or unlinking from
460         * predecessor. This is called during traversals when value
461         * field seen to be null.
462         * @param b predecessor
463         * @param f successor
464         */
465        void helpDelete(Node<K,V> b, Node<K,V> f) {
466            /*
467             * Rechecking links and then doing only one of the
468             * help-out stages per call tends to minimize CAS
469             * interference among helping threads.
470             */
471            if (f == next && this == b.next) {
472                if (f == null || f.value != f) // not already marked
473                    casNext(f, new Node<K,V>(f));
474                else
475                    b.casNext(this, f.next);
476            }
477        }
478
479        /**
480         * Returns value if this node contains a valid key-value pair,
481         * else null.
482         * @return this node's value if it isn't a marker or header or
483         * is deleted, else null
484         */
485        V getValidValue() {
486            Object v = value;
487            if (v == this || v == BASE_HEADER)
488                return null;
489            @SuppressWarnings("unchecked") V vv = (V)v;
490            return vv;
491        }
492
493        /**
494         * Creates and returns a new SimpleImmutableEntry holding current
495         * mapping if this node holds a valid value, else null.
496         * @return new entry or null
497         */
498        AbstractMap.SimpleImmutableEntry<K,V> createSnapshot() {
499            Object v = value;
500            if (v == null || v == this || v == BASE_HEADER)
501                return null;
502            @SuppressWarnings("unchecked") V vv = (V)v;
503            return new AbstractMap.SimpleImmutableEntry<K,V>(key, vv);
504        }
505
506        // Unsafe mechanics
507
508        private static final sun.misc.Unsafe U = sun.misc.Unsafe.getUnsafe();
509        private static final long VALUE;
510        private static final long NEXT;
511
512        static {
513            try {
514                VALUE = U.objectFieldOffset
515                    (Node.class.getDeclaredField("value"));
516                NEXT = U.objectFieldOffset
517                    (Node.class.getDeclaredField("next"));
518            } catch (ReflectiveOperationException e) {
519                throw new Error(e);
520            }
521        }
522    }
523
524    /* ---------------- Indexing -------------- */
525
526    /**
527     * Index nodes represent the levels of the skip list.  Note that
528     * even though both Nodes and Indexes have forward-pointing
529     * fields, they have different types and are handled in different
530     * ways, that can't nicely be captured by placing field in a
531     * shared abstract class.
532     */
533    static class Index<K,V> {
534        final Node<K,V> node;
535        final Index<K,V> down;
536        volatile Index<K,V> right;
537
538        /**
539         * Creates index node with given values.
540         */
541        Index(Node<K,V> node, Index<K,V> down, Index<K,V> right) {
542            this.node = node;
543            this.down = down;
544            this.right = right;
545        }
546
547        /**
548         * compareAndSet right field.
549         */
550        final boolean casRight(Index<K,V> cmp, Index<K,V> val) {
551            return U.compareAndSwapObject(this, RIGHT, cmp, val);
552        }
553
554        /**
555         * Returns true if the node this indexes has been deleted.
556         * @return true if indexed node is known to be deleted
557         */
558        final boolean indexesDeletedNode() {
559            return node.value == null;
560        }
561
562        /**
563         * Tries to CAS newSucc as successor.  To minimize races with
564         * unlink that may lose this index node, if the node being
565         * indexed is known to be deleted, it doesn't try to link in.
566         * @param succ the expected current successor
567         * @param newSucc the new successor
568         * @return true if successful
569         */
570        final boolean link(Index<K,V> succ, Index<K,V> newSucc) {
571            Node<K,V> n = node;
572            newSucc.right = succ;
573            return n.value != null && casRight(succ, newSucc);
574        }
575
576        /**
577         * Tries to CAS right field to skip over apparent successor
578         * succ.  Fails (forcing a retraversal by caller) if this node
579         * is known to be deleted.
580         * @param succ the expected current successor
581         * @return true if successful
582         */
583        final boolean unlink(Index<K,V> succ) {
584            return node.value != null && casRight(succ, succ.right);
585        }
586
587        // Unsafe mechanics
588        private static final sun.misc.Unsafe U = sun.misc.Unsafe.getUnsafe();
589        private static final long RIGHT;
590        static {
591            try {
592                RIGHT = U.objectFieldOffset
593                    (Index.class.getDeclaredField("right"));
594            } catch (ReflectiveOperationException e) {
595                throw new Error(e);
596            }
597        }
598    }
599
600    /* ---------------- Head nodes -------------- */
601
602    /**
603     * Nodes heading each level keep track of their level.
604     */
605    static final class HeadIndex<K,V> extends Index<K,V> {
606        final int level;
607        HeadIndex(Node<K,V> node, Index<K,V> down, Index<K,V> right, int level) {
608            super(node, down, right);
609            this.level = level;
610        }
611    }
612
613    /* ---------------- Comparison utilities -------------- */
614
615    /**
616     * Compares using comparator or natural ordering if null.
617     * Called only by methods that have performed required type checks.
618     */
619    @SuppressWarnings({"unchecked", "rawtypes"})
620    static final int cpr(Comparator c, Object x, Object y) {
621        return (c != null) ? c.compare(x, y) : ((Comparable)x).compareTo(y);
622    }
623
624    /* ---------------- Traversal -------------- */
625
626    /**
627     * Returns a base-level node with key strictly less than given key,
628     * or the base-level header if there is no such node.  Also
629     * unlinks indexes to deleted nodes found along the way.  Callers
630     * rely on this side-effect of clearing indices to deleted nodes.
631     * @param key the key
632     * @return a predecessor of key
633     */
634    private Node<K,V> findPredecessor(Object key, Comparator<? super K> cmp) {
635        if (key == null)
636            throw new NullPointerException(); // don't postpone errors
637        for (;;) {
638            for (Index<K,V> q = head, r = q.right, d;;) {
639                if (r != null) {
640                    Node<K,V> n = r.node;
641                    K k = n.key;
642                    if (n.value == null) {
643                        if (!q.unlink(r))
644                            break;           // restart
645                        r = q.right;         // reread r
646                        continue;
647                    }
648                    if (cpr(cmp, key, k) > 0) {
649                        q = r;
650                        r = r.right;
651                        continue;
652                    }
653                }
654                if ((d = q.down) == null)
655                    return q.node;
656                q = d;
657                r = d.right;
658            }
659        }
660    }
661
662    /**
663     * Returns node holding key or null if no such, clearing out any
664     * deleted nodes seen along the way.  Repeatedly traverses at
665     * base-level looking for key starting at predecessor returned
666     * from findPredecessor, processing base-level deletions as
667     * encountered. Some callers rely on this side-effect of clearing
668     * deleted nodes.
669     *
670     * Restarts occur, at traversal step centered on node n, if:
671     *
672     *   (1) After reading n's next field, n is no longer assumed
673     *       predecessor b's current successor, which means that
674     *       we don't have a consistent 3-node snapshot and so cannot
675     *       unlink any subsequent deleted nodes encountered.
676     *
677     *   (2) n's value field is null, indicating n is deleted, in
678     *       which case we help out an ongoing structural deletion
679     *       before retrying.  Even though there are cases where such
680     *       unlinking doesn't require restart, they aren't sorted out
681     *       here because doing so would not usually outweigh cost of
682     *       restarting.
683     *
684     *   (3) n is a marker or n's predecessor's value field is null,
685     *       indicating (among other possibilities) that
686     *       findPredecessor returned a deleted node. We can't unlink
687     *       the node because we don't know its predecessor, so rely
688     *       on another call to findPredecessor to notice and return
689     *       some earlier predecessor, which it will do. This check is
690     *       only strictly needed at beginning of loop, (and the
691     *       b.value check isn't strictly needed at all) but is done
692     *       each iteration to help avoid contention with other
693     *       threads by callers that will fail to be able to change
694     *       links, and so will retry anyway.
695     *
696     * The traversal loops in doPut, doRemove, and findNear all
697     * include the same three kinds of checks. And specialized
698     * versions appear in findFirst, and findLast and their variants.
699     * They can't easily share code because each uses the reads of
700     * fields held in locals occurring in the orders they were
701     * performed.
702     *
703     * @param key the key
704     * @return node holding key, or null if no such
705     */
706    private Node<K,V> findNode(Object key) {
707        if (key == null)
708            throw new NullPointerException(); // don't postpone errors
709        Comparator<? super K> cmp = comparator;
710        outer: for (;;) {
711            for (Node<K,V> b = findPredecessor(key, cmp), n = b.next;;) {
712                Object v; int c;
713                if (n == null)
714                    break outer;
715                Node<K,V> f = n.next;
716                if (n != b.next)                // inconsistent read
717                    break;
718                if ((v = n.value) == null) {    // n is deleted
719                    n.helpDelete(b, f);
720                    break;
721                }
722                if (b.value == null || v == n)  // b is deleted
723                    break;
724                if ((c = cpr(cmp, key, n.key)) == 0)
725                    return n;
726                if (c < 0)
727                    break outer;
728                b = n;
729                n = f;
730            }
731        }
732        return null;
733    }
734
735    /**
736     * Gets value for key. Almost the same as findNode, but returns
737     * the found value (to avoid retries during re-reads)
738     *
739     * @param key the key
740     * @return the value, or null if absent
741     */
742    private V doGet(Object key) {
743        if (key == null)
744            throw new NullPointerException();
745        Comparator<? super K> cmp = comparator;
746        outer: for (;;) {
747            for (Node<K,V> b = findPredecessor(key, cmp), n = b.next;;) {
748                Object v; int c;
749                if (n == null)
750                    break outer;
751                Node<K,V> f = n.next;
752                if (n != b.next)                // inconsistent read
753                    break;
754                if ((v = n.value) == null) {    // n is deleted
755                    n.helpDelete(b, f);
756                    break;
757                }
758                if (b.value == null || v == n)  // b is deleted
759                    break;
760                if ((c = cpr(cmp, key, n.key)) == 0) {
761                    @SuppressWarnings("unchecked") V vv = (V)v;
762                    return vv;
763                }
764                if (c < 0)
765                    break outer;
766                b = n;
767                n = f;
768            }
769        }
770        return null;
771    }
772
773    /* ---------------- Insertion -------------- */
774
775    /**
776     * Main insertion method.  Adds element if not present, or
777     * replaces value if present and onlyIfAbsent is false.
778     * @param key the key
779     * @param value the value that must be associated with key
780     * @param onlyIfAbsent if should not insert if already present
781     * @return the old value, or null if newly inserted
782     */
783    private V doPut(K key, V value, boolean onlyIfAbsent) {
784        Node<K,V> z;             // added node
785        if (key == null)
786            throw new NullPointerException();
787        Comparator<? super K> cmp = comparator;
788        outer: for (;;) {
789            for (Node<K,V> b = findPredecessor(key, cmp), n = b.next;;) {
790                if (n != null) {
791                    Object v; int c;
792                    Node<K,V> f = n.next;
793                    if (n != b.next)               // inconsistent read
794                        break;
795                    if ((v = n.value) == null) {   // n is deleted
796                        n.helpDelete(b, f);
797                        break;
798                    }
799                    if (b.value == null || v == n) // b is deleted
800                        break;
801                    if ((c = cpr(cmp, key, n.key)) > 0) {
802                        b = n;
803                        n = f;
804                        continue;
805                    }
806                    if (c == 0) {
807                        if (onlyIfAbsent || n.casValue(v, value)) {
808                            @SuppressWarnings("unchecked") V vv = (V)v;
809                            return vv;
810                        }
811                        break; // restart if lost race to replace value
812                    }
813                    // else c < 0; fall through
814                }
815
816                z = new Node<K,V>(key, value, n);
817                if (!b.casNext(n, z))
818                    break;         // restart if lost race to append to b
819                break outer;
820            }
821        }
822
823        int rnd = ThreadLocalRandom.nextSecondarySeed();
824        if ((rnd & 0x80000001) == 0) { // test highest and lowest bits
825            int level = 1, max;
826            while (((rnd >>>= 1) & 1) != 0)
827                ++level;
828            Index<K,V> idx = null;
829            HeadIndex<K,V> h = head;
830            if (level <= (max = h.level)) {
831                for (int i = 1; i <= level; ++i)
832                    idx = new Index<K,V>(z, idx, null);
833            }
834            else { // try to grow by one level
835                level = max + 1; // hold in array and later pick the one to use
836                @SuppressWarnings("unchecked")Index<K,V>[] idxs =
837                    (Index<K,V>[])new Index<?,?>[level+1];
838                for (int i = 1; i <= level; ++i)
839                    idxs[i] = idx = new Index<K,V>(z, idx, null);
840                for (;;) {
841                    h = head;
842                    int oldLevel = h.level;
843                    if (level <= oldLevel) // lost race to add level
844                        break;
845                    HeadIndex<K,V> newh = h;
846                    Node<K,V> oldbase = h.node;
847                    for (int j = oldLevel+1; j <= level; ++j)
848                        newh = new HeadIndex<K,V>(oldbase, newh, idxs[j], j);
849                    if (casHead(h, newh)) {
850                        h = newh;
851                        idx = idxs[level = oldLevel];
852                        break;
853                    }
854                }
855            }
856            // find insertion points and splice in
857            splice: for (int insertionLevel = level;;) {
858                int j = h.level;
859                for (Index<K,V> q = h, r = q.right, t = idx;;) {
860                    if (q == null || t == null)
861                        break splice;
862                    if (r != null) {
863                        Node<K,V> n = r.node;
864                        // compare before deletion check avoids needing recheck
865                        int c = cpr(cmp, key, n.key);
866                        if (n.value == null) {
867                            if (!q.unlink(r))
868                                break;
869                            r = q.right;
870                            continue;
871                        }
872                        if (c > 0) {
873                            q = r;
874                            r = r.right;
875                            continue;
876                        }
877                    }
878
879                    if (j == insertionLevel) {
880                        if (!q.link(r, t))
881                            break; // restart
882                        if (t.node.value == null) {
883                            findNode(key);
884                            break splice;
885                        }
886                        if (--insertionLevel == 0)
887                            break splice;
888                    }
889
890                    if (--j >= insertionLevel && j < level)
891                        t = t.down;
892                    q = q.down;
893                    r = q.right;
894                }
895            }
896        }
897        return null;
898    }
899
900    /* ---------------- Deletion -------------- */
901
902    /**
903     * Main deletion method. Locates node, nulls value, appends a
904     * deletion marker, unlinks predecessor, removes associated index
905     * nodes, and possibly reduces head index level.
906     *
907     * Index nodes are cleared out simply by calling findPredecessor.
908     * which unlinks indexes to deleted nodes found along path to key,
909     * which will include the indexes to this node.  This is done
910     * unconditionally. We can't check beforehand whether there are
911     * index nodes because it might be the case that some or all
912     * indexes hadn't been inserted yet for this node during initial
913     * search for it, and we'd like to ensure lack of garbage
914     * retention, so must call to be sure.
915     *
916     * @param key the key
917     * @param value if non-null, the value that must be
918     * associated with key
919     * @return the node, or null if not found
920     */
921    final V doRemove(Object key, Object value) {
922        if (key == null)
923            throw new NullPointerException();
924        Comparator<? super K> cmp = comparator;
925        outer: for (;;) {
926            for (Node<K,V> b = findPredecessor(key, cmp), n = b.next;;) {
927                Object v; int c;
928                if (n == null)
929                    break outer;
930                Node<K,V> f = n.next;
931                if (n != b.next)                    // inconsistent read
932                    break;
933                if ((v = n.value) == null) {        // n is deleted
934                    n.helpDelete(b, f);
935                    break;
936                }
937                if (b.value == null || v == n)      // b is deleted
938                    break;
939                if ((c = cpr(cmp, key, n.key)) < 0)
940                    break outer;
941                if (c > 0) {
942                    b = n;
943                    n = f;
944                    continue;
945                }
946                if (value != null && !value.equals(v))
947                    break outer;
948                if (!n.casValue(v, null))
949                    break;
950                if (!n.appendMarker(f) || !b.casNext(n, f))
951                    findNode(key);                  // retry via findNode
952                else {
953                    findPredecessor(key, cmp);      // clean index
954                    if (head.right == null)
955                        tryReduceLevel();
956                }
957                @SuppressWarnings("unchecked") V vv = (V)v;
958                return vv;
959            }
960        }
961        return null;
962    }
963
964    /**
965     * Possibly reduce head level if it has no nodes.  This method can
966     * (rarely) make mistakes, in which case levels can disappear even
967     * though they are about to contain index nodes. This impacts
968     * performance, not correctness.  To minimize mistakes as well as
969     * to reduce hysteresis, the level is reduced by one only if the
970     * topmost three levels look empty. Also, if the removed level
971     * looks non-empty after CAS, we try to change it back quick
972     * before anyone notices our mistake! (This trick works pretty
973     * well because this method will practically never make mistakes
974     * unless current thread stalls immediately before first CAS, in
975     * which case it is very unlikely to stall again immediately
976     * afterwards, so will recover.)
977     *
978     * We put up with all this rather than just let levels grow
979     * because otherwise, even a small map that has undergone a large
980     * number of insertions and removals will have a lot of levels,
981     * slowing down access more than would an occasional unwanted
982     * reduction.
983     */
984    private void tryReduceLevel() {
985        HeadIndex<K,V> h = head;
986        HeadIndex<K,V> d;
987        HeadIndex<K,V> e;
988        if (h.level > 3 &&
989            (d = (HeadIndex<K,V>)h.down) != null &&
990            (e = (HeadIndex<K,V>)d.down) != null &&
991            e.right == null &&
992            d.right == null &&
993            h.right == null &&
994            casHead(h, d) && // try to set
995            h.right != null) // recheck
996            casHead(d, h);   // try to backout
997    }
998
999    /* ---------------- Finding and removing first element -------------- */
1000
1001    /**
1002     * Specialized variant of findNode to get first valid node.
1003     * @return first node or null if empty
1004     */
1005    final Node<K,V> findFirst() {
1006        for (Node<K,V> b, n;;) {
1007            if ((n = (b = head.node).next) == null)
1008                return null;
1009            if (n.value != null)
1010                return n;
1011            n.helpDelete(b, n.next);
1012        }
1013    }
1014
1015    /**
1016     * Removes first entry; returns its snapshot.
1017     * @return null if empty, else snapshot of first entry
1018     */
1019    private Map.Entry<K,V> doRemoveFirstEntry() {
1020        for (Node<K,V> b, n;;) {
1021            if ((n = (b = head.node).next) == null)
1022                return null;
1023            Node<K,V> f = n.next;
1024            if (n != b.next)
1025                continue;
1026            Object v = n.value;
1027            if (v == null) {
1028                n.helpDelete(b, f);
1029                continue;
1030            }
1031            if (!n.casValue(v, null))
1032                continue;
1033            if (!n.appendMarker(f) || !b.casNext(n, f))
1034                findFirst(); // retry
1035            clearIndexToFirst();
1036            @SuppressWarnings("unchecked") V vv = (V)v;
1037            return new AbstractMap.SimpleImmutableEntry<K,V>(n.key, vv);
1038        }
1039    }
1040
1041    /**
1042     * Clears out index nodes associated with deleted first entry.
1043     */
1044    private void clearIndexToFirst() {
1045        for (;;) {
1046            for (Index<K,V> q = head;;) {
1047                Index<K,V> r = q.right;
1048                if (r != null && r.indexesDeletedNode() && !q.unlink(r))
1049                    break;
1050                if ((q = q.down) == null) {
1051                    if (head.right == null)
1052                        tryReduceLevel();
1053                    return;
1054                }
1055            }
1056        }
1057    }
1058
1059    /**
1060     * Removes last entry; returns its snapshot.
1061     * Specialized variant of doRemove.
1062     * @return null if empty, else snapshot of last entry
1063     */
1064    private Map.Entry<K,V> doRemoveLastEntry() {
1065        for (;;) {
1066            Node<K,V> b = findPredecessorOfLast();
1067            Node<K,V> n = b.next;
1068            if (n == null) {
1069                if (b.isBaseHeader())               // empty
1070                    return null;
1071                else
1072                    continue; // all b's successors are deleted; retry
1073            }
1074            for (;;) {
1075                Node<K,V> f = n.next;
1076                if (n != b.next)                    // inconsistent read
1077                    break;
1078                Object v = n.value;
1079                if (v == null) {                    // n is deleted
1080                    n.helpDelete(b, f);
1081                    break;
1082                }
1083                if (b.value == null || v == n)      // b is deleted
1084                    break;
1085                if (f != null) {
1086                    b = n;
1087                    n = f;
1088                    continue;
1089                }
1090                if (!n.casValue(v, null))
1091                    break;
1092                K key = n.key;
1093                if (!n.appendMarker(f) || !b.casNext(n, f))
1094                    findNode(key);                  // retry via findNode
1095                else {                              // clean index
1096                    findPredecessor(key, comparator);
1097                    if (head.right == null)
1098                        tryReduceLevel();
1099                }
1100                @SuppressWarnings("unchecked") V vv = (V)v;
1101                return new AbstractMap.SimpleImmutableEntry<K,V>(key, vv);
1102            }
1103        }
1104    }
1105
1106    /* ---------------- Finding and removing last element -------------- */
1107
1108    /**
1109     * Specialized version of find to get last valid node.
1110     * @return last node or null if empty
1111     */
1112    final Node<K,V> findLast() {
1113        /*
1114         * findPredecessor can't be used to traverse index level
1115         * because this doesn't use comparisons.  So traversals of
1116         * both levels are folded together.
1117         */
1118        Index<K,V> q = head;
1119        for (;;) {
1120            Index<K,V> d, r;
1121            if ((r = q.right) != null) {
1122                if (r.indexesDeletedNode()) {
1123                    q.unlink(r);
1124                    q = head; // restart
1125                }
1126                else
1127                    q = r;
1128            } else if ((d = q.down) != null) {
1129                q = d;
1130            } else {
1131                for (Node<K,V> b = q.node, n = b.next;;) {
1132                    if (n == null)
1133                        return b.isBaseHeader() ? null : b;
1134                    Node<K,V> f = n.next;            // inconsistent read
1135                    if (n != b.next)
1136                        break;
1137                    Object v = n.value;
1138                    if (v == null) {                 // n is deleted
1139                        n.helpDelete(b, f);
1140                        break;
1141                    }
1142                    if (b.value == null || v == n)      // b is deleted
1143                        break;
1144                    b = n;
1145                    n = f;
1146                }
1147                q = head; // restart
1148            }
1149        }
1150    }
1151
1152    /**
1153     * Specialized variant of findPredecessor to get predecessor of last
1154     * valid node.  Needed when removing the last entry.  It is possible
1155     * that all successors of returned node will have been deleted upon
1156     * return, in which case this method can be retried.
1157     * @return likely predecessor of last node
1158     */
1159    private Node<K,V> findPredecessorOfLast() {
1160        for (;;) {
1161            for (Index<K,V> q = head;;) {
1162                Index<K,V> d, r;
1163                if ((r = q.right) != null) {
1164                    if (r.indexesDeletedNode()) {
1165                        q.unlink(r);
1166                        break;    // must restart
1167                    }
1168                    // proceed as far across as possible without overshooting
1169                    if (r.node.next != null) {
1170                        q = r;
1171                        continue;
1172                    }
1173                }
1174                if ((d = q.down) != null)
1175                    q = d;
1176                else
1177                    return q.node;
1178            }
1179        }
1180    }
1181
1182    /* ---------------- Relational operations -------------- */
1183
1184    // Control values OR'ed as arguments to findNear
1185
1186    private static final int EQ = 1;
1187    private static final int LT = 2;
1188    private static final int GT = 0; // Actually checked as !LT
1189
1190    /**
1191     * Utility for ceiling, floor, lower, higher methods.
1192     * @param key the key
1193     * @param rel the relation -- OR'ed combination of EQ, LT, GT
1194     * @return nearest node fitting relation, or null if no such
1195     */
1196    final Node<K,V> findNear(K key, int rel, Comparator<? super K> cmp) {
1197        if (key == null)
1198            throw new NullPointerException();
1199        for (;;) {
1200            for (Node<K,V> b = findPredecessor(key, cmp), n = b.next;;) {
1201                Object v;
1202                if (n == null)
1203                    return ((rel & LT) == 0 || b.isBaseHeader()) ? null : b;
1204                Node<K,V> f = n.next;
1205                if (n != b.next)                  // inconsistent read
1206                    break;
1207                if ((v = n.value) == null) {      // n is deleted
1208                    n.helpDelete(b, f);
1209                    break;
1210                }
1211                if (b.value == null || v == n)      // b is deleted
1212                    break;
1213                int c = cpr(cmp, key, n.key);
1214                if ((c == 0 && (rel & EQ) != 0) ||
1215                    (c <  0 && (rel & LT) == 0))
1216                    return n;
1217                if ( c <= 0 && (rel & LT) != 0)
1218                    return b.isBaseHeader() ? null : b;
1219                b = n;
1220                n = f;
1221            }
1222        }
1223    }
1224
1225    /**
1226     * Returns SimpleImmutableEntry for results of findNear.
1227     * @param key the key
1228     * @param rel the relation -- OR'ed combination of EQ, LT, GT
1229     * @return Entry fitting relation, or null if no such
1230     */
1231    final AbstractMap.SimpleImmutableEntry<K,V> getNear(K key, int rel) {
1232        Comparator<? super K> cmp = comparator;
1233        for (;;) {
1234            Node<K,V> n = findNear(key, rel, cmp);
1235            if (n == null)
1236                return null;
1237            AbstractMap.SimpleImmutableEntry<K,V> e = n.createSnapshot();
1238            if (e != null)
1239                return e;
1240        }
1241    }
1242
1243    /* ---------------- Constructors -------------- */
1244
1245    /**
1246     * Constructs a new, empty map, sorted according to the
1247     * {@linkplain Comparable natural ordering} of the keys.
1248     */
1249    public ConcurrentSkipListMap() {
1250        this.comparator = null;
1251        initialize();
1252    }
1253
1254    /**
1255     * Constructs a new, empty map, sorted according to the specified
1256     * comparator.
1257     *
1258     * @param comparator the comparator that will be used to order this map.
1259     *        If {@code null}, the {@linkplain Comparable natural
1260     *        ordering} of the keys will be used.
1261     */
1262    public ConcurrentSkipListMap(Comparator<? super K> comparator) {
1263        this.comparator = comparator;
1264        initialize();
1265    }
1266
1267    /**
1268     * Constructs a new map containing the same mappings as the given map,
1269     * sorted according to the {@linkplain Comparable natural ordering} of
1270     * the keys.
1271     *
1272     * @param  m the map whose mappings are to be placed in this map
1273     * @throws ClassCastException if the keys in {@code m} are not
1274     *         {@link Comparable}, or are not mutually comparable
1275     * @throws NullPointerException if the specified map or any of its keys
1276     *         or values are null
1277     */
1278    public ConcurrentSkipListMap(Map<? extends K, ? extends V> m) {
1279        this.comparator = null;
1280        initialize();
1281        putAll(m);
1282    }
1283
1284    /**
1285     * Constructs a new map containing the same mappings and using the
1286     * same ordering as the specified sorted map.
1287     *
1288     * @param m the sorted map whose mappings are to be placed in this
1289     *        map, and whose comparator is to be used to sort this map
1290     * @throws NullPointerException if the specified sorted map or any of
1291     *         its keys or values are null
1292     */
1293    public ConcurrentSkipListMap(SortedMap<K, ? extends V> m) {
1294        this.comparator = m.comparator();
1295        initialize();
1296        buildFromSorted(m);
1297    }
1298
1299    /**
1300     * Returns a shallow copy of this {@code ConcurrentSkipListMap}
1301     * instance. (The keys and values themselves are not cloned.)
1302     *
1303     * @return a shallow copy of this map
1304     */
1305    public ConcurrentSkipListMap<K,V> clone() {
1306        try {
1307            @SuppressWarnings("unchecked")
1308            ConcurrentSkipListMap<K,V> clone =
1309                (ConcurrentSkipListMap<K,V>) super.clone();
1310            clone.initialize();
1311            clone.buildFromSorted(this);
1312            return clone;
1313        } catch (CloneNotSupportedException e) {
1314            throw new InternalError();
1315        }
1316    }
1317
1318    /**
1319     * Streamlined bulk insertion to initialize from elements of
1320     * given sorted map.  Call only from constructor or clone
1321     * method.
1322     */
1323    private void buildFromSorted(SortedMap<K, ? extends V> map) {
1324        if (map == null)
1325            throw new NullPointerException();
1326
1327        HeadIndex<K,V> h = head;
1328        Node<K,V> basepred = h.node;
1329
1330        // Track the current rightmost node at each level. Uses an
1331        // ArrayList to avoid committing to initial or maximum level.
1332        ArrayList<Index<K,V>> preds = new ArrayList<>();
1333
1334        // initialize
1335        for (int i = 0; i <= h.level; ++i)
1336            preds.add(null);
1337        Index<K,V> q = h;
1338        for (int i = h.level; i > 0; --i) {
1339            preds.set(i, q);
1340            q = q.down;
1341        }
1342
1343        Iterator<? extends Map.Entry<? extends K, ? extends V>> it =
1344            map.entrySet().iterator();
1345        while (it.hasNext()) {
1346            Map.Entry<? extends K, ? extends V> e = it.next();
1347            int rnd = ThreadLocalRandom.current().nextInt();
1348            int j = 0;
1349            if ((rnd & 0x80000001) == 0) {
1350                do {
1351                    ++j;
1352                } while (((rnd >>>= 1) & 1) != 0);
1353                if (j > h.level) j = h.level + 1;
1354            }
1355            K k = e.getKey();
1356            V v = e.getValue();
1357            if (k == null || v == null)
1358                throw new NullPointerException();
1359            Node<K,V> z = new Node<K,V>(k, v, null);
1360            basepred.next = z;
1361            basepred = z;
1362            if (j > 0) {
1363                Index<K,V> idx = null;
1364                for (int i = 1; i <= j; ++i) {
1365                    idx = new Index<K,V>(z, idx, null);
1366                    if (i > h.level)
1367                        h = new HeadIndex<K,V>(h.node, h, idx, i);
1368
1369                    if (i < preds.size()) {
1370                        preds.get(i).right = idx;
1371                        preds.set(i, idx);
1372                    } else
1373                        preds.add(idx);
1374                }
1375            }
1376        }
1377        head = h;
1378    }
1379
1380    /* ---------------- Serialization -------------- */
1381
1382    /**
1383     * Saves this map to a stream (that is, serializes it).
1384     *
1385     * @param s the stream
1386     * @throws java.io.IOException if an I/O error occurs
1387     * @serialData The key (Object) and value (Object) for each
1388     * key-value mapping represented by the map, followed by
1389     * {@code null}. The key-value mappings are emitted in key-order
1390     * (as determined by the Comparator, or by the keys' natural
1391     * ordering if no Comparator).
1392     */
1393    private void writeObject(java.io.ObjectOutputStream s)
1394        throws java.io.IOException {
1395        // Write out the Comparator and any hidden stuff
1396        s.defaultWriteObject();
1397
1398        // Write out keys and values (alternating)
1399        for (Node<K,V> n = findFirst(); n != null; n = n.next) {
1400            V v = n.getValidValue();
1401            if (v != null) {
1402                s.writeObject(n.key);
1403                s.writeObject(v);
1404            }
1405        }
1406        s.writeObject(null);
1407    }
1408
1409    /**
1410     * Reconstitutes this map from a stream (that is, deserializes it).
1411     * @param s the stream
1412     * @throws ClassNotFoundException if the class of a serialized object
1413     *         could not be found
1414     * @throws java.io.IOException if an I/O error occurs
1415     */
1416    @SuppressWarnings("unchecked")
1417    private void readObject(final java.io.ObjectInputStream s)
1418        throws java.io.IOException, ClassNotFoundException {
1419        // Read in the Comparator and any hidden stuff
1420        s.defaultReadObject();
1421        // Reset transients
1422        initialize();
1423
1424        /*
1425         * This is nearly identical to buildFromSorted, but is
1426         * distinct because readObject calls can't be nicely adapted
1427         * as the kind of iterator needed by buildFromSorted. (They
1428         * can be, but doing so requires type cheats and/or creation
1429         * of adapter classes.) It is simpler to just adapt the code.
1430         */
1431
1432        HeadIndex<K,V> h = head;
1433        Node<K,V> basepred = h.node;
1434        ArrayList<Index<K,V>> preds = new ArrayList<>();
1435        for (int i = 0; i <= h.level; ++i)
1436            preds.add(null);
1437        Index<K,V> q = h;
1438        for (int i = h.level; i > 0; --i) {
1439            preds.set(i, q);
1440            q = q.down;
1441        }
1442
1443        for (;;) {
1444            Object k = s.readObject();
1445            if (k == null)
1446                break;
1447            Object v = s.readObject();
1448            if (v == null)
1449                throw new NullPointerException();
1450            K key = (K) k;
1451            V val = (V) v;
1452            int rnd = ThreadLocalRandom.current().nextInt();
1453            int j = 0;
1454            if ((rnd & 0x80000001) == 0) {
1455                do {
1456                    ++j;
1457                } while (((rnd >>>= 1) & 1) != 0);
1458                if (j > h.level) j = h.level + 1;
1459            }
1460            Node<K,V> z = new Node<K,V>(key, val, null);
1461            basepred.next = z;
1462            basepred = z;
1463            if (j > 0) {
1464                Index<K,V> idx = null;
1465                for (int i = 1; i <= j; ++i) {
1466                    idx = new Index<K,V>(z, idx, null);
1467                    if (i > h.level)
1468                        h = new HeadIndex<K,V>(h.node, h, idx, i);
1469
1470                    if (i < preds.size()) {
1471                        preds.get(i).right = idx;
1472                        preds.set(i, idx);
1473                    } else
1474                        preds.add(idx);
1475                }
1476            }
1477        }
1478        head = h;
1479    }
1480
1481    /* ------ Map API methods ------ */
1482
1483    /**
1484     * Returns {@code true} if this map contains a mapping for the specified
1485     * key.
1486     *
1487     * @param key key whose presence in this map is to be tested
1488     * @return {@code true} if this map contains a mapping for the specified key
1489     * @throws ClassCastException if the specified key cannot be compared
1490     *         with the keys currently in the map
1491     * @throws NullPointerException if the specified key is null
1492     */
1493    public boolean containsKey(Object key) {
1494        return doGet(key) != null;
1495    }
1496
1497    /**
1498     * Returns the value to which the specified key is mapped,
1499     * or {@code null} if this map contains no mapping for the key.
1500     *
1501     * <p>More formally, if this map contains a mapping from a key
1502     * {@code k} to a value {@code v} such that {@code key} compares
1503     * equal to {@code k} according to the map's ordering, then this
1504     * method returns {@code v}; otherwise it returns {@code null}.
1505     * (There can be at most one such mapping.)
1506     *
1507     * @throws ClassCastException if the specified key cannot be compared
1508     *         with the keys currently in the map
1509     * @throws NullPointerException if the specified key is null
1510     */
1511    public V get(Object key) {
1512        return doGet(key);
1513    }
1514
1515    /**
1516     * Returns the value to which the specified key is mapped,
1517     * or the given defaultValue if this map contains no mapping for the key.
1518     *
1519     * @param key the key
1520     * @param defaultValue the value to return if this map contains
1521     * no mapping for the given key
1522     * @return the mapping for the key, if present; else the defaultValue
1523     * @throws NullPointerException if the specified key is null
1524     * @since 1.8
1525     */
1526    public V getOrDefault(Object key, V defaultValue) {
1527        V v;
1528        return (v = doGet(key)) == null ? defaultValue : v;
1529    }
1530
1531    /**
1532     * Associates the specified value with the specified key in this map.
1533     * If the map previously contained a mapping for the key, the old
1534     * value is replaced.
1535     *
1536     * @param key key with which the specified value is to be associated
1537     * @param value value to be associated with the specified key
1538     * @return the previous value associated with the specified key, or
1539     *         {@code null} if there was no mapping for the key
1540     * @throws ClassCastException if the specified key cannot be compared
1541     *         with the keys currently in the map
1542     * @throws NullPointerException if the specified key or value is null
1543     */
1544    public V put(K key, V value) {
1545        if (value == null)
1546            throw new NullPointerException();
1547        return doPut(key, value, false);
1548    }
1549
1550    /**
1551     * Removes the mapping for the specified key from this map if present.
1552     *
1553     * @param  key key for which mapping should be removed
1554     * @return the previous value associated with the specified key, or
1555     *         {@code null} if there was no mapping for the key
1556     * @throws ClassCastException if the specified key cannot be compared
1557     *         with the keys currently in the map
1558     * @throws NullPointerException if the specified key is null
1559     */
1560    public V remove(Object key) {
1561        return doRemove(key, null);
1562    }
1563
1564    /**
1565     * Returns {@code true} if this map maps one or more keys to the
1566     * specified value.  This operation requires time linear in the
1567     * map size. Additionally, it is possible for the map to change
1568     * during execution of this method, in which case the returned
1569     * result may be inaccurate.
1570     *
1571     * @param value value whose presence in this map is to be tested
1572     * @return {@code true} if a mapping to {@code value} exists;
1573     *         {@code false} otherwise
1574     * @throws NullPointerException if the specified value is null
1575     */
1576    public boolean containsValue(Object value) {
1577        if (value == null)
1578            throw new NullPointerException();
1579        for (Node<K,V> n = findFirst(); n != null; n = n.next) {
1580            V v = n.getValidValue();
1581            if (v != null && value.equals(v))
1582                return true;
1583        }
1584        return false;
1585    }
1586
1587    /**
1588     * Returns the number of key-value mappings in this map.  If this map
1589     * contains more than {@code Integer.MAX_VALUE} elements, it
1590     * returns {@code Integer.MAX_VALUE}.
1591     *
1592     * <p>Beware that, unlike in most collections, this method is
1593     * <em>NOT</em> a constant-time operation. Because of the
1594     * asynchronous nature of these maps, determining the current
1595     * number of elements requires traversing them all to count them.
1596     * Additionally, it is possible for the size to change during
1597     * execution of this method, in which case the returned result
1598     * will be inaccurate. Thus, this method is typically not very
1599     * useful in concurrent applications.
1600     *
1601     * @return the number of elements in this map
1602     */
1603    public int size() {
1604        long count = 0;
1605        for (Node<K,V> n = findFirst(); n != null; n = n.next) {
1606            if (n.getValidValue() != null)
1607                ++count;
1608        }
1609        return (count >= Integer.MAX_VALUE) ? Integer.MAX_VALUE : (int) count;
1610    }
1611
1612    /**
1613     * Returns {@code true} if this map contains no key-value mappings.
1614     * @return {@code true} if this map contains no key-value mappings
1615     */
1616    public boolean isEmpty() {
1617        return findFirst() == null;
1618    }
1619
1620    /**
1621     * Removes all of the mappings from this map.
1622     */
1623    public void clear() {
1624        initialize();
1625    }
1626
1627    /**
1628     * If the specified key is not already associated with a value,
1629     * attempts to compute its value using the given mapping function
1630     * and enters it into this map unless {@code null}.  The function
1631     * is <em>NOT</em> guaranteed to be applied once atomically only
1632     * if the value is not present.
1633     *
1634     * @param key key with which the specified value is to be associated
1635     * @param mappingFunction the function to compute a value
1636     * @return the current (existing or computed) value associated with
1637     *         the specified key, or null if the computed value is null
1638     * @throws NullPointerException if the specified key is null
1639     *         or the mappingFunction is null
1640     * @since 1.8
1641     */
1642    public V computeIfAbsent(K key,
1643                             Function<? super K, ? extends V> mappingFunction) {
1644        if (key == null || mappingFunction == null)
1645            throw new NullPointerException();
1646        V v, p, r;
1647        if ((v = doGet(key)) == null &&
1648            (r = mappingFunction.apply(key)) != null)
1649            v = (p = doPut(key, r, true)) == null ? r : p;
1650        return v;
1651    }
1652
1653    /**
1654     * If the value for the specified key is present, attempts to
1655     * compute a new mapping given the key and its current mapped
1656     * value. The function is <em>NOT</em> guaranteed to be applied
1657     * once atomically.
1658     *
1659     * @param key key with which a value may be associated
1660     * @param remappingFunction the function to compute a value
1661     * @return the new value associated with the specified key, or null if none
1662     * @throws NullPointerException if the specified key is null
1663     *         or the remappingFunction is null
1664     * @since 1.8
1665     */
1666    public V computeIfPresent(K key,
1667                              BiFunction<? super K, ? super V, ? extends V> remappingFunction) {
1668        if (key == null || remappingFunction == null)
1669            throw new NullPointerException();
1670        Node<K,V> n; Object v;
1671        while ((n = findNode(key)) != null) {
1672            if ((v = n.value) != null) {
1673                @SuppressWarnings("unchecked") V vv = (V) v;
1674                V r = remappingFunction.apply(key, vv);
1675                if (r != null) {
1676                    if (n.casValue(vv, r))
1677                        return r;
1678                }
1679                else if (doRemove(key, vv) != null)
1680                    break;
1681            }
1682        }
1683        return null;
1684    }
1685
1686    /**
1687     * Attempts to compute a mapping for the specified key and its
1688     * current mapped value (or {@code null} if there is no current
1689     * mapping). The function is <em>NOT</em> guaranteed to be applied
1690     * once atomically.
1691     *
1692     * @param key key with which the specified value is to be associated
1693     * @param remappingFunction the function to compute a value
1694     * @return the new value associated with the specified key, or null if none
1695     * @throws NullPointerException if the specified key is null
1696     *         or the remappingFunction is null
1697     * @since 1.8
1698     */
1699    public V compute(K key,
1700                     BiFunction<? super K, ? super V, ? extends V> remappingFunction) {
1701        if (key == null || remappingFunction == null)
1702            throw new NullPointerException();
1703        for (;;) {
1704            Node<K,V> n; Object v; V r;
1705            if ((n = findNode(key)) == null) {
1706                if ((r = remappingFunction.apply(key, null)) == null)
1707                    break;
1708                if (doPut(key, r, true) == null)
1709                    return r;
1710            }
1711            else if ((v = n.value) != null) {
1712                @SuppressWarnings("unchecked") V vv = (V) v;
1713                if ((r = remappingFunction.apply(key, vv)) != null) {
1714                    if (n.casValue(vv, r))
1715                        return r;
1716                }
1717                else if (doRemove(key, vv) != null)
1718                    break;
1719            }
1720        }
1721        return null;
1722    }
1723
1724    /**
1725     * If the specified key is not already associated with a value,
1726     * associates it with the given value.  Otherwise, replaces the
1727     * value with the results of the given remapping function, or
1728     * removes if {@code null}. The function is <em>NOT</em>
1729     * guaranteed to be applied once atomically.
1730     *
1731     * @param key key with which the specified value is to be associated
1732     * @param value the value to use if absent
1733     * @param remappingFunction the function to recompute a value if present
1734     * @return the new value associated with the specified key, or null if none
1735     * @throws NullPointerException if the specified key or value is null
1736     *         or the remappingFunction is null
1737     * @since 1.8
1738     */
1739    public V merge(K key, V value,
1740                   BiFunction<? super V, ? super V, ? extends V> remappingFunction) {
1741        if (key == null || value == null || remappingFunction == null)
1742            throw new NullPointerException();
1743        for (;;) {
1744            Node<K,V> n; Object v; V r;
1745            if ((n = findNode(key)) == null) {
1746                if (doPut(key, value, true) == null)
1747                    return value;
1748            }
1749            else if ((v = n.value) != null) {
1750                @SuppressWarnings("unchecked") V vv = (V) v;
1751                if ((r = remappingFunction.apply(vv, value)) != null) {
1752                    if (n.casValue(vv, r))
1753                        return r;
1754                }
1755                else if (doRemove(key, vv) != null)
1756                    return null;
1757            }
1758        }
1759    }
1760
1761    /* ---------------- View methods -------------- */
1762
1763    /*
1764     * Note: Lazy initialization works for views because view classes
1765     * are stateless/immutable so it doesn't matter wrt correctness if
1766     * more than one is created (which will only rarely happen).  Even
1767     * so, the following idiom conservatively ensures that the method
1768     * returns the one it created if it does so, not one created by
1769     * another racing thread.
1770     */
1771
1772    /**
1773     * Returns a {@link NavigableSet} view of the keys contained in this map.
1774     *
1775     * <p>The set's iterator returns the keys in ascending order.
1776     * The set's spliterator additionally reports {@link Spliterator#CONCURRENT},
1777     * {@link Spliterator#NONNULL}, {@link Spliterator#SORTED} and
1778     * {@link Spliterator#ORDERED}, with an encounter order that is ascending
1779     * key order.  The spliterator's comparator (see
1780     * {@link java.util.Spliterator#getComparator()}) is {@code null} if
1781     * the map's comparator (see {@link #comparator()}) is {@code null}.
1782     * Otherwise, the spliterator's comparator is the same as or imposes the
1783     * same total ordering as the map's comparator.
1784     *
1785     * <p>The set is backed by the map, so changes to the map are
1786     * reflected in the set, and vice-versa.  The set supports element
1787     * removal, which removes the corresponding mapping from the map,
1788     * via the {@code Iterator.remove}, {@code Set.remove},
1789     * {@code removeAll}, {@code retainAll}, and {@code clear}
1790     * operations.  It does not support the {@code add} or {@code addAll}
1791     * operations.
1792     *
1793     * <p>The view's iterators and spliterators are
1794     * <a href="package-summary.html#Weakly"><i>weakly consistent</i></a>.
1795     *
1796     * <p>This method is equivalent to method {@code navigableKeySet}.
1797     *
1798     * @return a navigable set view of the keys in this map
1799     */
1800    public NavigableSet<K> keySet() {
1801        KeySet<K,V> ks = keySet;
1802        return (ks != null) ? ks : (keySet = new KeySet<>(this));
1803    }
1804
1805    public NavigableSet<K> navigableKeySet() {
1806        KeySet<K,V> ks = keySet;
1807        return (ks != null) ? ks : (keySet = new KeySet<>(this));
1808    }
1809
1810    /**
1811     * Returns a {@link Collection} view of the values contained in this map.
1812     * <p>The collection's iterator returns the values in ascending order
1813     * of the corresponding keys. The collections's spliterator additionally
1814     * reports {@link Spliterator#CONCURRENT}, {@link Spliterator#NONNULL} and
1815     * {@link Spliterator#ORDERED}, with an encounter order that is ascending
1816     * order of the corresponding keys.
1817     *
1818     * <p>The collection is backed by the map, so changes to the map are
1819     * reflected in the collection, and vice-versa.  The collection
1820     * supports element removal, which removes the corresponding
1821     * mapping from the map, via the {@code Iterator.remove},
1822     * {@code Collection.remove}, {@code removeAll},
1823     * {@code retainAll} and {@code clear} operations.  It does not
1824     * support the {@code add} or {@code addAll} operations.
1825     *
1826     * <p>The view's iterators and spliterators are
1827     * <a href="package-summary.html#Weakly"><i>weakly consistent</i></a>.
1828     */
1829    public Collection<V> values() {
1830        Values<K,V> vs = values;
1831        return (vs != null) ? vs : (values = new Values<>(this));
1832    }
1833
1834    /**
1835     * Returns a {@link Set} view of the mappings contained in this map.
1836     *
1837     * <p>The set's iterator returns the entries in ascending key order.  The
1838     * set's spliterator additionally reports {@link Spliterator#CONCURRENT},
1839     * {@link Spliterator#NONNULL}, {@link Spliterator#SORTED} and
1840     * {@link Spliterator#ORDERED}, with an encounter order that is ascending
1841     * key order.
1842     *
1843     * <p>The set is backed by the map, so changes to the map are
1844     * reflected in the set, and vice-versa.  The set supports element
1845     * removal, which removes the corresponding mapping from the map,
1846     * via the {@code Iterator.remove}, {@code Set.remove},
1847     * {@code removeAll}, {@code retainAll} and {@code clear}
1848     * operations.  It does not support the {@code add} or
1849     * {@code addAll} operations.
1850     *
1851     * <p>The view's iterators and spliterators are
1852     * <a href="package-summary.html#Weakly"><i>weakly consistent</i></a>.
1853     *
1854     * <p>The {@code Map.Entry} elements traversed by the {@code iterator}
1855     * or {@code spliterator} do <em>not</em> support the {@code setValue}
1856     * operation.
1857     *
1858     * @return a set view of the mappings contained in this map,
1859     *         sorted in ascending key order
1860     */
1861    public Set<Map.Entry<K,V>> entrySet() {
1862        EntrySet<K,V> es = entrySet;
1863        return (es != null) ? es : (entrySet = new EntrySet<K,V>(this));
1864    }
1865
1866    public ConcurrentNavigableMap<K,V> descendingMap() {
1867        ConcurrentNavigableMap<K,V> dm = descendingMap;
1868        return (dm != null) ? dm : (descendingMap = new SubMap<K,V>
1869                                    (this, null, false, null, false, true));
1870    }
1871
1872    public NavigableSet<K> descendingKeySet() {
1873        return descendingMap().navigableKeySet();
1874    }
1875
1876    /* ---------------- AbstractMap Overrides -------------- */
1877
1878    /**
1879     * Compares the specified object with this map for equality.
1880     * Returns {@code true} if the given object is also a map and the
1881     * two maps represent the same mappings.  More formally, two maps
1882     * {@code m1} and {@code m2} represent the same mappings if
1883     * {@code m1.entrySet().equals(m2.entrySet())}.  This
1884     * operation may return misleading results if either map is
1885     * concurrently modified during execution of this method.
1886     *
1887     * @param o object to be compared for equality with this map
1888     * @return {@code true} if the specified object is equal to this map
1889     */
1890    public boolean equals(Object o) {
1891        if (o == this)
1892            return true;
1893        if (!(o instanceof Map))
1894            return false;
1895        Map<?,?> m = (Map<?,?>) o;
1896        try {
1897            for (Map.Entry<K,V> e : this.entrySet())
1898                if (! e.getValue().equals(m.get(e.getKey())))
1899                    return false;
1900            for (Map.Entry<?,?> e : m.entrySet()) {
1901                Object k = e.getKey();
1902                Object v = e.getValue();
1903                if (k == null || v == null || !v.equals(get(k)))
1904                    return false;
1905            }
1906            return true;
1907        } catch (ClassCastException unused) {
1908            return false;
1909        } catch (NullPointerException unused) {
1910            return false;
1911        }
1912    }
1913
1914    /* ------ ConcurrentMap API methods ------ */
1915
1916    /**
1917     * {@inheritDoc}
1918     *
1919     * @return the previous value associated with the specified key,
1920     *         or {@code null} if there was no mapping for the key
1921     * @throws ClassCastException if the specified key cannot be compared
1922     *         with the keys currently in the map
1923     * @throws NullPointerException if the specified key or value is null
1924     */
1925    public V putIfAbsent(K key, V value) {
1926        if (value == null)
1927            throw new NullPointerException();
1928        return doPut(key, value, true);
1929    }
1930
1931    /**
1932     * {@inheritDoc}
1933     *
1934     * @throws ClassCastException if the specified key cannot be compared
1935     *         with the keys currently in the map
1936     * @throws NullPointerException if the specified key is null
1937     */
1938    public boolean remove(Object key, Object value) {
1939        if (key == null)
1940            throw new NullPointerException();
1941        return value != null && doRemove(key, value) != null;
1942    }
1943
1944    /**
1945     * {@inheritDoc}
1946     *
1947     * @throws ClassCastException if the specified key cannot be compared
1948     *         with the keys currently in the map
1949     * @throws NullPointerException if any of the arguments are null
1950     */
1951    public boolean replace(K key, V oldValue, V newValue) {
1952        if (key == null || oldValue == null || newValue == null)
1953            throw new NullPointerException();
1954        for (;;) {
1955            Node<K,V> n; Object v;
1956            if ((n = findNode(key)) == null)
1957                return false;
1958            if ((v = n.value) != null) {
1959                if (!oldValue.equals(v))
1960                    return false;
1961                if (n.casValue(v, newValue))
1962                    return true;
1963            }
1964        }
1965    }
1966
1967    /**
1968     * {@inheritDoc}
1969     *
1970     * @return the previous value associated with the specified key,
1971     *         or {@code null} if there was no mapping for the key
1972     * @throws ClassCastException if the specified key cannot be compared
1973     *         with the keys currently in the map
1974     * @throws NullPointerException if the specified key or value is null
1975     */
1976    public V replace(K key, V value) {
1977        if (key == null || value == null)
1978            throw new NullPointerException();
1979        for (;;) {
1980            Node<K,V> n; Object v;
1981            if ((n = findNode(key)) == null)
1982                return null;
1983            if ((v = n.value) != null && n.casValue(v, value)) {
1984                @SuppressWarnings("unchecked") V vv = (V)v;
1985                return vv;
1986            }
1987        }
1988    }
1989
1990    /* ------ SortedMap API methods ------ */
1991
1992    public Comparator<? super K> comparator() {
1993        return comparator;
1994    }
1995
1996    /**
1997     * @throws NoSuchElementException {@inheritDoc}
1998     */
1999    public K firstKey() {
2000        Node<K,V> n = findFirst();
2001        if (n == null)
2002            throw new NoSuchElementException();
2003        return n.key;
2004    }
2005
2006    /**
2007     * @throws NoSuchElementException {@inheritDoc}
2008     */
2009    public K lastKey() {
2010        Node<K,V> n = findLast();
2011        if (n == null)
2012            throw new NoSuchElementException();
2013        return n.key;
2014    }
2015
2016    /**
2017     * @throws ClassCastException {@inheritDoc}
2018     * @throws NullPointerException if {@code fromKey} or {@code toKey} is null
2019     * @throws IllegalArgumentException {@inheritDoc}
2020     */
2021    public ConcurrentNavigableMap<K,V> subMap(K fromKey,
2022                                              boolean fromInclusive,
2023                                              K toKey,
2024                                              boolean toInclusive) {
2025        if (fromKey == null || toKey == null)
2026            throw new NullPointerException();
2027        return new SubMap<K,V>
2028            (this, fromKey, fromInclusive, toKey, toInclusive, false);
2029    }
2030
2031    /**
2032     * @throws ClassCastException {@inheritDoc}
2033     * @throws NullPointerException if {@code toKey} is null
2034     * @throws IllegalArgumentException {@inheritDoc}
2035     */
2036    public ConcurrentNavigableMap<K,V> headMap(K toKey,
2037                                               boolean inclusive) {
2038        if (toKey == null)
2039            throw new NullPointerException();
2040        return new SubMap<K,V>
2041            (this, null, false, toKey, inclusive, false);
2042    }
2043
2044    /**
2045     * @throws ClassCastException {@inheritDoc}
2046     * @throws NullPointerException if {@code fromKey} is null
2047     * @throws IllegalArgumentException {@inheritDoc}
2048     */
2049    public ConcurrentNavigableMap<K,V> tailMap(K fromKey,
2050                                               boolean inclusive) {
2051        if (fromKey == null)
2052            throw new NullPointerException();
2053        return new SubMap<K,V>
2054            (this, fromKey, inclusive, null, false, false);
2055    }
2056
2057    /**
2058     * @throws ClassCastException {@inheritDoc}
2059     * @throws NullPointerException if {@code fromKey} or {@code toKey} is null
2060     * @throws IllegalArgumentException {@inheritDoc}
2061     */
2062    public ConcurrentNavigableMap<K,V> subMap(K fromKey, K toKey) {
2063        return subMap(fromKey, true, toKey, false);
2064    }
2065
2066    /**
2067     * @throws ClassCastException {@inheritDoc}
2068     * @throws NullPointerException if {@code toKey} is null
2069     * @throws IllegalArgumentException {@inheritDoc}
2070     */
2071    public ConcurrentNavigableMap<K,V> headMap(K toKey) {
2072        return headMap(toKey, false);
2073    }
2074
2075    /**
2076     * @throws ClassCastException {@inheritDoc}
2077     * @throws NullPointerException if {@code fromKey} is null
2078     * @throws IllegalArgumentException {@inheritDoc}
2079     */
2080    public ConcurrentNavigableMap<K,V> tailMap(K fromKey) {
2081        return tailMap(fromKey, true);
2082    }
2083
2084    /* ---------------- Relational operations -------------- */
2085
2086    /**
2087     * Returns a key-value mapping associated with the greatest key
2088     * strictly less than the given key, or {@code null} if there is
2089     * no such key. The returned entry does <em>not</em> support the
2090     * {@code Entry.setValue} method.
2091     *
2092     * @throws ClassCastException {@inheritDoc}
2093     * @throws NullPointerException if the specified key is null
2094     */
2095    public Map.Entry<K,V> lowerEntry(K key) {
2096        return getNear(key, LT);
2097    }
2098
2099    /**
2100     * @throws ClassCastException {@inheritDoc}
2101     * @throws NullPointerException if the specified key is null
2102     */
2103    public K lowerKey(K key) {
2104        Node<K,V> n = findNear(key, LT, comparator);
2105        return (n == null) ? null : n.key;
2106    }
2107
2108    /**
2109     * Returns a key-value mapping associated with the greatest key
2110     * less than or equal to the given key, or {@code null} if there
2111     * is no such key. The returned entry does <em>not</em> support
2112     * the {@code Entry.setValue} method.
2113     *
2114     * @param key the key
2115     * @throws ClassCastException {@inheritDoc}
2116     * @throws NullPointerException if the specified key is null
2117     */
2118    public Map.Entry<K,V> floorEntry(K key) {
2119        return getNear(key, LT|EQ);
2120    }
2121
2122    /**
2123     * @param key the key
2124     * @throws ClassCastException {@inheritDoc}
2125     * @throws NullPointerException if the specified key is null
2126     */
2127    public K floorKey(K key) {
2128        Node<K,V> n = findNear(key, LT|EQ, comparator);
2129        return (n == null) ? null : n.key;
2130    }
2131
2132    /**
2133     * Returns a key-value mapping associated with the least key
2134     * greater than or equal to the given key, or {@code null} if
2135     * there is no such entry. The returned entry does <em>not</em>
2136     * support the {@code Entry.setValue} method.
2137     *
2138     * @throws ClassCastException {@inheritDoc}
2139     * @throws NullPointerException if the specified key is null
2140     */
2141    public Map.Entry<K,V> ceilingEntry(K key) {
2142        return getNear(key, GT|EQ);
2143    }
2144
2145    /**
2146     * @throws ClassCastException {@inheritDoc}
2147     * @throws NullPointerException if the specified key is null
2148     */
2149    public K ceilingKey(K key) {
2150        Node<K,V> n = findNear(key, GT|EQ, comparator);
2151        return (n == null) ? null : n.key;
2152    }
2153
2154    /**
2155     * Returns a key-value mapping associated with the least key
2156     * strictly greater than the given key, or {@code null} if there
2157     * is no such key. The returned entry does <em>not</em> support
2158     * the {@code Entry.setValue} method.
2159     *
2160     * @param key the key
2161     * @throws ClassCastException {@inheritDoc}
2162     * @throws NullPointerException if the specified key is null
2163     */
2164    public Map.Entry<K,V> higherEntry(K key) {
2165        return getNear(key, GT);
2166    }
2167
2168    /**
2169     * @param key the key
2170     * @throws ClassCastException {@inheritDoc}
2171     * @throws NullPointerException if the specified key is null
2172     */
2173    public K higherKey(K key) {
2174        Node<K,V> n = findNear(key, GT, comparator);
2175        return (n == null) ? null : n.key;
2176    }
2177
2178    /**
2179     * Returns a key-value mapping associated with the least
2180     * key in this map, or {@code null} if the map is empty.
2181     * The returned entry does <em>not</em> support
2182     * the {@code Entry.setValue} method.
2183     */
2184    public Map.Entry<K,V> firstEntry() {
2185        for (;;) {
2186            Node<K,V> n = findFirst();
2187            if (n == null)
2188                return null;
2189            AbstractMap.SimpleImmutableEntry<K,V> e = n.createSnapshot();
2190            if (e != null)
2191                return e;
2192        }
2193    }
2194
2195    /**
2196     * Returns a key-value mapping associated with the greatest
2197     * key in this map, or {@code null} if the map is empty.
2198     * The returned entry does <em>not</em> support
2199     * the {@code Entry.setValue} method.
2200     */
2201    public Map.Entry<K,V> lastEntry() {
2202        for (;;) {
2203            Node<K,V> n = findLast();
2204            if (n == null)
2205                return null;
2206            AbstractMap.SimpleImmutableEntry<K,V> e = n.createSnapshot();
2207            if (e != null)
2208                return e;
2209        }
2210    }
2211
2212    /**
2213     * Removes and returns a key-value mapping associated with
2214     * the least key in this map, or {@code null} if the map is empty.
2215     * The returned entry does <em>not</em> support
2216     * the {@code Entry.setValue} method.
2217     */
2218    public Map.Entry<K,V> pollFirstEntry() {
2219        return doRemoveFirstEntry();
2220    }
2221
2222    /**
2223     * Removes and returns a key-value mapping associated with
2224     * the greatest key in this map, or {@code null} if the map is empty.
2225     * The returned entry does <em>not</em> support
2226     * the {@code Entry.setValue} method.
2227     */
2228    public Map.Entry<K,V> pollLastEntry() {
2229        return doRemoveLastEntry();
2230    }
2231
2232
2233    /* ---------------- Iterators -------------- */
2234
2235    /**
2236     * Base of iterator classes:
2237     */
2238    abstract class Iter<T> implements Iterator<T> {
2239        /** the last node returned by next() */
2240        Node<K,V> lastReturned;
2241        /** the next node to return from next(); */
2242        Node<K,V> next;
2243        /** Cache of next value field to maintain weak consistency */
2244        V nextValue;
2245
2246        /** Initializes ascending iterator for entire range. */
2247        Iter() {
2248            while ((next = findFirst()) != null) {
2249                Object x = next.value;
2250                if (x != null && x != next) {
2251                    @SuppressWarnings("unchecked") V vv = (V)x;
2252                    nextValue = vv;
2253                    break;
2254                }
2255            }
2256        }
2257
2258        public final boolean hasNext() {
2259            return next != null;
2260        }
2261
2262        /** Advances next to higher entry. */
2263        final void advance() {
2264            if (next == null)
2265                throw new NoSuchElementException();
2266            lastReturned = next;
2267            while ((next = next.next) != null) {
2268                Object x = next.value;
2269                if (x != null && x != next) {
2270                    @SuppressWarnings("unchecked") V vv = (V)x;
2271                    nextValue = vv;
2272                    break;
2273                }
2274            }
2275        }
2276
2277        public void remove() {
2278            Node<K,V> l = lastReturned;
2279            if (l == null)
2280                throw new IllegalStateException();
2281            // It would not be worth all of the overhead to directly
2282            // unlink from here. Using remove is fast enough.
2283            ConcurrentSkipListMap.this.remove(l.key);
2284            lastReturned = null;
2285        }
2286
2287    }
2288
2289    final class ValueIterator extends Iter<V> {
2290        public V next() {
2291            V v = nextValue;
2292            advance();
2293            return v;
2294        }
2295    }
2296
2297    final class KeyIterator extends Iter<K> {
2298        public K next() {
2299            Node<K,V> n = next;
2300            advance();
2301            return n.key;
2302        }
2303    }
2304
2305    final class EntryIterator extends Iter<Map.Entry<K,V>> {
2306        public Map.Entry<K,V> next() {
2307            Node<K,V> n = next;
2308            V v = nextValue;
2309            advance();
2310            return new AbstractMap.SimpleImmutableEntry<K,V>(n.key, v);
2311        }
2312    }
2313
2314    /* ---------------- View Classes -------------- */
2315
2316    /*
2317     * View classes are static, delegating to a ConcurrentNavigableMap
2318     * to allow use by SubMaps, which outweighs the ugliness of
2319     * needing type-tests for Iterator methods.
2320     */
2321
2322    static final <E> List<E> toList(Collection<E> c) {
2323        // Using size() here would be a pessimization.
2324        ArrayList<E> list = new ArrayList<E>();
2325        for (E e : c)
2326            list.add(e);
2327        return list;
2328    }
2329
2330    static final class KeySet<K,V>
2331            extends AbstractSet<K> implements NavigableSet<K> {
2332        final ConcurrentNavigableMap<K,V> m;
2333        KeySet(ConcurrentNavigableMap<K,V> map) { m = map; }
2334        public int size() { return m.size(); }
2335        public boolean isEmpty() { return m.isEmpty(); }
2336        public boolean contains(Object o) { return m.containsKey(o); }
2337        public boolean remove(Object o) { return m.remove(o) != null; }
2338        public void clear() { m.clear(); }
2339        public K lower(K e) { return m.lowerKey(e); }
2340        public K floor(K e) { return m.floorKey(e); }
2341        public K ceiling(K e) { return m.ceilingKey(e); }
2342        public K higher(K e) { return m.higherKey(e); }
2343        public Comparator<? super K> comparator() { return m.comparator(); }
2344        public K first() { return m.firstKey(); }
2345        public K last() { return m.lastKey(); }
2346        public K pollFirst() {
2347            Map.Entry<K,V> e = m.pollFirstEntry();
2348            return (e == null) ? null : e.getKey();
2349        }
2350        public K pollLast() {
2351            Map.Entry<K,V> e = m.pollLastEntry();
2352            return (e == null) ? null : e.getKey();
2353        }
2354        public Iterator<K> iterator() {
2355            return (m instanceof ConcurrentSkipListMap)
2356                ? ((ConcurrentSkipListMap<K,V>)m).new KeyIterator()
2357                : ((SubMap<K,V>)m).new SubMapKeyIterator();
2358        }
2359        public boolean equals(Object o) {
2360            if (o == this)
2361                return true;
2362            if (!(o instanceof Set))
2363                return false;
2364            Collection<?> c = (Collection<?>) o;
2365            try {
2366                return containsAll(c) && c.containsAll(this);
2367            } catch (ClassCastException unused) {
2368                return false;
2369            } catch (NullPointerException unused) {
2370                return false;
2371            }
2372        }
2373        public Object[] toArray()     { return toList(this).toArray();  }
2374        public <T> T[] toArray(T[] a) { return toList(this).toArray(a); }
2375        public Iterator<K> descendingIterator() {
2376            return descendingSet().iterator();
2377        }
2378        public NavigableSet<K> subSet(K fromElement,
2379                                      boolean fromInclusive,
2380                                      K toElement,
2381                                      boolean toInclusive) {
2382            return new KeySet<>(m.subMap(fromElement, fromInclusive,
2383                                         toElement,   toInclusive));
2384        }
2385        public NavigableSet<K> headSet(K toElement, boolean inclusive) {
2386            return new KeySet<>(m.headMap(toElement, inclusive));
2387        }
2388        public NavigableSet<K> tailSet(K fromElement, boolean inclusive) {
2389            return new KeySet<>(m.tailMap(fromElement, inclusive));
2390        }
2391        public NavigableSet<K> subSet(K fromElement, K toElement) {
2392            return subSet(fromElement, true, toElement, false);
2393        }
2394        public NavigableSet<K> headSet(K toElement) {
2395            return headSet(toElement, false);
2396        }
2397        public NavigableSet<K> tailSet(K fromElement) {
2398            return tailSet(fromElement, true);
2399        }
2400        public NavigableSet<K> descendingSet() {
2401            return new KeySet<>(m.descendingMap());
2402        }
2403
2404        public Spliterator<K> spliterator() {
2405            return (m instanceof ConcurrentSkipListMap)
2406                ? ((ConcurrentSkipListMap<K,V>)m).keySpliterator()
2407                : ((SubMap<K,V>)m).new SubMapKeyIterator();
2408        }
2409    }
2410
2411    static final class Values<K,V> extends AbstractCollection<V> {
2412        final ConcurrentNavigableMap<K,V> m;
2413        Values(ConcurrentNavigableMap<K,V> map) {
2414            m = map;
2415        }
2416        public Iterator<V> iterator() {
2417            return (m instanceof ConcurrentSkipListMap)
2418                ? ((ConcurrentSkipListMap<K,V>)m).new ValueIterator()
2419                : ((SubMap<K,V>)m).new SubMapValueIterator();
2420        }
2421        public int size() { return m.size(); }
2422        public boolean isEmpty() { return m.isEmpty(); }
2423        public boolean contains(Object o) { return m.containsValue(o); }
2424        public void clear() { m.clear(); }
2425        public Object[] toArray()     { return toList(this).toArray();  }
2426        public <T> T[] toArray(T[] a) { return toList(this).toArray(a); }
2427
2428        public Spliterator<V> spliterator() {
2429            return (m instanceof ConcurrentSkipListMap)
2430                ? ((ConcurrentSkipListMap<K,V>)m).valueSpliterator()
2431                : ((SubMap<K,V>)m).new SubMapValueIterator();
2432        }
2433
2434        public boolean removeIf(Predicate<? super V> filter) {
2435            if (filter == null) throw new NullPointerException();
2436            if (m instanceof ConcurrentSkipListMap)
2437                return ((ConcurrentSkipListMap<K,V>)m).removeValueIf(filter);
2438            // else use iterator
2439            Iterator<Map.Entry<K,V>> it =
2440                ((SubMap<K,V>)m).new SubMapEntryIterator();
2441            boolean removed = false;
2442            while (it.hasNext()) {
2443                Map.Entry<K,V> e = it.next();
2444                V v = e.getValue();
2445                if (filter.test(v) && m.remove(e.getKey(), v))
2446                    removed = true;
2447            }
2448            return removed;
2449        }
2450    }
2451
2452    static final class EntrySet<K,V> extends AbstractSet<Map.Entry<K,V>> {
2453        final ConcurrentNavigableMap<K,V> m;
2454        EntrySet(ConcurrentNavigableMap<K,V> map) {
2455            m = map;
2456        }
2457        public Iterator<Map.Entry<K,V>> iterator() {
2458            return (m instanceof ConcurrentSkipListMap)
2459                ? ((ConcurrentSkipListMap<K,V>)m).new EntryIterator()
2460                : ((SubMap<K,V>)m).new SubMapEntryIterator();
2461        }
2462
2463        public boolean contains(Object o) {
2464            if (!(o instanceof Map.Entry))
2465                return false;
2466            Map.Entry<?,?> e = (Map.Entry<?,?>)o;
2467            V v = m.get(e.getKey());
2468            return v != null && v.equals(e.getValue());
2469        }
2470        public boolean remove(Object o) {
2471            if (!(o instanceof Map.Entry))
2472                return false;
2473            Map.Entry<?,?> e = (Map.Entry<?,?>)o;
2474            return m.remove(e.getKey(),
2475                            e.getValue());
2476        }
2477        public boolean isEmpty() {
2478            return m.isEmpty();
2479        }
2480        public int size() {
2481            return m.size();
2482        }
2483        public void clear() {
2484            m.clear();
2485        }
2486        public boolean equals(Object o) {
2487            if (o == this)
2488                return true;
2489            if (!(o instanceof Set))
2490                return false;
2491            Collection<?> c = (Collection<?>) o;
2492            try {
2493                return containsAll(c) && c.containsAll(this);
2494            } catch (ClassCastException unused) {
2495                return false;
2496            } catch (NullPointerException unused) {
2497                return false;
2498            }
2499        }
2500        public Object[] toArray()     { return toList(this).toArray();  }
2501        public <T> T[] toArray(T[] a) { return toList(this).toArray(a); }
2502
2503        public Spliterator<Map.Entry<K,V>> spliterator() {
2504            return (m instanceof ConcurrentSkipListMap)
2505                ? ((ConcurrentSkipListMap<K,V>)m).entrySpliterator()
2506                : ((SubMap<K,V>)m).new SubMapEntryIterator();
2507        }
2508        public boolean removeIf(Predicate<? super Entry<K,V>> filter) {
2509            if (filter == null) throw new NullPointerException();
2510            if (m instanceof ConcurrentSkipListMap)
2511                return ((ConcurrentSkipListMap<K,V>)m).removeEntryIf(filter);
2512            // else use iterator
2513            Iterator<Map.Entry<K,V>> it =
2514                ((SubMap<K,V>)m).new SubMapEntryIterator();
2515            boolean removed = false;
2516            while (it.hasNext()) {
2517                Map.Entry<K,V> e = it.next();
2518                if (filter.test(e) && m.remove(e.getKey(), e.getValue()))
2519                    removed = true;
2520            }
2521            return removed;
2522        }
2523    }
2524
2525    /**
2526     * Submaps returned by {@link ConcurrentSkipListMap} submap operations
2527     * represent a subrange of mappings of their underlying maps.
2528     * Instances of this class support all methods of their underlying
2529     * maps, differing in that mappings outside their range are ignored,
2530     * and attempts to add mappings outside their ranges result in {@link
2531     * IllegalArgumentException}.  Instances of this class are constructed
2532     * only using the {@code subMap}, {@code headMap}, and {@code tailMap}
2533     * methods of their underlying maps.
2534     *
2535     * @serial include
2536     */
2537    static final class SubMap<K,V> extends AbstractMap<K,V>
2538        implements ConcurrentNavigableMap<K,V>, Cloneable, Serializable {
2539        private static final long serialVersionUID = -7647078645895051609L;
2540
2541        /** Underlying map */
2542        final ConcurrentSkipListMap<K,V> m;
2543        /** lower bound key, or null if from start */
2544        private final K lo;
2545        /** upper bound key, or null if to end */
2546        private final K hi;
2547        /** inclusion flag for lo */
2548        private final boolean loInclusive;
2549        /** inclusion flag for hi */
2550        private final boolean hiInclusive;
2551        /** direction */
2552        final boolean isDescending;
2553
2554        // Lazily initialized view holders
2555        private transient KeySet<K,V> keySetView;
2556        private transient Set<Map.Entry<K,V>> entrySetView;
2557        private transient Collection<V> valuesView;
2558
2559        /**
2560         * Creates a new submap, initializing all fields.
2561         */
2562        SubMap(ConcurrentSkipListMap<K,V> map,
2563               K fromKey, boolean fromInclusive,
2564               K toKey, boolean toInclusive,
2565               boolean isDescending) {
2566            Comparator<? super K> cmp = map.comparator;
2567            if (fromKey != null && toKey != null &&
2568                cpr(cmp, fromKey, toKey) > 0)
2569                throw new IllegalArgumentException("inconsistent range");
2570            this.m = map;
2571            this.lo = fromKey;
2572            this.hi = toKey;
2573            this.loInclusive = fromInclusive;
2574            this.hiInclusive = toInclusive;
2575            this.isDescending = isDescending;
2576        }
2577
2578        /* ----------------  Utilities -------------- */
2579
2580        boolean tooLow(Object key, Comparator<? super K> cmp) {
2581            int c;
2582            return (lo != null && ((c = cpr(cmp, key, lo)) < 0 ||
2583                                   (c == 0 && !loInclusive)));
2584        }
2585
2586        boolean tooHigh(Object key, Comparator<? super K> cmp) {
2587            int c;
2588            return (hi != null && ((c = cpr(cmp, key, hi)) > 0 ||
2589                                   (c == 0 && !hiInclusive)));
2590        }
2591
2592        boolean inBounds(Object key, Comparator<? super K> cmp) {
2593            return !tooLow(key, cmp) && !tooHigh(key, cmp);
2594        }
2595
2596        void checkKeyBounds(K key, Comparator<? super K> cmp) {
2597            if (key == null)
2598                throw new NullPointerException();
2599            if (!inBounds(key, cmp))
2600                throw new IllegalArgumentException("key out of range");
2601        }
2602
2603        /**
2604         * Returns true if node key is less than upper bound of range.
2605         */
2606        boolean isBeforeEnd(ConcurrentSkipListMap.Node<K,V> n,
2607                            Comparator<? super K> cmp) {
2608            if (n == null)
2609                return false;
2610            if (hi == null)
2611                return true;
2612            K k = n.key;
2613            if (k == null) // pass by markers and headers
2614                return true;
2615            int c = cpr(cmp, k, hi);
2616            if (c > 0 || (c == 0 && !hiInclusive))
2617                return false;
2618            return true;
2619        }
2620
2621        /**
2622         * Returns lowest node. This node might not be in range, so
2623         * most usages need to check bounds.
2624         */
2625        ConcurrentSkipListMap.Node<K,V> loNode(Comparator<? super K> cmp) {
2626            if (lo == null)
2627                return m.findFirst();
2628            else if (loInclusive)
2629                return m.findNear(lo, GT|EQ, cmp);
2630            else
2631                return m.findNear(lo, GT, cmp);
2632        }
2633
2634        /**
2635         * Returns highest node. This node might not be in range, so
2636         * most usages need to check bounds.
2637         */
2638        ConcurrentSkipListMap.Node<K,V> hiNode(Comparator<? super K> cmp) {
2639            if (hi == null)
2640                return m.findLast();
2641            else if (hiInclusive)
2642                return m.findNear(hi, LT|EQ, cmp);
2643            else
2644                return m.findNear(hi, LT, cmp);
2645        }
2646
2647        /**
2648         * Returns lowest absolute key (ignoring directionality).
2649         */
2650        K lowestKey() {
2651            Comparator<? super K> cmp = m.comparator;
2652            ConcurrentSkipListMap.Node<K,V> n = loNode(cmp);
2653            if (isBeforeEnd(n, cmp))
2654                return n.key;
2655            else
2656                throw new NoSuchElementException();
2657        }
2658
2659        /**
2660         * Returns highest absolute key (ignoring directionality).
2661         */
2662        K highestKey() {
2663            Comparator<? super K> cmp = m.comparator;
2664            ConcurrentSkipListMap.Node<K,V> n = hiNode(cmp);
2665            if (n != null) {
2666                K last = n.key;
2667                if (inBounds(last, cmp))
2668                    return last;
2669            }
2670            throw new NoSuchElementException();
2671        }
2672
2673        Map.Entry<K,V> lowestEntry() {
2674            Comparator<? super K> cmp = m.comparator;
2675            for (;;) {
2676                ConcurrentSkipListMap.Node<K,V> n = loNode(cmp);
2677                if (!isBeforeEnd(n, cmp))
2678                    return null;
2679                Map.Entry<K,V> e = n.createSnapshot();
2680                if (e != null)
2681                    return e;
2682            }
2683        }
2684
2685        Map.Entry<K,V> highestEntry() {
2686            Comparator<? super K> cmp = m.comparator;
2687            for (;;) {
2688                ConcurrentSkipListMap.Node<K,V> n = hiNode(cmp);
2689                if (n == null || !inBounds(n.key, cmp))
2690                    return null;
2691                Map.Entry<K,V> e = n.createSnapshot();
2692                if (e != null)
2693                    return e;
2694            }
2695        }
2696
2697        Map.Entry<K,V> removeLowest() {
2698            Comparator<? super K> cmp = m.comparator;
2699            for (;;) {
2700                Node<K,V> n = loNode(cmp);
2701                if (n == null)
2702                    return null;
2703                K k = n.key;
2704                if (!inBounds(k, cmp))
2705                    return null;
2706                V v = m.doRemove(k, null);
2707                if (v != null)
2708                    return new AbstractMap.SimpleImmutableEntry<K,V>(k, v);
2709            }
2710        }
2711
2712        Map.Entry<K,V> removeHighest() {
2713            Comparator<? super K> cmp = m.comparator;
2714            for (;;) {
2715                Node<K,V> n = hiNode(cmp);
2716                if (n == null)
2717                    return null;
2718                K k = n.key;
2719                if (!inBounds(k, cmp))
2720                    return null;
2721                V v = m.doRemove(k, null);
2722                if (v != null)
2723                    return new AbstractMap.SimpleImmutableEntry<K,V>(k, v);
2724            }
2725        }
2726
2727        /**
2728         * Submap version of ConcurrentSkipListMap.getNearEntry.
2729         */
2730        Map.Entry<K,V> getNearEntry(K key, int rel) {
2731            Comparator<? super K> cmp = m.comparator;
2732            if (isDescending) { // adjust relation for direction
2733                if ((rel & LT) == 0)
2734                    rel |= LT;
2735                else
2736                    rel &= ~LT;
2737            }
2738            if (tooLow(key, cmp))
2739                return ((rel & LT) != 0) ? null : lowestEntry();
2740            if (tooHigh(key, cmp))
2741                return ((rel & LT) != 0) ? highestEntry() : null;
2742            for (;;) {
2743                Node<K,V> n = m.findNear(key, rel, cmp);
2744                if (n == null || !inBounds(n.key, cmp))
2745                    return null;
2746                K k = n.key;
2747                V v = n.getValidValue();
2748                if (v != null)
2749                    return new AbstractMap.SimpleImmutableEntry<K,V>(k, v);
2750            }
2751        }
2752
2753        // Almost the same as getNearEntry, except for keys
2754        K getNearKey(K key, int rel) {
2755            Comparator<? super K> cmp = m.comparator;
2756            if (isDescending) { // adjust relation for direction
2757                if ((rel & LT) == 0)
2758                    rel |= LT;
2759                else
2760                    rel &= ~LT;
2761            }
2762            if (tooLow(key, cmp)) {
2763                if ((rel & LT) == 0) {
2764                    ConcurrentSkipListMap.Node<K,V> n = loNode(cmp);
2765                    if (isBeforeEnd(n, cmp))
2766                        return n.key;
2767                }
2768                return null;
2769            }
2770            if (tooHigh(key, cmp)) {
2771                if ((rel & LT) != 0) {
2772                    ConcurrentSkipListMap.Node<K,V> n = hiNode(cmp);
2773                    if (n != null) {
2774                        K last = n.key;
2775                        if (inBounds(last, cmp))
2776                            return last;
2777                    }
2778                }
2779                return null;
2780            }
2781            for (;;) {
2782                Node<K,V> n = m.findNear(key, rel, cmp);
2783                if (n == null || !inBounds(n.key, cmp))
2784                    return null;
2785                K k = n.key;
2786                V v = n.getValidValue();
2787                if (v != null)
2788                    return k;
2789            }
2790        }
2791
2792        /* ----------------  Map API methods -------------- */
2793
2794        public boolean containsKey(Object key) {
2795            if (key == null) throw new NullPointerException();
2796            return inBounds(key, m.comparator) && m.containsKey(key);
2797        }
2798
2799        public V get(Object key) {
2800            if (key == null) throw new NullPointerException();
2801            return (!inBounds(key, m.comparator)) ? null : m.get(key);
2802        }
2803
2804        public V put(K key, V value) {
2805            checkKeyBounds(key, m.comparator);
2806            return m.put(key, value);
2807        }
2808
2809        public V remove(Object key) {
2810            return (!inBounds(key, m.comparator)) ? null : m.remove(key);
2811        }
2812
2813        public int size() {
2814            Comparator<? super K> cmp = m.comparator;
2815            long count = 0;
2816            for (ConcurrentSkipListMap.Node<K,V> n = loNode(cmp);
2817                 isBeforeEnd(n, cmp);
2818                 n = n.next) {
2819                if (n.getValidValue() != null)
2820                    ++count;
2821            }
2822            return count >= Integer.MAX_VALUE ? Integer.MAX_VALUE : (int)count;
2823        }
2824
2825        public boolean isEmpty() {
2826            Comparator<? super K> cmp = m.comparator;
2827            return !isBeforeEnd(loNode(cmp), cmp);
2828        }
2829
2830        public boolean containsValue(Object value) {
2831            if (value == null)
2832                throw new NullPointerException();
2833            Comparator<? super K> cmp = m.comparator;
2834            for (ConcurrentSkipListMap.Node<K,V> n = loNode(cmp);
2835                 isBeforeEnd(n, cmp);
2836                 n = n.next) {
2837                V v = n.getValidValue();
2838                if (v != null && value.equals(v))
2839                    return true;
2840            }
2841            return false;
2842        }
2843
2844        public void clear() {
2845            Comparator<? super K> cmp = m.comparator;
2846            for (ConcurrentSkipListMap.Node<K,V> n = loNode(cmp);
2847                 isBeforeEnd(n, cmp);
2848                 n = n.next) {
2849                if (n.getValidValue() != null)
2850                    m.remove(n.key);
2851            }
2852        }
2853
2854        /* ----------------  ConcurrentMap API methods -------------- */
2855
2856        public V putIfAbsent(K key, V value) {
2857            checkKeyBounds(key, m.comparator);
2858            return m.putIfAbsent(key, value);
2859        }
2860
2861        public boolean remove(Object key, Object value) {
2862            return inBounds(key, m.comparator) && m.remove(key, value);
2863        }
2864
2865        public boolean replace(K key, V oldValue, V newValue) {
2866            checkKeyBounds(key, m.comparator);
2867            return m.replace(key, oldValue, newValue);
2868        }
2869
2870        public V replace(K key, V value) {
2871            checkKeyBounds(key, m.comparator);
2872            return m.replace(key, value);
2873        }
2874
2875        /* ----------------  SortedMap API methods -------------- */
2876
2877        public Comparator<? super K> comparator() {
2878            Comparator<? super K> cmp = m.comparator();
2879            if (isDescending)
2880                return Collections.reverseOrder(cmp);
2881            else
2882                return cmp;
2883        }
2884
2885        /**
2886         * Utility to create submaps, where given bounds override
2887         * unbounded(null) ones and/or are checked against bounded ones.
2888         */
2889        SubMap<K,V> newSubMap(K fromKey, boolean fromInclusive,
2890                              K toKey, boolean toInclusive) {
2891            Comparator<? super K> cmp = m.comparator;
2892            if (isDescending) { // flip senses
2893                K tk = fromKey;
2894                fromKey = toKey;
2895                toKey = tk;
2896                boolean ti = fromInclusive;
2897                fromInclusive = toInclusive;
2898                toInclusive = ti;
2899            }
2900            if (lo != null) {
2901                if (fromKey == null) {
2902                    fromKey = lo;
2903                    fromInclusive = loInclusive;
2904                }
2905                else {
2906                    int c = cpr(cmp, fromKey, lo);
2907                    if (c < 0 || (c == 0 && !loInclusive && fromInclusive))
2908                        throw new IllegalArgumentException("key out of range");
2909                }
2910            }
2911            if (hi != null) {
2912                if (toKey == null) {
2913                    toKey = hi;
2914                    toInclusive = hiInclusive;
2915                }
2916                else {
2917                    int c = cpr(cmp, toKey, hi);
2918                    if (c > 0 || (c == 0 && !hiInclusive && toInclusive))
2919                        throw new IllegalArgumentException("key out of range");
2920                }
2921            }
2922            return new SubMap<K,V>(m, fromKey, fromInclusive,
2923                                   toKey, toInclusive, isDescending);
2924        }
2925
2926        public SubMap<K,V> subMap(K fromKey, boolean fromInclusive,
2927                                  K toKey, boolean toInclusive) {
2928            if (fromKey == null || toKey == null)
2929                throw new NullPointerException();
2930            return newSubMap(fromKey, fromInclusive, toKey, toInclusive);
2931        }
2932
2933        public SubMap<K,V> headMap(K toKey, boolean inclusive) {
2934            if (toKey == null)
2935                throw new NullPointerException();
2936            return newSubMap(null, false, toKey, inclusive);
2937        }
2938
2939        public SubMap<K,V> tailMap(K fromKey, boolean inclusive) {
2940            if (fromKey == null)
2941                throw new NullPointerException();
2942            return newSubMap(fromKey, inclusive, null, false);
2943        }
2944
2945        public SubMap<K,V> subMap(K fromKey, K toKey) {
2946            return subMap(fromKey, true, toKey, false);
2947        }
2948
2949        public SubMap<K,V> headMap(K toKey) {
2950            return headMap(toKey, false);
2951        }
2952
2953        public SubMap<K,V> tailMap(K fromKey) {
2954            return tailMap(fromKey, true);
2955        }
2956
2957        public SubMap<K,V> descendingMap() {
2958            return new SubMap<K,V>(m, lo, loInclusive,
2959                                   hi, hiInclusive, !isDescending);
2960        }
2961
2962        /* ----------------  Relational methods -------------- */
2963
2964        public Map.Entry<K,V> ceilingEntry(K key) {
2965            return getNearEntry(key, GT|EQ);
2966        }
2967
2968        public K ceilingKey(K key) {
2969            return getNearKey(key, GT|EQ);
2970        }
2971
2972        public Map.Entry<K,V> lowerEntry(K key) {
2973            return getNearEntry(key, LT);
2974        }
2975
2976        public K lowerKey(K key) {
2977            return getNearKey(key, LT);
2978        }
2979
2980        public Map.Entry<K,V> floorEntry(K key) {
2981            return getNearEntry(key, LT|EQ);
2982        }
2983
2984        public K floorKey(K key) {
2985            return getNearKey(key, LT|EQ);
2986        }
2987
2988        public Map.Entry<K,V> higherEntry(K key) {
2989            return getNearEntry(key, GT);
2990        }
2991
2992        public K higherKey(K key) {
2993            return getNearKey(key, GT);
2994        }
2995
2996        public K firstKey() {
2997            return isDescending ? highestKey() : lowestKey();
2998        }
2999
3000        public K lastKey() {
3001            return isDescending ? lowestKey() : highestKey();
3002        }
3003
3004        public Map.Entry<K,V> firstEntry() {
3005            return isDescending ? highestEntry() : lowestEntry();
3006        }
3007
3008        public Map.Entry<K,V> lastEntry() {
3009            return isDescending ? lowestEntry() : highestEntry();
3010        }
3011
3012        public Map.Entry<K,V> pollFirstEntry() {
3013            return isDescending ? removeHighest() : removeLowest();
3014        }
3015
3016        public Map.Entry<K,V> pollLastEntry() {
3017            return isDescending ? removeLowest() : removeHighest();
3018        }
3019
3020        /* ---------------- Submap Views -------------- */
3021
3022        public NavigableSet<K> keySet() {
3023            KeySet<K,V> ks = keySetView;
3024            return (ks != null) ? ks : (keySetView = new KeySet<>(this));
3025        }
3026
3027        public NavigableSet<K> navigableKeySet() {
3028            KeySet<K,V> ks = keySetView;
3029            return (ks != null) ? ks : (keySetView = new KeySet<>(this));
3030        }
3031
3032        public Collection<V> values() {
3033            Collection<V> vs = valuesView;
3034            return (vs != null) ? vs : (valuesView = new Values<>(this));
3035        }
3036
3037        public Set<Map.Entry<K,V>> entrySet() {
3038            Set<Map.Entry<K,V>> es = entrySetView;
3039            return (es != null) ? es : (entrySetView = new EntrySet<K,V>(this));
3040        }
3041
3042        public NavigableSet<K> descendingKeySet() {
3043            return descendingMap().navigableKeySet();
3044        }
3045
3046        /**
3047         * Variant of main Iter class to traverse through submaps.
3048         * Also serves as back-up Spliterator for views.
3049         */
3050        abstract class SubMapIter<T> implements Iterator<T>, Spliterator<T> {
3051            /** the last node returned by next() */
3052            Node<K,V> lastReturned;
3053            /** the next node to return from next(); */
3054            Node<K,V> next;
3055            /** Cache of next value field to maintain weak consistency */
3056            V nextValue;
3057
3058            SubMapIter() {
3059                Comparator<? super K> cmp = m.comparator;
3060                for (;;) {
3061                    next = isDescending ? hiNode(cmp) : loNode(cmp);
3062                    if (next == null)
3063                        break;
3064                    Object x = next.value;
3065                    if (x != null && x != next) {
3066                        if (! inBounds(next.key, cmp))
3067                            next = null;
3068                        else {
3069                            @SuppressWarnings("unchecked") V vv = (V)x;
3070                            nextValue = vv;
3071                        }
3072                        break;
3073                    }
3074                }
3075            }
3076
3077            public final boolean hasNext() {
3078                return next != null;
3079            }
3080
3081            final void advance() {
3082                if (next == null)
3083                    throw new NoSuchElementException();
3084                lastReturned = next;
3085                if (isDescending)
3086                    descend();
3087                else
3088                    ascend();
3089            }
3090
3091            private void ascend() {
3092                Comparator<? super K> cmp = m.comparator;
3093                for (;;) {
3094                    next = next.next;
3095                    if (next == null)
3096                        break;
3097                    Object x = next.value;
3098                    if (x != null && x != next) {
3099                        if (tooHigh(next.key, cmp))
3100                            next = null;
3101                        else {
3102                            @SuppressWarnings("unchecked") V vv = (V)x;
3103                            nextValue = vv;
3104                        }
3105                        break;
3106                    }
3107                }
3108            }
3109
3110            private void descend() {
3111                Comparator<? super K> cmp = m.comparator;
3112                for (;;) {
3113                    next = m.findNear(lastReturned.key, LT, cmp);
3114                    if (next == null)
3115                        break;
3116                    Object x = next.value;
3117                    if (x != null && x != next) {
3118                        if (tooLow(next.key, cmp))
3119                            next = null;
3120                        else {
3121                            @SuppressWarnings("unchecked") V vv = (V)x;
3122                            nextValue = vv;
3123                        }
3124                        break;
3125                    }
3126                }
3127            }
3128
3129            public void remove() {
3130                Node<K,V> l = lastReturned;
3131                if (l == null)
3132                    throw new IllegalStateException();
3133                m.remove(l.key);
3134                lastReturned = null;
3135            }
3136
3137            public Spliterator<T> trySplit() {
3138                return null;
3139            }
3140
3141            public boolean tryAdvance(Consumer<? super T> action) {
3142                if (hasNext()) {
3143                    action.accept(next());
3144                    return true;
3145                }
3146                return false;
3147            }
3148
3149            public void forEachRemaining(Consumer<? super T> action) {
3150                while (hasNext())
3151                    action.accept(next());
3152            }
3153
3154            public long estimateSize() {
3155                return Long.MAX_VALUE;
3156            }
3157
3158        }
3159
3160        final class SubMapValueIterator extends SubMapIter<V> {
3161            public V next() {
3162                V v = nextValue;
3163                advance();
3164                return v;
3165            }
3166            public int characteristics() {
3167                return 0;
3168            }
3169        }
3170
3171        final class SubMapKeyIterator extends SubMapIter<K> {
3172            public K next() {
3173                Node<K,V> n = next;
3174                advance();
3175                return n.key;
3176            }
3177            public int characteristics() {
3178                return Spliterator.DISTINCT | Spliterator.ORDERED |
3179                    Spliterator.SORTED;
3180            }
3181            public final Comparator<? super K> getComparator() {
3182                return SubMap.this.comparator();
3183            }
3184        }
3185
3186        final class SubMapEntryIterator extends SubMapIter<Map.Entry<K,V>> {
3187            public Map.Entry<K,V> next() {
3188                Node<K,V> n = next;
3189                V v = nextValue;
3190                advance();
3191                return new AbstractMap.SimpleImmutableEntry<K,V>(n.key, v);
3192            }
3193            public int characteristics() {
3194                return Spliterator.DISTINCT;
3195            }
3196        }
3197    }
3198
3199    // default Map method overrides
3200
3201    public void forEach(BiConsumer<? super K, ? super V> action) {
3202        if (action == null) throw new NullPointerException();
3203        V v;
3204        for (Node<K,V> n = findFirst(); n != null; n = n.next) {
3205            if ((v = n.getValidValue()) != null)
3206                action.accept(n.key, v);
3207        }
3208    }
3209
3210    public void replaceAll(BiFunction<? super K, ? super V, ? extends V> function) {
3211        if (function == null) throw new NullPointerException();
3212        V v;
3213        for (Node<K,V> n = findFirst(); n != null; n = n.next) {
3214            while ((v = n.getValidValue()) != null) {
3215                V r = function.apply(n.key, v);
3216                if (r == null) throw new NullPointerException();
3217                if (n.casValue(v, r))
3218                    break;
3219            }
3220        }
3221    }
3222
3223    /**
3224     * Helper method for EntrySet.removeIf.
3225     */
3226    boolean removeEntryIf(Predicate<? super Entry<K,V>> function) {
3227        if (function == null) throw new NullPointerException();
3228        boolean removed = false;
3229        for (Node<K,V> n = findFirst(); n != null; n = n.next) {
3230            V v;
3231            if ((v = n.getValidValue()) != null) {
3232                K k = n.key;
3233                Map.Entry<K,V> e = new AbstractMap.SimpleImmutableEntry<>(k, v);
3234                if (function.test(e) && remove(k, v))
3235                    removed = true;
3236            }
3237        }
3238        return removed;
3239    }
3240
3241    /**
3242     * Helper method for Values.removeIf.
3243     */
3244    boolean removeValueIf(Predicate<? super V> function) {
3245        if (function == null) throw new NullPointerException();
3246        boolean removed = false;
3247        for (Node<K,V> n = findFirst(); n != null; n = n.next) {
3248            V v;
3249            if ((v = n.getValidValue()) != null) {
3250                K k = n.key;
3251                if (function.test(v) && remove(k, v))
3252                    removed = true;
3253            }
3254        }
3255        return removed;
3256    }
3257
3258    /**
3259     * Base class providing common structure for Spliterators.
3260     * (Although not all that much common functionality; as usual for
3261     * view classes, details annoyingly vary in key, value, and entry
3262     * subclasses in ways that are not worth abstracting out for
3263     * internal classes.)
3264     *
3265     * The basic split strategy is to recursively descend from top
3266     * level, row by row, descending to next row when either split
3267     * off, or the end of row is encountered. Control of the number of
3268     * splits relies on some statistical estimation: The expected
3269     * remaining number of elements of a skip list when advancing
3270     * either across or down decreases by about 25%. To make this
3271     * observation useful, we need to know initial size, which we
3272     * don't. But we can just use Integer.MAX_VALUE so that we
3273     * don't prematurely zero out while splitting.
3274     */
3275    abstract static class CSLMSpliterator<K,V> {
3276        final Comparator<? super K> comparator;
3277        final K fence;     // exclusive upper bound for keys, or null if to end
3278        Index<K,V> row;    // the level to split out
3279        Node<K,V> current; // current traversal node; initialize at origin
3280        int est;           // pseudo-size estimate
3281        CSLMSpliterator(Comparator<? super K> comparator, Index<K,V> row,
3282                        Node<K,V> origin, K fence, int est) {
3283            this.comparator = comparator; this.row = row;
3284            this.current = origin; this.fence = fence; this.est = est;
3285        }
3286
3287        public final long estimateSize() { return (long)est; }
3288    }
3289
3290    static final class KeySpliterator<K,V> extends CSLMSpliterator<K,V>
3291        implements Spliterator<K> {
3292        KeySpliterator(Comparator<? super K> comparator, Index<K,V> row,
3293                       Node<K,V> origin, K fence, int est) {
3294            super(comparator, row, origin, fence, est);
3295        }
3296
3297        public KeySpliterator<K,V> trySplit() {
3298            Node<K,V> e; K ek;
3299            Comparator<? super K> cmp = comparator;
3300            K f = fence;
3301            if ((e = current) != null && (ek = e.key) != null) {
3302                for (Index<K,V> q = row; q != null; q = row = q.down) {
3303                    Index<K,V> s; Node<K,V> b, n; K sk;
3304                    if ((s = q.right) != null && (b = s.node) != null &&
3305                        (n = b.next) != null && n.value != null &&
3306                        (sk = n.key) != null && cpr(cmp, sk, ek) > 0 &&
3307                        (f == null || cpr(cmp, sk, f) < 0)) {
3308                        current = n;
3309                        Index<K,V> r = q.down;
3310                        row = (s.right != null) ? s : s.down;
3311                        est -= est >>> 2;
3312                        return new KeySpliterator<K,V>(cmp, r, e, sk, est);
3313                    }
3314                }
3315            }
3316            return null;
3317        }
3318
3319        public void forEachRemaining(Consumer<? super K> action) {
3320            if (action == null) throw new NullPointerException();
3321            Comparator<? super K> cmp = comparator;
3322            K f = fence;
3323            Node<K,V> e = current;
3324            current = null;
3325            for (; e != null; e = e.next) {
3326                K k; Object v;
3327                if ((k = e.key) != null && f != null && cpr(cmp, f, k) <= 0)
3328                    break;
3329                if ((v = e.value) != null && v != e)
3330                    action.accept(k);
3331            }
3332        }
3333
3334        public boolean tryAdvance(Consumer<? super K> action) {
3335            if (action == null) throw new NullPointerException();
3336            Comparator<? super K> cmp = comparator;
3337            K f = fence;
3338            Node<K,V> e = current;
3339            for (; e != null; e = e.next) {
3340                K k; Object v;
3341                if ((k = e.key) != null && f != null && cpr(cmp, f, k) <= 0) {
3342                    e = null;
3343                    break;
3344                }
3345                if ((v = e.value) != null && v != e) {
3346                    current = e.next;
3347                    action.accept(k);
3348                    return true;
3349                }
3350            }
3351            current = e;
3352            return false;
3353        }
3354
3355        public int characteristics() {
3356            return Spliterator.DISTINCT | Spliterator.SORTED |
3357                Spliterator.ORDERED | Spliterator.CONCURRENT |
3358                Spliterator.NONNULL;
3359        }
3360
3361        public final Comparator<? super K> getComparator() {
3362            return comparator;
3363        }
3364    }
3365    // factory method for KeySpliterator
3366    final KeySpliterator<K,V> keySpliterator() {
3367        Comparator<? super K> cmp = comparator;
3368        for (;;) { // ensure h corresponds to origin p
3369            HeadIndex<K,V> h; Node<K,V> p;
3370            Node<K,V> b = (h = head).node;
3371            if ((p = b.next) == null || p.value != null)
3372                return new KeySpliterator<K,V>(cmp, h, p, null, (p == null) ?
3373                                               0 : Integer.MAX_VALUE);
3374            p.helpDelete(b, p.next);
3375        }
3376    }
3377
3378    static final class ValueSpliterator<K,V> extends CSLMSpliterator<K,V>
3379        implements Spliterator<V> {
3380        ValueSpliterator(Comparator<? super K> comparator, Index<K,V> row,
3381                       Node<K,V> origin, K fence, int est) {
3382            super(comparator, row, origin, fence, est);
3383        }
3384
3385        public ValueSpliterator<K,V> trySplit() {
3386            Node<K,V> e; K ek;
3387            Comparator<? super K> cmp = comparator;
3388            K f = fence;
3389            if ((e = current) != null && (ek = e.key) != null) {
3390                for (Index<K,V> q = row; q != null; q = row = q.down) {
3391                    Index<K,V> s; Node<K,V> b, n; K sk;
3392                    if ((s = q.right) != null && (b = s.node) != null &&
3393                        (n = b.next) != null && n.value != null &&
3394                        (sk = n.key) != null && cpr(cmp, sk, ek) > 0 &&
3395                        (f == null || cpr(cmp, sk, f) < 0)) {
3396                        current = n;
3397                        Index<K,V> r = q.down;
3398                        row = (s.right != null) ? s : s.down;
3399                        est -= est >>> 2;
3400                        return new ValueSpliterator<K,V>(cmp, r, e, sk, est);
3401                    }
3402                }
3403            }
3404            return null;
3405        }
3406
3407        public void forEachRemaining(Consumer<? super V> action) {
3408            if (action == null) throw new NullPointerException();
3409            Comparator<? super K> cmp = comparator;
3410            K f = fence;
3411            Node<K,V> e = current;
3412            current = null;
3413            for (; e != null; e = e.next) {
3414                K k; Object v;
3415                if ((k = e.key) != null && f != null && cpr(cmp, f, k) <= 0)
3416                    break;
3417                if ((v = e.value) != null && v != e) {
3418                    @SuppressWarnings("unchecked") V vv = (V)v;
3419                    action.accept(vv);
3420                }
3421            }
3422        }
3423
3424        public boolean tryAdvance(Consumer<? super V> action) {
3425            if (action == null) throw new NullPointerException();
3426            Comparator<? super K> cmp = comparator;
3427            K f = fence;
3428            Node<K,V> e = current;
3429            for (; e != null; e = e.next) {
3430                K k; Object v;
3431                if ((k = e.key) != null && f != null && cpr(cmp, f, k) <= 0) {
3432                    e = null;
3433                    break;
3434                }
3435                if ((v = e.value) != null && v != e) {
3436                    current = e.next;
3437                    @SuppressWarnings("unchecked") V vv = (V)v;
3438                    action.accept(vv);
3439                    return true;
3440                }
3441            }
3442            current = e;
3443            return false;
3444        }
3445
3446        public int characteristics() {
3447            return Spliterator.CONCURRENT | Spliterator.ORDERED |
3448                Spliterator.NONNULL;
3449        }
3450    }
3451
3452    // Almost the same as keySpliterator()
3453    final ValueSpliterator<K,V> valueSpliterator() {
3454        Comparator<? super K> cmp = comparator;
3455        for (;;) {
3456            HeadIndex<K,V> h; Node<K,V> p;
3457            Node<K,V> b = (h = head).node;
3458            if ((p = b.next) == null || p.value != null)
3459                return new ValueSpliterator<K,V>(cmp, h, p, null, (p == null) ?
3460                                                 0 : Integer.MAX_VALUE);
3461            p.helpDelete(b, p.next);
3462        }
3463    }
3464
3465    static final class EntrySpliterator<K,V> extends CSLMSpliterator<K,V>
3466        implements Spliterator<Map.Entry<K,V>> {
3467        EntrySpliterator(Comparator<? super K> comparator, Index<K,V> row,
3468                         Node<K,V> origin, K fence, int est) {
3469            super(comparator, row, origin, fence, est);
3470        }
3471
3472        public EntrySpliterator<K,V> trySplit() {
3473            Node<K,V> e; K ek;
3474            Comparator<? super K> cmp = comparator;
3475            K f = fence;
3476            if ((e = current) != null && (ek = e.key) != null) {
3477                for (Index<K,V> q = row; q != null; q = row = q.down) {
3478                    Index<K,V> s; Node<K,V> b, n; K sk;
3479                    if ((s = q.right) != null && (b = s.node) != null &&
3480                        (n = b.next) != null && n.value != null &&
3481                        (sk = n.key) != null && cpr(cmp, sk, ek) > 0 &&
3482                        (f == null || cpr(cmp, sk, f) < 0)) {
3483                        current = n;
3484                        Index<K,V> r = q.down;
3485                        row = (s.right != null) ? s : s.down;
3486                        est -= est >>> 2;
3487                        return new EntrySpliterator<K,V>(cmp, r, e, sk, est);
3488                    }
3489                }
3490            }
3491            return null;
3492        }
3493
3494        public void forEachRemaining(Consumer<? super Map.Entry<K,V>> action) {
3495            if (action == null) throw new NullPointerException();
3496            Comparator<? super K> cmp = comparator;
3497            K f = fence;
3498            Node<K,V> e = current;
3499            current = null;
3500            for (; e != null; e = e.next) {
3501                K k; Object v;
3502                if ((k = e.key) != null && f != null && cpr(cmp, f, k) <= 0)
3503                    break;
3504                if ((v = e.value) != null && v != e) {
3505                    @SuppressWarnings("unchecked") V vv = (V)v;
3506                    action.accept
3507                        (new AbstractMap.SimpleImmutableEntry<K,V>(k, vv));
3508                }
3509            }
3510        }
3511
3512        public boolean tryAdvance(Consumer<? super Map.Entry<K,V>> action) {
3513            if (action == null) throw new NullPointerException();
3514            Comparator<? super K> cmp = comparator;
3515            K f = fence;
3516            Node<K,V> e = current;
3517            for (; e != null; e = e.next) {
3518                K k; Object v;
3519                if ((k = e.key) != null && f != null && cpr(cmp, f, k) <= 0) {
3520                    e = null;
3521                    break;
3522                }
3523                if ((v = e.value) != null && v != e) {
3524                    current = e.next;
3525                    @SuppressWarnings("unchecked") V vv = (V)v;
3526                    action.accept
3527                        (new AbstractMap.SimpleImmutableEntry<K,V>(k, vv));
3528                    return true;
3529                }
3530            }
3531            current = e;
3532            return false;
3533        }
3534
3535        public int characteristics() {
3536            return Spliterator.DISTINCT | Spliterator.SORTED |
3537                Spliterator.ORDERED | Spliterator.CONCURRENT |
3538                Spliterator.NONNULL;
3539        }
3540
3541        public final Comparator<Map.Entry<K,V>> getComparator() {
3542            // Adapt or create a key-based comparator
3543            if (comparator != null) {
3544                return Map.Entry.comparingByKey(comparator);
3545            }
3546            else {
3547                return (Comparator<Map.Entry<K,V>> & Serializable) (e1, e2) -> {
3548                    @SuppressWarnings("unchecked")
3549                    Comparable<? super K> k1 = (Comparable<? super K>) e1.getKey();
3550                    return k1.compareTo(e2.getKey());
3551                };
3552            }
3553        }
3554    }
3555
3556    // Almost the same as keySpliterator()
3557    final EntrySpliterator<K,V> entrySpliterator() {
3558        Comparator<? super K> cmp = comparator;
3559        for (;;) { // almost same as key version
3560            HeadIndex<K,V> h; Node<K,V> p;
3561            Node<K,V> b = (h = head).node;
3562            if ((p = b.next) == null || p.value != null)
3563                return new EntrySpliterator<K,V>(cmp, h, p, null, (p == null) ?
3564                                                 0 : Integer.MAX_VALUE);
3565            p.helpDelete(b, p.next);
3566        }
3567    }
3568
3569    // Unsafe mechanics
3570    private static final sun.misc.Unsafe U = sun.misc.Unsafe.getUnsafe();
3571    private static final long HEAD;
3572    static {
3573        try {
3574            HEAD = U.objectFieldOffset
3575                (ConcurrentSkipListMap.class.getDeclaredField("head"));
3576        } catch (ReflectiveOperationException e) {
3577            throw new Error(e);
3578        }
3579    }
3580}
3581