ViewVC Help
View File | Revision Log | Show Annotations | Download File | Root Listing
root/jsr166/jsr166/src/main/java/util/concurrent/ConcurrentSkipListMap.java
Revision: 1.1
Committed: Tue Dec 28 12:14:13 2004 UTC (19 years, 5 months ago) by dl
Branch: MAIN
Log Message:
Prepare jsr166x classes for Mustang integration

File Contents

# User Rev Content
1 dl 1.1 /*
2     * Written by Doug Lea with assistance from members of JCP JSR-166
3     * Expert Group and released to the public domain, as explained at
4     * http://creativecommons.org/licenses/publicdomain
5     */
6    
7     package java.util.concurrent;
8     import java.util.*;
9     import java.util.concurrent.atomic.*;
10    
11     /**
12     * A scalable {@link ConcurrentNavigableMap} implementation. This
13     * class maintains a map in ascending key order, sorted according to
14     * the <i>natural order</i> for the key's class (see {@link
15     * Comparable}), or by the {@link Comparator} provided at creation
16     * time, depending on which constructor is used.
17     *
18     * <p>This class implements a concurrent variant of <a
19     * href="http://www.cs.umd.edu/~pugh/">SkipLists</a> providing
20     * expected average <i>log(n)</i> time cost for the
21     * <tt>containsKey</tt>, <tt>get</tt>, <tt>put</tt> and
22     * <tt>remove</tt> operations and their variants. Insertion, removal,
23     * update, and access operations safely execute concurrently by
24     * multiple threads. Iterators are <i>weakly consistent</i>, returning
25     * elements reflecting the state of the map at some point at or since
26     * the creation of the iterator. They do <em>not</em> throw {@link
27     * ConcurrentModificationException}, and may proceed concurrently with
28     * other operations. Ascending key ordered views and their iterators
29     * are faster than descending ones.
30     *
31     * <p>All <tt>Map.Entry</tt> pairs returned by methods in this class
32     * and its views represent snapshots of mappings at the time they were
33     * produced. They do <em>not</em> support the <tt>Entry.setValue</tt>
34     * method. (Note however that it is possible to change mappings in the
35     * associated map using <tt>put</tt>, <tt>putIfAbsent</tt>, or
36     * <tt>replace</tt>, depending on exactly which effect you need.)
37     *
38     * <p>Beware that, unlike in most collections, the <tt>size</tt>
39     * method is <em>not</em> a constant-time operation. Because of the
40     * asynchronous nature of these maps, determining the current number
41     * of elements requires a traversal of the elements. Additionally,
42     * the bulk operations <tt>putAll</tt>, <tt>equals</tt>, and
43     * <tt>clear</tt> are <em>not</em> guaranteed to be performed
44     * atomically. For example, an iterator operating concurrently with a
45     * <tt>putAll</tt> operation might view only some of the added
46     * elements.
47     *
48     * <p>This class and its views and iterators implement all of the
49     * <em>optional</em> methods of the {@link Map} and {@link Iterator}
50     * interfaces. Like most other concurrent collections, this class does
51     * not permit the use of <tt>null</tt> keys or values because some
52     * null return values cannot be reliably distinguished from the
53     * absence of elements.
54     *
55     * @author Doug Lea
56     * @param <K> the type of keys maintained by this map
57     * @param <V> the type of mapped values
58     */
59     public class ConcurrentSkipListMap<K,V> extends AbstractMap<K,V>
60     implements ConcurrentNavigableMap<K,V>,
61     Cloneable,
62     java.io.Serializable {
63     /*
64     * This class implements a tree-like two-dimensionally linked skip
65     * list in which the index levels are represented in separate
66     * nodes from the base nodes holding data. There are two reasons
67     * for taking this approach instead of the usual array-based
68     * structure: 1) Array based implementations seem to encounter
69     * more complexity and overhead 2) We can use cheaper algorithms
70     * for the heavily-traversed index lists than can be used for the
71     * base lists. Here's a picture of some of the basics for a
72     * possible list with 2 levels of index:
73     *
74     * Head nodes Index nodes
75     * +-+ right +-+ +-+
76     * |2|---------------->| |--------------------->| |->null
77     * +-+ +-+ +-+
78     * | down | |
79     * v v v
80     * +-+ +-+ +-+ +-+ +-+ +-+
81     * |1|----------->| |->| |------>| |----------->| |------>| |->null
82     * +-+ +-+ +-+ +-+ +-+ +-+
83     * v | | | | |
84     * Nodes next v v v v v
85     * +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+
86     * | |->|A|->|B|->|C|->|D|->|E|->|F|->|G|->|H|->|I|->|J|->|K|->null
87     * +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+
88     *
89     * The base lists use a variant of the HM linked ordered set
90     * algorithm. See Tim Harris, "A pragmatic implementation of
91     * non-blocking linked lists"
92     * http://www.cl.cam.ac.uk/~tlh20/publications.html and Maged
93     * Michael "High Performance Dynamic Lock-Free Hash Tables and
94     * List-Based Sets"
95     * http://www.research.ibm.com/people/m/michael/pubs.htm. The
96     * basic idea in these lists is to mark the "next" pointers of
97     * deleted nodes when deleting to avoid conflicts with concurrent
98     * insertions, and when traversing to keep track of triples
99     * (predecessor, node, successor) in order to detect when and how
100     * to unlink these deleted nodes.
101     *
102     * Rather than using mark-bits to mark list deletions (which can
103     * be slow and space-intensive using AtomicMarkedReference), nodes
104     * use direct CAS'able next pointers. On deletion, instead of
105     * marking a pointer, they splice in another node that can be
106     * thought of as standing for a marked pointer (indicating this by
107     * using otherwise impossible field values). Using plain nodes
108     * acts roughly like "boxed" implementations of marked pointers,
109     * but uses new nodes only when nodes are deleted, not for every
110     * link. This requires less space and supports faster
111     * traversal. Even if marked references were better supported by
112     * JVMs, traversal using this technique might still be faster
113     * because any search need only read ahead one more node than
114     * otherwise required (to check for trailing marker) rather than
115     * unmasking mark bits or whatever on each read.
116     *
117     * This approach maintains the essential property needed in the HM
118     * algorithm of changing the next-pointer of a deleted node so
119     * that any other CAS of it will fail, but implements the idea by
120     * changing the pointer to point to a different node, not by
121     * marking it. While it would be possible to further squeeze
122     * space by defining marker nodes not to have key/value fields, it
123     * isn't worth the extra type-testing overhead. The deletion
124     * markers are rarely encountered during traversal and are
125     * normally quickly garbage collected. (Note that this technique
126     * would not work well in systems without garbage collection.)
127     *
128     * In addition to using deletion markers, the lists also use
129     * nullness of value fields to indicate deletion, in a style
130     * similar to typical lazy-deletion schemes. If a node's value is
131     * null, then it is considered logically deleted and ignored even
132     * though it is still reachable. This maintains proper control of
133     * concurrent replace vs delete operations -- an attempted replace
134     * must fail if a delete beat it by nulling field, and a delete
135     * must return the last non-null value held in the field. (Note:
136     * Null, rather than some special marker, is used for value fields
137     * here because it just so happens to mesh with the Map API
138     * requirement that method get returns null if there is no
139     * mapping, which allows nodes to remain concurrently readable
140     * even when deleted. Using any other marker value here would be
141     * messy at best.)
142     *
143     * Here's the sequence of events for a deletion of node n with
144     * predecessor b and successor f, initially:
145     *
146     * +------+ +------+ +------+
147     * ... | b |------>| n |----->| f | ...
148     * +------+ +------+ +------+
149     *
150     * 1. CAS n's value field from non-null to null.
151     * From this point on, no public operations encountering
152     * the node consider this mapping to exist. However, other
153     * ongoing insertions and deletions might still modify
154     * n's next pointer.
155     *
156     * 2. CAS n's next pointer to point to a new marker node.
157     * From this point on, no other nodes can be appended to n.
158     * which avoids deletion errors in CAS-based linked lists.
159     *
160     * +------+ +------+ +------+ +------+
161     * ... | b |------>| n |----->|marker|------>| f | ...
162     * +------+ +------+ +------+ +------+
163     *
164     * 3. CAS b's next pointer over both n and its marker.
165     * From this point on, no new traversals will encounter n,
166     * and it can eventually be GCed.
167     * +------+ +------+
168     * ... | b |----------------------------------->| f | ...
169     * +------+ +------+
170     *
171     * A failure at step 1 leads to simple retry due to a lost race
172     * with another operation. Steps 2-3 can fail because some other
173     * thread noticed during a traversal a node with null value and
174     * helped out by marking and/or unlinking. This helping-out
175     * ensures that no thread can become stuck waiting for progress of
176     * the deleting thread. The use of marker nodes slightly
177     * complicates helping-out code because traversals must track
178     * consistent reads of up to four nodes (b, n, marker, f), not
179     * just (b, n, f), although the next field of a marker is
180     * immutable, and once a next field is CAS'ed to point to a
181     * marker, it never again changes, so this requires less care.
182     *
183     * Skip lists add indexing to this scheme, so that the base-level
184     * traversals start close to the locations being found, inserted
185     * or deleted -- usually base level traversals only traverse a few
186     * nodes. This doesn't change the basic algorithm except for the
187     * need to make sure base traversals start at predecessors (here,
188     * b) that are not (structurally) deleted, otherwise retrying
189     * after processing the deletion.
190     *
191     * Index levels are maintained as lists with volatile next fields,
192     * using CAS to link and unlink. Races are allowed in index-list
193     * operations that can (rarely) fail to link in a new index node
194     * or delete one. (We can't do this of course for data nodes.)
195     * However, even when this happens, the index lists remain sorted,
196     * so correctly serve as indices. This can impact performance,
197     * but since skip lists are probabilistic anyway, the net result
198     * is that under contention, the effective "p" value may be lower
199     * than its nominal value. And race windows are kept small enough
200     * that in practice these failures are rare, even under a lot of
201     * contention.
202     *
203     * The fact that retries (for both base and index lists) are
204     * relatively cheap due to indexing allows some minor
205     * simplifications of retry logic. Traversal restarts are
206     * performed after most "helping-out" CASes. This isn't always
207     * strictly necessary, but the implicit backoffs tend to help
208     * reduce other downstream failed CAS's enough to outweigh restart
209     * cost. This worsens the worst case, but seems to improve even
210     * highly contended cases.
211     *
212     * Unlike most skip-list implementations, index insertion and
213     * deletion here require a separate traversal pass occuring after
214     * the base-level action, to add or remove index nodes. This adds
215     * to single-threaded overhead, but improves contended
216     * multithreaded performance by narrowing interference windows,
217     * and allows deletion to ensure that all index nodes will be made
218     * unreachable upon return from a public remove operation, thus
219     * avoiding unwanted garbage retention. This is more important
220     * here than in some other data structures because we cannot null
221     * out node fields referencing user keys since they might still be
222     * read by other ongoing traversals.
223     *
224     * Indexing uses skip list parameters that maintain good search
225     * performance while using sparser-than-usual indices: The
226     * hardwired parameters k=1, p=0.5 (see method randomLevel) mean
227     * that about one-quarter of the nodes have indices. Of those that
228     * do, half have one level, a quarter have two, and so on (see
229     * Pugh's Skip List Cookbook, sec 3.4). The expected total space
230     * requirement for a map is slightly less than for the current
231     * implementation of java.util.TreeMap.
232     *
233     * Changing the level of the index (i.e, the height of the
234     * tree-like structure) also uses CAS. The head index has initial
235     * level/height of one. Creation of an index with height greater
236     * than the current level adds a level to the head index by
237     * CAS'ing on a new top-most head. To maintain good performance
238     * after a lot of removals, deletion methods heuristically try to
239     * reduce the height if the topmost levels appear to be empty.
240     * This may encounter races in which it possible (but rare) to
241     * reduce and "lose" a level just as it is about to contain an
242     * index (that will then never be encountered). This does no
243     * structural harm, and in practice appears to be a better option
244     * than allowing unrestrained growth of levels.
245     *
246     * The code for all this is more verbose than you'd like. Most
247     * operations entail locating an element (or position to insert an
248     * element). The code to do this can't be nicely factored out
249     * because subsequent uses require a snapshot of predecessor
250     * and/or successor and/or value fields which can't be returned
251     * all at once, at least not without creating yet another object
252     * to hold them -- creating such little objects is an especially
253     * bad idea for basic internal search operations because it adds
254     * to GC overhead. (This is one of the few times I've wished Java
255     * had macros.) Instead, some traversal code is interleaved within
256     * insertion and removal operations. The control logic to handle
257     * all the retry conditions is sometimes twisty. Most search is
258     * broken into 2 parts. findPredecessor() searches index nodes
259     * only, returning a base-level predecessor of the key. findNode()
260     * finishes out the base-level search. Even with this factoring,
261     * there is a fair amount of near-duplication of code to handle
262     * variants.
263     *
264     * For explanation of algorithms sharing at least a couple of
265     * features with this one, see Mikhail Fomitchev's thesis
266     * (http://www.cs.yorku.ca/~mikhail/), Keir Fraser's thesis
267     * (http://www.cl.cam.ac.uk/users/kaf24/), and Håkan Sundell's
268     * thesis (http://www.cs.chalmers.se/~phs/).
269     *
270     * Given the use of tree-like index nodes, you might wonder why
271     * this doesn't use some kind of search tree instead, which would
272     * support somewhat faster search operations. The reason is that
273     * there are no known efficient lock-free insertion and deletion
274     * algorithms for search trees. The immutability of the "down"
275     * links of index nodes (as opposed to mutable "left" fields in
276     * true trees) makes this tractable using only CAS operations.
277     *
278     * Notation guide for local variables
279     * Node: b, n, f for predecessor, node, successor
280     * Index: q, r, d for index node, right, down.
281     * t for another index node
282     * Head: h
283     * Levels: j
284     * Keys: k, key
285     * Values: v, value
286     * Comparisons: c
287     */
288    
289     private static final long serialVersionUID = -8627078645895051609L;
290    
291     /**
292     * Special value used to identify base-level header
293     */
294     private static final Object BASE_HEADER = new Object();
295    
296     /**
297     * The topmost head index of the skiplist.
298     */
299     private transient volatile HeadIndex<K,V> head;
300    
301     /**
302     * The Comparator used to maintain order in this Map, or null
303     * if using natural order.
304     * @serial
305     */
306     private final Comparator<? super K> comparator;
307    
308     /**
309     * Seed for simple random number generator. Not volatile since it
310     * doesn't matter too much if different threads don't see updates.
311     */
312     private transient int randomSeed;
313    
314     /** Lazily initialized key set */
315     private transient KeySet keySet;
316     /** Lazily initialized entry set */
317     private transient EntrySet entrySet;
318     /** Lazily initialized values collection */
319     private transient Values values;
320     /** Lazily initialized descending key set */
321     private transient DescendingKeySet descendingKeySet;
322     /** Lazily initialized descending entry set */
323     private transient DescendingEntrySet descendingEntrySet;
324    
325     /**
326     * Initialize or reset state. Needed by constructors, clone,
327     * clear, readObject. and ConcurrentSkipListSet.clone.
328     * (Note that comparator must be separately initialized.)
329     */
330     final void initialize() {
331     keySet = null;
332     entrySet = null;
333     values = null;
334     descendingEntrySet = null;
335     descendingKeySet = null;
336     randomSeed = (int) System.nanoTime();
337     head = new HeadIndex<K,V>(new Node<K,V>(null, BASE_HEADER, null),
338     null, null, 1);
339     }
340    
341     /** Updater for casHead */
342     private static final
343     AtomicReferenceFieldUpdater<ConcurrentSkipListMap, HeadIndex>
344     headUpdater = AtomicReferenceFieldUpdater.newUpdater
345     (ConcurrentSkipListMap.class, HeadIndex.class, "head");
346    
347     /**
348     * compareAndSet head node
349     */
350     private boolean casHead(HeadIndex<K,V> cmp, HeadIndex<K,V> val) {
351     return headUpdater.compareAndSet(this, cmp, val);
352     }
353    
354     /* ---------------- Nodes -------------- */
355    
356     /**
357     * Nodes hold keys and values, and are singly linked in sorted
358     * order, possibly with some intervening marker nodes. The list is
359     * headed by a dummy node accessible as head.node. The value field
360     * is declared only as Object because it takes special non-V
361     * values for marker and header nodes.
362     */
363     static final class Node<K,V> {
364     final K key;
365     volatile Object value;
366     volatile Node<K,V> next;
367    
368     /**
369     * Creates a new regular node.
370     */
371     Node(K key, Object value, Node<K,V> next) {
372     this.key = key;
373     this.value = value;
374     this.next = next;
375     }
376    
377     /**
378     * Creates a new marker node. A marker is distinguished by
379     * having its value field point to itself. Marker nodes also
380     * have null keys, a fact that is exploited in a few places,
381     * but this doesn't distinguish markers from the base-level
382     * header node (head.node), which also has a null key.
383     */
384     Node(Node<K,V> next) {
385     this.key = null;
386     this.value = this;
387     this.next = next;
388     }
389    
390     /** Updater for casNext */
391     static final AtomicReferenceFieldUpdater<Node, Node>
392     nextUpdater = AtomicReferenceFieldUpdater.newUpdater
393     (Node.class, Node.class, "next");
394    
395     /** Updater for casValue */
396     static final AtomicReferenceFieldUpdater<Node, Object>
397     valueUpdater = AtomicReferenceFieldUpdater.newUpdater
398     (Node.class, Object.class, "value");
399    
400     /**
401     * compareAndSet value field
402     */
403     boolean casValue(Object cmp, Object val) {
404     return valueUpdater.compareAndSet(this, cmp, val);
405     }
406    
407     /**
408     * compareAndSet next field
409     */
410     boolean casNext(Node<K,V> cmp, Node<K,V> val) {
411     return nextUpdater.compareAndSet(this, cmp, val);
412     }
413    
414     /**
415     * Return true if this node is a marker. This method isn't
416     * actually called in an any current code checking for markers
417     * because callers will have already read value field and need
418     * to use that read (not another done here) and so directly
419     * test if value points to node.
420     * @param n a possibly null reference to a node
421     * @return true if this node is a marker node
422     */
423     boolean isMarker() {
424     return value == this;
425     }
426    
427     /**
428     * Return true if this node is the header of base-level list.
429     * @return true if this node is header node
430     */
431     boolean isBaseHeader() {
432     return value == BASE_HEADER;
433     }
434    
435     /**
436     * Tries to append a deletion marker to this node.
437     * @param f the assumed current successor of this node
438     * @return true if successful
439     */
440     boolean appendMarker(Node<K,V> f) {
441     return casNext(f, new Node<K,V>(f));
442     }
443    
444     /**
445     * Helps out a deletion by appending marker or unlinking from
446     * predecessor. This is called during traversals when value
447     * field seen to be null.
448     * @param b predecessor
449     * @param f successor
450     */
451     void helpDelete(Node<K,V> b, Node<K,V> f) {
452     /*
453     * Rechecking links and then doing only one of the
454     * help-out stages per call tends to minimize CAS
455     * interference among helping threads.
456     */
457     if (f == next && this == b.next) {
458     if (f == null || f.value != f) // not already marked
459     appendMarker(f);
460     else
461     b.casNext(this, f.next);
462     }
463     }
464    
465     /**
466     * Return value if this node contains a valid key-value pair,
467     * else null.
468     * @return this node's value if it isn't a marker or header or
469     * is deleted, else null.
470     */
471     V getValidValue() {
472     Object v = value;
473     if (v == this || v == BASE_HEADER)
474     return null;
475     return (V)v;
476     }
477    
478     /**
479     * Create and return a new SnapshotEntry holding current
480     * mapping if this node holds a valid value, else null
481     * @return new entry or null
482     */
483     SnapshotEntry<K,V> createSnapshot() {
484     V v = getValidValue();
485     if (v == null)
486     return null;
487     return new SnapshotEntry(key, v);
488     }
489     }
490    
491     /* ---------------- Indexing -------------- */
492    
493     /**
494     * Index nodes represent the levels of the skip list. To improve
495     * search performance, keys of the underlying nodes are cached.
496     * Note that even though both Nodes and Indexes have
497     * forward-pointing fields, they have different types and are
498     * handled in different ways, that can't nicely be captured by
499     * placing field in a shared abstract class.
500     */
501     static class Index<K,V> {
502     final K key;
503     final Node<K,V> node;
504     final Index<K,V> down;
505     volatile Index<K,V> right;
506    
507     /**
508     * Creates index node with given values
509     */
510     Index(Node<K,V> node, Index<K,V> down, Index<K,V> right) {
511     this.node = node;
512     this.key = node.key;
513     this.down = down;
514     this.right = right;
515     }
516    
517     /** Updater for casRight */
518     static final AtomicReferenceFieldUpdater<Index, Index>
519     rightUpdater = AtomicReferenceFieldUpdater.newUpdater
520     (Index.class, Index.class, "right");
521    
522     /**
523     * compareAndSet right field
524     */
525     final boolean casRight(Index<K,V> cmp, Index<K,V> val) {
526     return rightUpdater.compareAndSet(this, cmp, val);
527     }
528    
529     /**
530     * Returns true if the node this indexes has been deleted.
531     * @return true if indexed node is known to be deleted
532     */
533     final boolean indexesDeletedNode() {
534     return node.value == null;
535     }
536    
537     /**
538     * Tries to CAS newSucc as successor. To minimize races with
539     * unlink that may lose this index node, if the node being
540     * indexed is known to be deleted, it doesn't try to link in.
541     * @param succ the expected current successor
542     * @param newSucc the new successor
543     * @return true if successful
544     */
545     final boolean link(Index<K,V> succ, Index<K,V> newSucc) {
546     Node<K,V> n = node;
547     newSucc.right = succ;
548     return n.value != null && casRight(succ, newSucc);
549     }
550    
551     /**
552     * Tries to CAS right field to skip over apparent successor
553     * succ. Fails (forcing a retraversal by caller) if this node
554     * is known to be deleted.
555     * @param succ the expected current successor
556     * @return true if successful
557     */
558     final boolean unlink(Index<K,V> succ) {
559     return !indexesDeletedNode() && casRight(succ, succ.right);
560     }
561     }
562    
563     /* ---------------- Head nodes -------------- */
564    
565     /**
566     * Nodes heading each level keep track of their level.
567     */
568     static final class HeadIndex<K,V> extends Index<K,V> {
569     final int level;
570     HeadIndex(Node<K,V> node, Index<K,V> down, Index<K,V> right, int level) {
571     super(node, down, right);
572     this.level = level;
573     }
574     }
575    
576     /* ---------------- Map.Entry support -------------- */
577    
578     /**
579     * An immutable representation of a key-value mapping as it
580     * existed at some point in time. This class does <em>not</em>
581     * support the <tt>Map.Entry.setValue</tt> method.
582     */
583     static class SnapshotEntry<K,V> implements Map.Entry<K,V> {
584     private final K key;
585     private final V value;
586    
587     /**
588     * Creates a new entry representing the given key and value.
589     * @param key the key
590     * @param value the value
591     */
592     SnapshotEntry(K key, V value) {
593     this.key = key;
594     this.value = value;
595     }
596    
597     /**
598     * Returns the key corresponding to this entry.
599     *
600     * @return the key corresponding to this entry.
601     */
602     public K getKey() {
603     return key;
604     }
605    
606     /**
607     * Returns the value corresponding to this entry.
608     *
609     * @return the value corresponding to this entry.
610     */
611     public V getValue() {
612     return value;
613     }
614    
615     /**
616     * Always fails, throwing <tt>UnsupportedOperationException</tt>.
617     * @throws UnsupportedOperationException always.
618     */
619     public V setValue(V value) {
620     throw new UnsupportedOperationException();
621     }
622    
623     // inherit javadoc
624     public boolean equals(Object o) {
625     if (!(o instanceof Map.Entry))
626     return false;
627     Map.Entry e = (Map.Entry)o;
628     // As mandated by Map.Entry spec:
629     return ((key==null ?
630     e.getKey()==null : key.equals(e.getKey())) &&
631     (value==null ?
632     e.getValue()==null : value.equals(e.getValue())));
633     }
634    
635    
636     // inherit javadoc
637     public int hashCode() {
638     // As mandated by Map.Entry spec:
639     return ((key==null ? 0 : key.hashCode()) ^
640     (value==null ? 0 : value.hashCode()));
641     }
642    
643     /**
644     * Returns a String consisting of the key followed by an
645     * equals sign (<tt>"="</tt>) followed by the associated
646     * value.
647     * @return a String representation of this entry.
648     */
649     public String toString() {
650     return getKey() + "=" + getValue();
651     }
652     }
653    
654     /* ---------------- Comparison utilities -------------- */
655    
656     /**
657     * Represents a key with a comparator as a Comparable.
658     *
659     * Because most sorted collections seem to use natural order on
660     * Comparables (Strings, Integers, etc), most internal methods are
661     * geared to use them. This is generally faster than checking
662     * per-comparison whether to use comparator or comparable because
663     * it doesn't require a (Comparable) cast for each comparison.
664     * (Optimizers can only sometimes remove such redundant checks
665     * themselves.) When Comparators are used,
666     * ComparableUsingComparators are created so that they act in the
667     * same way as natural orderings. This penalizes use of
668     * Comparators vs Comparables, which seems like the right
669     * tradeoff.
670     */
671     static final class ComparableUsingComparator<K> implements Comparable<K> {
672     final K actualKey;
673     final Comparator<? super K> cmp;
674     ComparableUsingComparator(K key, Comparator<? super K> cmp) {
675     this.actualKey = key;
676     this.cmp = cmp;
677     }
678     public int compareTo(K k2) {
679     return cmp.compare(actualKey, k2);
680     }
681     }
682    
683     /**
684     * If using comparator, return a ComparableUsingComparator, else
685     * cast key as Comparator, which may cause ClassCastException,
686     * which is propagated back to caller.
687     */
688     private Comparable<K> comparable(Object key) throws ClassCastException {
689     if (key == null)
690     throw new NullPointerException();
691     return (comparator != null)
692     ? new ComparableUsingComparator(key, comparator)
693     : (Comparable<K>)key;
694     }
695    
696     /**
697     * Compare using comparator or natural ordering. Used when the
698     * ComparableUsingComparator approach doesn't apply.
699     */
700     int compare(K k1, K k2) throws ClassCastException {
701     Comparator<? super K> cmp = comparator;
702     if (cmp != null)
703     return cmp.compare(k1, k2);
704     else
705     return ((Comparable<K>)k1).compareTo(k2);
706     }
707    
708     /**
709     * Return true if given key greater than or equal to least and
710     * strictly less than fence, bypassing either test if least or
711     * fence oare null. Needed mainly in submap operations.
712     */
713     boolean inHalfOpenRange(K key, K least, K fence) {
714     if (key == null)
715     throw new NullPointerException();
716     return ((least == null || compare(key, least) >= 0) &&
717     (fence == null || compare(key, fence) < 0));
718     }
719    
720     /**
721     * Return true if given key greater than or equal to least and less
722     * or equal to fence. Needed mainly in submap operations.
723     */
724     boolean inOpenRange(K key, K least, K fence) {
725     if (key == null)
726     throw new NullPointerException();
727     return ((least == null || compare(key, least) >= 0) &&
728     (fence == null || compare(key, fence) <= 0));
729     }
730    
731     /* ---------------- Traversal -------------- */
732    
733     /**
734     * Return a base-level node with key strictly less than given key,
735     * or the base-level header if there is no such node. Also
736     * unlinks indexes to deleted nodes found along the way. Callers
737     * rely on this side-effect of clearing indices to deleted nodes.
738     * @param key the key
739     * @return a predecessor of key
740     */
741     private Node<K,V> findPredecessor(Comparable<K> key) {
742     for (;;) {
743     Index<K,V> q = head;
744     for (;;) {
745     Index<K,V> d, r;
746     if ((r = q.right) != null) {
747     if (r.indexesDeletedNode()) {
748     if (q.unlink(r))
749     continue; // reread r
750     else
751     break; // restart
752     }
753     if (key.compareTo(r.key) > 0) {
754     q = r;
755     continue;
756     }
757     }
758     if ((d = q.down) != null)
759     q = d;
760     else
761     return q.node;
762     }
763     }
764     }
765    
766     /**
767     * Return node holding key or null if no such, clearing out any
768     * deleted nodes seen along the way. Repeatedly traverses at
769     * base-level looking for key starting at predecessor returned
770     * from findPredecessor, processing base-level deletions as
771     * encountered. Some callers rely on this side-effect of clearing
772     * deleted nodes.
773     *
774     * Restarts occur, at traversal step centered on node n, if:
775     *
776     * (1) After reading n's next field, n is no longer assumed
777     * predecessor b's current successor, which means that
778     * we don't have a consistent 3-node snapshot and so cannot
779     * unlink any subsequent deleted nodes encountered.
780     *
781     * (2) n's value field is null, indicating n is deleted, in
782     * which case we help out an ongoing structural deletion
783     * before retrying. Even though there are cases where such
784     * unlinking doesn't require restart, they aren't sorted out
785     * here because doing so would not usually outweigh cost of
786     * restarting.
787     *
788     * (3) n is a marker or n's predecessor's value field is null,
789     * indicating (among other possibilities) that
790     * findPredecessor returned a deleted node. We can't unlink
791     * the node because we don't know its predecessor, so rely
792     * on another call to findPredecessor to notice and return
793     * some earlier predecessor, which it will do. This check is
794     * only strictly needed at beginning of loop, (and the
795     * b.value check isn't strictly needed at all) but is done
796     * each iteration to help avoid contention with other
797     * threads by callers that will fail to be able to change
798     * links, and so will retry anyway.
799     *
800     * The traversal loops in doPut, doRemove, and findNear all
801     * include the same three kinds of checks. And specialized
802     * versions appear in doRemoveFirst, doRemoveLast, findFirst, and
803     * findLast. They can't easily share code because each uses the
804     * reads of fields held in locals occurring in the orders they
805     * were performed.
806     *
807     * @param key the key
808     * @return node holding key, or null if no such.
809     */
810     private Node<K,V> findNode(Comparable<K> key) {
811     for (;;) {
812     Node<K,V> b = findPredecessor(key);
813     Node<K,V> n = b.next;
814     for (;;) {
815     if (n == null)
816     return null;
817     Node<K,V> f = n.next;
818     if (n != b.next) // inconsistent read
819     break;
820     Object v = n.value;
821     if (v == null) { // n is deleted
822     n.helpDelete(b, f);
823     break;
824     }
825     if (v == n || b.value == null) // b is deleted
826     break;
827     int c = key.compareTo(n.key);
828     if (c < 0)
829     return null;
830     if (c == 0)
831     return n;
832     b = n;
833     n = f;
834     }
835     }
836     }
837    
838     /**
839     * Specialized variant of findNode to perform Map.get. Does a weak
840     * traversal, not bothering to fix any deleted index nodes,
841     * returning early if it happens to see key in index, and passing
842     * over any deleted base nodes, falling back to getUsingFindNode
843     * only if it would otherwise return value from an ongoing
844     * deletion. Also uses "bound" to eliminate need for some
845     * comparisons (see Pugh Cookbook). Also folds uses of null checks
846     * and node-skipping because markers have null keys.
847     * @param okey the key
848     * @return the value, or null if absent
849     */
850     private V doGet(Object okey) {
851     Comparable<K> key = comparable(okey);
852     K bound = null;
853     Index<K,V> q = head;
854     for (;;) {
855     K rk;
856     Index<K,V> d, r;
857     if ((r = q.right) != null &&
858     (rk = r.key) != null && rk != bound) {
859     int c = key.compareTo(rk);
860     if (c > 0) {
861     q = r;
862     continue;
863     }
864     if (c == 0) {
865     Object v = r.node.value;
866     return (v != null)? (V)v : getUsingFindNode(key);
867     }
868     bound = rk;
869     }
870     if ((d = q.down) != null)
871     q = d;
872     else {
873     for (Node<K,V> n = q.node.next; n != null; n = n.next) {
874     K nk = n.key;
875     if (nk != null) {
876     int c = key.compareTo(nk);
877     if (c == 0) {
878     Object v = n.value;
879     return (v != null)? (V)v : getUsingFindNode(key);
880     }
881     if (c < 0)
882     return null;
883     }
884     }
885     return null;
886     }
887     }
888     }
889    
890     /**
891     * Perform map.get via findNode. Used as a backup if doGet
892     * encounters an in-progress deletion.
893     * @param key the key
894     * @return the value, or null if absent
895     */
896     private V getUsingFindNode(Comparable<K> key) {
897     /*
898     * Loop needed here and elsewhere in case value field goes
899     * null just as it is about to be returned, in which case we
900     * lost a race with a deletion, so must retry.
901     */
902     for (;;) {
903     Node<K,V> n = findNode(key);
904     if (n == null)
905     return null;
906     Object v = n.value;
907     if (v != null)
908     return (V)v;
909     }
910     }
911    
912     /* ---------------- Insertion -------------- */
913    
914     /**
915     * Main insertion method. Adds element if not present, or
916     * replaces value if present and onlyIfAbsent is false.
917     * @param kkey the key
918     * @param value the value that must be associated with key
919     * @param onlyIfAbsent if should not insert if already present
920     * @return the old value, or null if newly inserted
921     */
922     private V doPut(K kkey, V value, boolean onlyIfAbsent) {
923     Comparable<K> key = comparable(kkey);
924     for (;;) {
925     Node<K,V> b = findPredecessor(key);
926     Node<K,V> n = b.next;
927     for (;;) {
928     if (n != null) {
929     Node<K,V> f = n.next;
930     if (n != b.next) // inconsistent read
931     break;;
932     Object v = n.value;
933     if (v == null) { // n is deleted
934     n.helpDelete(b, f);
935     break;
936     }
937     if (v == n || b.value == null) // b is deleted
938     break;
939     int c = key.compareTo(n.key);
940     if (c > 0) {
941     b = n;
942     n = f;
943     continue;
944     }
945     if (c == 0) {
946     if (onlyIfAbsent || n.casValue(v, value))
947     return (V)v;
948     else
949     break; // restart if lost race to replace value
950     }
951     // else c < 0; fall through
952     }
953    
954     Node<K,V> z = new Node<K,V>(kkey, value, n);
955     if (!b.casNext(n, z))
956     break; // restart if lost race to append to b
957     int level = randomLevel();
958     if (level > 0)
959     insertIndex(z, level);
960     return null;
961     }
962     }
963     }
964    
965     /**
966     * Return a random level for inserting a new node.
967     * Hardwired to k=1, p=0.5, max 31.
968     *
969     * This uses a cheap pseudo-random function that according to
970     * http://home1.gte.net/deleyd/random/random4.html was used in
971     * Turbo Pascal. It seems the fastest usable one here. The low
972     * bits are apparently not very random (the original used only
973     * upper 16 bits) so we traverse from highest bit down (i.e., test
974     * sign), thus hardly ever use lower bits.
975     */
976     private int randomLevel() {
977     int level = 0;
978     int r = randomSeed;
979     randomSeed = r * 134775813 + 1;
980     if (r < 0) {
981     while ((r <<= 1) > 0)
982     ++level;
983     }
984     return level;
985     }
986    
987     /**
988     * Create and add index nodes for given node.
989     * @param z the node
990     * @param level the level of the index
991     */
992     private void insertIndex(Node<K,V> z, int level) {
993     HeadIndex<K,V> h = head;
994     int max = h.level;
995    
996     if (level <= max) {
997     Index<K,V> idx = null;
998     for (int i = 1; i <= level; ++i)
999     idx = new Index<K,V>(z, idx, null);
1000     addIndex(idx, h, level);
1001    
1002     } else { // Add a new level
1003     /*
1004     * To reduce interference by other threads checking for
1005     * empty levels in tryReduceLevel, new levels are added
1006     * with initialized right pointers. Which in turn requires
1007     * keeping levels in an array to access them while
1008     * creating new head index nodes from the opposite
1009     * direction.
1010     */
1011     level = max + 1;
1012     Index<K,V>[] idxs = (Index<K,V>[])new Index[level+1];
1013     Index<K,V> idx = null;
1014     for (int i = 1; i <= level; ++i)
1015     idxs[i] = idx = new Index<K,V>(z, idx, null);
1016    
1017     HeadIndex<K,V> oldh;
1018     int k;
1019     for (;;) {
1020     oldh = head;
1021     int oldLevel = oldh.level;
1022     if (level <= oldLevel) { // lost race to add level
1023     k = level;
1024     break;
1025     }
1026     HeadIndex<K,V> newh = oldh;
1027     Node<K,V> oldbase = oldh.node;
1028     for (int j = oldLevel+1; j <= level; ++j)
1029     newh = new HeadIndex<K,V>(oldbase, newh, idxs[j], j);
1030     if (casHead(oldh, newh)) {
1031     k = oldLevel;
1032     break;
1033     }
1034     }
1035     addIndex(idxs[k], oldh, k);
1036     }
1037     }
1038    
1039     /**
1040     * Add given index nodes from given level down to 1.
1041     * @param idx the topmost index node being inserted
1042     * @param h the value of head to use to insert. This must be
1043     * snapshotted by callers to provide correct insertion level
1044     * @param indexLevel the level of the index
1045     */
1046     private void addIndex(Index<K,V> idx, HeadIndex<K,V> h, int indexLevel) {
1047     // Track next level to insert in case of retries
1048     int insertionLevel = indexLevel;
1049     Comparable<K> key = comparable(idx.key);
1050    
1051     // Similar to findPredecessor, but adding index nodes along
1052     // path to key.
1053     for (;;) {
1054     Index<K,V> q = h;
1055     Index<K,V> t = idx;
1056     int j = h.level;
1057     for (;;) {
1058     Index<K,V> r = q.right;
1059     if (r != null) {
1060     // compare before deletion check avoids needing recheck
1061     int c = key.compareTo(r.key);
1062     if (r.indexesDeletedNode()) {
1063     if (q.unlink(r))
1064     continue;
1065     else
1066     break;
1067     }
1068     if (c > 0) {
1069     q = r;
1070     continue;
1071     }
1072     }
1073    
1074     if (j == insertionLevel) {
1075     // Don't insert index if node already deleted
1076     if (t.indexesDeletedNode()) {
1077     findNode(key); // cleans up
1078     return;
1079     }
1080     if (!q.link(r, t))
1081     break; // restart
1082     if (--insertionLevel == 0) {
1083     // need final deletion check before return
1084     if (t.indexesDeletedNode())
1085     findNode(key);
1086     return;
1087     }
1088     }
1089    
1090     if (j > insertionLevel && j <= indexLevel)
1091     t = t.down;
1092     q = q.down;
1093     --j;
1094     }
1095     }
1096     }
1097    
1098     /* ---------------- Deletion -------------- */
1099    
1100     /**
1101     * Main deletion method. Locates node, nulls value, appends a
1102     * deletion marker, unlinks predecessor, removes associated index
1103     * nodes, and possibly reduces head index level.
1104     *
1105     * Index nodes are cleared out simply by calling findPredecessor.
1106     * which unlinks indexes to deleted nodes found along path to key,
1107     * which will include the indexes to this node. This is done
1108     * unconditionally. We can't check beforehand whether there are
1109     * index nodes because it might be the case that some or all
1110     * indexes hadn't been inserted yet for this node during initial
1111     * search for it, and we'd like to ensure lack of garbage
1112     * retention, so must call to be sure.
1113     *
1114     * @param okey the key
1115     * @param value if non-null, the value that must be
1116     * associated with key
1117     * @return the node, or null if not found
1118     */
1119     private V doRemove(Object okey, Object value) {
1120     Comparable<K> key = comparable(okey);
1121     for (;;) {
1122     Node<K,V> b = findPredecessor(key);
1123     Node<K,V> n = b.next;
1124     for (;;) {
1125     if (n == null)
1126     return null;
1127     Node<K,V> f = n.next;
1128     if (n != b.next) // inconsistent read
1129     break;
1130     Object v = n.value;
1131     if (v == null) { // n is deleted
1132     n.helpDelete(b, f);
1133     break;
1134     }
1135     if (v == n || b.value == null) // b is deleted
1136     break;
1137     int c = key.compareTo(n.key);
1138     if (c < 0)
1139     return null;
1140     if (c > 0) {
1141     b = n;
1142     n = f;
1143     continue;
1144     }
1145     if (value != null && !value.equals(v))
1146     return null;
1147     if (!n.casValue(v, null))
1148     break;
1149     if (!n.appendMarker(f) || !b.casNext(n, f))
1150     findNode(key); // Retry via findNode
1151     else {
1152     findPredecessor(key); // Clean index
1153     if (head.right == null)
1154     tryReduceLevel();
1155     }
1156     return (V)v;
1157     }
1158     }
1159     }
1160    
1161     /**
1162     * Possibly reduce head level if it has no nodes. This method can
1163     * (rarely) make mistakes, in which case levels can disappear even
1164     * though they are about to contain index nodes. This impacts
1165     * performance, not correctness. To minimize mistakes as well as
1166     * to reduce hysteresis, the level is reduced by one only if the
1167     * topmost three levels look empty. Also, if the removed level
1168     * looks non-empty after CAS, we try to change it back quick
1169     * before anyone notices our mistake! (This trick works pretty
1170     * well because this method will practically never make mistakes
1171     * unless current thread stalls immediately before first CAS, in
1172     * which case it is very unlikely to stall again immediately
1173     * afterwards, so will recover.)
1174     *
1175     * We put up with all this rather than just let levels grow
1176     * because otherwise, even a small map that has undergone a large
1177     * number of insertions and removals will have a lot of levels,
1178     * slowing down access more than would an occasional unwanted
1179     * reduction.
1180     */
1181     private void tryReduceLevel() {
1182     HeadIndex<K,V> h = head;
1183     HeadIndex<K,V> d;
1184     HeadIndex<K,V> e;
1185     if (h.level > 3 &&
1186     (d = (HeadIndex<K,V>)h.down) != null &&
1187     (e = (HeadIndex<K,V>)d.down) != null &&
1188     e.right == null &&
1189     d.right == null &&
1190     h.right == null &&
1191     casHead(h, d) && // try to set
1192     h.right != null) // recheck
1193     casHead(d, h); // try to backout
1194     }
1195    
1196     /**
1197     * Version of remove with boolean return. Needed by view classes
1198     */
1199     boolean removep(Object key) {
1200     return doRemove(key, null) != null;
1201     }
1202    
1203     /* ---------------- Finding and removing first element -------------- */
1204    
1205     /**
1206     * Specialized variant of findNode to get first valid node
1207     * @return first node or null if empty
1208     */
1209     Node<K,V> findFirst() {
1210     for (;;) {
1211     Node<K,V> b = head.node;
1212     Node<K,V> n = b.next;
1213     if (n == null)
1214     return null;
1215     if (n.value != null)
1216     return n;
1217     n.helpDelete(b, n.next);
1218     }
1219     }
1220    
1221     /**
1222     * Remove first entry; return either its key or a snapshot.
1223     * @param keyOnly if true return key, else return SnapshotEntry
1224     * (This is a little ugly, but avoids code duplication.)
1225     * @return null if empty, first key if keyOnly true, else key,value entry
1226     */
1227     Object doRemoveFirst(boolean keyOnly) {
1228     for (;;) {
1229     Node<K,V> b = head.node;
1230     Node<K,V> n = b.next;
1231     if (n == null)
1232     return null;
1233     Node<K,V> f = n.next;
1234     if (n != b.next)
1235     continue;
1236     Object v = n.value;
1237     if (v == null) {
1238     n.helpDelete(b, f);
1239     continue;
1240     }
1241     if (!n.casValue(v, null))
1242     continue;
1243     if (!n.appendMarker(f) || !b.casNext(n, f))
1244     findFirst(); // retry
1245     clearIndexToFirst();
1246     K key = n.key;
1247     return (keyOnly)? key : new SnapshotEntry<K,V>(key, (V)v);
1248     }
1249     }
1250    
1251     /**
1252     * Clear out index nodes associated with deleted first entry.
1253     * Needed by doRemoveFirst
1254     */
1255     private void clearIndexToFirst() {
1256     for (;;) {
1257     Index<K,V> q = head;
1258     for (;;) {
1259     Index<K,V> r = q.right;
1260     if (r != null && r.indexesDeletedNode() && !q.unlink(r))
1261     break;
1262     if ((q = q.down) == null) {
1263     if (head.right == null)
1264     tryReduceLevel();
1265     return;
1266     }
1267     }
1268     }
1269     }
1270    
1271     /**
1272     * Remove first entry; return key or null if empty.
1273     */
1274     K pollFirstKey() {
1275     return (K)doRemoveFirst(true);
1276     }
1277    
1278     /* ---------------- Finding and removing last element -------------- */
1279    
1280     /**
1281     * Specialized version of find to get last valid node
1282     * @return last node or null if empty
1283     */
1284     Node<K,V> findLast() {
1285     /*
1286     * findPredecessor can't be used to traverse index level
1287     * because this doesn't use comparisons. So traversals of
1288     * both levels are folded together.
1289     */
1290     Index<K,V> q = head;
1291     for (;;) {
1292     Index<K,V> d, r;
1293     if ((r = q.right) != null) {
1294     if (r.indexesDeletedNode()) {
1295     q.unlink(r);
1296     q = head; // restart
1297     }
1298     else
1299     q = r;
1300     } else if ((d = q.down) != null) {
1301     q = d;
1302     } else {
1303     Node<K,V> b = q.node;
1304     Node<K,V> n = b.next;
1305     for (;;) {
1306     if (n == null)
1307     return (b.isBaseHeader())? null : b;
1308     Node<K,V> f = n.next; // inconsistent read
1309     if (n != b.next)
1310     break;
1311     Object v = n.value;
1312     if (v == null) { // n is deleted
1313     n.helpDelete(b, f);
1314     break;
1315     }
1316     if (v == n || b.value == null) // b is deleted
1317     break;
1318     b = n;
1319     n = f;
1320     }
1321     q = head; // restart
1322     }
1323     }
1324     }
1325    
1326    
1327     /**
1328     * Specialized version of doRemove for last entry.
1329     * @param keyOnly if true return key, else return SnapshotEntry
1330     * @return null if empty, last key if keyOnly true, else key,value entry
1331     */
1332     Object doRemoveLast(boolean keyOnly) {
1333     for (;;) {
1334     Node<K,V> b = findPredecessorOfLast();
1335     Node<K,V> n = b.next;
1336     if (n == null) {
1337     if (b.isBaseHeader()) // empty
1338     return null;
1339     else
1340     continue; // all b's successors are deleted; retry
1341     }
1342     for (;;) {
1343     Node<K,V> f = n.next;
1344     if (n != b.next) // inconsistent read
1345     break;
1346     Object v = n.value;
1347     if (v == null) { // n is deleted
1348     n.helpDelete(b, f);
1349     break;
1350     }
1351     if (v == n || b.value == null) // b is deleted
1352     break;
1353     if (f != null) {
1354     b = n;
1355     n = f;
1356     continue;
1357     }
1358     if (!n.casValue(v, null))
1359     break;
1360     K key = n.key;
1361     Comparable<K> ck = comparable(key);
1362     if (!n.appendMarker(f) || !b.casNext(n, f))
1363     findNode(ck); // Retry via findNode
1364     else {
1365     findPredecessor(ck); // Clean index
1366     if (head.right == null)
1367     tryReduceLevel();
1368     }
1369     return (keyOnly)? key : new SnapshotEntry<K,V>(key, (V)v);
1370     }
1371     }
1372     }
1373    
1374     /**
1375     * Specialized variant of findPredecessor to get predecessor of
1376     * last valid node. Needed by doRemoveLast. It is possible that
1377     * all successors of returned node will have been deleted upon
1378     * return, in which case this method can be retried.
1379     * @return likely predecessor of last node.
1380     */
1381     private Node<K,V> findPredecessorOfLast() {
1382     for (;;) {
1383     Index<K,V> q = head;
1384     for (;;) {
1385     Index<K,V> d, r;
1386     if ((r = q.right) != null) {
1387     if (r.indexesDeletedNode()) {
1388     q.unlink(r);
1389     break; // must restart
1390     }
1391     // proceed as far across as possible without overshooting
1392     if (r.node.next != null) {
1393     q = r;
1394     continue;
1395     }
1396     }
1397     if ((d = q.down) != null)
1398     q = d;
1399     else
1400     return q.node;
1401     }
1402     }
1403     }
1404    
1405     /**
1406     * Remove last entry; return key or null if empty.
1407     */
1408     K pollLastKey() {
1409     return (K)doRemoveLast(true);
1410     }
1411    
1412     /* ---------------- Relational operations -------------- */
1413    
1414     // Control values OR'ed as arguments to findNear
1415    
1416     private static final int EQ = 1;
1417     private static final int LT = 2;
1418     private static final int GT = 0; // Actually checked as !LT
1419    
1420     /**
1421     * Utility for ceiling, floor, lower, higher methods.
1422     * @param kkey the key
1423     * @param rel the relation -- OR'ed combination of EQ, LT, GT
1424     * @return nearest node fitting relation, or null if no such
1425     */
1426     Node<K,V> findNear(K kkey, int rel) {
1427     Comparable<K> key = comparable(kkey);
1428     for (;;) {
1429     Node<K,V> b = findPredecessor(key);
1430     Node<K,V> n = b.next;
1431     for (;;) {
1432     if (n == null)
1433     return ((rel & LT) == 0 || b.isBaseHeader())? null : b;
1434     Node<K,V> f = n.next;
1435     if (n != b.next) // inconsistent read
1436     break;
1437     Object v = n.value;
1438     if (v == null) { // n is deleted
1439     n.helpDelete(b, f);
1440     break;
1441     }
1442     if (v == n || b.value == null) // b is deleted
1443     break;
1444     int c = key.compareTo(n.key);
1445     if ((c == 0 && (rel & EQ) != 0) ||
1446     (c < 0 && (rel & LT) == 0))
1447     return n;
1448     if ( c <= 0 && (rel & LT) != 0)
1449     return (b.isBaseHeader())? null : b;
1450     b = n;
1451     n = f;
1452     }
1453     }
1454     }
1455    
1456     /**
1457     * Return SnapshotEntry for results of findNear.
1458     * @param kkey the key
1459     * @param rel the relation -- OR'ed combination of EQ, LT, GT
1460     * @return Entry fitting relation, or null if no such
1461     */
1462     SnapshotEntry<K,V> getNear(K kkey, int rel) {
1463     for (;;) {
1464     Node<K,V> n = findNear(kkey, rel);
1465     if (n == null)
1466     return null;
1467     SnapshotEntry<K,V> e = n.createSnapshot();
1468     if (e != null)
1469     return e;
1470     }
1471     }
1472    
1473     /**
1474     * Return ceiling, or first node if key is <tt>null</tt>
1475     */
1476     Node<K,V> findCeiling(K key) {
1477     return (key == null)? findFirst() : findNear(key, GT|EQ);
1478     }
1479    
1480     /**
1481     * Return lower node, or last node if key is <tt>null</tt>
1482     */
1483     Node<K,V> findLower(K key) {
1484     return (key == null)? findLast() : findNear(key, LT);
1485     }
1486    
1487     /**
1488     * Return SnapshotEntry or key for results of findNear ofter screening
1489     * to ensure result is in given range. Needed by submaps.
1490     * @param kkey the key
1491     * @param rel the relation -- OR'ed combination of EQ, LT, GT
1492     * @param least minimum allowed key value
1493     * @param fence key greater than maximum allowed key value
1494     * @param keyOnly if true return key, else return SnapshotEntry
1495     * @return Key or Entry fitting relation, or <tt>null</tt> if no such
1496     */
1497     Object getNear(K kkey, int rel, K least, K fence, boolean keyOnly) {
1498     K key = kkey;
1499     // Don't return keys less than least
1500     if ((rel & LT) == 0) {
1501     if (compare(key, least) < 0) {
1502     key = least;
1503     rel = rel | EQ;
1504     }
1505     }
1506    
1507     for (;;) {
1508     Node<K,V> n = findNear(key, rel);
1509     if (n == null || !inHalfOpenRange(n.key, least, fence))
1510     return null;
1511     K k = n.key;
1512     V v = n.getValidValue();
1513     if (v != null)
1514     return keyOnly? k : new SnapshotEntry<K,V>(k, v);
1515     }
1516     }
1517    
1518     /**
1519     * Find and remove least element of subrange.
1520     * @param least minimum allowed key value
1521     * @param fence key greater than maximum allowed key value
1522     * @param keyOnly if true return key, else return SnapshotEntry
1523     * @return least Key or Entry, or <tt>null</tt> if no such
1524     */
1525     Object removeFirstEntryOfSubrange(K least, K fence, boolean keyOnly) {
1526     for (;;) {
1527     Node<K,V> n = findCeiling(least);
1528     if (n == null)
1529     return null;
1530     K k = n.key;
1531     if (fence != null && compare(k, fence) >= 0)
1532     return null;
1533     V v = doRemove(k, null);
1534     if (v != null)
1535     return (keyOnly)? k : new SnapshotEntry<K,V>(k, v);
1536     }
1537     }
1538    
1539     /**
1540     * Find and remove greatest element of subrange.
1541     * @param least minimum allowed key value
1542     * @param fence key greater than maximum allowed key value
1543     * @param keyOnly if true return key, else return SnapshotEntry
1544     * @return least Key or Entry, or <tt>null</tt> if no such
1545     */
1546     Object removeLastEntryOfSubrange(K least, K fence, boolean keyOnly) {
1547     for (;;) {
1548     Node<K,V> n = findLower(fence);
1549     if (n == null)
1550     return null;
1551     K k = n.key;
1552     if (least != null && compare(k, least) < 0)
1553     return null;
1554     V v = doRemove(k, null);
1555     if (v != null)
1556     return (keyOnly)? k : new SnapshotEntry<K,V>(k, v);
1557     }
1558     }
1559    
1560     /* ---------------- Constructors -------------- */
1561    
1562     /**
1563     * Constructs a new empty map, sorted according to the keys' natural
1564     * order.
1565     */
1566     public ConcurrentSkipListMap() {
1567     this.comparator = null;
1568     initialize();
1569     }
1570    
1571     /**
1572     * Constructs a new empty map, sorted according to the given comparator.
1573     *
1574     * @param c the comparator that will be used to sort this map. A
1575     * <tt>null</tt> value indicates that the keys' <i>natural
1576     * ordering</i> should be used.
1577     */
1578     public ConcurrentSkipListMap(Comparator<? super K> c) {
1579     this.comparator = c;
1580     initialize();
1581     }
1582    
1583     /**
1584     * Constructs a new map containing the same mappings as the given map,
1585     * sorted according to the keys' <i>natural order</i>.
1586     *
1587     * @param m the map whose mappings are to be placed in this map.
1588     * @throws ClassCastException if the keys in m are not Comparable, or
1589     * are not mutually comparable.
1590     * @throws NullPointerException if the specified map is <tt>null</tt>.
1591     */
1592     public ConcurrentSkipListMap(Map<? extends K, ? extends V> m) {
1593     this.comparator = null;
1594     initialize();
1595     putAll(m);
1596     }
1597    
1598     /**
1599     * Constructs a new map containing the same mappings as the given
1600     * <tt>SortedMap</tt>, sorted according to the same ordering.
1601     * @param m the sorted map whose mappings are to be placed in this
1602     * map, and whose comparator is to be used to sort this map.
1603     * @throws NullPointerException if the specified sorted map is
1604     * <tt>null</tt>.
1605     */
1606     public ConcurrentSkipListMap(SortedMap<K, ? extends V> m) {
1607     this.comparator = m.comparator();
1608     initialize();
1609     buildFromSorted(m);
1610     }
1611    
1612     /**
1613     * Returns a shallow copy of this <tt>Map</tt> instance. (The keys and
1614     * values themselves are not cloned.)
1615     *
1616     * @return a shallow copy of this Map.
1617     */
1618     public Object clone() {
1619     ConcurrentSkipListMap<K,V> clone = null;
1620     try {
1621     clone = (ConcurrentSkipListMap<K,V>) super.clone();
1622     } catch (CloneNotSupportedException e) {
1623     throw new InternalError();
1624     }
1625    
1626     clone.initialize();
1627     clone.buildFromSorted(this);
1628     return clone;
1629     }
1630    
1631     /**
1632     * Streamlined bulk insertion to initialize from elements of
1633     * given sorted map. Call only from constructor or clone
1634     * method.
1635     */
1636     private void buildFromSorted(SortedMap<K, ? extends V> map) {
1637     if (map == null)
1638     throw new NullPointerException();
1639    
1640     HeadIndex<K,V> h = head;
1641     Node<K,V> basepred = h.node;
1642    
1643     // Track the current rightmost node at each level. Uses an
1644     // ArrayList to avoid committing to initial or maximum level.
1645     ArrayList<Index<K,V>> preds = new ArrayList<Index<K,V>>();
1646    
1647     // initialize
1648     for (int i = 0; i <= h.level; ++i)
1649     preds.add(null);
1650     Index<K,V> q = h;
1651     for (int i = h.level; i > 0; --i) {
1652     preds.set(i, q);
1653     q = q.down;
1654     }
1655    
1656     Iterator<? extends Map.Entry<? extends K, ? extends V>> it =
1657     map.entrySet().iterator();
1658     while (it.hasNext()) {
1659     Map.Entry<? extends K, ? extends V> e = it.next();
1660     int j = randomLevel();
1661     if (j > h.level) j = h.level + 1;
1662     K k = e.getKey();
1663     V v = e.getValue();
1664     if (k == null || v == null)
1665     throw new NullPointerException();
1666     Node<K,V> z = new Node<K,V>(k, v, null);
1667     basepred.next = z;
1668     basepred = z;
1669     if (j > 0) {
1670     Index<K,V> idx = null;
1671     for (int i = 1; i <= j; ++i) {
1672     idx = new Index<K,V>(z, idx, null);
1673     if (i > h.level)
1674     h = new HeadIndex<K,V>(h.node, h, idx, i);
1675    
1676     if (i < preds.size()) {
1677     preds.get(i).right = idx;
1678     preds.set(i, idx);
1679     } else
1680     preds.add(idx);
1681     }
1682     }
1683     }
1684     head = h;
1685     }
1686    
1687     /* ---------------- Serialization -------------- */
1688    
1689     /**
1690     * Save the state of the <tt>Map</tt> instance to a stream.
1691     *
1692     * @serialData The key (Object) and value (Object) for each
1693     * key-value mapping represented by the Map, followed by
1694     * <tt>null</tt>. The key-value mappings are emitted in key-order
1695     * (as determined by the Comparator, or by the keys' natural
1696     * ordering if no Comparator).
1697     */
1698     private void writeObject(java.io.ObjectOutputStream s)
1699     throws java.io.IOException {
1700     // Write out the Comparator and any hidden stuff
1701     s.defaultWriteObject();
1702    
1703     // Write out keys and values (alternating)
1704     for (Node<K,V> n = findFirst(); n != null; n = n.next) {
1705     V v = n.getValidValue();
1706     if (v != null) {
1707     s.writeObject(n.key);
1708     s.writeObject(v);
1709     }
1710     }
1711     s.writeObject(null);
1712     }
1713    
1714     /**
1715     * Reconstitute the <tt>Map</tt> instance from a stream.
1716     */
1717     private void readObject(final java.io.ObjectInputStream s)
1718     throws java.io.IOException, ClassNotFoundException {
1719     // Read in the Comparator and any hidden stuff
1720     s.defaultReadObject();
1721     // Reset transients
1722     initialize();
1723    
1724     /*
1725     * This is nearly identical to buildFromSorted, but is
1726     * distinct because readObject calls can't be nicely adapted
1727     * as the kind of iterator needed by buildFromSorted. (They
1728     * can be, but doing so requires type cheats and/or creation
1729     * of adaptor classes.) It is simpler to just adapt the code.
1730     */
1731    
1732     HeadIndex<K,V> h = head;
1733     Node<K,V> basepred = h.node;
1734     ArrayList<Index<K,V>> preds = new ArrayList<Index<K,V>>();
1735     for (int i = 0; i <= h.level; ++i)
1736     preds.add(null);
1737     Index<K,V> q = h;
1738     for (int i = h.level; i > 0; --i) {
1739     preds.set(i, q);
1740     q = q.down;
1741     }
1742    
1743     for (;;) {
1744     Object k = s.readObject();
1745     if (k == null)
1746     break;
1747     Object v = s.readObject();
1748     if (v == null)
1749     throw new NullPointerException();
1750     K key = (K) k;
1751     V val = (V) v;
1752     int j = randomLevel();
1753     if (j > h.level) j = h.level + 1;
1754     Node<K,V> z = new Node<K,V>(key, val, null);
1755     basepred.next = z;
1756     basepred = z;
1757     if (j > 0) {
1758     Index<K,V> idx = null;
1759     for (int i = 1; i <= j; ++i) {
1760     idx = new Index<K,V>(z, idx, null);
1761     if (i > h.level)
1762     h = new HeadIndex<K,V>(h.node, h, idx, i);
1763    
1764     if (i < preds.size()) {
1765     preds.get(i).right = idx;
1766     preds.set(i, idx);
1767     } else
1768     preds.add(idx);
1769     }
1770     }
1771     }
1772     head = h;
1773     }
1774    
1775     /* ------ Map API methods ------ */
1776    
1777     /**
1778     * Returns <tt>true</tt> if this map contains a mapping for the specified
1779     * key.
1780     * @param key key whose presence in this map is to be tested.
1781     * @return <tt>true</tt> if this map contains a mapping for the
1782     * specified key.
1783     * @throws ClassCastException if the key cannot be compared with the keys
1784     * currently in the map.
1785     * @throws NullPointerException if the key is <tt>null</tt>.
1786     */
1787     public boolean containsKey(Object key) {
1788     return doGet(key) != null;
1789     }
1790    
1791     /**
1792     * Returns the value to which this map maps the specified key. Returns
1793     * <tt>null</tt> if the map contains no mapping for this key.
1794     *
1795     * @param key key whose associated value is to be returned.
1796     * @return the value to which this map maps the specified key, or
1797     * <tt>null</tt> if the map contains no mapping for the key.
1798     * @throws ClassCastException if the key cannot be compared with the keys
1799     * currently in the map.
1800     * @throws NullPointerException if the key is <tt>null</tt>.
1801     */
1802     public V get(Object key) {
1803     return doGet(key);
1804     }
1805    
1806     /**
1807     * Associates the specified value with the specified key in this map.
1808     * If the map previously contained a mapping for this key, the old
1809     * value is replaced.
1810     *
1811     * @param key key with which the specified value is to be associated.
1812     * @param value value to be associated with the specified key.
1813     *
1814     * @return previous value associated with specified key, or <tt>null</tt>
1815     * if there was no mapping for key.
1816     * @throws ClassCastException if the key cannot be compared with the keys
1817     * currently in the map.
1818     * @throws NullPointerException if the key or value are <tt>null</tt>.
1819     */
1820     public V put(K key, V value) {
1821     if (value == null)
1822     throw new NullPointerException();
1823     return doPut(key, value, false);
1824     }
1825    
1826     /**
1827     * Removes the mapping for this key from this Map if present.
1828     *
1829     * @param key key for which mapping should be removed
1830     * @return previous value associated with specified key, or <tt>null</tt>
1831     * if there was no mapping for key.
1832     *
1833     * @throws ClassCastException if the key cannot be compared with the keys
1834     * currently in the map.
1835     * @throws NullPointerException if the key is <tt>null</tt>.
1836     */
1837     public V remove(Object key) {
1838     return doRemove(key, null);
1839     }
1840    
1841     /**
1842     * Returns <tt>true</tt> if this map maps one or more keys to the
1843     * specified value. This operation requires time linear in the
1844     * Map size.
1845     *
1846     * @param value value whose presence in this Map is to be tested.
1847     * @return <tt>true</tt> if a mapping to <tt>value</tt> exists;
1848     * <tt>false</tt> otherwise.
1849     * @throws NullPointerException if the value is <tt>null</tt>.
1850     */
1851     public boolean containsValue(Object value) {
1852     if (value == null)
1853     throw new NullPointerException();
1854     for (Node<K,V> n = findFirst(); n != null; n = n.next) {
1855     V v = n.getValidValue();
1856     if (v != null && value.equals(v))
1857     return true;
1858     }
1859     return false;
1860     }
1861    
1862     /**
1863     * Returns the number of elements in this map. If this map
1864     * contains more than <tt>Integer.MAX_VALUE</tt> elements, it
1865     * returns <tt>Integer.MAX_VALUE</tt>.
1866     *
1867     * <p>Beware that, unlike in most collections, this method is
1868     * <em>NOT</em> a constant-time operation. Because of the
1869     * asynchronous nature of these maps, determining the current
1870     * number of elements requires traversing them all to count them.
1871     * Additionally, it is possible for the size to change during
1872     * execution of this method, in which case the returned result
1873     * will be inaccurate. Thus, this method is typically not very
1874     * useful in concurrent applications.
1875     *
1876     * @return the number of elements in this map.
1877     */
1878     public int size() {
1879     long count = 0;
1880     for (Node<K,V> n = findFirst(); n != null; n = n.next) {
1881     if (n.getValidValue() != null)
1882     ++count;
1883     }
1884     return (count >= Integer.MAX_VALUE)? Integer.MAX_VALUE : (int)count;
1885     }
1886    
1887     /**
1888     * Returns <tt>true</tt> if this map contains no key-value mappings.
1889     * @return <tt>true</tt> if this map contains no key-value mappings.
1890     */
1891     public boolean isEmpty() {
1892     return findFirst() == null;
1893     }
1894    
1895     /**
1896     * Removes all mappings from this map.
1897     */
1898     public void clear() {
1899     initialize();
1900     }
1901    
1902     /**
1903     * Returns a set view of the keys contained in this map. The set is
1904     * backed by the map, so changes to the map are reflected in the set, and
1905     * vice-versa. The set supports element removal, which removes the
1906     * corresponding mapping from this map, via the <tt>Iterator.remove</tt>,
1907     * <tt>Set.remove</tt>, <tt>removeAll</tt>, <tt>retainAll</tt>, and
1908     * <tt>clear</tt> operations. It does not support the <tt>add</tt> or
1909     * <tt>addAll</tt> operations.
1910     * The view's <tt>iterator</tt> is a "weakly consistent" iterator that
1911     * will never throw {@link java.util.ConcurrentModificationException},
1912     * and guarantees to traverse elements as they existed upon
1913     * construction of the iterator, and may (but is not guaranteed to)
1914     * reflect any modifications subsequent to construction.
1915     *
1916     * @return a set view of the keys contained in this map.
1917     */
1918     public Set<K> keySet() {
1919     /*
1920     * Note: Lazy intialization works here and for other views
1921     * because view classes are stateless/immutable so it doesn't
1922     * matter wrt correctness if more than one is created (which
1923     * will only rarely happen). Even so, the following idiom
1924     * conservatively ensures that the method returns the one it
1925     * created if it does so, not one created by another racing
1926     * thread.
1927     */
1928     KeySet ks = keySet;
1929     return (ks != null) ? ks : (keySet = new KeySet());
1930     }
1931    
1932     /**
1933     * Returns a set view of the keys contained in this map in
1934     * descending order. The set is backed by the map, so changes to
1935     * the map are reflected in the set, and vice-versa. The set
1936     * supports element removal, which removes the corresponding
1937     * mapping from this map, via the <tt>Iterator.remove</tt>,
1938     * <tt>Set.remove</tt>, <tt>removeAll</tt>, <tt>retainAll</tt>,
1939     * and <tt>clear</tt> operations. It does not support the
1940     * <tt>add</tt> or <tt>addAll</tt> operations. The view's
1941     * <tt>iterator</tt> is a "weakly consistent" iterator that will
1942     * never throw {@link java.util.ConcurrentModificationException},
1943     * and guarantees to traverse elements as they existed upon
1944     * construction of the iterator, and may (but is not guaranteed
1945     * to) reflect any modifications subsequent to construction.
1946     *
1947     * @return a set view of the keys contained in this map.
1948     */
1949     public Set<K> descendingKeySet() {
1950     /*
1951     * Note: Lazy intialization works here and for other views
1952     * because view classes are stateless/immutable so it doesn't
1953     * matter wrt correctness if more than one is created (which
1954     * will only rarely happen). Even so, the following idiom
1955     * conservatively ensures that the method returns the one it
1956     * created if it does so, not one created by another racing
1957     * thread.
1958     */
1959     DescendingKeySet ks = descendingKeySet;
1960     return (ks != null) ? ks : (descendingKeySet = new DescendingKeySet());
1961     }
1962    
1963     /**
1964     * Returns a collection view of the values contained in this map.
1965     * The collection is backed by the map, so changes to the map are
1966     * reflected in the collection, and vice-versa. The collection
1967     * supports element removal, which removes the corresponding
1968     * mapping from this map, via the <tt>Iterator.remove</tt>,
1969     * <tt>Collection.remove</tt>, <tt>removeAll</tt>,
1970     * <tt>retainAll</tt>, and <tt>clear</tt> operations. It does not
1971     * support the <tt>add</tt> or <tt>addAll</tt> operations. The
1972     * view's <tt>iterator</tt> is a "weakly consistent" iterator that
1973     * will never throw {@link
1974     * java.util.ConcurrentModificationException}, and guarantees to
1975     * traverse elements as they existed upon construction of the
1976     * iterator, and may (but is not guaranteed to) reflect any
1977     * modifications subsequent to construction.
1978     *
1979     * @return a collection view of the values contained in this map.
1980     */
1981     public Collection<V> values() {
1982     Values vs = values;
1983     return (vs != null) ? vs : (values = new Values());
1984     }
1985    
1986     /**
1987     * Returns a collection view of the mappings contained in this
1988     * map. Each element in the returned collection is a
1989     * <tt>Map.Entry</tt>. The collection is backed by the map, so
1990     * changes to the map are reflected in the collection, and
1991     * vice-versa. The collection supports element removal, which
1992     * removes the corresponding mapping from the map, via the
1993     * <tt>Iterator.remove</tt>, <tt>Collection.remove</tt>,
1994     * <tt>removeAll</tt>, <tt>retainAll</tt>, and <tt>clear</tt>
1995     * operations. It does not support the <tt>add</tt> or
1996     * <tt>addAll</tt> operations. The view's <tt>iterator</tt> is a
1997     * "weakly consistent" iterator that will never throw {@link
1998     * java.util.ConcurrentModificationException}, and guarantees to
1999     * traverse elements as they existed upon construction of the
2000     * iterator, and may (but is not guaranteed to) reflect any
2001     * modifications subsequent to construction. The
2002     * <tt>Map.Entry</tt> elements returned by
2003     * <tt>iterator.next()</tt> do <em>not</em> support the
2004     * <tt>setValue</tt> operation.
2005     *
2006     * @return a collection view of the mappings contained in this map.
2007     */
2008     public Set<Map.Entry<K,V>> entrySet() {
2009     EntrySet es = entrySet;
2010     return (es != null) ? es : (entrySet = new EntrySet());
2011     }
2012    
2013     /**
2014     * Returns a collection view of the mappings contained in this
2015     * map, in descending order. Each element in the returned
2016     * collection is a <tt>Map.Entry</tt>. The collection is backed
2017     * by the map, so changes to the map are reflected in the
2018     * collection, and vice-versa. The collection supports element
2019     * removal, which removes the corresponding mapping from the map,
2020     * via the <tt>Iterator.remove</tt>, <tt>Collection.remove</tt>,
2021     * <tt>removeAll</tt>, <tt>retainAll</tt>, and <tt>clear</tt>
2022     * operations. It does not support the <tt>add</tt> or
2023     * <tt>addAll</tt> operations. The view's <tt>iterator</tt> is a
2024     * "weakly consistent" iterator that will never throw {@link
2025     * java.util.ConcurrentModificationException}, and guarantees to
2026     * traverse elements as they existed upon construction of the
2027     * iterator, and may (but is not guaranteed to) reflect any
2028     * modifications subsequent to construction. The
2029     * <tt>Map.Entry</tt> elements returned by
2030     * <tt>iterator.next()</tt> do <em>not</em> support the
2031     * <tt>setValue</tt> operation.
2032     *
2033     * @return a collection view of the mappings contained in this map.
2034     */
2035     public Set<Map.Entry<K,V>> descendingEntrySet() {
2036     DescendingEntrySet es = descendingEntrySet;
2037     return (es != null) ? es : (descendingEntrySet = new DescendingEntrySet());
2038     }
2039    
2040     /* ---------------- AbstractMap Overrides -------------- */
2041    
2042     /**
2043     * Compares the specified object with this map for equality.
2044     * Returns <tt>true</tt> if the given object is also a map and the
2045     * two maps represent the same mappings. More formally, two maps
2046     * <tt>t1</tt> and <tt>t2</tt> represent the same mappings if
2047     * <tt>t1.keySet().equals(t2.keySet())</tt> and for every key
2048     * <tt>k</tt> in <tt>t1.keySet()</tt>, <tt> (t1.get(k)==null ?
2049     * t2.get(k)==null : t1.get(k).equals(t2.get(k))) </tt>. This
2050     * operation may return misleading results if either map is
2051     * concurrently modified during execution of this method.
2052     *
2053     * @param o object to be compared for equality with this map.
2054     * @return <tt>true</tt> if the specified object is equal to this map.
2055     */
2056     public boolean equals(Object o) {
2057     if (o == this)
2058     return true;
2059     if (!(o instanceof Map))
2060     return false;
2061     Map<K,V> t = (Map<K,V>) o;
2062     try {
2063     return (containsAllMappings(this, t) &&
2064     containsAllMappings(t, this));
2065     } catch(ClassCastException unused) {
2066     return false;
2067     } catch(NullPointerException unused) {
2068     return false;
2069     }
2070     }
2071    
2072     /**
2073     * Helper for equals -- check for containment, avoiding nulls.
2074     */
2075     static <K,V> boolean containsAllMappings(Map<K,V> a, Map<K,V> b) {
2076     Iterator<Entry<K,V>> it = b.entrySet().iterator();
2077     while (it.hasNext()) {
2078     Entry<K,V> e = it.next();
2079     Object k = e.getKey();
2080     Object v = e.getValue();
2081     if (k == null || v == null || !v.equals(a.get(k)))
2082     return false;
2083     }
2084     return true;
2085     }
2086    
2087     /* ------ ConcurrentMap API methods ------ */
2088    
2089     /**
2090     * If the specified key is not already associated
2091     * with a value, associate it with the given value.
2092     * This is equivalent to
2093     * <pre>
2094     * if (!map.containsKey(key))
2095     * return map.put(key, value);
2096     * else
2097     * return map.get(key);
2098     * </pre>
2099     * except that the action is performed atomically.
2100     * @param key key with which the specified value is to be associated.
2101     * @param value value to be associated with the specified key.
2102     * @return previous value associated with specified key, or <tt>null</tt>
2103     * if there was no mapping for key.
2104     *
2105     * @throws ClassCastException if the key cannot be compared with the keys
2106     * currently in the map.
2107     * @throws NullPointerException if the key or value are <tt>null</tt>.
2108     */
2109     public V putIfAbsent(K key, V value) {
2110     if (value == null)
2111     throw new NullPointerException();
2112     return doPut(key, value, true);
2113     }
2114    
2115     /**
2116     * Remove entry for key only if currently mapped to given value.
2117     * Acts as
2118     * <pre>
2119     * if ((map.containsKey(key) && map.get(key).equals(value)) {
2120     * map.remove(key);
2121     * return true;
2122     * } else return false;
2123     * </pre>
2124     * except that the action is performed atomically.
2125     * @param key key with which the specified value is associated.
2126     * @param value value associated with the specified key.
2127     * @return true if the value was removed, false otherwise
2128     * @throws ClassCastException if the key cannot be compared with the keys
2129     * currently in the map.
2130     * @throws NullPointerException if the key or value are <tt>null</tt>.
2131     */
2132     public boolean remove(Object key, Object value) {
2133     if (value == null)
2134     throw new NullPointerException();
2135     return doRemove(key, value) != null;
2136     }
2137    
2138     /**
2139     * Replace entry for key only if currently mapped to given value.
2140     * Acts as
2141     * <pre>
2142     * if ((map.containsKey(key) && map.get(key).equals(oldValue)) {
2143     * map.put(key, newValue);
2144     * return true;
2145     * } else return false;
2146     * </pre>
2147     * except that the action is performed atomically.
2148     * @param key key with which the specified value is associated.
2149     * @param oldValue value expected to be associated with the specified key.
2150     * @param newValue value to be associated with the specified key.
2151     * @return true if the value was replaced
2152     * @throws ClassCastException if the key cannot be compared with the keys
2153     * currently in the map.
2154     * @throws NullPointerException if key, oldValue or newValue are
2155     * <tt>null</tt>.
2156     */
2157     public boolean replace(K key, V oldValue, V newValue) {
2158     if (oldValue == null || newValue == null)
2159     throw new NullPointerException();
2160     Comparable<K> k = comparable(key);
2161     for (;;) {
2162     Node<K,V> n = findNode(k);
2163     if (n == null)
2164     return false;
2165     Object v = n.value;
2166     if (v != null) {
2167     if (!oldValue.equals(v))
2168     return false;
2169     if (n.casValue(v, newValue))
2170     return true;
2171     }
2172     }
2173     }
2174    
2175     /**
2176     * Replace entry for key only if currently mapped to some value.
2177     * Acts as
2178     * <pre>
2179     * if ((map.containsKey(key)) {
2180     * return map.put(key, value);
2181     * } else return null;
2182     * </pre>
2183     * except that the action is performed atomically.
2184     * @param key key with which the specified value is associated.
2185     * @param value value to be associated with the specified key.
2186     * @return previous value associated with specified key, or <tt>null</tt>
2187     * if there was no mapping for key.
2188     * @throws ClassCastException if the key cannot be compared with the keys
2189     * currently in the map.
2190     * @throws NullPointerException if the key or value are <tt>null</tt>.
2191     */
2192     public V replace(K key, V value) {
2193     if (value == null)
2194     throw new NullPointerException();
2195     Comparable<K> k = comparable(key);
2196     for (;;) {
2197     Node<K,V> n = findNode(k);
2198     if (n == null)
2199     return null;
2200     Object v = n.value;
2201     if (v != null && n.casValue(v, value))
2202     return (V)v;
2203     }
2204     }
2205    
2206     /* ------ SortedMap API methods ------ */
2207    
2208     /**
2209     * Returns the comparator used to order this map, or <tt>null</tt>
2210     * if this map uses its keys' natural order.
2211     *
2212     * @return the comparator associated with this map, or
2213     * <tt>null</tt> if it uses its keys' natural sort method.
2214     */
2215     public Comparator<? super K> comparator() {
2216     return comparator;
2217     }
2218    
2219     /**
2220     * Returns the first (lowest) key currently in this map.
2221     *
2222     * @return the first (lowest) key currently in this map.
2223     * @throws NoSuchElementException Map is empty.
2224     */
2225     public K firstKey() {
2226     Node<K,V> n = findFirst();
2227     if (n == null)
2228     throw new NoSuchElementException();
2229     return n.key;
2230     }
2231    
2232     /**
2233     * Returns the last (highest) key currently in this map.
2234     *
2235     * @return the last (highest) key currently in this map.
2236     * @throws NoSuchElementException Map is empty.
2237     */
2238     public K lastKey() {
2239     Node<K,V> n = findLast();
2240     if (n == null)
2241     throw new NoSuchElementException();
2242     return n.key;
2243     }
2244    
2245     /**
2246     * Returns a view of the portion of this map whose keys range from
2247     * <tt>fromKey</tt>, inclusive, to <tt>toKey</tt>, exclusive. (If
2248     * <tt>fromKey</tt> and <tt>toKey</tt> are equal, the returned sorted map
2249     * is empty.) The returned sorted map is backed by this map, so changes
2250     * in the returned sorted map are reflected in this map, and vice-versa.
2251    
2252     * @param fromKey low endpoint (inclusive) of the subMap.
2253     * @param toKey high endpoint (exclusive) of the subMap.
2254     *
2255     * @return a view of the portion of this map whose keys range from
2256     * <tt>fromKey</tt>, inclusive, to <tt>toKey</tt>, exclusive.
2257     *
2258     * @throws ClassCastException if <tt>fromKey</tt> and <tt>toKey</tt>
2259     * cannot be compared to one another using this map's comparator
2260     * (or, if the map has no comparator, using natural ordering).
2261     * @throws IllegalArgumentException if <tt>fromKey</tt> is greater than
2262     * <tt>toKey</tt>.
2263     * @throws NullPointerException if <tt>fromKey</tt> or <tt>toKey</tt> is
2264     * <tt>null</tt>.
2265     */
2266     public ConcurrentNavigableMap<K,V> subMap(K fromKey, K toKey) {
2267     if (fromKey == null || toKey == null)
2268     throw new NullPointerException();
2269     return new ConcurrentSkipListSubMap(this, fromKey, toKey);
2270     }
2271    
2272     /**
2273     * Returns a view of the portion of this map whose keys are
2274     * strictly less than <tt>toKey</tt>. The returned sorted map is
2275     * backed by this map, so changes in the returned sorted map are
2276     * reflected in this map, and vice-versa.
2277     * @param toKey high endpoint (exclusive) of the headMap.
2278     * @return a view of the portion of this map whose keys are
2279     * strictly less than <tt>toKey</tt>.
2280     *
2281     * @throws ClassCastException if <tt>toKey</tt> is not compatible
2282     * with this map's comparator (or, if the map has no comparator,
2283     * if <tt>toKey</tt> does not implement <tt>Comparable</tt>).
2284     * @throws NullPointerException if <tt>toKey</tt> is <tt>null</tt>.
2285     */
2286     public ConcurrentNavigableMap<K,V> headMap(K toKey) {
2287     if (toKey == null)
2288     throw new NullPointerException();
2289     return new ConcurrentSkipListSubMap(this, null, toKey);
2290     }
2291    
2292     /**
2293     * Returns a view of the portion of this map whose keys are
2294     * greater than or equal to <tt>fromKey</tt>. The returned sorted
2295     * map is backed by this map, so changes in the returned sorted
2296     * map are reflected in this map, and vice-versa.
2297     * @param fromKey low endpoint (inclusive) of the tailMap.
2298     * @return a view of the portion of this map whose keys are
2299     * greater than or equal to <tt>fromKey</tt>.
2300     * @throws ClassCastException if <tt>fromKey</tt> is not
2301     * compatible with this map's comparator (or, if the map has no
2302     * comparator, if <tt>fromKey</tt> does not implement
2303     * <tt>Comparable</tt>).
2304     * @throws NullPointerException if <tt>fromKey</tt> is <tt>null</tt>.
2305     */
2306     public ConcurrentNavigableMap<K,V> tailMap(K fromKey) {
2307     if (fromKey == null)
2308     throw new NullPointerException();
2309     return new ConcurrentSkipListSubMap(this, fromKey, null);
2310     }
2311    
2312     /* ---------------- Relational operations -------------- */
2313    
2314     /**
2315     * Returns a key-value mapping associated with the least key
2316     * greater than or equal to the given key, or <tt>null</tt> if
2317     * there is no such entry. The returned entry does <em>not</em>
2318     * support the <tt>Entry.setValue</tt> method.
2319     *
2320     * @param key the key.
2321     * @return an Entry associated with ceiling of given key, or
2322     * <tt>null</tt> if there is no such Entry.
2323     * @throws ClassCastException if key cannot be compared with the
2324     * keys currently in the map.
2325     * @throws NullPointerException if key is <tt>null</tt>.
2326     */
2327     public Map.Entry<K,V> ceilingEntry(K key) {
2328     return getNear(key, GT|EQ);
2329     }
2330    
2331     /**
2332     * Returns least key greater than or equal to the given key, or
2333     * <tt>null</tt> if there is no such key.
2334     *
2335     * @param key the key.
2336     * @return the ceiling key, or <tt>null</tt>
2337     * if there is no such key.
2338     * @throws ClassCastException if key cannot be compared with the keys
2339     * currently in the map.
2340     * @throws NullPointerException if key is <tt>null</tt>.
2341     */
2342     public K ceilingKey(K key) {
2343     Node<K,V> n = findNear(key, GT|EQ);
2344     return (n == null)? null : n.key;
2345     }
2346    
2347     /**
2348     * Returns a key-value mapping associated with the greatest
2349     * key strictly less than the given key, or <tt>null</tt> if there is no
2350     * such entry. The returned entry does <em>not</em> support
2351     * the <tt>Entry.setValue</tt> method.
2352     *
2353     * @param key the key.
2354     * @return an Entry with greatest key less than the given
2355     * key, or <tt>null</tt> if there is no such Entry.
2356     * @throws ClassCastException if key cannot be compared with the keys
2357     * currently in the map.
2358     * @throws NullPointerException if key is <tt>null</tt>.
2359     */
2360     public Map.Entry<K,V> lowerEntry(K key) {
2361     return getNear(key, LT);
2362     }
2363    
2364     /**
2365     * Returns the greatest key strictly less than the given key, or
2366     * <tt>null</tt> if there is no such key.
2367     *
2368     * @param key the key.
2369     * @return the greatest key less than the given
2370     * key, or <tt>null</tt> if there is no such key.
2371     * @throws ClassCastException if key cannot be compared with the keys
2372     * currently in the map.
2373     * @throws NullPointerException if key is <tt>null</tt>.
2374     */
2375     public K lowerKey(K key) {
2376     Node<K,V> n = findNear(key, LT);
2377     return (n == null)? null : n.key;
2378     }
2379    
2380     /**
2381     * Returns a key-value mapping associated with the greatest key
2382     * less than or equal to the given key, or <tt>null</tt> if there
2383     * is no such entry. The returned entry does <em>not</em> support
2384     * the <tt>Entry.setValue</tt> method.
2385     *
2386     * @param key the key.
2387     * @return an Entry associated with floor of given key, or <tt>null</tt>
2388     * if there is no such Entry.
2389     * @throws ClassCastException if key cannot be compared with the keys
2390     * currently in the map.
2391     * @throws NullPointerException if key is <tt>null</tt>.
2392     */
2393     public Map.Entry<K,V> floorEntry(K key) {
2394     return getNear(key, LT|EQ);
2395     }
2396    
2397     /**
2398     * Returns the greatest key
2399     * less than or equal to the given key, or <tt>null</tt> if there
2400     * is no such key.
2401     *
2402     * @param key the key.
2403     * @return the floor of given key, or <tt>null</tt> if there is no
2404     * such key.
2405     * @throws ClassCastException if key cannot be compared with the keys
2406     * currently in the map.
2407     * @throws NullPointerException if key is <tt>null</tt>.
2408     */
2409     public K floorKey(K key) {
2410     Node<K,V> n = findNear(key, LT|EQ);
2411     return (n == null)? null : n.key;
2412     }
2413    
2414     /**
2415     * Returns a key-value mapping associated with the least key
2416     * strictly greater than the given key, or <tt>null</tt> if there
2417     * is no such entry. The returned entry does <em>not</em> support
2418     * the <tt>Entry.setValue</tt> method.
2419     *
2420     * @param key the key.
2421     * @return an Entry with least key greater than the given key, or
2422     * <tt>null</tt> if there is no such Entry.
2423     * @throws ClassCastException if key cannot be compared with the keys
2424     * currently in the map.
2425     * @throws NullPointerException if key is <tt>null</tt>.
2426     */
2427     public Map.Entry<K,V> higherEntry(K key) {
2428     return getNear(key, GT);
2429     }
2430    
2431     /**
2432     * Returns the least key strictly greater than the given key, or
2433     * <tt>null</tt> if there is no such key.
2434     *
2435     * @param key the key.
2436     * @return the least key greater than the given key, or
2437     * <tt>null</tt> if there is no such key.
2438     * @throws ClassCastException if key cannot be compared with the keys
2439     * currently in the map.
2440     * @throws NullPointerException if key is <tt>null</tt>.
2441     */
2442     public K higherKey(K key) {
2443     Node<K,V> n = findNear(key, GT);
2444     return (n == null)? null : n.key;
2445     }
2446    
2447     /**
2448     * Returns a key-value mapping associated with the least
2449     * key in this map, or <tt>null</tt> if the map is empty.
2450     * The returned entry does <em>not</em> support
2451     * the <tt>Entry.setValue</tt> method.
2452     *
2453     * @return an Entry with least key, or <tt>null</tt>
2454     * if the map is empty.
2455     */
2456     public Map.Entry<K,V> firstEntry() {
2457     for (;;) {
2458     Node<K,V> n = findFirst();
2459     if (n == null)
2460     return null;
2461     SnapshotEntry<K,V> e = n.createSnapshot();
2462     if (e != null)
2463     return e;
2464     }
2465     }
2466    
2467     /**
2468     * Returns a key-value mapping associated with the greatest
2469     * key in this map, or <tt>null</tt> if the map is empty.
2470     * The returned entry does <em>not</em> support
2471     * the <tt>Entry.setValue</tt> method.
2472     *
2473     * @return an Entry with greatest key, or <tt>null</tt>
2474     * if the map is empty.
2475     */
2476     public Map.Entry<K,V> lastEntry() {
2477     for (;;) {
2478     Node<K,V> n = findLast();
2479     if (n == null)
2480     return null;
2481     SnapshotEntry<K,V> e = n.createSnapshot();
2482     if (e != null)
2483     return e;
2484     }
2485     }
2486    
2487     /**
2488     * Removes and returns a key-value mapping associated with
2489     * the least key in this map, or <tt>null</tt> if the map is empty.
2490     * The returned entry does <em>not</em> support
2491     * the <tt>Entry.setValue</tt> method.
2492     *
2493     * @return the removed first entry of this map, or <tt>null</tt>
2494     * if the map is empty.
2495     */
2496     public Map.Entry<K,V> pollFirstEntry() {
2497     return (SnapshotEntry<K,V>)doRemoveFirst(false);
2498     }
2499    
2500     /**
2501     * Removes and returns a key-value mapping associated with
2502     * the greatest key in this map, or <tt>null</tt> if the map is empty.
2503     * The returned entry does <em>not</em> support
2504     * the <tt>Entry.setValue</tt> method.
2505     *
2506     * @return the removed last entry of this map, or <tt>null</tt>
2507     * if the map is empty.
2508     */
2509     public Map.Entry<K,V> pollLastEntry() {
2510     return (SnapshotEntry<K,V>)doRemoveLast(false);
2511     }
2512    
2513    
2514     /* ---------------- Iterators -------------- */
2515    
2516     /**
2517     * Base of ten kinds of iterator classes:
2518     * ascending: {map, submap} X {key, value, entry}
2519     * descending: {map, submap} X {key, entry}
2520     */
2521     abstract class Iter {
2522     /** the last node returned by next() */
2523     Node<K,V> last;
2524     /** the next node to return from next(); */
2525     Node<K,V> next;
2526     /** Cache of next value field to maintain weak consistency */
2527     Object nextValue;
2528    
2529     Iter() {}
2530    
2531     public final boolean hasNext() {
2532     return next != null;
2533     }
2534    
2535     /** initialize ascending iterator for entire range */
2536     final void initAscending() {
2537     for (;;) {
2538     next = findFirst();
2539     if (next == null)
2540     break;
2541     nextValue = next.value;
2542     if (nextValue != null && nextValue != next)
2543     break;
2544     }
2545     }
2546    
2547     /**
2548     * initialize ascending iterator starting at given least key,
2549     * or first node if least is <tt>null</tt>, but not greater or
2550     * equal to fence, or end if fence is <tt>null</tt>.
2551     */
2552     final void initAscending(K least, K fence) {
2553     for (;;) {
2554     next = findCeiling(least);
2555     if (next == null)
2556     break;
2557     nextValue = next.value;
2558     if (nextValue != null && nextValue != next) {
2559     if (fence != null && compare(fence, next.key) <= 0) {
2560     next = null;
2561     nextValue = null;
2562     }
2563     break;
2564     }
2565     }
2566     }
2567     /** advance next to higher entry */
2568     final void ascend() {
2569     if ((last = next) == null)
2570     throw new NoSuchElementException();
2571     for (;;) {
2572     next = next.next;
2573     if (next == null)
2574     break;
2575     nextValue = next.value;
2576     if (nextValue != null && nextValue != next)
2577     break;
2578     }
2579     }
2580    
2581     /**
2582     * Version of ascend for submaps to stop at fence
2583     */
2584     final void ascend(K fence) {
2585     if ((last = next) == null)
2586     throw new NoSuchElementException();
2587     for (;;) {
2588     next = next.next;
2589     if (next == null)
2590     break;
2591     nextValue = next.value;
2592     if (nextValue != null && nextValue != next) {
2593     if (fence != null && compare(fence, next.key) <= 0) {
2594     next = null;
2595     nextValue = null;
2596     }
2597     break;
2598     }
2599     }
2600     }
2601    
2602     /** initialize descending iterator for entire range */
2603     final void initDescending() {
2604     for (;;) {
2605     next = findLast();
2606     if (next == null)
2607     break;
2608     nextValue = next.value;
2609     if (nextValue != null && nextValue != next)
2610     break;
2611     }
2612     }
2613    
2614     /**
2615     * initialize descending iterator starting at key less
2616     * than or equal to given fence key, or
2617     * last node if fence is <tt>null</tt>, but not less than
2618     * least, or beginning if lest is <tt>null</tt>.
2619     */
2620     final void initDescending(K least, K fence) {
2621     for (;;) {
2622     next = findLower(fence);
2623     if (next == null)
2624     break;
2625     nextValue = next.value;
2626     if (nextValue != null && nextValue != next) {
2627     if (least != null && compare(least, next.key) > 0) {
2628     next = null;
2629     nextValue = null;
2630     }
2631     break;
2632     }
2633     }
2634     }
2635    
2636     /** advance next to lower entry */
2637     final void descend() {
2638     if ((last = next) == null)
2639     throw new NoSuchElementException();
2640     K k = last.key;
2641     for (;;) {
2642     next = findNear(k, LT);
2643     if (next == null)
2644     break;
2645     nextValue = next.value;
2646     if (nextValue != null && nextValue != next)
2647     break;
2648     }
2649     }
2650    
2651     /**
2652     * Version of descend for submaps to stop at least
2653     */
2654     final void descend(K least) {
2655     if ((last = next) == null)
2656     throw new NoSuchElementException();
2657     K k = last.key;
2658     for (;;) {
2659     next = findNear(k, LT);
2660     if (next == null)
2661     break;
2662     nextValue = next.value;
2663     if (nextValue != null && nextValue != next) {
2664     if (least != null && compare(least, next.key) > 0) {
2665     next = null;
2666     nextValue = null;
2667     }
2668     break;
2669     }
2670     }
2671     }
2672    
2673     public void remove() {
2674     Node<K,V> l = last;
2675     if (l == null)
2676     throw new IllegalStateException();
2677     // It would not be worth all of the overhead to directly
2678     // unlink from here. Using remove is fast enough.
2679     ConcurrentSkipListMap.this.remove(l.key);
2680     }
2681    
2682     }
2683    
2684     final class ValueIterator extends Iter implements Iterator<V> {
2685     ValueIterator() {
2686     initAscending();
2687     }
2688     public V next() {
2689     Object v = nextValue;
2690     ascend();
2691     return (V)v;
2692     }
2693     }
2694    
2695     final class KeyIterator extends Iter implements Iterator<K> {
2696     KeyIterator() {
2697     initAscending();
2698     }
2699     public K next() {
2700     Node<K,V> n = next;
2701     ascend();
2702     return n.key;
2703     }
2704     }
2705    
2706     class SubMapValueIterator extends Iter implements Iterator<V> {
2707     final K fence;
2708     SubMapValueIterator(K least, K fence) {
2709     initAscending(least, fence);
2710     this.fence = fence;
2711     }
2712    
2713     public V next() {
2714     Object v = nextValue;
2715     ascend(fence);
2716     return (V)v;
2717     }
2718     }
2719    
2720     final class SubMapKeyIterator extends Iter implements Iterator<K> {
2721     final K fence;
2722     SubMapKeyIterator(K least, K fence) {
2723     initAscending(least, fence);
2724     this.fence = fence;
2725     }
2726    
2727     public K next() {
2728     Node<K,V> n = next;
2729     ascend(fence);
2730     return n.key;
2731     }
2732     }
2733    
2734     final class DescendingKeyIterator extends Iter implements Iterator<K> {
2735     DescendingKeyIterator() {
2736     initDescending();
2737     }
2738     public K next() {
2739     Node<K,V> n = next;
2740     descend();
2741     return n.key;
2742     }
2743     }
2744    
2745     final class DescendingSubMapKeyIterator extends Iter implements Iterator<K> {
2746     final K least;
2747     DescendingSubMapKeyIterator(K least, K fence) {
2748     initDescending(least, fence);
2749     this.least = least;
2750     }
2751    
2752     public K next() {
2753     Node<K,V> n = next;
2754     descend(least);
2755     return n.key;
2756     }
2757     }
2758    
2759     /**
2760     * Entry iterators use the same trick as in ConcurrentHashMap and
2761     * elsewhere of using the iterator itself to represent entries,
2762     * thus avoiding having to create entry objects in next().
2763     */
2764     abstract class EntryIter extends Iter implements Map.Entry<K,V> {
2765     /** Cache of last value returned */
2766     Object lastValue;
2767    
2768     EntryIter() {
2769     }
2770    
2771     public K getKey() {
2772     Node<K,V> l = last;
2773     if (l == null)
2774     throw new IllegalStateException();
2775     return l.key;
2776     }
2777    
2778     public V getValue() {
2779     Object v = lastValue;
2780     if (last == null || v == null)
2781     throw new IllegalStateException();
2782     return (V)v;
2783     }
2784    
2785     public V setValue(V value) {
2786     throw new UnsupportedOperationException();
2787     }
2788    
2789     public boolean equals(Object o) {
2790     // If not acting as entry, just use default.
2791     if (last == null)
2792     return super.equals(o);
2793     if (!(o instanceof Map.Entry))
2794     return false;
2795     Map.Entry e = (Map.Entry)o;
2796     return (getKey().equals(e.getKey()) &&
2797     getValue().equals(e.getValue()));
2798     }
2799    
2800     public int hashCode() {
2801     // If not acting as entry, just use default.
2802     if (last == null)
2803     return super.hashCode();
2804     return getKey().hashCode() ^ getValue().hashCode();
2805     }
2806    
2807     public String toString() {
2808     // If not acting as entry, just use default.
2809     if (last == null)
2810     return super.toString();
2811     return getKey() + "=" + getValue();
2812     }
2813     }
2814    
2815     final class EntryIterator extends EntryIter
2816     implements Iterator<Map.Entry<K,V>> {
2817     EntryIterator() {
2818     initAscending();
2819     }
2820     public Map.Entry<K,V> next() {
2821     lastValue = nextValue;
2822     ascend();
2823     return this;
2824     }
2825     }
2826    
2827     final class SubMapEntryIterator extends EntryIter
2828     implements Iterator<Map.Entry<K,V>> {
2829     final K fence;
2830     SubMapEntryIterator(K least, K fence) {
2831     initAscending(least, fence);
2832     this.fence = fence;
2833     }
2834    
2835     public Map.Entry<K,V> next() {
2836     lastValue = nextValue;
2837     ascend(fence);
2838     return this;
2839     }
2840     }
2841    
2842     final class DescendingEntryIterator extends EntryIter
2843     implements Iterator<Map.Entry<K,V>> {
2844     DescendingEntryIterator() {
2845     initDescending();
2846     }
2847     public Map.Entry<K,V> next() {
2848     lastValue = nextValue;
2849     descend();
2850     return this;
2851     }
2852     }
2853    
2854     final class DescendingSubMapEntryIterator extends EntryIter
2855     implements Iterator<Map.Entry<K,V>> {
2856     final K least;
2857     DescendingSubMapEntryIterator(K least, K fence) {
2858     initDescending(least, fence);
2859     this.least = least;
2860     }
2861    
2862     public Map.Entry<K,V> next() {
2863     lastValue = nextValue;
2864     descend(least);
2865     return this;
2866     }
2867     }
2868    
2869     // Factory methods for iterators needed by submaps and/or
2870     // ConcurrentSkipListSet
2871    
2872     Iterator<K> keyIterator() {
2873     return new KeyIterator();
2874     }
2875    
2876     Iterator<K> descendingKeyIterator() {
2877     return new DescendingKeyIterator();
2878     }
2879    
2880     SubMapEntryIterator subMapEntryIterator(K least, K fence) {
2881     return new SubMapEntryIterator(least, fence);
2882     }
2883    
2884     DescendingSubMapEntryIterator descendingSubMapEntryIterator(K least, K fence) {
2885     return new DescendingSubMapEntryIterator(least, fence);
2886     }
2887    
2888     SubMapKeyIterator subMapKeyIterator(K least, K fence) {
2889     return new SubMapKeyIterator(least, fence);
2890     }
2891    
2892     DescendingSubMapKeyIterator descendingSubMapKeyIterator(K least, K fence) {
2893     return new DescendingSubMapKeyIterator(least, fence);
2894     }
2895    
2896     SubMapValueIterator subMapValueIterator(K least, K fence) {
2897     return new SubMapValueIterator(least, fence);
2898     }
2899    
2900     /* ---------------- Views -------------- */
2901    
2902     class KeySet extends AbstractSet<K> {
2903     public Iterator<K> iterator() {
2904     return new KeyIterator();
2905     }
2906     public boolean isEmpty() {
2907     return ConcurrentSkipListMap.this.isEmpty();
2908     }
2909     public int size() {
2910     return ConcurrentSkipListMap.this.size();
2911     }
2912     public boolean contains(Object o) {
2913     return ConcurrentSkipListMap.this.containsKey(o);
2914     }
2915     public boolean remove(Object o) {
2916     return ConcurrentSkipListMap.this.removep(o);
2917     }
2918     public void clear() {
2919     ConcurrentSkipListMap.this.clear();
2920     }
2921     public Object[] toArray() {
2922     Collection<K> c = new ArrayList<K>();
2923     for (Iterator<K> i = iterator(); i.hasNext(); )
2924     c.add(i.next());
2925     return c.toArray();
2926     }
2927     public <T> T[] toArray(T[] a) {
2928     Collection<K> c = new ArrayList<K>();
2929     for (Iterator<K> i = iterator(); i.hasNext(); )
2930     c.add(i.next());
2931     return c.toArray(a);
2932     }
2933     }
2934    
2935     class DescendingKeySet extends KeySet {
2936     public Iterator<K> iterator() {
2937     return new DescendingKeyIterator();
2938     }
2939     }
2940    
2941     final class Values extends AbstractCollection<V> {
2942     public Iterator<V> iterator() {
2943     return new ValueIterator();
2944     }
2945     public boolean isEmpty() {
2946     return ConcurrentSkipListMap.this.isEmpty();
2947     }
2948     public int size() {
2949     return ConcurrentSkipListMap.this.size();
2950     }
2951     public boolean contains(Object o) {
2952     return ConcurrentSkipListMap.this.containsValue(o);
2953     }
2954     public void clear() {
2955     ConcurrentSkipListMap.this.clear();
2956     }
2957     public Object[] toArray() {
2958     Collection<V> c = new ArrayList<V>();
2959     for (Iterator<V> i = iterator(); i.hasNext(); )
2960     c.add(i.next());
2961     return c.toArray();
2962     }
2963     public <T> T[] toArray(T[] a) {
2964     Collection<V> c = new ArrayList<V>();
2965     for (Iterator<V> i = iterator(); i.hasNext(); )
2966     c.add(i.next());
2967     return c.toArray(a);
2968     }
2969     }
2970    
2971     class EntrySet extends AbstractSet<Map.Entry<K,V>> {
2972     public Iterator<Map.Entry<K,V>> iterator() {
2973     return new EntryIterator();
2974     }
2975     public boolean contains(Object o) {
2976     if (!(o instanceof Map.Entry))
2977     return false;
2978     Map.Entry<K,V> e = (Map.Entry<K,V>)o;
2979     V v = ConcurrentSkipListMap.this.get(e.getKey());
2980     return v != null && v.equals(e.getValue());
2981     }
2982     public boolean remove(Object o) {
2983     if (!(o instanceof Map.Entry))
2984     return false;
2985     Map.Entry<K,V> e = (Map.Entry<K,V>)o;
2986     return ConcurrentSkipListMap.this.remove(e.getKey(),
2987     e.getValue());
2988     }
2989     public boolean isEmpty() {
2990     return ConcurrentSkipListMap.this.isEmpty();
2991     }
2992     public int size() {
2993     return ConcurrentSkipListMap.this.size();
2994     }
2995     public void clear() {
2996     ConcurrentSkipListMap.this.clear();
2997     }
2998    
2999     public Object[] toArray() {
3000     Collection<Map.Entry<K,V>> c = new ArrayList<Map.Entry<K,V>>();
3001     for (Map.Entry e : this)
3002     c.add(new SnapshotEntry(e.getKey(), e.getValue()));
3003     return c.toArray();
3004     }
3005     public <T> T[] toArray(T[] a) {
3006     Collection<Map.Entry<K,V>> c = new ArrayList<Map.Entry<K,V>>();
3007     for (Map.Entry e : this)
3008     c.add(new SnapshotEntry(e.getKey(), e.getValue()));
3009     return c.toArray(a);
3010     }
3011     }
3012    
3013     class DescendingEntrySet extends EntrySet {
3014     public Iterator<Map.Entry<K,V>> iterator() {
3015     return new DescendingEntryIterator();
3016     }
3017     }
3018    
3019     /**
3020     * Submaps returned by {@link ConcurrentSkipListMap} submap operations
3021     * represent a subrange of mappings of their underlying
3022     * maps. Instances of this class support all methods of their
3023     * underlying maps, differing in that mappings outside their range are
3024     * ignored, and attempts to add mappings outside their ranges result
3025     * in {@link IllegalArgumentException}. Instances of this class are
3026     * constructed only using the <tt>subMap</tt>, <tt>headMap</tt>, and
3027     * <tt>tailMap</tt> methods of their underlying maps.
3028     */
3029     static class ConcurrentSkipListSubMap<K,V> extends AbstractMap<K,V>
3030     implements ConcurrentNavigableMap<K,V>, java.io.Serializable {
3031    
3032     private static final long serialVersionUID = -7647078645895051609L;
3033    
3034     /** Underlying map */
3035     private final ConcurrentSkipListMap<K,V> m;
3036     /** lower bound key, or null if from start */
3037     private final K least;
3038     /** upper fence key, or null if to end */
3039     private final K fence;
3040     // Lazily initialized view holders
3041     private transient Set<K> keySetView;
3042     private transient Set<Map.Entry<K,V>> entrySetView;
3043     private transient Collection<V> valuesView;
3044     private transient Set<K> descendingKeySetView;
3045     private transient Set<Map.Entry<K,V>> descendingEntrySetView;
3046    
3047     /**
3048     * Creates a new submap.
3049     * @param least inclusive least value, or <tt>null</tt> if from start
3050     * @param fence exclusive upper bound or <tt>null</tt> if to end
3051     * @throws IllegalArgumentException if least and fence nonnull
3052     * and least greater than fence
3053     */
3054     ConcurrentSkipListSubMap(ConcurrentSkipListMap<K,V> map,
3055     K least, K fence) {
3056     if (least != null &&
3057     fence != null &&
3058     map.compare(least, fence) > 0)
3059     throw new IllegalArgumentException("inconsistent range");
3060     this.m = map;
3061     this.least = least;
3062     this.fence = fence;
3063     }
3064    
3065     /* ---------------- Utilities -------------- */
3066    
3067     boolean inHalfOpenRange(K key) {
3068     return m.inHalfOpenRange(key, least, fence);
3069     }
3070    
3071     boolean inOpenRange(K key) {
3072     return m.inOpenRange(key, least, fence);
3073     }
3074    
3075     ConcurrentSkipListMap.Node<K,V> firstNode() {
3076     return m.findCeiling(least);
3077     }
3078    
3079     ConcurrentSkipListMap.Node<K,V> lastNode() {
3080     return m.findLower(fence);
3081     }
3082    
3083     boolean isBeforeEnd(ConcurrentSkipListMap.Node<K,V> n) {
3084     return (n != null &&
3085     (fence == null ||
3086     n.key == null || // pass by markers and headers
3087     m.compare(fence, n.key) > 0));
3088     }
3089    
3090     void checkKey(K key) throws IllegalArgumentException {
3091     if (!inHalfOpenRange(key))
3092     throw new IllegalArgumentException("key out of range");
3093     }
3094    
3095     /**
3096     * Returns underlying map. Needed by ConcurrentSkipListSet
3097     * @return the backing map
3098     */
3099     ConcurrentSkipListMap<K,V> getMap() {
3100     return m;
3101     }
3102    
3103     /**
3104     * Returns least key. Needed by ConcurrentSkipListSet
3105     * @return least key or <tt>null</tt> if from start
3106     */
3107     K getLeast() {
3108     return least;
3109     }
3110    
3111     /**
3112     * Returns fence key. Needed by ConcurrentSkipListSet
3113     * @return fence key or <tt>null</tt> of to end
3114     */
3115     K getFence() {
3116     return fence;
3117     }
3118    
3119    
3120     /* ---------------- Map API methods -------------- */
3121    
3122     public boolean containsKey(Object key) {
3123     K k = (K)key;
3124     return inHalfOpenRange(k) && m.containsKey(k);
3125     }
3126    
3127     public V get(Object key) {
3128     K k = (K)key;
3129     return ((!inHalfOpenRange(k)) ? null : m.get(k));
3130     }
3131    
3132     public V put(K key, V value) {
3133     checkKey(key);
3134     return m.put(key, value);
3135     }
3136    
3137     public V remove(Object key) {
3138     K k = (K)key;
3139     return (!inHalfOpenRange(k))? null : m.remove(k);
3140     }
3141    
3142     public int size() {
3143     long count = 0;
3144     for (ConcurrentSkipListMap.Node<K,V> n = firstNode();
3145     isBeforeEnd(n);
3146     n = n.next) {
3147     if (n.getValidValue() != null)
3148     ++count;
3149     }
3150     return count >= Integer.MAX_VALUE? Integer.MAX_VALUE : (int)count;
3151     }
3152    
3153     public boolean isEmpty() {
3154     return !isBeforeEnd(firstNode());
3155     }
3156    
3157     public boolean containsValue(Object value) {
3158     if (value == null)
3159     throw new NullPointerException();
3160     for (ConcurrentSkipListMap.Node<K,V> n = firstNode();
3161     isBeforeEnd(n);
3162     n = n.next) {
3163     V v = n.getValidValue();
3164     if (v != null && value.equals(v))
3165     return true;
3166     }
3167     return false;
3168     }
3169    
3170     public void clear() {
3171     for (ConcurrentSkipListMap.Node<K,V> n = firstNode();
3172     isBeforeEnd(n);
3173     n = n.next) {
3174     if (n.getValidValue() != null)
3175     m.remove(n.key);
3176     }
3177     }
3178    
3179     /* ---------------- ConcurrentMap API methods -------------- */
3180    
3181     public V putIfAbsent(K key, V value) {
3182     checkKey(key);
3183     return m.putIfAbsent(key, value);
3184     }
3185    
3186     public boolean remove(Object key, Object value) {
3187     K k = (K)key;
3188     return inHalfOpenRange(k) && m.remove(k, value);
3189     }
3190    
3191     public boolean replace(K key, V oldValue, V newValue) {
3192     checkKey(key);
3193     return m.replace(key, oldValue, newValue);
3194     }
3195    
3196     public V replace(K key, V value) {
3197     checkKey(key);
3198     return m.replace(key, value);
3199     }
3200    
3201     /* ---------------- SortedMap API methods -------------- */
3202    
3203     public Comparator<? super K> comparator() {
3204     return m.comparator();
3205     }
3206    
3207     public K firstKey() {
3208     ConcurrentSkipListMap.Node<K,V> n = firstNode();
3209     if (isBeforeEnd(n))
3210     return n.key;
3211     else
3212     throw new NoSuchElementException();
3213     }
3214    
3215     public K lastKey() {
3216     ConcurrentSkipListMap.Node<K,V> n = lastNode();
3217     if (n != null) {
3218     K last = n.key;
3219     if (inHalfOpenRange(last))
3220     return last;
3221     }
3222     throw new NoSuchElementException();
3223     }
3224    
3225     public ConcurrentNavigableMap<K,V> subMap(K fromKey, K toKey) {
3226     if (fromKey == null || toKey == null)
3227     throw new NullPointerException();
3228     if (!inOpenRange(fromKey) || !inOpenRange(toKey))
3229     throw new IllegalArgumentException("key out of range");
3230     return new ConcurrentSkipListSubMap(m, fromKey, toKey);
3231     }
3232    
3233     public ConcurrentNavigableMap<K,V> headMap(K toKey) {
3234     if (toKey == null)
3235     throw new NullPointerException();
3236     if (!inOpenRange(toKey))
3237     throw new IllegalArgumentException("key out of range");
3238     return new ConcurrentSkipListSubMap(m, least, toKey);
3239     }
3240    
3241     public ConcurrentNavigableMap<K,V> tailMap(K fromKey) {
3242     if (fromKey == null)
3243     throw new NullPointerException();
3244     if (!inOpenRange(fromKey))
3245     throw new IllegalArgumentException("key out of range");
3246     return new ConcurrentSkipListSubMap(m, fromKey, fence);
3247     }
3248    
3249     /* ---------------- Relational methods -------------- */
3250    
3251     public Map.Entry<K,V> ceilingEntry(K key) {
3252     return (SnapshotEntry<K,V>)
3253     m.getNear(key, m.GT|m.EQ, least, fence, false);
3254     }
3255    
3256     public K ceilingKey(K key) {
3257     return (K)
3258     m.getNear(key, m.GT|m.EQ, least, fence, true);
3259     }
3260    
3261     public Map.Entry<K,V> lowerEntry(K key) {
3262     return (SnapshotEntry<K,V>)
3263     m.getNear(key, m.LT, least, fence, false);
3264     }
3265    
3266     public K lowerKey(K key) {
3267     return (K)
3268     m.getNear(key, m.LT, least, fence, true);
3269     }
3270    
3271     public Map.Entry<K,V> floorEntry(K key) {
3272     return (SnapshotEntry<K,V>)
3273     m.getNear(key, m.LT|m.EQ, least, fence, false);
3274     }
3275    
3276     public K floorKey(K key) {
3277     return (K)
3278     m.getNear(key, m.LT|m.EQ, least, fence, true);
3279     }
3280    
3281    
3282     public Map.Entry<K,V> higherEntry(K key) {
3283     return (SnapshotEntry<K,V>)
3284     m.getNear(key, m.GT, least, fence, false);
3285     }
3286    
3287     public K higherKey(K key) {
3288     return (K)
3289     m.getNear(key, m.GT, least, fence, true);
3290     }
3291    
3292     public Map.Entry<K,V> firstEntry() {
3293     for (;;) {
3294     ConcurrentSkipListMap.Node<K,V> n = firstNode();
3295     if (!isBeforeEnd(n))
3296     return null;
3297     Map.Entry<K,V> e = n.createSnapshot();
3298     if (e != null)
3299     return e;
3300     }
3301     }
3302    
3303     public Map.Entry<K,V> lastEntry() {
3304     for (;;) {
3305     ConcurrentSkipListMap.Node<K,V> n = lastNode();
3306     if (n == null || !inHalfOpenRange(n.key))
3307     return null;
3308     Map.Entry<K,V> e = n.createSnapshot();
3309     if (e != null)
3310     return e;
3311     }
3312     }
3313    
3314     public Map.Entry<K,V> pollFirstEntry() {
3315     return (SnapshotEntry<K,V>)
3316     m.removeFirstEntryOfSubrange(least, fence, false);
3317     }
3318    
3319     public Map.Entry<K,V> pollLastEntry() {
3320     return (SnapshotEntry<K,V>)
3321     m.removeLastEntryOfSubrange(least, fence, false);
3322     }
3323    
3324     /* ---------------- Submap Views -------------- */
3325    
3326     public Set<K> keySet() {
3327     Set<K> ks = keySetView;
3328     return (ks != null) ? ks : (keySetView = new KeySetView());
3329     }
3330    
3331     class KeySetView extends AbstractSet<K> {
3332     public Iterator<K> iterator() {
3333     return m.subMapKeyIterator(least, fence);
3334     }
3335     public int size() {
3336     return ConcurrentSkipListSubMap.this.size();
3337     }
3338     public boolean isEmpty() {
3339     return ConcurrentSkipListSubMap.this.isEmpty();
3340     }
3341     public boolean contains(Object k) {
3342     return ConcurrentSkipListSubMap.this.containsKey(k);
3343     }
3344     public Object[] toArray() {
3345     Collection<K> c = new ArrayList<K>();
3346     for (Iterator<K> i = iterator(); i.hasNext(); )
3347     c.add(i.next());
3348     return c.toArray();
3349     }
3350     public <T> T[] toArray(T[] a) {
3351     Collection<K> c = new ArrayList<K>();
3352     for (Iterator<K> i = iterator(); i.hasNext(); )
3353     c.add(i.next());
3354     return c.toArray(a);
3355     }
3356     }
3357    
3358     public Set<K> descendingKeySet() {
3359     Set<K> ks = descendingKeySetView;
3360     return (ks != null) ? ks : (descendingKeySetView = new DescendingKeySetView());
3361     }
3362    
3363     class DescendingKeySetView extends KeySetView {
3364     public Iterator<K> iterator() {
3365     return m.descendingSubMapKeyIterator(least, fence);
3366     }
3367     }
3368    
3369     public Collection<V> values() {
3370     Collection<V> vs = valuesView;
3371     return (vs != null) ? vs : (valuesView = new ValuesView());
3372     }
3373    
3374     class ValuesView extends AbstractCollection<V> {
3375     public Iterator<V> iterator() {
3376     return m.subMapValueIterator(least, fence);
3377     }
3378     public int size() {
3379     return ConcurrentSkipListSubMap.this.size();
3380     }
3381     public boolean isEmpty() {
3382     return ConcurrentSkipListSubMap.this.isEmpty();
3383     }
3384     public boolean contains(Object v) {
3385     return ConcurrentSkipListSubMap.this.containsValue(v);
3386     }
3387     public Object[] toArray() {
3388     Collection<V> c = new ArrayList<V>();
3389     for (Iterator<V> i = iterator(); i.hasNext(); )
3390     c.add(i.next());
3391     return c.toArray();
3392     }
3393     public <T> T[] toArray(T[] a) {
3394     Collection<V> c = new ArrayList<V>();
3395     for (Iterator<V> i = iterator(); i.hasNext(); )
3396     c.add(i.next());
3397     return c.toArray(a);
3398     }
3399     }
3400    
3401     public Set<Map.Entry<K,V>> entrySet() {
3402     Set<Map.Entry<K,V>> es = entrySetView;
3403     return (es != null) ? es : (entrySetView = new EntrySetView());
3404     }
3405    
3406     class EntrySetView extends AbstractSet<Map.Entry<K,V>> {
3407     public Iterator<Map.Entry<K,V>> iterator() {
3408     return m.subMapEntryIterator(least, fence);
3409     }
3410     public int size() {
3411     return ConcurrentSkipListSubMap.this.size();
3412     }
3413     public boolean isEmpty() {
3414     return ConcurrentSkipListSubMap.this.isEmpty();
3415     }
3416     public boolean contains(Object o) {
3417     if (!(o instanceof Map.Entry))
3418     return false;
3419     Map.Entry<K,V> e = (Map.Entry<K,V>) o;
3420     K key = e.getKey();
3421     if (!inHalfOpenRange(key))
3422     return false;
3423     V v = m.get(key);
3424     return v != null && v.equals(e.getValue());
3425     }
3426     public boolean remove(Object o) {
3427     if (!(o instanceof Map.Entry))
3428     return false;
3429     Map.Entry<K,V> e = (Map.Entry<K,V>) o;
3430     K key = e.getKey();
3431     if (!inHalfOpenRange(key))
3432     return false;
3433     return m.remove(key, e.getValue());
3434     }
3435     public Object[] toArray() {
3436     Collection<Map.Entry<K,V>> c = new ArrayList<Map.Entry<K,V>>();
3437     for (Map.Entry e : this)
3438     c.add(new SnapshotEntry(e.getKey(), e.getValue()));
3439     return c.toArray();
3440     }
3441     public <T> T[] toArray(T[] a) {
3442     Collection<Map.Entry<K,V>> c = new ArrayList<Map.Entry<K,V>>();
3443     for (Map.Entry e : this)
3444     c.add(new SnapshotEntry(e.getKey(), e.getValue()));
3445     return c.toArray(a);
3446     }
3447     }
3448    
3449     public Set<Map.Entry<K,V>> descendingEntrySet() {
3450     Set<Map.Entry<K,V>> es = descendingEntrySetView;
3451     return (es != null) ? es : (descendingEntrySetView = new DescendingEntrySetView());
3452     }
3453    
3454     class DescendingEntrySetView extends EntrySetView {
3455     public Iterator<Map.Entry<K,V>> iterator() {
3456     return m.descendingSubMapEntryIterator(least, fence);
3457     }
3458     }
3459     }
3460     }