ViewVC Help
View File | Revision Log | Show Annotations | Download File | Root Listing
root/jsr166/jsr166/src/main/java/util/concurrent/ConcurrentSkipListMap.java
Revision: 1.63
Committed: Sat Dec 4 17:44:20 2010 UTC (13 years, 6 months ago) by dl
Branch: MAIN
Changes since 1.62: +1 -57 lines
Log Message:
Remove broken bypass

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 jsr166 1.22 * A scalable concurrent {@link ConcurrentNavigableMap} implementation.
13     * The map is sorted according to the {@linkplain Comparable natural
14     * ordering} of its keys, or by a {@link Comparator} provided at map
15     * creation time, depending on which constructor is used.
16 dl 1.1 *
17     * <p>This class implements a concurrent variant of <a
18     * href="http://www.cs.umd.edu/~pugh/">SkipLists</a> providing
19     * expected average <i>log(n)</i> time cost for the
20     * <tt>containsKey</tt>, <tt>get</tt>, <tt>put</tt> and
21     * <tt>remove</tt> operations and their variants. Insertion, removal,
22     * update, and access operations safely execute concurrently by
23 jsr166 1.51 * multiple threads. Iterators are <i>weakly consistent</i>, returning
24 dl 1.1 * elements reflecting the state of the map at some point at or since
25     * the creation of the iterator. They do <em>not</em> throw {@link
26     * ConcurrentModificationException}, and may proceed concurrently with
27     * other operations. Ascending key ordered views and their iterators
28     * are faster than descending ones.
29     *
30     * <p>All <tt>Map.Entry</tt> pairs returned by methods in this class
31     * and its views represent snapshots of mappings at the time they were
32     * produced. They do <em>not</em> support the <tt>Entry.setValue</tt>
33     * method. (Note however that it is possible to change mappings in the
34     * associated map using <tt>put</tt>, <tt>putIfAbsent</tt>, or
35     * <tt>replace</tt>, depending on exactly which effect you need.)
36     *
37     * <p>Beware that, unlike in most collections, the <tt>size</tt>
38     * method is <em>not</em> a constant-time operation. Because of the
39     * asynchronous nature of these maps, determining the current number
40     * of elements requires a traversal of the elements. Additionally,
41     * the bulk operations <tt>putAll</tt>, <tt>equals</tt>, and
42     * <tt>clear</tt> are <em>not</em> guaranteed to be performed
43     * atomically. For example, an iterator operating concurrently with a
44     * <tt>putAll</tt> operation might view only some of the added
45     * elements.
46     *
47     * <p>This class and its views and iterators implement all of the
48     * <em>optional</em> methods of the {@link Map} and {@link Iterator}
49     * interfaces. Like most other concurrent collections, this class does
50 jsr166 1.22 * <em>not</em> permit the use of <tt>null</tt> keys or values because some
51     * null return values cannot be reliably distinguished from the absence of
52     * elements.
53 dl 1.1 *
54 jsr166 1.21 * <p>This class is a member of the
55 jsr166 1.51 * <a href="{@docRoot}/../technotes/guides/collections/index.html">
56 jsr166 1.21 * Java Collections Framework</a>.
57     *
58 dl 1.1 * @author Doug Lea
59     * @param <K> the type of keys maintained by this map
60 dl 1.9 * @param <V> the type of mapped values
61 jsr166 1.20 * @since 1.6
62 dl 1.1 */
63 dl 1.9 public class ConcurrentSkipListMap<K,V> extends AbstractMap<K,V>
64 dl 1.1 implements ConcurrentNavigableMap<K,V>,
65 dl 1.9 Cloneable,
66 dl 1.1 java.io.Serializable {
67     /*
68     * This class implements a tree-like two-dimensionally linked skip
69     * list in which the index levels are represented in separate
70     * nodes from the base nodes holding data. There are two reasons
71     * for taking this approach instead of the usual array-based
72     * structure: 1) Array based implementations seem to encounter
73     * more complexity and overhead 2) We can use cheaper algorithms
74     * for the heavily-traversed index lists than can be used for the
75     * base lists. Here's a picture of some of the basics for a
76     * possible list with 2 levels of index:
77     *
78     * Head nodes Index nodes
79 dl 1.9 * +-+ right +-+ +-+
80 dl 1.1 * |2|---------------->| |--------------------->| |->null
81 dl 1.9 * +-+ +-+ +-+
82 dl 1.1 * | down | |
83     * v v v
84 dl 1.9 * +-+ +-+ +-+ +-+ +-+ +-+
85 dl 1.1 * |1|----------->| |->| |------>| |----------->| |------>| |->null
86 dl 1.9 * +-+ +-+ +-+ +-+ +-+ +-+
87 dl 1.1 * v | | | | |
88     * Nodes next v v v v v
89 dl 1.9 * +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+
90 dl 1.1 * | |->|A|->|B|->|C|->|D|->|E|->|F|->|G|->|H|->|I|->|J|->|K|->null
91 dl 1.9 * +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+
92 dl 1.1 *
93     * The base lists use a variant of the HM linked ordered set
94     * algorithm. See Tim Harris, "A pragmatic implementation of
95     * non-blocking linked lists"
96     * http://www.cl.cam.ac.uk/~tlh20/publications.html and Maged
97     * Michael "High Performance Dynamic Lock-Free Hash Tables and
98     * List-Based Sets"
99     * http://www.research.ibm.com/people/m/michael/pubs.htm. The
100     * basic idea in these lists is to mark the "next" pointers of
101     * deleted nodes when deleting to avoid conflicts with concurrent
102     * insertions, and when traversing to keep track of triples
103     * (predecessor, node, successor) in order to detect when and how
104     * to unlink these deleted nodes.
105     *
106     * Rather than using mark-bits to mark list deletions (which can
107     * be slow and space-intensive using AtomicMarkedReference), nodes
108     * use direct CAS'able next pointers. On deletion, instead of
109     * marking a pointer, they splice in another node that can be
110     * thought of as standing for a marked pointer (indicating this by
111     * using otherwise impossible field values). Using plain nodes
112     * acts roughly like "boxed" implementations of marked pointers,
113     * but uses new nodes only when nodes are deleted, not for every
114     * link. This requires less space and supports faster
115     * traversal. Even if marked references were better supported by
116     * JVMs, traversal using this technique might still be faster
117     * because any search need only read ahead one more node than
118     * otherwise required (to check for trailing marker) rather than
119     * unmasking mark bits or whatever on each read.
120     *
121     * This approach maintains the essential property needed in the HM
122     * algorithm of changing the next-pointer of a deleted node so
123     * that any other CAS of it will fail, but implements the idea by
124     * changing the pointer to point to a different node, not by
125     * marking it. While it would be possible to further squeeze
126     * space by defining marker nodes not to have key/value fields, it
127     * isn't worth the extra type-testing overhead. The deletion
128     * markers are rarely encountered during traversal and are
129     * normally quickly garbage collected. (Note that this technique
130     * would not work well in systems without garbage collection.)
131     *
132     * In addition to using deletion markers, the lists also use
133     * nullness of value fields to indicate deletion, in a style
134     * similar to typical lazy-deletion schemes. If a node's value is
135     * null, then it is considered logically deleted and ignored even
136     * though it is still reachable. This maintains proper control of
137     * concurrent replace vs delete operations -- an attempted replace
138     * must fail if a delete beat it by nulling field, and a delete
139     * must return the last non-null value held in the field. (Note:
140     * Null, rather than some special marker, is used for value fields
141     * here because it just so happens to mesh with the Map API
142     * requirement that method get returns null if there is no
143     * mapping, which allows nodes to remain concurrently readable
144     * even when deleted. Using any other marker value here would be
145     * messy at best.)
146     *
147     * Here's the sequence of events for a deletion of node n with
148     * predecessor b and successor f, initially:
149     *
150 dl 1.9 * +------+ +------+ +------+
151 dl 1.1 * ... | b |------>| n |----->| f | ...
152 dl 1.9 * +------+ +------+ +------+
153 dl 1.1 *
154     * 1. CAS n's value field from non-null to null.
155     * From this point on, no public operations encountering
156     * the node consider this mapping to exist. However, other
157     * ongoing insertions and deletions might still modify
158     * n's next pointer.
159     *
160     * 2. CAS n's next pointer to point to a new marker node.
161     * From this point on, no other nodes can be appended to n.
162     * which avoids deletion errors in CAS-based linked lists.
163     *
164     * +------+ +------+ +------+ +------+
165     * ... | b |------>| n |----->|marker|------>| f | ...
166 dl 1.9 * +------+ +------+ +------+ +------+
167 dl 1.1 *
168     * 3. CAS b's next pointer over both n and its marker.
169     * From this point on, no new traversals will encounter n,
170     * and it can eventually be GCed.
171     * +------+ +------+
172     * ... | b |----------------------------------->| f | ...
173 dl 1.9 * +------+ +------+
174     *
175 dl 1.1 * A failure at step 1 leads to simple retry due to a lost race
176     * with another operation. Steps 2-3 can fail because some other
177     * thread noticed during a traversal a node with null value and
178     * helped out by marking and/or unlinking. This helping-out
179     * ensures that no thread can become stuck waiting for progress of
180     * the deleting thread. The use of marker nodes slightly
181     * complicates helping-out code because traversals must track
182     * consistent reads of up to four nodes (b, n, marker, f), not
183     * just (b, n, f), although the next field of a marker is
184     * immutable, and once a next field is CAS'ed to point to a
185     * marker, it never again changes, so this requires less care.
186     *
187     * Skip lists add indexing to this scheme, so that the base-level
188     * traversals start close to the locations being found, inserted
189     * or deleted -- usually base level traversals only traverse a few
190     * nodes. This doesn't change the basic algorithm except for the
191     * need to make sure base traversals start at predecessors (here,
192     * b) that are not (structurally) deleted, otherwise retrying
193 dl 1.9 * after processing the deletion.
194 dl 1.1 *
195     * Index levels are maintained as lists with volatile next fields,
196     * using CAS to link and unlink. Races are allowed in index-list
197     * operations that can (rarely) fail to link in a new index node
198     * or delete one. (We can't do this of course for data nodes.)
199     * However, even when this happens, the index lists remain sorted,
200     * so correctly serve as indices. This can impact performance,
201     * but since skip lists are probabilistic anyway, the net result
202     * is that under contention, the effective "p" value may be lower
203     * than its nominal value. And race windows are kept small enough
204     * that in practice these failures are rare, even under a lot of
205     * contention.
206     *
207     * The fact that retries (for both base and index lists) are
208     * relatively cheap due to indexing allows some minor
209     * simplifications of retry logic. Traversal restarts are
210     * performed after most "helping-out" CASes. This isn't always
211     * strictly necessary, but the implicit backoffs tend to help
212     * reduce other downstream failed CAS's enough to outweigh restart
213     * cost. This worsens the worst case, but seems to improve even
214     * highly contended cases.
215     *
216     * Unlike most skip-list implementations, index insertion and
217     * deletion here require a separate traversal pass occuring after
218     * the base-level action, to add or remove index nodes. This adds
219     * to single-threaded overhead, but improves contended
220     * multithreaded performance by narrowing interference windows,
221     * and allows deletion to ensure that all index nodes will be made
222     * unreachable upon return from a public remove operation, thus
223     * avoiding unwanted garbage retention. This is more important
224     * here than in some other data structures because we cannot null
225     * out node fields referencing user keys since they might still be
226     * read by other ongoing traversals.
227     *
228     * Indexing uses skip list parameters that maintain good search
229     * performance while using sparser-than-usual indices: The
230     * hardwired parameters k=1, p=0.5 (see method randomLevel) mean
231     * that about one-quarter of the nodes have indices. Of those that
232     * do, half have one level, a quarter have two, and so on (see
233     * Pugh's Skip List Cookbook, sec 3.4). The expected total space
234     * requirement for a map is slightly less than for the current
235     * implementation of java.util.TreeMap.
236     *
237     * Changing the level of the index (i.e, the height of the
238     * tree-like structure) also uses CAS. The head index has initial
239     * level/height of one. Creation of an index with height greater
240     * than the current level adds a level to the head index by
241     * CAS'ing on a new top-most head. To maintain good performance
242     * after a lot of removals, deletion methods heuristically try to
243     * reduce the height if the topmost levels appear to be empty.
244     * This may encounter races in which it possible (but rare) to
245     * reduce and "lose" a level just as it is about to contain an
246     * index (that will then never be encountered). This does no
247     * structural harm, and in practice appears to be a better option
248     * than allowing unrestrained growth of levels.
249     *
250     * The code for all this is more verbose than you'd like. Most
251     * operations entail locating an element (or position to insert an
252     * element). The code to do this can't be nicely factored out
253     * because subsequent uses require a snapshot of predecessor
254     * and/or successor and/or value fields which can't be returned
255     * all at once, at least not without creating yet another object
256     * to hold them -- creating such little objects is an especially
257     * bad idea for basic internal search operations because it adds
258     * to GC overhead. (This is one of the few times I've wished Java
259     * had macros.) Instead, some traversal code is interleaved within
260     * insertion and removal operations. The control logic to handle
261     * all the retry conditions is sometimes twisty. Most search is
262     * broken into 2 parts. findPredecessor() searches index nodes
263     * only, returning a base-level predecessor of the key. findNode()
264     * finishes out the base-level search. Even with this factoring,
265     * there is a fair amount of near-duplication of code to handle
266     * variants.
267     *
268     * For explanation of algorithms sharing at least a couple of
269     * features with this one, see Mikhail Fomitchev's thesis
270     * (http://www.cs.yorku.ca/~mikhail/), Keir Fraser's thesis
271 dl 1.4 * (http://www.cl.cam.ac.uk/users/kaf24/), and Hakan Sundell's
272 dl 1.1 * thesis (http://www.cs.chalmers.se/~phs/).
273     *
274     * Given the use of tree-like index nodes, you might wonder why
275     * this doesn't use some kind of search tree instead, which would
276     * support somewhat faster search operations. The reason is that
277     * there are no known efficient lock-free insertion and deletion
278     * algorithms for search trees. The immutability of the "down"
279     * links of index nodes (as opposed to mutable "left" fields in
280     * true trees) makes this tractable using only CAS operations.
281     *
282     * Notation guide for local variables
283     * Node: b, n, f for predecessor, node, successor
284     * Index: q, r, d for index node, right, down.
285     * t for another index node
286     * Head: h
287     * Levels: j
288     * Keys: k, key
289     * Values: v, value
290     * Comparisons: c
291     */
292    
293     private static final long serialVersionUID = -8627078645895051609L;
294    
295     /**
296 dl 1.40 * Generates the initial random seed for the cheaper per-instance
297     * random number generators used in randomLevel.
298     */
299     private static final Random seedGenerator = new Random();
300    
301     /**
302 dl 1.1 * Special value used to identify base-level header
303 dl 1.9 */
304 dl 1.1 private static final Object BASE_HEADER = new Object();
305    
306     /**
307 dl 1.9 * The topmost head index of the skiplist.
308 dl 1.1 */
309     private transient volatile HeadIndex<K,V> head;
310    
311     /**
312 jsr166 1.10 * The comparator used to maintain order in this map, or null
313 jsr166 1.22 * if using natural ordering.
314 dl 1.1 * @serial
315     */
316     private final Comparator<? super K> comparator;
317    
318     /**
319     * Seed for simple random number generator. Not volatile since it
320     * doesn't matter too much if different threads don't see updates.
321     */
322     private transient int randomSeed;
323    
324     /** Lazily initialized key set */
325     private transient KeySet keySet;
326     /** Lazily initialized entry set */
327     private transient EntrySet entrySet;
328     /** Lazily initialized values collection */
329     private transient Values values;
330     /** Lazily initialized descending key set */
331 dl 1.46 private transient ConcurrentNavigableMap<K,V> descendingMap;
332 dl 1.1
333     /**
334 jsr166 1.13 * Initializes or resets state. Needed by constructors, clone,
335 dl 1.1 * clear, readObject. and ConcurrentSkipListSet.clone.
336     * (Note that comparator must be separately initialized.)
337     */
338     final void initialize() {
339     keySet = null;
340 dl 1.9 entrySet = null;
341 dl 1.1 values = null;
342 dl 1.46 descendingMap = null;
343 dl 1.40 randomSeed = seedGenerator.nextInt() | 0x0100; // ensure nonzero
344 dl 1.1 head = new HeadIndex<K,V>(new Node<K,V>(null, BASE_HEADER, null),
345     null, null, 1);
346     }
347    
348     /**
349     * compareAndSet head node
350     */
351     private boolean casHead(HeadIndex<K,V> cmp, HeadIndex<K,V> val) {
352 dl 1.59 return UNSAFE.compareAndSwapObject(this, headOffset, cmp, val);
353 dl 1.1 }
354    
355     /* ---------------- Nodes -------------- */
356    
357     /**
358     * Nodes hold keys and values, and are singly linked in sorted
359     * order, possibly with some intervening marker nodes. The list is
360     * headed by a dummy node accessible as head.node. The value field
361     * is declared only as Object because it takes special non-V
362     * values for marker and header nodes.
363     */
364     static final class Node<K,V> {
365     final K key;
366     volatile Object value;
367     volatile Node<K,V> next;
368    
369     /**
370     * Creates a new regular node.
371     */
372     Node(K key, Object value, Node<K,V> next) {
373     this.key = key;
374     this.value = value;
375     this.next = next;
376     }
377    
378     /**
379     * Creates a new marker node. A marker is distinguished by
380     * having its value field point to itself. Marker nodes also
381     * have null keys, a fact that is exploited in a few places,
382     * but this doesn't distinguish markers from the base-level
383     * header node (head.node), which also has a null key.
384     */
385     Node(Node<K,V> next) {
386     this.key = null;
387     this.value = this;
388     this.next = next;
389     }
390    
391     /**
392     * compareAndSet value field
393     */
394     boolean casValue(Object cmp, Object val) {
395 dl 1.59 return UNSAFE.compareAndSwapObject(this, valueOffset, cmp, val);
396 dl 1.1 }
397 jsr166 1.60
398 dl 1.1 /**
399     * compareAndSet next field
400     */
401     boolean casNext(Node<K,V> cmp, Node<K,V> val) {
402 dl 1.59 return UNSAFE.compareAndSwapObject(this, nextOffset, cmp, val);
403 dl 1.1 }
404    
405     /**
406 jsr166 1.10 * Returns true if this node is a marker. This method isn't
407     * actually called in any current code checking for markers
408 dl 1.1 * because callers will have already read value field and need
409     * to use that read (not another done here) and so directly
410     * test if value points to node.
411     * @param n a possibly null reference to a node
412     * @return true if this node is a marker node
413     */
414     boolean isMarker() {
415     return value == this;
416     }
417    
418     /**
419 jsr166 1.10 * Returns true if this node is the header of base-level list.
420 dl 1.1 * @return true if this node is header node
421     */
422     boolean isBaseHeader() {
423     return value == BASE_HEADER;
424     }
425    
426     /**
427     * Tries to append a deletion marker to this node.
428     * @param f the assumed current successor of this node
429     * @return true if successful
430     */
431     boolean appendMarker(Node<K,V> f) {
432     return casNext(f, new Node<K,V>(f));
433     }
434    
435     /**
436     * Helps out a deletion by appending marker or unlinking from
437     * predecessor. This is called during traversals when value
438     * field seen to be null.
439     * @param b predecessor
440     * @param f successor
441     */
442     void helpDelete(Node<K,V> b, Node<K,V> f) {
443     /*
444     * Rechecking links and then doing only one of the
445     * help-out stages per call tends to minimize CAS
446     * interference among helping threads.
447     */
448     if (f == next && this == b.next) {
449     if (f == null || f.value != f) // not already marked
450     appendMarker(f);
451     else
452     b.casNext(this, f.next);
453     }
454     }
455    
456     /**
457 jsr166 1.11 * Returns value if this node contains a valid key-value pair,
458 dl 1.9 * else null.
459 dl 1.1 * @return this node's value if it isn't a marker or header or
460     * is deleted, else null.
461     */
462     V getValidValue() {
463     Object v = value;
464     if (v == this || v == BASE_HEADER)
465     return null;
466     return (V)v;
467     }
468    
469     /**
470 jsr166 1.10 * Creates and returns a new SimpleImmutableEntry holding current
471     * mapping if this node holds a valid value, else null.
472 dl 1.1 * @return new entry or null
473     */
474 dl 1.2 AbstractMap.SimpleImmutableEntry<K,V> createSnapshot() {
475 dl 1.1 V v = getValidValue();
476     if (v == null)
477     return null;
478 dl 1.24 return new AbstractMap.SimpleImmutableEntry<K,V>(key, v);
479 dl 1.1 }
480 dl 1.59
481     // Unsafe mechanics
482     private static final sun.misc.Unsafe UNSAFE = sun.misc.Unsafe.getUnsafe();
483     private static final long valueOffset =
484     objectFieldOffset(UNSAFE, "value", Node.class);
485     private static final long nextOffset =
486     objectFieldOffset(UNSAFE, "next", Node.class);
487    
488 dl 1.1 }
489    
490     /* ---------------- Indexing -------------- */
491    
492     /**
493 dl 1.40 * Index nodes represent the levels of the skip list. Note that
494     * even though both Nodes and Indexes have forward-pointing
495     * fields, they have different types and are handled in different
496     * ways, that can't nicely be captured by placing field in a
497     * shared abstract class.
498 dl 1.1 */
499     static class Index<K,V> {
500     final Node<K,V> node;
501     final Index<K,V> down;
502     volatile Index<K,V> right;
503    
504     /**
505 jsr166 1.10 * Creates index node with given values.
506 dl 1.9 */
507 dl 1.1 Index(Node<K,V> node, Index<K,V> down, Index<K,V> right) {
508     this.node = node;
509     this.down = down;
510     this.right = right;
511     }
512    
513     /**
514     * compareAndSet right field
515     */
516     final boolean casRight(Index<K,V> cmp, Index<K,V> val) {
517 dl 1.59 return UNSAFE.compareAndSwapObject(this, rightOffset, cmp, val);
518 dl 1.1 }
519    
520     /**
521     * Returns true if the node this indexes has been deleted.
522     * @return true if indexed node is known to be deleted
523     */
524     final boolean indexesDeletedNode() {
525     return node.value == null;
526     }
527    
528     /**
529     * Tries to CAS newSucc as successor. To minimize races with
530     * unlink that may lose this index node, if the node being
531     * indexed is known to be deleted, it doesn't try to link in.
532     * @param succ the expected current successor
533     * @param newSucc the new successor
534     * @return true if successful
535     */
536     final boolean link(Index<K,V> succ, Index<K,V> newSucc) {
537     Node<K,V> n = node;
538 dl 1.9 newSucc.right = succ;
539 dl 1.1 return n.value != null && casRight(succ, newSucc);
540     }
541    
542     /**
543     * Tries to CAS right field to skip over apparent successor
544     * succ. Fails (forcing a retraversal by caller) if this node
545     * is known to be deleted.
546     * @param succ the expected current successor
547     * @return true if successful
548     */
549     final boolean unlink(Index<K,V> succ) {
550     return !indexesDeletedNode() && casRight(succ, succ.right);
551     }
552 dl 1.59
553     // Unsafe mechanics
554     private static final sun.misc.Unsafe UNSAFE = sun.misc.Unsafe.getUnsafe();
555     private static final long rightOffset =
556     objectFieldOffset(UNSAFE, "right", Index.class);
557    
558 dl 1.1 }
559    
560     /* ---------------- Head nodes -------------- */
561    
562     /**
563     * Nodes heading each level keep track of their level.
564     */
565     static final class HeadIndex<K,V> extends Index<K,V> {
566     final int level;
567     HeadIndex(Node<K,V> node, Index<K,V> down, Index<K,V> right, int level) {
568     super(node, down, right);
569     this.level = level;
570     }
571 dl 1.9 }
572 dl 1.1
573     /* ---------------- Comparison utilities -------------- */
574    
575     /**
576     * Represents a key with a comparator as a Comparable.
577     *
578 jsr166 1.22 * Because most sorted collections seem to use natural ordering on
579 dl 1.1 * Comparables (Strings, Integers, etc), most internal methods are
580     * geared to use them. This is generally faster than checking
581     * per-comparison whether to use comparator or comparable because
582     * it doesn't require a (Comparable) cast for each comparison.
583     * (Optimizers can only sometimes remove such redundant checks
584     * themselves.) When Comparators are used,
585     * ComparableUsingComparators are created so that they act in the
586     * same way as natural orderings. This penalizes use of
587     * Comparators vs Comparables, which seems like the right
588     * tradeoff.
589     */
590     static final class ComparableUsingComparator<K> implements Comparable<K> {
591     final K actualKey;
592     final Comparator<? super K> cmp;
593     ComparableUsingComparator(K key, Comparator<? super K> cmp) {
594     this.actualKey = key;
595     this.cmp = cmp;
596     }
597     public int compareTo(K k2) {
598     return cmp.compare(actualKey, k2);
599     }
600     }
601    
602     /**
603     * If using comparator, return a ComparableUsingComparator, else
604 jsr166 1.50 * cast key as Comparable, which may cause ClassCastException,
605 dl 1.1 * which is propagated back to caller.
606     */
607 jsr166 1.62 private Comparable<? super K> comparable(Object key)
608     throws ClassCastException {
609 dl 1.9 if (key == null)
610 dl 1.1 throw new NullPointerException();
611 dl 1.24 if (comparator != null)
612     return new ComparableUsingComparator<K>((K)key, comparator);
613     else
614     return (Comparable<? super K>)key;
615 dl 1.1 }
616    
617     /**
618 jsr166 1.10 * Compares using comparator or natural ordering. Used when the
619 dl 1.1 * ComparableUsingComparator approach doesn't apply.
620     */
621     int compare(K k1, K k2) throws ClassCastException {
622     Comparator<? super K> cmp = comparator;
623     if (cmp != null)
624     return cmp.compare(k1, k2);
625     else
626 jsr166 1.18 return ((Comparable<? super K>)k1).compareTo(k2);
627 dl 1.1 }
628    
629     /**
630 jsr166 1.10 * Returns true if given key greater than or equal to least and
631 dl 1.1 * strictly less than fence, bypassing either test if least or
632 dl 1.5 * fence are null. Needed mainly in submap operations.
633 dl 1.1 */
634     boolean inHalfOpenRange(K key, K least, K fence) {
635 dl 1.9 if (key == null)
636 dl 1.1 throw new NullPointerException();
637     return ((least == null || compare(key, least) >= 0) &&
638     (fence == null || compare(key, fence) < 0));
639     }
640    
641     /**
642 jsr166 1.10 * Returns true if given key greater than or equal to least and less
643 dl 1.1 * or equal to fence. Needed mainly in submap operations.
644     */
645     boolean inOpenRange(K key, K least, K fence) {
646 dl 1.9 if (key == null)
647 dl 1.1 throw new NullPointerException();
648     return ((least == null || compare(key, least) >= 0) &&
649     (fence == null || compare(key, fence) <= 0));
650     }
651    
652     /* ---------------- Traversal -------------- */
653    
654     /**
655 jsr166 1.10 * Returns a base-level node with key strictly less than given key,
656 dl 1.1 * or the base-level header if there is no such node. Also
657     * unlinks indexes to deleted nodes found along the way. Callers
658     * rely on this side-effect of clearing indices to deleted nodes.
659     * @param key the key
660 dl 1.9 * @return a predecessor of key
661 dl 1.1 */
662 dl 1.9 private Node<K,V> findPredecessor(Comparable<? super K> key) {
663 jsr166 1.41 if (key == null)
664 dl 1.40 throw new NullPointerException(); // don't postpone errors
665 dl 1.1 for (;;) {
666     Index<K,V> q = head;
667 dl 1.40 Index<K,V> r = q.right;
668 dl 1.1 for (;;) {
669 dl 1.40 if (r != null) {
670     Node<K,V> n = r.node;
671     K k = n.key;
672     if (n.value == null) {
673     if (!q.unlink(r))
674     break; // restart
675     r = q.right; // reread r
676     continue;
677 dl 1.1 }
678 dl 1.40 if (key.compareTo(k) > 0) {
679 dl 1.1 q = r;
680 dl 1.40 r = r.right;
681 dl 1.1 continue;
682     }
683     }
684 dl 1.40 Index<K,V> d = q.down;
685     if (d != null) {
686 dl 1.1 q = d;
687 dl 1.40 r = d.right;
688     } else
689 dl 1.1 return q.node;
690     }
691     }
692     }
693    
694     /**
695 jsr166 1.10 * Returns node holding key or null if no such, clearing out any
696 dl 1.1 * deleted nodes seen along the way. Repeatedly traverses at
697     * base-level looking for key starting at predecessor returned
698     * from findPredecessor, processing base-level deletions as
699     * encountered. Some callers rely on this side-effect of clearing
700     * deleted nodes.
701     *
702     * Restarts occur, at traversal step centered on node n, if:
703     *
704     * (1) After reading n's next field, n is no longer assumed
705     * predecessor b's current successor, which means that
706     * we don't have a consistent 3-node snapshot and so cannot
707     * unlink any subsequent deleted nodes encountered.
708     *
709     * (2) n's value field is null, indicating n is deleted, in
710     * which case we help out an ongoing structural deletion
711     * before retrying. Even though there are cases where such
712     * unlinking doesn't require restart, they aren't sorted out
713     * here because doing so would not usually outweigh cost of
714     * restarting.
715     *
716 dl 1.9 * (3) n is a marker or n's predecessor's value field is null,
717 dl 1.1 * indicating (among other possibilities) that
718     * findPredecessor returned a deleted node. We can't unlink
719     * the node because we don't know its predecessor, so rely
720     * on another call to findPredecessor to notice and return
721     * some earlier predecessor, which it will do. This check is
722     * only strictly needed at beginning of loop, (and the
723     * b.value check isn't strictly needed at all) but is done
724     * each iteration to help avoid contention with other
725     * threads by callers that will fail to be able to change
726     * links, and so will retry anyway.
727     *
728     * The traversal loops in doPut, doRemove, and findNear all
729     * include the same three kinds of checks. And specialized
730 dl 1.31 * versions appear in findFirst, and findLast and their
731     * variants. They can't easily share code because each uses the
732 dl 1.1 * reads of fields held in locals occurring in the orders they
733     * were performed.
734 dl 1.9 *
735 dl 1.1 * @param key the key
736 jsr166 1.22 * @return node holding key, or null if no such
737 dl 1.1 */
738 dl 1.9 private Node<K,V> findNode(Comparable<? super K> key) {
739 dl 1.1 for (;;) {
740     Node<K,V> b = findPredecessor(key);
741     Node<K,V> n = b.next;
742     for (;;) {
743 dl 1.9 if (n == null)
744 dl 1.1 return null;
745     Node<K,V> f = n.next;
746     if (n != b.next) // inconsistent read
747     break;
748     Object v = n.value;
749     if (v == null) { // n is deleted
750     n.helpDelete(b, f);
751     break;
752     }
753     if (v == n || b.value == null) // b is deleted
754     break;
755     int c = key.compareTo(n.key);
756 dl 1.40 if (c == 0)
757     return n;
758 dl 1.1 if (c < 0)
759     return null;
760     b = n;
761     n = f;
762     }
763     }
764     }
765    
766 dl 1.9 /**
767 dl 1.63 * Get value for key using findNode
768 dl 1.1 * @param okey the key
769     * @return the value, or null if absent
770     */
771     private V doGet(Object okey) {
772 dl 1.9 Comparable<? super K> key = comparable(okey);
773 dl 1.1 /*
774     * Loop needed here and elsewhere in case value field goes
775     * null just as it is about to be returned, in which case we
776     * lost a race with a deletion, so must retry.
777     */
778     for (;;) {
779     Node<K,V> n = findNode(key);
780     if (n == null)
781     return null;
782     Object v = n.value;
783     if (v != null)
784     return (V)v;
785     }
786     }
787    
788     /* ---------------- Insertion -------------- */
789    
790     /**
791     * Main insertion method. Adds element if not present, or
792     * replaces value if present and onlyIfAbsent is false.
793 dl 1.9 * @param kkey the key
794 dl 1.1 * @param value the value that must be associated with key
795     * @param onlyIfAbsent if should not insert if already present
796     * @return the old value, or null if newly inserted
797     */
798     private V doPut(K kkey, V value, boolean onlyIfAbsent) {
799 dl 1.9 Comparable<? super K> key = comparable(kkey);
800 dl 1.1 for (;;) {
801     Node<K,V> b = findPredecessor(key);
802     Node<K,V> n = b.next;
803     for (;;) {
804     if (n != null) {
805     Node<K,V> f = n.next;
806     if (n != b.next) // inconsistent read
807 jsr166 1.57 break;
808 dl 1.1 Object v = n.value;
809     if (v == null) { // n is deleted
810     n.helpDelete(b, f);
811     break;
812     }
813     if (v == n || b.value == null) // b is deleted
814     break;
815     int c = key.compareTo(n.key);
816     if (c > 0) {
817     b = n;
818     n = f;
819     continue;
820     }
821     if (c == 0) {
822     if (onlyIfAbsent || n.casValue(v, value))
823     return (V)v;
824     else
825     break; // restart if lost race to replace value
826     }
827     // else c < 0; fall through
828     }
829 dl 1.9
830 dl 1.1 Node<K,V> z = new Node<K,V>(kkey, value, n);
831 dl 1.9 if (!b.casNext(n, z))
832 dl 1.1 break; // restart if lost race to append to b
833 dl 1.9 int level = randomLevel();
834     if (level > 0)
835 dl 1.1 insertIndex(z, level);
836     return null;
837     }
838     }
839     }
840    
841     /**
842 jsr166 1.10 * Returns a random level for inserting a new node.
843 dl 1.35 * Hardwired to k=1, p=0.5, max 31 (see above and
844 dl 1.34 * Pugh's "Skip List Cookbook", sec 3.4).
845 dl 1.1 *
846 dl 1.33 * This uses the simplest of the generators described in George
847     * Marsaglia's "Xorshift RNGs" paper. This is not a high-quality
848 dl 1.40 * generator but is acceptable here.
849 dl 1.1 */
850     private int randomLevel() {
851 dl 1.40 int x = randomSeed;
852     x ^= x << 13;
853 dl 1.33 x ^= x >>> 17;
854 dl 1.40 randomSeed = x ^= x << 5;
855 dl 1.58 if ((x & 0x80000001) != 0) // test highest and lowest bits
856 dl 1.40 return 0;
857     int level = 1;
858     while (((x >>>= 1) & 1) != 0) ++level;
859 dl 1.1 return level;
860     }
861    
862     /**
863 jsr166 1.11 * Creates and adds index nodes for the given node.
864 dl 1.1 * @param z the node
865     * @param level the level of the index
866     */
867     private void insertIndex(Node<K,V> z, int level) {
868     HeadIndex<K,V> h = head;
869     int max = h.level;
870    
871     if (level <= max) {
872     Index<K,V> idx = null;
873     for (int i = 1; i <= level; ++i)
874     idx = new Index<K,V>(z, idx, null);
875     addIndex(idx, h, level);
876    
877     } else { // Add a new level
878     /*
879     * To reduce interference by other threads checking for
880     * empty levels in tryReduceLevel, new levels are added
881     * with initialized right pointers. Which in turn requires
882     * keeping levels in an array to access them while
883     * creating new head index nodes from the opposite
884     * direction.
885     */
886     level = max + 1;
887     Index<K,V>[] idxs = (Index<K,V>[])new Index[level+1];
888     Index<K,V> idx = null;
889 dl 1.9 for (int i = 1; i <= level; ++i)
890 dl 1.1 idxs[i] = idx = new Index<K,V>(z, idx, null);
891    
892     HeadIndex<K,V> oldh;
893     int k;
894     for (;;) {
895     oldh = head;
896     int oldLevel = oldh.level;
897     if (level <= oldLevel) { // lost race to add level
898     k = level;
899     break;
900     }
901     HeadIndex<K,V> newh = oldh;
902     Node<K,V> oldbase = oldh.node;
903 dl 1.9 for (int j = oldLevel+1; j <= level; ++j)
904 dl 1.1 newh = new HeadIndex<K,V>(oldbase, newh, idxs[j], j);
905     if (casHead(oldh, newh)) {
906     k = oldLevel;
907     break;
908     }
909     }
910     addIndex(idxs[k], oldh, k);
911     }
912     }
913    
914     /**
915 jsr166 1.10 * Adds given index nodes from given level down to 1.
916 dl 1.1 * @param idx the topmost index node being inserted
917     * @param h the value of head to use to insert. This must be
918     * snapshotted by callers to provide correct insertion level
919     * @param indexLevel the level of the index
920     */
921     private void addIndex(Index<K,V> idx, HeadIndex<K,V> h, int indexLevel) {
922     // Track next level to insert in case of retries
923     int insertionLevel = indexLevel;
924 dl 1.40 Comparable<? super K> key = comparable(idx.node.key);
925     if (key == null) throw new NullPointerException();
926 dl 1.1
927     // Similar to findPredecessor, but adding index nodes along
928     // path to key.
929     for (;;) {
930 dl 1.40 int j = h.level;
931 dl 1.1 Index<K,V> q = h;
932 dl 1.40 Index<K,V> r = q.right;
933 dl 1.1 Index<K,V> t = idx;
934     for (;;) {
935     if (r != null) {
936 dl 1.40 Node<K,V> n = r.node;
937 dl 1.1 // compare before deletion check avoids needing recheck
938 dl 1.40 int c = key.compareTo(n.key);
939     if (n.value == null) {
940     if (!q.unlink(r))
941 dl 1.9 break;
942 dl 1.40 r = q.right;
943     continue;
944 dl 1.1 }
945     if (c > 0) {
946     q = r;
947 dl 1.40 r = r.right;
948 dl 1.1 continue;
949     }
950     }
951    
952     if (j == insertionLevel) {
953     // Don't insert index if node already deleted
954     if (t.indexesDeletedNode()) {
955     findNode(key); // cleans up
956     return;
957     }
958 dl 1.9 if (!q.link(r, t))
959 dl 1.1 break; // restart
960     if (--insertionLevel == 0) {
961     // need final deletion check before return
962 dl 1.9 if (t.indexesDeletedNode())
963     findNode(key);
964 dl 1.1 return;
965     }
966     }
967    
968 dl 1.40 if (--j >= insertionLevel && j < indexLevel)
969 dl 1.1 t = t.down;
970     q = q.down;
971 dl 1.40 r = q.right;
972 dl 1.1 }
973     }
974     }
975    
976     /* ---------------- Deletion -------------- */
977    
978     /**
979     * Main deletion method. Locates node, nulls value, appends a
980     * deletion marker, unlinks predecessor, removes associated index
981     * nodes, and possibly reduces head index level.
982     *
983     * Index nodes are cleared out simply by calling findPredecessor.
984     * which unlinks indexes to deleted nodes found along path to key,
985     * which will include the indexes to this node. This is done
986     * unconditionally. We can't check beforehand whether there are
987     * index nodes because it might be the case that some or all
988     * indexes hadn't been inserted yet for this node during initial
989     * search for it, and we'd like to ensure lack of garbage
990 dl 1.9 * retention, so must call to be sure.
991 dl 1.1 *
992     * @param okey the key
993     * @param value if non-null, the value that must be
994     * associated with key
995     * @return the node, or null if not found
996     */
997 dl 1.46 final V doRemove(Object okey, Object value) {
998 dl 1.9 Comparable<? super K> key = comparable(okey);
999     for (;;) {
1000 dl 1.1 Node<K,V> b = findPredecessor(key);
1001     Node<K,V> n = b.next;
1002     for (;;) {
1003 dl 1.9 if (n == null)
1004 dl 1.1 return null;
1005     Node<K,V> f = n.next;
1006     if (n != b.next) // inconsistent read
1007     break;
1008     Object v = n.value;
1009     if (v == null) { // n is deleted
1010     n.helpDelete(b, f);
1011     break;
1012     }
1013     if (v == n || b.value == null) // b is deleted
1014     break;
1015     int c = key.compareTo(n.key);
1016     if (c < 0)
1017     return null;
1018     if (c > 0) {
1019     b = n;
1020     n = f;
1021     continue;
1022     }
1023 dl 1.9 if (value != null && !value.equals(v))
1024     return null;
1025     if (!n.casValue(v, null))
1026 dl 1.1 break;
1027 dl 1.9 if (!n.appendMarker(f) || !b.casNext(n, f))
1028 dl 1.1 findNode(key); // Retry via findNode
1029     else {
1030     findPredecessor(key); // Clean index
1031 dl 1.9 if (head.right == null)
1032 dl 1.1 tryReduceLevel();
1033     }
1034     return (V)v;
1035     }
1036     }
1037     }
1038    
1039     /**
1040     * Possibly reduce head level if it has no nodes. This method can
1041     * (rarely) make mistakes, in which case levels can disappear even
1042     * though they are about to contain index nodes. This impacts
1043     * performance, not correctness. To minimize mistakes as well as
1044     * to reduce hysteresis, the level is reduced by one only if the
1045     * topmost three levels look empty. Also, if the removed level
1046     * looks non-empty after CAS, we try to change it back quick
1047     * before anyone notices our mistake! (This trick works pretty
1048     * well because this method will practically never make mistakes
1049     * unless current thread stalls immediately before first CAS, in
1050     * which case it is very unlikely to stall again immediately
1051     * afterwards, so will recover.)
1052     *
1053     * We put up with all this rather than just let levels grow
1054     * because otherwise, even a small map that has undergone a large
1055     * number of insertions and removals will have a lot of levels,
1056     * slowing down access more than would an occasional unwanted
1057     * reduction.
1058     */
1059     private void tryReduceLevel() {
1060     HeadIndex<K,V> h = head;
1061     HeadIndex<K,V> d;
1062     HeadIndex<K,V> e;
1063     if (h.level > 3 &&
1064 dl 1.9 (d = (HeadIndex<K,V>)h.down) != null &&
1065     (e = (HeadIndex<K,V>)d.down) != null &&
1066     e.right == null &&
1067     d.right == null &&
1068 dl 1.1 h.right == null &&
1069     casHead(h, d) && // try to set
1070     h.right != null) // recheck
1071     casHead(d, h); // try to backout
1072     }
1073    
1074     /* ---------------- Finding and removing first element -------------- */
1075    
1076     /**
1077 jsr166 1.22 * Specialized variant of findNode to get first valid node.
1078 dl 1.1 * @return first node or null if empty
1079     */
1080     Node<K,V> findFirst() {
1081     for (;;) {
1082     Node<K,V> b = head.node;
1083     Node<K,V> n = b.next;
1084     if (n == null)
1085     return null;
1086 dl 1.9 if (n.value != null)
1087 dl 1.1 return n;
1088     n.helpDelete(b, n.next);
1089     }
1090     }
1091    
1092     /**
1093 dl 1.25 * Removes first entry; returns its snapshot.
1094 jsr166 1.28 * @return null if empty, else snapshot of first entry
1095 dl 1.1 */
1096 dl 1.25 Map.Entry<K,V> doRemoveFirstEntry() {
1097 dl 1.9 for (;;) {
1098 dl 1.1 Node<K,V> b = head.node;
1099     Node<K,V> n = b.next;
1100 dl 1.9 if (n == null)
1101 dl 1.1 return null;
1102     Node<K,V> f = n.next;
1103     if (n != b.next)
1104     continue;
1105     Object v = n.value;
1106     if (v == null) {
1107     n.helpDelete(b, f);
1108     continue;
1109     }
1110     if (!n.casValue(v, null))
1111     continue;
1112     if (!n.appendMarker(f) || !b.casNext(n, f))
1113     findFirst(); // retry
1114     clearIndexToFirst();
1115 dl 1.30 return new AbstractMap.SimpleImmutableEntry<K,V>(n.key, (V)v);
1116 jsr166 1.55 }
1117 dl 1.1 }
1118    
1119     /**
1120 jsr166 1.10 * Clears out index nodes associated with deleted first entry.
1121 dl 1.1 */
1122     private void clearIndexToFirst() {
1123     for (;;) {
1124     Index<K,V> q = head;
1125     for (;;) {
1126     Index<K,V> r = q.right;
1127     if (r != null && r.indexesDeletedNode() && !q.unlink(r))
1128 dl 1.9 break;
1129 dl 1.1 if ((q = q.down) == null) {
1130 dl 1.9 if (head.right == null)
1131 dl 1.1 tryReduceLevel();
1132     return;
1133     }
1134     }
1135     }
1136     }
1137    
1138    
1139     /* ---------------- Finding and removing last element -------------- */
1140    
1141     /**
1142 jsr166 1.10 * Specialized version of find to get last valid node.
1143 dl 1.1 * @return last node or null if empty
1144     */
1145     Node<K,V> findLast() {
1146     /*
1147     * findPredecessor can't be used to traverse index level
1148     * because this doesn't use comparisons. So traversals of
1149     * both levels are folded together.
1150     */
1151     Index<K,V> q = head;
1152     for (;;) {
1153     Index<K,V> d, r;
1154     if ((r = q.right) != null) {
1155     if (r.indexesDeletedNode()) {
1156     q.unlink(r);
1157     q = head; // restart
1158 dl 1.9 }
1159 dl 1.1 else
1160     q = r;
1161     } else if ((d = q.down) != null) {
1162     q = d;
1163     } else {
1164     Node<K,V> b = q.node;
1165     Node<K,V> n = b.next;
1166     for (;;) {
1167 dl 1.9 if (n == null)
1168 jsr166 1.61 return b.isBaseHeader() ? null : b;
1169 dl 1.1 Node<K,V> f = n.next; // inconsistent read
1170     if (n != b.next)
1171     break;
1172     Object v = n.value;
1173     if (v == null) { // n is deleted
1174     n.helpDelete(b, f);
1175     break;
1176     }
1177     if (v == n || b.value == null) // b is deleted
1178     break;
1179     b = n;
1180     n = f;
1181     }
1182     q = head; // restart
1183     }
1184     }
1185     }
1186    
1187 dl 1.31 /**
1188 jsr166 1.32 * Specialized variant of findPredecessor to get predecessor of last
1189     * valid node. Needed when removing the last entry. It is possible
1190     * that all successors of returned node will have been deleted upon
1191 dl 1.31 * return, in which case this method can be retried.
1192     * @return likely predecessor of last node
1193     */
1194     private Node<K,V> findPredecessorOfLast() {
1195     for (;;) {
1196     Index<K,V> q = head;
1197     for (;;) {
1198     Index<K,V> d, r;
1199     if ((r = q.right) != null) {
1200     if (r.indexesDeletedNode()) {
1201     q.unlink(r);
1202     break; // must restart
1203     }
1204     // proceed as far across as possible without overshooting
1205     if (r.node.next != null) {
1206     q = r;
1207     continue;
1208     }
1209     }
1210     if ((d = q.down) != null)
1211     q = d;
1212     else
1213     return q.node;
1214     }
1215     }
1216     }
1217 dl 1.1
1218     /**
1219 jsr166 1.32 * Removes last entry; returns its snapshot.
1220     * Specialized variant of doRemove.
1221     * @return null if empty, else snapshot of last entry
1222 dl 1.1 */
1223 dl 1.31 Map.Entry<K,V> doRemoveLastEntry() {
1224 dl 1.1 for (;;) {
1225 dl 1.31 Node<K,V> b = findPredecessorOfLast();
1226     Node<K,V> n = b.next;
1227     if (n == null) {
1228     if (b.isBaseHeader()) // empty
1229     return null;
1230     else
1231     continue; // all b's successors are deleted; retry
1232     }
1233 dl 1.1 for (;;) {
1234 dl 1.31 Node<K,V> f = n.next;
1235     if (n != b.next) // inconsistent read
1236     break;
1237     Object v = n.value;
1238     if (v == null) { // n is deleted
1239     n.helpDelete(b, f);
1240     break;
1241     }
1242     if (v == n || b.value == null) // b is deleted
1243     break;
1244     if (f != null) {
1245     b = n;
1246     n = f;
1247     continue;
1248     }
1249     if (!n.casValue(v, null))
1250     break;
1251     K key = n.key;
1252     Comparable<? super K> ck = comparable(key);
1253     if (!n.appendMarker(f) || !b.casNext(n, f))
1254     findNode(ck); // Retry via findNode
1255     else {
1256     findPredecessor(ck); // Clean index
1257     if (head.right == null)
1258     tryReduceLevel();
1259 dl 1.1 }
1260 dl 1.31 return new AbstractMap.SimpleImmutableEntry<K,V>(key, (V)v);
1261 dl 1.1 }
1262     }
1263     }
1264    
1265     /* ---------------- Relational operations -------------- */
1266    
1267     // Control values OR'ed as arguments to findNear
1268    
1269     private static final int EQ = 1;
1270     private static final int LT = 2;
1271     private static final int GT = 0; // Actually checked as !LT
1272    
1273     /**
1274     * Utility for ceiling, floor, lower, higher methods.
1275     * @param kkey the key
1276     * @param rel the relation -- OR'ed combination of EQ, LT, GT
1277     * @return nearest node fitting relation, or null if no such
1278     */
1279     Node<K,V> findNear(K kkey, int rel) {
1280 dl 1.9 Comparable<? super K> key = comparable(kkey);
1281 dl 1.1 for (;;) {
1282     Node<K,V> b = findPredecessor(key);
1283     Node<K,V> n = b.next;
1284     for (;;) {
1285 dl 1.9 if (n == null)
1286 jsr166 1.61 return ((rel & LT) == 0 || b.isBaseHeader()) ? null : b;
1287 dl 1.1 Node<K,V> f = n.next;
1288     if (n != b.next) // inconsistent read
1289     break;
1290     Object v = n.value;
1291     if (v == null) { // n is deleted
1292     n.helpDelete(b, f);
1293     break;
1294     }
1295     if (v == n || b.value == null) // b is deleted
1296     break;
1297     int c = key.compareTo(n.key);
1298     if ((c == 0 && (rel & EQ) != 0) ||
1299     (c < 0 && (rel & LT) == 0))
1300     return n;
1301     if ( c <= 0 && (rel & LT) != 0)
1302 jsr166 1.61 return b.isBaseHeader() ? null : b;
1303 dl 1.1 b = n;
1304     n = f;
1305     }
1306     }
1307     }
1308    
1309     /**
1310 jsr166 1.10 * Returns SimpleImmutableEntry for results of findNear.
1311 dl 1.40 * @param key the key
1312 dl 1.1 * @param rel the relation -- OR'ed combination of EQ, LT, GT
1313     * @return Entry fitting relation, or null if no such
1314     */
1315 dl 1.40 AbstractMap.SimpleImmutableEntry<K,V> getNear(K key, int rel) {
1316 dl 1.1 for (;;) {
1317 dl 1.40 Node<K,V> n = findNear(key, rel);
1318 dl 1.1 if (n == null)
1319     return null;
1320 dl 1.2 AbstractMap.SimpleImmutableEntry<K,V> e = n.createSnapshot();
1321 dl 1.1 if (e != null)
1322     return e;
1323     }
1324     }
1325    
1326 jsr166 1.53
1327 dl 1.1 /* ---------------- Constructors -------------- */
1328    
1329     /**
1330 jsr166 1.22 * Constructs a new, empty map, sorted according to the
1331     * {@linkplain Comparable natural ordering} of the keys.
1332 dl 1.1 */
1333     public ConcurrentSkipListMap() {
1334     this.comparator = null;
1335     initialize();
1336     }
1337    
1338     /**
1339 jsr166 1.22 * Constructs a new, empty map, sorted according to the specified
1340     * comparator.
1341 dl 1.1 *
1342 jsr166 1.22 * @param comparator the comparator that will be used to order this map.
1343     * If <tt>null</tt>, the {@linkplain Comparable natural
1344     * ordering} of the keys will be used.
1345 dl 1.1 */
1346 jsr166 1.22 public ConcurrentSkipListMap(Comparator<? super K> comparator) {
1347     this.comparator = comparator;
1348 dl 1.1 initialize();
1349     }
1350    
1351     /**
1352     * Constructs a new map containing the same mappings as the given map,
1353 jsr166 1.22 * sorted according to the {@linkplain Comparable natural ordering} of
1354     * the keys.
1355 dl 1.1 *
1356 jsr166 1.22 * @param m the map whose mappings are to be placed in this map
1357     * @throws ClassCastException if the keys in <tt>m</tt> are not
1358     * {@link Comparable}, or are not mutually comparable
1359     * @throws NullPointerException if the specified map or any of its keys
1360     * or values are null
1361 dl 1.1 */
1362     public ConcurrentSkipListMap(Map<? extends K, ? extends V> m) {
1363     this.comparator = null;
1364     initialize();
1365     putAll(m);
1366     }
1367    
1368     /**
1369 jsr166 1.22 * Constructs a new map containing the same mappings and using the
1370     * same ordering as the specified sorted map.
1371     *
1372 dl 1.1 * @param m the sorted map whose mappings are to be placed in this
1373 jsr166 1.22 * map, and whose comparator is to be used to sort this map
1374     * @throws NullPointerException if the specified sorted map or any of
1375     * its keys or values are null
1376 dl 1.1 */
1377     public ConcurrentSkipListMap(SortedMap<K, ? extends V> m) {
1378     this.comparator = m.comparator();
1379     initialize();
1380     buildFromSorted(m);
1381     }
1382    
1383     /**
1384 jsr166 1.22 * Returns a shallow copy of this <tt>ConcurrentSkipListMap</tt>
1385     * instance. (The keys and values themselves are not cloned.)
1386 dl 1.1 *
1387 jsr166 1.22 * @return a shallow copy of this map
1388 dl 1.1 */
1389 jsr166 1.16 public ConcurrentSkipListMap<K,V> clone() {
1390 dl 1.1 ConcurrentSkipListMap<K,V> clone = null;
1391     try {
1392     clone = (ConcurrentSkipListMap<K,V>) super.clone();
1393     } catch (CloneNotSupportedException e) {
1394     throw new InternalError();
1395     }
1396    
1397     clone.initialize();
1398     clone.buildFromSorted(this);
1399     return clone;
1400     }
1401    
1402     /**
1403     * Streamlined bulk insertion to initialize from elements of
1404     * given sorted map. Call only from constructor or clone
1405     * method.
1406     */
1407     private void buildFromSorted(SortedMap<K, ? extends V> map) {
1408     if (map == null)
1409     throw new NullPointerException();
1410    
1411     HeadIndex<K,V> h = head;
1412     Node<K,V> basepred = h.node;
1413    
1414     // Track the current rightmost node at each level. Uses an
1415     // ArrayList to avoid committing to initial or maximum level.
1416     ArrayList<Index<K,V>> preds = new ArrayList<Index<K,V>>();
1417    
1418     // initialize
1419 dl 1.9 for (int i = 0; i <= h.level; ++i)
1420 dl 1.1 preds.add(null);
1421     Index<K,V> q = h;
1422     for (int i = h.level; i > 0; --i) {
1423     preds.set(i, q);
1424     q = q.down;
1425     }
1426    
1427 dl 1.9 Iterator<? extends Map.Entry<? extends K, ? extends V>> it =
1428 dl 1.1 map.entrySet().iterator();
1429     while (it.hasNext()) {
1430     Map.Entry<? extends K, ? extends V> e = it.next();
1431     int j = randomLevel();
1432     if (j > h.level) j = h.level + 1;
1433     K k = e.getKey();
1434     V v = e.getValue();
1435     if (k == null || v == null)
1436     throw new NullPointerException();
1437     Node<K,V> z = new Node<K,V>(k, v, null);
1438     basepred.next = z;
1439     basepred = z;
1440     if (j > 0) {
1441     Index<K,V> idx = null;
1442     for (int i = 1; i <= j; ++i) {
1443     idx = new Index<K,V>(z, idx, null);
1444 dl 1.9 if (i > h.level)
1445 dl 1.1 h = new HeadIndex<K,V>(h.node, h, idx, i);
1446    
1447     if (i < preds.size()) {
1448     preds.get(i).right = idx;
1449     preds.set(i, idx);
1450     } else
1451     preds.add(idx);
1452     }
1453     }
1454     }
1455     head = h;
1456     }
1457    
1458     /* ---------------- Serialization -------------- */
1459    
1460     /**
1461 jsr166 1.10 * Save the state of this map to a stream.
1462 dl 1.1 *
1463     * @serialData The key (Object) and value (Object) for each
1464 jsr166 1.10 * key-value mapping represented by the map, followed by
1465 dl 1.1 * <tt>null</tt>. The key-value mappings are emitted in key-order
1466     * (as determined by the Comparator, or by the keys' natural
1467     * ordering if no Comparator).
1468     */
1469     private void writeObject(java.io.ObjectOutputStream s)
1470     throws java.io.IOException {
1471     // Write out the Comparator and any hidden stuff
1472     s.defaultWriteObject();
1473    
1474     // Write out keys and values (alternating)
1475     for (Node<K,V> n = findFirst(); n != null; n = n.next) {
1476     V v = n.getValidValue();
1477     if (v != null) {
1478     s.writeObject(n.key);
1479     s.writeObject(v);
1480     }
1481     }
1482     s.writeObject(null);
1483     }
1484    
1485     /**
1486 jsr166 1.10 * Reconstitute the map from a stream.
1487 dl 1.1 */
1488     private void readObject(final java.io.ObjectInputStream s)
1489     throws java.io.IOException, ClassNotFoundException {
1490     // Read in the Comparator and any hidden stuff
1491     s.defaultReadObject();
1492     // Reset transients
1493     initialize();
1494    
1495 dl 1.9 /*
1496 dl 1.1 * This is nearly identical to buildFromSorted, but is
1497     * distinct because readObject calls can't be nicely adapted
1498     * as the kind of iterator needed by buildFromSorted. (They
1499     * can be, but doing so requires type cheats and/or creation
1500     * of adaptor classes.) It is simpler to just adapt the code.
1501     */
1502    
1503     HeadIndex<K,V> h = head;
1504     Node<K,V> basepred = h.node;
1505     ArrayList<Index<K,V>> preds = new ArrayList<Index<K,V>>();
1506 dl 1.9 for (int i = 0; i <= h.level; ++i)
1507 dl 1.1 preds.add(null);
1508     Index<K,V> q = h;
1509     for (int i = h.level; i > 0; --i) {
1510     preds.set(i, q);
1511     q = q.down;
1512     }
1513    
1514     for (;;) {
1515     Object k = s.readObject();
1516     if (k == null)
1517     break;
1518     Object v = s.readObject();
1519 dl 1.9 if (v == null)
1520 dl 1.1 throw new NullPointerException();
1521     K key = (K) k;
1522     V val = (V) v;
1523     int j = randomLevel();
1524     if (j > h.level) j = h.level + 1;
1525     Node<K,V> z = new Node<K,V>(key, val, null);
1526     basepred.next = z;
1527     basepred = z;
1528     if (j > 0) {
1529     Index<K,V> idx = null;
1530     for (int i = 1; i <= j; ++i) {
1531     idx = new Index<K,V>(z, idx, null);
1532 dl 1.9 if (i > h.level)
1533 dl 1.1 h = new HeadIndex<K,V>(h.node, h, idx, i);
1534    
1535     if (i < preds.size()) {
1536     preds.get(i).right = idx;
1537     preds.set(i, idx);
1538     } else
1539     preds.add(idx);
1540     }
1541     }
1542     }
1543     head = h;
1544     }
1545    
1546     /* ------ Map API methods ------ */
1547    
1548     /**
1549     * Returns <tt>true</tt> if this map contains a mapping for the specified
1550     * key.
1551 jsr166 1.22 *
1552     * @param key key whose presence in this map is to be tested
1553     * @return <tt>true</tt> if this map contains a mapping for the specified key
1554     * @throws ClassCastException if the specified key cannot be compared
1555     * with the keys currently in the map
1556     * @throws NullPointerException if the specified key is null
1557 dl 1.1 */
1558     public boolean containsKey(Object key) {
1559     return doGet(key) != null;
1560     }
1561    
1562     /**
1563 jsr166 1.42 * Returns the value to which the specified key is mapped,
1564     * or {@code null} if this map contains no mapping for the key.
1565     *
1566     * <p>More formally, if this map contains a mapping from a key
1567     * {@code k} to a value {@code v} such that {@code key} compares
1568     * equal to {@code k} according to the map's ordering, then this
1569     * method returns {@code v}; otherwise it returns {@code null}.
1570     * (There can be at most one such mapping.)
1571 dl 1.1 *
1572 jsr166 1.22 * @throws ClassCastException if the specified key cannot be compared
1573     * with the keys currently in the map
1574     * @throws NullPointerException if the specified key is null
1575 dl 1.1 */
1576     public V get(Object key) {
1577     return doGet(key);
1578     }
1579    
1580     /**
1581     * Associates the specified value with the specified key in this map.
1582 jsr166 1.22 * If the map previously contained a mapping for the key, the old
1583 dl 1.1 * value is replaced.
1584     *
1585 jsr166 1.22 * @param key key with which the specified value is to be associated
1586     * @param value value to be associated with the specified key
1587     * @return the previous value associated with the specified key, or
1588     * <tt>null</tt> if there was no mapping for the key
1589     * @throws ClassCastException if the specified key cannot be compared
1590     * with the keys currently in the map
1591     * @throws NullPointerException if the specified key or value is null
1592 dl 1.1 */
1593     public V put(K key, V value) {
1594 dl 1.9 if (value == null)
1595 dl 1.1 throw new NullPointerException();
1596     return doPut(key, value, false);
1597     }
1598    
1599     /**
1600 jsr166 1.36 * Removes the mapping for the specified key from this map if present.
1601 dl 1.1 *
1602     * @param key key for which mapping should be removed
1603 jsr166 1.22 * @return the previous value associated with the specified key, or
1604     * <tt>null</tt> if there was no mapping for the key
1605     * @throws ClassCastException if the specified key cannot be compared
1606     * with the keys currently in the map
1607     * @throws NullPointerException if the specified key is null
1608 dl 1.1 */
1609     public V remove(Object key) {
1610     return doRemove(key, null);
1611     }
1612    
1613     /**
1614     * Returns <tt>true</tt> if this map maps one or more keys to the
1615     * specified value. This operation requires time linear in the
1616 jsr166 1.10 * map size.
1617 dl 1.1 *
1618 jsr166 1.22 * @param value value whose presence in this map is to be tested
1619     * @return <tt>true</tt> if a mapping to <tt>value</tt> exists;
1620     * <tt>false</tt> otherwise
1621     * @throws NullPointerException if the specified value is null
1622 dl 1.9 */
1623 dl 1.1 public boolean containsValue(Object value) {
1624 dl 1.9 if (value == null)
1625 dl 1.1 throw new NullPointerException();
1626     for (Node<K,V> n = findFirst(); n != null; n = n.next) {
1627     V v = n.getValidValue();
1628     if (v != null && value.equals(v))
1629     return true;
1630     }
1631     return false;
1632     }
1633    
1634     /**
1635 dl 1.6 * Returns the number of key-value mappings in this map. If this map
1636 dl 1.1 * contains more than <tt>Integer.MAX_VALUE</tt> elements, it
1637     * returns <tt>Integer.MAX_VALUE</tt>.
1638     *
1639     * <p>Beware that, unlike in most collections, this method is
1640     * <em>NOT</em> a constant-time operation. Because of the
1641     * asynchronous nature of these maps, determining the current
1642     * number of elements requires traversing them all to count them.
1643     * Additionally, it is possible for the size to change during
1644     * execution of this method, in which case the returned result
1645     * will be inaccurate. Thus, this method is typically not very
1646     * useful in concurrent applications.
1647     *
1648 jsr166 1.22 * @return the number of elements in this map
1649 dl 1.1 */
1650     public int size() {
1651     long count = 0;
1652     for (Node<K,V> n = findFirst(); n != null; n = n.next) {
1653     if (n.getValidValue() != null)
1654     ++count;
1655     }
1656 jsr166 1.61 return (count >= Integer.MAX_VALUE) ? Integer.MAX_VALUE : (int) count;
1657 dl 1.1 }
1658    
1659     /**
1660     * Returns <tt>true</tt> if this map contains no key-value mappings.
1661 jsr166 1.22 * @return <tt>true</tt> if this map contains no key-value mappings
1662 dl 1.1 */
1663     public boolean isEmpty() {
1664     return findFirst() == null;
1665     }
1666    
1667     /**
1668 jsr166 1.22 * Removes all of the mappings from this map.
1669 dl 1.1 */
1670     public void clear() {
1671     initialize();
1672     }
1673    
1674 dl 1.46 /* ---------------- View methods -------------- */
1675    
1676     /*
1677     * Note: Lazy initialization works for views because view classes
1678     * are stateless/immutable so it doesn't matter wrt correctness if
1679     * more than one is created (which will only rarely happen). Even
1680     * so, the following idiom conservatively ensures that the method
1681     * returns the one it created if it does so, not one created by
1682     * another racing thread.
1683     */
1684    
1685 dl 1.1 /**
1686 jsr166 1.51 * Returns a {@link NavigableSet} view of the keys contained in this map.
1687 jsr166 1.22 * The set's iterator returns the keys in ascending order.
1688     * The set is backed by the map, so changes to the map are
1689     * reflected in the set, and vice-versa. The set supports element
1690     * removal, which removes the corresponding mapping from the map,
1691 jsr166 1.51 * via the {@code Iterator.remove}, {@code Set.remove},
1692     * {@code removeAll}, {@code retainAll}, and {@code clear}
1693     * operations. It does not support the {@code add} or {@code addAll}
1694 jsr166 1.22 * operations.
1695     *
1696 jsr166 1.51 * <p>The view's {@code iterator} is a "weakly consistent" iterator
1697 jsr166 1.22 * that will never throw {@link ConcurrentModificationException},
1698 dl 1.1 * and guarantees to traverse elements as they existed upon
1699     * construction of the iterator, and may (but is not guaranteed to)
1700     * reflect any modifications subsequent to construction.
1701     *
1702 jsr166 1.51 * <p>This method is equivalent to method {@code navigableKeySet}.
1703     *
1704     * @return a navigable set view of the keys in this map
1705 dl 1.1 */
1706 jsr166 1.51 public NavigableSet<K> keySet() {
1707 dl 1.1 KeySet ks = keySet;
1708 dl 1.46 return (ks != null) ? ks : (keySet = new KeySet(this));
1709 dl 1.1 }
1710    
1711 dl 1.46 public NavigableSet<K> navigableKeySet() {
1712     KeySet ks = keySet;
1713     return (ks != null) ? ks : (keySet = new KeySet(this));
1714 dl 1.1 }
1715    
1716     /**
1717 jsr166 1.22 * Returns a {@link Collection} view of the values contained in this map.
1718     * The collection's iterator returns the values in ascending order
1719     * of the corresponding keys.
1720 dl 1.1 * The collection is backed by the map, so changes to the map are
1721     * reflected in the collection, and vice-versa. The collection
1722     * supports element removal, which removes the corresponding
1723 jsr166 1.22 * mapping from the map, via the <tt>Iterator.remove</tt>,
1724 dl 1.1 * <tt>Collection.remove</tt>, <tt>removeAll</tt>,
1725 jsr166 1.22 * <tt>retainAll</tt> and <tt>clear</tt> operations. It does not
1726     * support the <tt>add</tt> or <tt>addAll</tt> operations.
1727 dl 1.1 *
1728 jsr166 1.22 * <p>The view's <tt>iterator</tt> is a "weakly consistent" iterator
1729     * that will never throw {@link ConcurrentModificationException},
1730     * and guarantees to traverse elements as they existed upon
1731     * construction of the iterator, and may (but is not guaranteed to)
1732     * reflect any modifications subsequent to construction.
1733 dl 1.1 */
1734     public Collection<V> values() {
1735     Values vs = values;
1736 dl 1.46 return (vs != null) ? vs : (values = new Values(this));
1737 dl 1.1 }
1738    
1739     /**
1740 jsr166 1.22 * Returns a {@link Set} view of the mappings contained in this map.
1741     * The set's iterator returns the entries in ascending key order.
1742     * The set is backed by the map, so changes to the map are
1743     * reflected in the set, and vice-versa. The set supports element
1744     * removal, which removes the corresponding mapping from the map,
1745     * via the <tt>Iterator.remove</tt>, <tt>Set.remove</tt>,
1746     * <tt>removeAll</tt>, <tt>retainAll</tt> and <tt>clear</tt>
1747 dl 1.1 * operations. It does not support the <tt>add</tt> or
1748 jsr166 1.22 * <tt>addAll</tt> operations.
1749     *
1750     * <p>The view's <tt>iterator</tt> is a "weakly consistent" iterator
1751     * that will never throw {@link ConcurrentModificationException},
1752     * and guarantees to traverse elements as they existed upon
1753     * construction of the iterator, and may (but is not guaranteed to)
1754     * reflect any modifications subsequent to construction.
1755     *
1756     * <p>The <tt>Map.Entry</tt> elements returned by
1757 dl 1.1 * <tt>iterator.next()</tt> do <em>not</em> support the
1758     * <tt>setValue</tt> operation.
1759     *
1760 jsr166 1.22 * @return a set view of the mappings contained in this map,
1761     * sorted in ascending key order
1762 dl 1.1 */
1763     public Set<Map.Entry<K,V>> entrySet() {
1764     EntrySet es = entrySet;
1765 dl 1.46 return (es != null) ? es : (entrySet = new EntrySet(this));
1766     }
1767    
1768     public ConcurrentNavigableMap<K,V> descendingMap() {
1769     ConcurrentNavigableMap<K,V> dm = descendingMap;
1770     return (dm != null) ? dm : (descendingMap = new SubMap<K,V>
1771     (this, null, false, null, false, true));
1772 dl 1.1 }
1773    
1774 dl 1.46 public NavigableSet<K> descendingKeySet() {
1775     return descendingMap().navigableKeySet();
1776 dl 1.1 }
1777    
1778     /* ---------------- AbstractMap Overrides -------------- */
1779    
1780     /**
1781     * Compares the specified object with this map for equality.
1782     * Returns <tt>true</tt> if the given object is also a map and the
1783     * two maps represent the same mappings. More formally, two maps
1784 jsr166 1.22 * <tt>m1</tt> and <tt>m2</tt> represent the same mappings if
1785 jsr166 1.39 * <tt>m1.entrySet().equals(m2.entrySet())</tt>. This
1786 dl 1.1 * operation may return misleading results if either map is
1787     * concurrently modified during execution of this method.
1788     *
1789 jsr166 1.22 * @param o object to be compared for equality with this map
1790     * @return <tt>true</tt> if the specified object is equal to this map
1791 dl 1.1 */
1792     public boolean equals(Object o) {
1793 jsr166 1.55 if (o == this)
1794     return true;
1795     if (!(o instanceof Map))
1796     return false;
1797     Map<?,?> m = (Map<?,?>) o;
1798 dl 1.1 try {
1799 jsr166 1.55 for (Map.Entry<K,V> e : this.entrySet())
1800     if (! e.getValue().equals(m.get(e.getKey())))
1801 dl 1.25 return false;
1802 jsr166 1.55 for (Map.Entry<?,?> e : m.entrySet()) {
1803 dl 1.25 Object k = e.getKey();
1804     Object v = e.getValue();
1805 jsr166 1.55 if (k == null || v == null || !v.equals(get(k)))
1806 dl 1.25 return false;
1807     }
1808     return true;
1809 jsr166 1.15 } catch (ClassCastException unused) {
1810 dl 1.1 return false;
1811 jsr166 1.15 } catch (NullPointerException unused) {
1812 dl 1.1 return false;
1813     }
1814     }
1815    
1816     /* ------ ConcurrentMap API methods ------ */
1817    
1818     /**
1819 jsr166 1.22 * {@inheritDoc}
1820     *
1821     * @return the previous value associated with the specified key,
1822     * or <tt>null</tt> if there was no mapping for the key
1823     * @throws ClassCastException if the specified key cannot be compared
1824     * with the keys currently in the map
1825     * @throws NullPointerException if the specified key or value is null
1826 dl 1.1 */
1827     public V putIfAbsent(K key, V value) {
1828 dl 1.9 if (value == null)
1829 dl 1.1 throw new NullPointerException();
1830     return doPut(key, value, true);
1831     }
1832    
1833     /**
1834 jsr166 1.22 * {@inheritDoc}
1835     *
1836     * @throws ClassCastException if the specified key cannot be compared
1837     * with the keys currently in the map
1838 dl 1.23 * @throws NullPointerException if the specified key is null
1839 dl 1.1 */
1840     public boolean remove(Object key, Object value) {
1841 dl 1.45 if (key == null)
1842     throw new NullPointerException();
1843 dl 1.9 if (value == null)
1844 dl 1.23 return false;
1845 dl 1.1 return doRemove(key, value) != null;
1846     }
1847    
1848     /**
1849 jsr166 1.22 * {@inheritDoc}
1850     *
1851     * @throws ClassCastException if the specified key cannot be compared
1852     * with the keys currently in the map
1853     * @throws NullPointerException if any of the arguments are null
1854 dl 1.1 */
1855     public boolean replace(K key, V oldValue, V newValue) {
1856 dl 1.9 if (oldValue == null || newValue == null)
1857 dl 1.1 throw new NullPointerException();
1858 dl 1.9 Comparable<? super K> k = comparable(key);
1859 dl 1.1 for (;;) {
1860     Node<K,V> n = findNode(k);
1861     if (n == null)
1862     return false;
1863     Object v = n.value;
1864     if (v != null) {
1865     if (!oldValue.equals(v))
1866     return false;
1867     if (n.casValue(v, newValue))
1868     return true;
1869     }
1870     }
1871     }
1872    
1873     /**
1874 jsr166 1.22 * {@inheritDoc}
1875     *
1876     * @return the previous value associated with the specified key,
1877     * or <tt>null</tt> if there was no mapping for the key
1878     * @throws ClassCastException if the specified key cannot be compared
1879     * with the keys currently in the map
1880     * @throws NullPointerException if the specified key or value is null
1881 dl 1.1 */
1882     public V replace(K key, V value) {
1883 dl 1.9 if (value == null)
1884 dl 1.1 throw new NullPointerException();
1885 dl 1.9 Comparable<? super K> k = comparable(key);
1886 dl 1.1 for (;;) {
1887     Node<K,V> n = findNode(k);
1888     if (n == null)
1889     return null;
1890     Object v = n.value;
1891     if (v != null && n.casValue(v, value))
1892     return (V)v;
1893     }
1894     }
1895    
1896     /* ------ SortedMap API methods ------ */
1897    
1898     public Comparator<? super K> comparator() {
1899     return comparator;
1900     }
1901    
1902     /**
1903 jsr166 1.22 * @throws NoSuchElementException {@inheritDoc}
1904 dl 1.1 */
1905 dl 1.9 public K firstKey() {
1906 dl 1.1 Node<K,V> n = findFirst();
1907     if (n == null)
1908     throw new NoSuchElementException();
1909     return n.key;
1910     }
1911    
1912     /**
1913 jsr166 1.22 * @throws NoSuchElementException {@inheritDoc}
1914 dl 1.1 */
1915     public K lastKey() {
1916     Node<K,V> n = findLast();
1917     if (n == null)
1918     throw new NoSuchElementException();
1919     return n.key;
1920     }
1921    
1922     /**
1923 jsr166 1.49 * @throws ClassCastException {@inheritDoc}
1924     * @throws NullPointerException if {@code fromKey} or {@code toKey} is null
1925 jsr166 1.22 * @throws IllegalArgumentException {@inheritDoc}
1926 dl 1.1 */
1927 dl 1.47 public ConcurrentNavigableMap<K,V> subMap(K fromKey,
1928     boolean fromInclusive,
1929     K toKey,
1930     boolean toInclusive) {
1931 dl 1.1 if (fromKey == null || toKey == null)
1932     throw new NullPointerException();
1933 dl 1.46 return new SubMap<K,V>
1934     (this, fromKey, fromInclusive, toKey, toInclusive, false);
1935 dl 1.1 }
1936    
1937     /**
1938 jsr166 1.49 * @throws ClassCastException {@inheritDoc}
1939     * @throws NullPointerException if {@code toKey} is null
1940 jsr166 1.22 * @throws IllegalArgumentException {@inheritDoc}
1941 dl 1.1 */
1942 dl 1.47 public ConcurrentNavigableMap<K,V> headMap(K toKey,
1943     boolean inclusive) {
1944 dl 1.1 if (toKey == null)
1945     throw new NullPointerException();
1946 dl 1.46 return new SubMap<K,V>
1947     (this, null, false, toKey, inclusive, false);
1948 dl 1.1 }
1949    
1950     /**
1951 jsr166 1.49 * @throws ClassCastException {@inheritDoc}
1952     * @throws NullPointerException if {@code fromKey} is null
1953 jsr166 1.22 * @throws IllegalArgumentException {@inheritDoc}
1954 dl 1.1 */
1955 dl 1.47 public ConcurrentNavigableMap<K,V> tailMap(K fromKey,
1956     boolean inclusive) {
1957 dl 1.6 if (fromKey == null)
1958     throw new NullPointerException();
1959 dl 1.46 return new SubMap<K,V>
1960     (this, fromKey, inclusive, null, false, false);
1961 dl 1.6 }
1962    
1963     /**
1964 jsr166 1.49 * @throws ClassCastException {@inheritDoc}
1965     * @throws NullPointerException if {@code fromKey} or {@code toKey} is null
1966 jsr166 1.22 * @throws IllegalArgumentException {@inheritDoc}
1967 dl 1.6 */
1968 dl 1.37 public ConcurrentNavigableMap<K,V> subMap(K fromKey, K toKey) {
1969 dl 1.47 return subMap(fromKey, true, toKey, false);
1970 dl 1.6 }
1971    
1972     /**
1973 jsr166 1.49 * @throws ClassCastException {@inheritDoc}
1974     * @throws NullPointerException if {@code toKey} is null
1975 jsr166 1.22 * @throws IllegalArgumentException {@inheritDoc}
1976 dl 1.6 */
1977 dl 1.37 public ConcurrentNavigableMap<K,V> headMap(K toKey) {
1978 dl 1.47 return headMap(toKey, false);
1979 dl 1.6 }
1980    
1981     /**
1982 jsr166 1.49 * @throws ClassCastException {@inheritDoc}
1983     * @throws NullPointerException if {@code fromKey} is null
1984 jsr166 1.22 * @throws IllegalArgumentException {@inheritDoc}
1985 dl 1.6 */
1986 dl 1.37 public ConcurrentNavigableMap<K,V> tailMap(K fromKey) {
1987 dl 1.47 return tailMap(fromKey, true);
1988 dl 1.1 }
1989    
1990     /* ---------------- Relational operations -------------- */
1991    
1992     /**
1993 jsr166 1.22 * Returns a key-value mapping associated with the greatest key
1994     * strictly less than the given key, or <tt>null</tt> if there is
1995     * no such key. The returned entry does <em>not</em> support the
1996     * <tt>Entry.setValue</tt> method.
1997 dl 1.9 *
1998 jsr166 1.22 * @throws ClassCastException {@inheritDoc}
1999     * @throws NullPointerException if the specified key is null
2000 dl 1.1 */
2001 jsr166 1.22 public Map.Entry<K,V> lowerEntry(K key) {
2002     return getNear(key, LT);
2003 dl 1.1 }
2004    
2005     /**
2006 jsr166 1.22 * @throws ClassCastException {@inheritDoc}
2007     * @throws NullPointerException if the specified key is null
2008 dl 1.1 */
2009 jsr166 1.22 public K lowerKey(K key) {
2010     Node<K,V> n = findNear(key, LT);
2011 jsr166 1.61 return (n == null) ? null : n.key;
2012 dl 1.1 }
2013    
2014     /**
2015 jsr166 1.22 * Returns a key-value mapping associated with the greatest key
2016     * less than or equal to the given key, or <tt>null</tt> if there
2017     * is no such key. The returned entry does <em>not</em> support
2018 dl 1.1 * the <tt>Entry.setValue</tt> method.
2019 dl 1.9 *
2020 jsr166 1.22 * @param key the key
2021     * @throws ClassCastException {@inheritDoc}
2022     * @throws NullPointerException if the specified key is null
2023 dl 1.1 */
2024 jsr166 1.22 public Map.Entry<K,V> floorEntry(K key) {
2025     return getNear(key, LT|EQ);
2026 dl 1.1 }
2027    
2028     /**
2029 jsr166 1.22 * @param key the key
2030     * @throws ClassCastException {@inheritDoc}
2031     * @throws NullPointerException if the specified key is null
2032 dl 1.1 */
2033 jsr166 1.22 public K floorKey(K key) {
2034     Node<K,V> n = findNear(key, LT|EQ);
2035 jsr166 1.61 return (n == null) ? null : n.key;
2036 dl 1.1 }
2037    
2038     /**
2039 jsr166 1.22 * Returns a key-value mapping associated with the least key
2040     * greater than or equal to the given key, or <tt>null</tt> if
2041     * there is no such entry. The returned entry does <em>not</em>
2042     * support the <tt>Entry.setValue</tt> method.
2043 dl 1.9 *
2044 jsr166 1.22 * @throws ClassCastException {@inheritDoc}
2045     * @throws NullPointerException if the specified key is null
2046 dl 1.1 */
2047 jsr166 1.22 public Map.Entry<K,V> ceilingEntry(K key) {
2048     return getNear(key, GT|EQ);
2049 dl 1.1 }
2050    
2051     /**
2052 jsr166 1.22 * @throws ClassCastException {@inheritDoc}
2053     * @throws NullPointerException if the specified key is null
2054 dl 1.1 */
2055 jsr166 1.22 public K ceilingKey(K key) {
2056     Node<K,V> n = findNear(key, GT|EQ);
2057 jsr166 1.61 return (n == null) ? null : n.key;
2058 dl 1.1 }
2059    
2060     /**
2061     * Returns a key-value mapping associated with the least key
2062     * strictly greater than the given key, or <tt>null</tt> if there
2063 jsr166 1.22 * is no such key. The returned entry does <em>not</em> support
2064 dl 1.1 * the <tt>Entry.setValue</tt> method.
2065 dl 1.9 *
2066 jsr166 1.22 * @param key the key
2067     * @throws ClassCastException {@inheritDoc}
2068     * @throws NullPointerException if the specified key is null
2069 dl 1.1 */
2070     public Map.Entry<K,V> higherEntry(K key) {
2071     return getNear(key, GT);
2072     }
2073    
2074     /**
2075 jsr166 1.22 * @param key the key
2076     * @throws ClassCastException {@inheritDoc}
2077     * @throws NullPointerException if the specified key is null
2078 dl 1.1 */
2079     public K higherKey(K key) {
2080     Node<K,V> n = findNear(key, GT);
2081 jsr166 1.61 return (n == null) ? null : n.key;
2082 dl 1.1 }
2083    
2084     /**
2085     * Returns a key-value mapping associated with the least
2086     * key in this map, or <tt>null</tt> if the map is empty.
2087     * The returned entry does <em>not</em> support
2088     * the <tt>Entry.setValue</tt> method.
2089     */
2090     public Map.Entry<K,V> firstEntry() {
2091     for (;;) {
2092     Node<K,V> n = findFirst();
2093 dl 1.9 if (n == null)
2094 dl 1.1 return null;
2095 dl 1.2 AbstractMap.SimpleImmutableEntry<K,V> e = n.createSnapshot();
2096 dl 1.1 if (e != null)
2097     return e;
2098     }
2099     }
2100    
2101     /**
2102     * Returns a key-value mapping associated with the greatest
2103     * key in this map, or <tt>null</tt> if the map is empty.
2104     * The returned entry does <em>not</em> support
2105     * the <tt>Entry.setValue</tt> method.
2106     */
2107     public Map.Entry<K,V> lastEntry() {
2108     for (;;) {
2109     Node<K,V> n = findLast();
2110 dl 1.9 if (n == null)
2111 dl 1.1 return null;
2112 dl 1.2 AbstractMap.SimpleImmutableEntry<K,V> e = n.createSnapshot();
2113 dl 1.1 if (e != null)
2114     return e;
2115     }
2116     }
2117    
2118     /**
2119     * Removes and returns a key-value mapping associated with
2120     * the least key in this map, or <tt>null</tt> if the map is empty.
2121     * The returned entry does <em>not</em> support
2122     * the <tt>Entry.setValue</tt> method.
2123     */
2124     public Map.Entry<K,V> pollFirstEntry() {
2125 dl 1.25 return doRemoveFirstEntry();
2126 dl 1.1 }
2127    
2128     /**
2129     * Removes and returns a key-value mapping associated with
2130     * the greatest key in this map, or <tt>null</tt> if the map is empty.
2131     * The returned entry does <em>not</em> support
2132     * the <tt>Entry.setValue</tt> method.
2133     */
2134     public Map.Entry<K,V> pollLastEntry() {
2135 dl 1.31 return doRemoveLastEntry();
2136 dl 1.1 }
2137    
2138    
2139     /* ---------------- Iterators -------------- */
2140    
2141     /**
2142 dl 1.46 * Base of iterator classes:
2143 dl 1.1 */
2144 dl 1.46 abstract class Iter<T> implements Iterator<T> {
2145 dl 1.1 /** the last node returned by next() */
2146 jsr166 1.52 Node<K,V> lastReturned;
2147 dl 1.1 /** the next node to return from next(); */
2148     Node<K,V> next;
2149 jsr166 1.55 /** Cache of next value field to maintain weak consistency */
2150     V nextValue;
2151 dl 1.1
2152 jsr166 1.13 /** Initializes ascending iterator for entire range. */
2153 dl 1.46 Iter() {
2154 dl 1.1 for (;;) {
2155 jsr166 1.55 next = findFirst();
2156 dl 1.1 if (next == null)
2157     break;
2158 jsr166 1.52 Object x = next.value;
2159     if (x != null && x != next) {
2160 jsr166 1.55 nextValue = (V) x;
2161 dl 1.1 break;
2162 jsr166 1.55 }
2163 dl 1.1 }
2164     }
2165    
2166 dl 1.46 public final boolean hasNext() {
2167     return next != null;
2168 dl 1.1 }
2169 dl 1.46
2170 jsr166 1.13 /** Advances next to higher entry. */
2171 dl 1.46 final void advance() {
2172 jsr166 1.54 if (next == null)
2173 dl 1.1 throw new NoSuchElementException();
2174 jsr166 1.55 lastReturned = next;
2175 dl 1.1 for (;;) {
2176 jsr166 1.55 next = next.next;
2177 dl 1.1 if (next == null)
2178     break;
2179 jsr166 1.52 Object x = next.value;
2180     if (x != null && x != next) {
2181 jsr166 1.55 nextValue = (V) x;
2182 dl 1.1 break;
2183 jsr166 1.55 }
2184 dl 1.1 }
2185     }
2186    
2187     public void remove() {
2188 jsr166 1.52 Node<K,V> l = lastReturned;
2189 dl 1.1 if (l == null)
2190     throw new IllegalStateException();
2191     // It would not be worth all of the overhead to directly
2192     // unlink from here. Using remove is fast enough.
2193     ConcurrentSkipListMap.this.remove(l.key);
2194 jsr166 1.55 lastReturned = null;
2195 dl 1.1 }
2196    
2197     }
2198    
2199 dl 1.46 final class ValueIterator extends Iter<V> {
2200 dl 1.9 public V next() {
2201 jsr166 1.52 V v = nextValue;
2202 dl 1.46 advance();
2203 jsr166 1.52 return v;
2204 dl 1.1 }
2205     }
2206    
2207 dl 1.46 final class KeyIterator extends Iter<K> {
2208 dl 1.9 public K next() {
2209 dl 1.1 Node<K,V> n = next;
2210 dl 1.46 advance();
2211 dl 1.1 return n.key;
2212     }
2213     }
2214    
2215 dl 1.46 final class EntryIterator extends Iter<Map.Entry<K,V>> {
2216     public Map.Entry<K,V> next() {
2217     Node<K,V> n = next;
2218 jsr166 1.52 V v = nextValue;
2219 dl 1.46 advance();
2220     return new AbstractMap.SimpleImmutableEntry<K,V>(n.key, v);
2221 dl 1.1 }
2222 dl 1.46 }
2223 dl 1.1
2224 dl 1.46 // Factory methods for iterators needed by ConcurrentSkipListSet etc
2225    
2226     Iterator<K> keyIterator() {
2227 dl 1.1 return new KeyIterator();
2228     }
2229    
2230 dl 1.46 Iterator<V> valueIterator() {
2231     return new ValueIterator();
2232 dl 1.1 }
2233    
2234 dl 1.46 Iterator<Map.Entry<K,V>> entryIterator() {
2235     return new EntryIterator();
2236 dl 1.1 }
2237    
2238 dl 1.46 /* ---------------- View Classes -------------- */
2239    
2240     /*
2241     * View classes are static, delegating to a ConcurrentNavigableMap
2242     * to allow use by SubMaps, which outweighs the ugliness of
2243     * needing type-tests for Iterator methods.
2244     */
2245    
2246 jsr166 1.53 static final <E> List<E> toList(Collection<E> c) {
2247 jsr166 1.55 // Using size() here would be a pessimization.
2248     List<E> list = new ArrayList<E>();
2249     for (E e : c)
2250     list.add(e);
2251     return list;
2252 jsr166 1.53 }
2253    
2254 jsr166 1.62 static final class KeySet<E>
2255     extends AbstractSet<E> implements NavigableSet<E> {
2256 dl 1.46 private final ConcurrentNavigableMap<E,Object> m;
2257     KeySet(ConcurrentNavigableMap<E,Object> map) { m = map; }
2258     public int size() { return m.size(); }
2259     public boolean isEmpty() { return m.isEmpty(); }
2260     public boolean contains(Object o) { return m.containsKey(o); }
2261     public boolean remove(Object o) { return m.remove(o) != null; }
2262     public void clear() { m.clear(); }
2263     public E lower(E e) { return m.lowerKey(e); }
2264     public E floor(E e) { return m.floorKey(e); }
2265     public E ceiling(E e) { return m.ceilingKey(e); }
2266     public E higher(E e) { return m.higherKey(e); }
2267     public Comparator<? super E> comparator() { return m.comparator(); }
2268     public E first() { return m.firstKey(); }
2269     public E last() { return m.lastKey(); }
2270     public E pollFirst() {
2271     Map.Entry<E,Object> e = m.pollFirstEntry();
2272 jsr166 1.61 return (e == null) ? null : e.getKey();
2273 dl 1.46 }
2274     public E pollLast() {
2275     Map.Entry<E,Object> e = m.pollLastEntry();
2276 jsr166 1.61 return (e == null) ? null : e.getKey();
2277 dl 1.46 }
2278     public Iterator<E> iterator() {
2279     if (m instanceof ConcurrentSkipListMap)
2280     return ((ConcurrentSkipListMap<E,Object>)m).keyIterator();
2281     else
2282     return ((ConcurrentSkipListMap.SubMap<E,Object>)m).keyIterator();
2283 dl 1.1 }
2284 dl 1.45 public boolean equals(Object o) {
2285     if (o == this)
2286     return true;
2287     if (!(o instanceof Set))
2288     return false;
2289     Collection<?> c = (Collection<?>) o;
2290     try {
2291     return containsAll(c) && c.containsAll(this);
2292     } catch (ClassCastException unused) {
2293     return false;
2294     } catch (NullPointerException unused) {
2295     return false;
2296     }
2297     }
2298 jsr166 1.55 public Object[] toArray() { return toList(this).toArray(); }
2299     public <T> T[] toArray(T[] a) { return toList(this).toArray(a); }
2300 dl 1.46 public Iterator<E> descendingIterator() {
2301     return descendingSet().iterator();
2302     }
2303 dl 1.47 public NavigableSet<E> subSet(E fromElement,
2304     boolean fromInclusive,
2305     E toElement,
2306     boolean toInclusive) {
2307 jsr166 1.56 return new KeySet<E>(m.subMap(fromElement, fromInclusive,
2308     toElement, toInclusive));
2309 dl 1.46 }
2310 dl 1.47 public NavigableSet<E> headSet(E toElement, boolean inclusive) {
2311 jsr166 1.56 return new KeySet<E>(m.headMap(toElement, inclusive));
2312 dl 1.46 }
2313 dl 1.47 public NavigableSet<E> tailSet(E fromElement, boolean inclusive) {
2314 jsr166 1.56 return new KeySet<E>(m.tailMap(fromElement, inclusive));
2315 dl 1.46 }
2316 jsr166 1.51 public NavigableSet<E> subSet(E fromElement, E toElement) {
2317 dl 1.47 return subSet(fromElement, true, toElement, false);
2318 dl 1.46 }
2319 jsr166 1.51 public NavigableSet<E> headSet(E toElement) {
2320 dl 1.47 return headSet(toElement, false);
2321 dl 1.46 }
2322 jsr166 1.51 public NavigableSet<E> tailSet(E fromElement) {
2323 dl 1.47 return tailSet(fromElement, true);
2324 dl 1.46 }
2325     public NavigableSet<E> descendingSet() {
2326 jsr166 1.56 return new KeySet(m.descendingMap());
2327 dl 1.46 }
2328 dl 1.1 }
2329    
2330 dl 1.46 static final class Values<E> extends AbstractCollection<E> {
2331     private final ConcurrentNavigableMap<Object, E> m;
2332     Values(ConcurrentNavigableMap<Object, E> map) {
2333     m = map;
2334 dl 1.1 }
2335 dl 1.46 public Iterator<E> iterator() {
2336     if (m instanceof ConcurrentSkipListMap)
2337     return ((ConcurrentSkipListMap<Object,E>)m).valueIterator();
2338     else
2339     return ((SubMap<Object,E>)m).valueIterator();
2340 dl 1.1 }
2341     public boolean isEmpty() {
2342 dl 1.46 return m.isEmpty();
2343 dl 1.1 }
2344     public int size() {
2345 dl 1.46 return m.size();
2346 dl 1.1 }
2347     public boolean contains(Object o) {
2348 dl 1.46 return m.containsValue(o);
2349 dl 1.1 }
2350     public void clear() {
2351 dl 1.46 m.clear();
2352 dl 1.1 }
2353 jsr166 1.55 public Object[] toArray() { return toList(this).toArray(); }
2354     public <T> T[] toArray(T[] a) { return toList(this).toArray(a); }
2355 dl 1.1 }
2356    
2357 dl 1.46 static final class EntrySet<K1,V1> extends AbstractSet<Map.Entry<K1,V1>> {
2358     private final ConcurrentNavigableMap<K1, V1> m;
2359     EntrySet(ConcurrentNavigableMap<K1, V1> map) {
2360     m = map;
2361 dl 1.1 }
2362 dl 1.46
2363     public Iterator<Map.Entry<K1,V1>> iterator() {
2364     if (m instanceof ConcurrentSkipListMap)
2365     return ((ConcurrentSkipListMap<K1,V1>)m).entryIterator();
2366     else
2367     return ((SubMap<K1,V1>)m).entryIterator();
2368     }
2369 dl 1.47
2370 dl 1.1 public boolean contains(Object o) {
2371     if (!(o instanceof Map.Entry))
2372     return false;
2373 dl 1.46 Map.Entry<K1,V1> e = (Map.Entry<K1,V1>)o;
2374     V1 v = m.get(e.getKey());
2375 dl 1.1 return v != null && v.equals(e.getValue());
2376     }
2377     public boolean remove(Object o) {
2378     if (!(o instanceof Map.Entry))
2379     return false;
2380 dl 1.46 Map.Entry<K1,V1> e = (Map.Entry<K1,V1>)o;
2381     return m.remove(e.getKey(),
2382 dl 1.47 e.getValue());
2383 dl 1.1 }
2384     public boolean isEmpty() {
2385 dl 1.46 return m.isEmpty();
2386 dl 1.1 }
2387     public int size() {
2388 dl 1.46 return m.size();
2389 dl 1.1 }
2390     public void clear() {
2391 dl 1.46 m.clear();
2392 dl 1.1 }
2393 dl 1.45 public boolean equals(Object o) {
2394     if (o == this)
2395     return true;
2396     if (!(o instanceof Set))
2397     return false;
2398     Collection<?> c = (Collection<?>) o;
2399     try {
2400     return containsAll(c) && c.containsAll(this);
2401     } catch (ClassCastException unused) {
2402     return false;
2403     } catch (NullPointerException unused) {
2404     return false;
2405     }
2406     }
2407 jsr166 1.55 public Object[] toArray() { return toList(this).toArray(); }
2408     public <T> T[] toArray(T[] a) { return toList(this).toArray(a); }
2409 dl 1.1 }
2410    
2411     /**
2412     * Submaps returned by {@link ConcurrentSkipListMap} submap operations
2413     * represent a subrange of mappings of their underlying
2414     * maps. Instances of this class support all methods of their
2415     * underlying maps, differing in that mappings outside their range are
2416     * ignored, and attempts to add mappings outside their ranges result
2417     * in {@link IllegalArgumentException}. Instances of this class are
2418     * constructed only using the <tt>subMap</tt>, <tt>headMap</tt>, and
2419     * <tt>tailMap</tt> methods of their underlying maps.
2420 jsr166 1.52 *
2421     * @serial include
2422 dl 1.1 */
2423 dl 1.46 static final class SubMap<K,V> extends AbstractMap<K,V>
2424     implements ConcurrentNavigableMap<K,V>, Cloneable,
2425     java.io.Serializable {
2426 dl 1.1 private static final long serialVersionUID = -7647078645895051609L;
2427    
2428     /** Underlying map */
2429     private final ConcurrentSkipListMap<K,V> m;
2430     /** lower bound key, or null if from start */
2431 dl 1.46 private final K lo;
2432     /** upper bound key, or null if to end */
2433     private final K hi;
2434     /** inclusion flag for lo */
2435     private final boolean loInclusive;
2436     /** inclusion flag for hi */
2437     private final boolean hiInclusive;
2438     /** direction */
2439     private final boolean isDescending;
2440    
2441 dl 1.1 // Lazily initialized view holders
2442 dl 1.46 private transient KeySet<K> keySetView;
2443 dl 1.1 private transient Set<Map.Entry<K,V>> entrySetView;
2444     private transient Collection<V> valuesView;
2445    
2446     /**
2447 dl 1.46 * Creates a new submap, initializing all fields
2448     */
2449     SubMap(ConcurrentSkipListMap<K,V> map,
2450     K fromKey, boolean fromInclusive,
2451     K toKey, boolean toInclusive,
2452     boolean isDescending) {
2453 dl 1.47 if (fromKey != null && toKey != null &&
2454 dl 1.46 map.compare(fromKey, toKey) > 0)
2455 dl 1.1 throw new IllegalArgumentException("inconsistent range");
2456     this.m = map;
2457 dl 1.46 this.lo = fromKey;
2458     this.hi = toKey;
2459     this.loInclusive = fromInclusive;
2460     this.hiInclusive = toInclusive;
2461     this.isDescending = isDescending;
2462 dl 1.1 }
2463    
2464     /* ---------------- Utilities -------------- */
2465    
2466 dl 1.46 private boolean tooLow(K key) {
2467     if (lo != null) {
2468     int c = m.compare(key, lo);
2469     if (c < 0 || (c == 0 && !loInclusive))
2470     return true;
2471     }
2472     return false;
2473 dl 1.1 }
2474    
2475 dl 1.46 private boolean tooHigh(K key) {
2476     if (hi != null) {
2477     int c = m.compare(key, hi);
2478     if (c > 0 || (c == 0 && !hiInclusive))
2479     return true;
2480     }
2481     return false;
2482 dl 1.1 }
2483    
2484 dl 1.46 private boolean inBounds(K key) {
2485     return !tooLow(key) && !tooHigh(key);
2486 dl 1.1 }
2487    
2488 dl 1.46 private void checkKeyBounds(K key) throws IllegalArgumentException {
2489     if (key == null)
2490     throw new NullPointerException();
2491     if (!inBounds(key))
2492     throw new IllegalArgumentException("key out of range");
2493 dl 1.1 }
2494    
2495 dl 1.46 /**
2496     * Returns true if node key is less than upper bound of range
2497     */
2498     private boolean isBeforeEnd(ConcurrentSkipListMap.Node<K,V> n) {
2499     if (n == null)
2500     return false;
2501     if (hi == null)
2502     return true;
2503     K k = n.key;
2504     if (k == null) // pass by markers and headers
2505     return true;
2506     int c = m.compare(k, hi);
2507     if (c > 0 || (c == 0 && !hiInclusive))
2508     return false;
2509     return true;
2510 dl 1.1 }
2511    
2512 dl 1.46 /**
2513     * Returns lowest node. This node might not be in range, so
2514     * most usages need to check bounds
2515     */
2516     private ConcurrentSkipListMap.Node<K,V> loNode() {
2517     if (lo == null)
2518     return m.findFirst();
2519     else if (loInclusive)
2520     return m.findNear(lo, m.GT|m.EQ);
2521     else
2522     return m.findNear(lo, m.GT);
2523 dl 1.1 }
2524    
2525     /**
2526 dl 1.46 * Returns highest node. This node might not be in range, so
2527     * most usages need to check bounds
2528 dl 1.1 */
2529 dl 1.46 private ConcurrentSkipListMap.Node<K,V> hiNode() {
2530     if (hi == null)
2531     return m.findLast();
2532     else if (hiInclusive)
2533     return m.findNear(hi, m.LT|m.EQ);
2534     else
2535     return m.findNear(hi, m.LT);
2536 dl 1.1 }
2537    
2538     /**
2539 dl 1.46 * Returns lowest absolute key (ignoring directonality)
2540 dl 1.1 */
2541 dl 1.46 private K lowestKey() {
2542     ConcurrentSkipListMap.Node<K,V> n = loNode();
2543     if (isBeforeEnd(n))
2544     return n.key;
2545     else
2546     throw new NoSuchElementException();
2547 dl 1.47 }
2548 dl 1.46
2549     /**
2550     * Returns highest absolute key (ignoring directonality)
2551     */
2552     private K highestKey() {
2553     ConcurrentSkipListMap.Node<K,V> n = hiNode();
2554     if (n != null) {
2555     K last = n.key;
2556     if (inBounds(last))
2557     return last;
2558     }
2559     throw new NoSuchElementException();
2560     }
2561    
2562     private Map.Entry<K,V> lowestEntry() {
2563     for (;;) {
2564     ConcurrentSkipListMap.Node<K,V> n = loNode();
2565     if (!isBeforeEnd(n))
2566     return null;
2567     Map.Entry<K,V> e = n.createSnapshot();
2568     if (e != null)
2569     return e;
2570     }
2571     }
2572    
2573     private Map.Entry<K,V> highestEntry() {
2574     for (;;) {
2575     ConcurrentSkipListMap.Node<K,V> n = hiNode();
2576     if (n == null || !inBounds(n.key))
2577     return null;
2578     Map.Entry<K,V> e = n.createSnapshot();
2579     if (e != null)
2580     return e;
2581     }
2582     }
2583    
2584     private Map.Entry<K,V> removeLowest() {
2585     for (;;) {
2586     Node<K,V> n = loNode();
2587     if (n == null)
2588     return null;
2589     K k = n.key;
2590     if (!inBounds(k))
2591     return null;
2592     V v = m.doRemove(k, null);
2593     if (v != null)
2594     return new AbstractMap.SimpleImmutableEntry<K,V>(k, v);
2595     }
2596     }
2597    
2598     private Map.Entry<K,V> removeHighest() {
2599     for (;;) {
2600     Node<K,V> n = hiNode();
2601     if (n == null)
2602     return null;
2603     K k = n.key;
2604     if (!inBounds(k))
2605     return null;
2606     V v = m.doRemove(k, null);
2607     if (v != null)
2608     return new AbstractMap.SimpleImmutableEntry<K,V>(k, v);
2609     }
2610 dl 1.1 }
2611    
2612     /**
2613 dl 1.46 * Submap version of ConcurrentSkipListMap.getNearEntry
2614 dl 1.1 */
2615 dl 1.46 private Map.Entry<K,V> getNearEntry(K key, int rel) {
2616     if (isDescending) { // adjust relation for direction
2617     if ((rel & m.LT) == 0)
2618     rel |= m.LT;
2619     else
2620     rel &= ~m.LT;
2621     }
2622     if (tooLow(key))
2623 jsr166 1.61 return ((rel & m.LT) != 0) ? null : lowestEntry();
2624 dl 1.46 if (tooHigh(key))
2625 jsr166 1.61 return ((rel & m.LT) != 0) ? highestEntry() : null;
2626 dl 1.46 for (;;) {
2627     Node<K,V> n = m.findNear(key, rel);
2628     if (n == null || !inBounds(n.key))
2629     return null;
2630     K k = n.key;
2631     V v = n.getValidValue();
2632     if (v != null)
2633     return new AbstractMap.SimpleImmutableEntry<K,V>(k, v);
2634     }
2635 dl 1.1 }
2636    
2637 jsr166 1.48 // Almost the same as getNearEntry, except for keys
2638 dl 1.46 private K getNearKey(K key, int rel) {
2639     if (isDescending) { // adjust relation for direction
2640     if ((rel & m.LT) == 0)
2641     rel |= m.LT;
2642     else
2643     rel &= ~m.LT;
2644     }
2645     if (tooLow(key)) {
2646     if ((rel & m.LT) == 0) {
2647     ConcurrentSkipListMap.Node<K,V> n = loNode();
2648     if (isBeforeEnd(n))
2649     return n.key;
2650     }
2651     return null;
2652     }
2653     if (tooHigh(key)) {
2654     if ((rel & m.LT) != 0) {
2655     ConcurrentSkipListMap.Node<K,V> n = hiNode();
2656     if (n != null) {
2657     K last = n.key;
2658     if (inBounds(last))
2659     return last;
2660     }
2661     }
2662     return null;
2663     }
2664     for (;;) {
2665     Node<K,V> n = m.findNear(key, rel);
2666     if (n == null || !inBounds(n.key))
2667     return null;
2668     K k = n.key;
2669     V v = n.getValidValue();
2670     if (v != null)
2671     return k;
2672     }
2673     }
2674 dl 1.1
2675     /* ---------------- Map API methods -------------- */
2676    
2677     public boolean containsKey(Object key) {
2678 dl 1.46 if (key == null) throw new NullPointerException();
2679 dl 1.1 K k = (K)key;
2680 dl 1.46 return inBounds(k) && m.containsKey(k);
2681 dl 1.1 }
2682    
2683     public V get(Object key) {
2684 dl 1.46 if (key == null) throw new NullPointerException();
2685 dl 1.1 K k = (K)key;
2686 dl 1.46 return ((!inBounds(k)) ? null : m.get(k));
2687 dl 1.1 }
2688    
2689     public V put(K key, V value) {
2690 dl 1.46 checkKeyBounds(key);
2691 dl 1.1 return m.put(key, value);
2692     }
2693    
2694     public V remove(Object key) {
2695     K k = (K)key;
2696 jsr166 1.61 return (!inBounds(k)) ? null : m.remove(k);
2697 dl 1.1 }
2698    
2699     public int size() {
2700     long count = 0;
2701 dl 1.46 for (ConcurrentSkipListMap.Node<K,V> n = loNode();
2702 dl 1.9 isBeforeEnd(n);
2703 dl 1.1 n = n.next) {
2704     if (n.getValidValue() != null)
2705     ++count;
2706     }
2707 jsr166 1.61 return count >= Integer.MAX_VALUE ? Integer.MAX_VALUE : (int)count;
2708 dl 1.1 }
2709    
2710     public boolean isEmpty() {
2711 dl 1.46 return !isBeforeEnd(loNode());
2712 dl 1.1 }
2713    
2714     public boolean containsValue(Object value) {
2715 dl 1.9 if (value == null)
2716 dl 1.1 throw new NullPointerException();
2717 dl 1.46 for (ConcurrentSkipListMap.Node<K,V> n = loNode();
2718 dl 1.9 isBeforeEnd(n);
2719 dl 1.1 n = n.next) {
2720     V v = n.getValidValue();
2721     if (v != null && value.equals(v))
2722     return true;
2723     }
2724     return false;
2725     }
2726    
2727     public void clear() {
2728 dl 1.46 for (ConcurrentSkipListMap.Node<K,V> n = loNode();
2729 dl 1.9 isBeforeEnd(n);
2730 dl 1.1 n = n.next) {
2731     if (n.getValidValue() != null)
2732     m.remove(n.key);
2733     }
2734     }
2735    
2736     /* ---------------- ConcurrentMap API methods -------------- */
2737    
2738     public V putIfAbsent(K key, V value) {
2739 dl 1.46 checkKeyBounds(key);
2740 dl 1.1 return m.putIfAbsent(key, value);
2741     }
2742    
2743     public boolean remove(Object key, Object value) {
2744     K k = (K)key;
2745 dl 1.46 return inBounds(k) && m.remove(k, value);
2746 dl 1.1 }
2747    
2748     public boolean replace(K key, V oldValue, V newValue) {
2749 dl 1.46 checkKeyBounds(key);
2750 dl 1.1 return m.replace(key, oldValue, newValue);
2751     }
2752    
2753     public V replace(K key, V value) {
2754 dl 1.46 checkKeyBounds(key);
2755 dl 1.1 return m.replace(key, value);
2756     }
2757    
2758     /* ---------------- SortedMap API methods -------------- */
2759    
2760     public Comparator<? super K> comparator() {
2761 dl 1.46 Comparator<? super K> cmp = m.comparator();
2762 jsr166 1.55 if (isDescending)
2763     return Collections.reverseOrder(cmp);
2764     else
2765     return cmp;
2766 dl 1.1 }
2767 dl 1.47
2768 dl 1.46 /**
2769     * Utility to create submaps, where given bounds override
2770     * unbounded(null) ones and/or are checked against bounded ones.
2771     */
2772 dl 1.47 private SubMap<K,V> newSubMap(K fromKey,
2773     boolean fromInclusive,
2774     K toKey,
2775 dl 1.46 boolean toInclusive) {
2776     if (isDescending) { // flip senses
2777 dl 1.47 K tk = fromKey;
2778     fromKey = toKey;
2779 dl 1.46 toKey = tk;
2780 dl 1.47 boolean ti = fromInclusive;
2781     fromInclusive = toInclusive;
2782 dl 1.46 toInclusive = ti;
2783     }
2784     if (lo != null) {
2785     if (fromKey == null) {
2786     fromKey = lo;
2787     fromInclusive = loInclusive;
2788     }
2789     else {
2790     int c = m.compare(fromKey, lo);
2791     if (c < 0 || (c == 0 && !loInclusive && fromInclusive))
2792     throw new IllegalArgumentException("key out of range");
2793     }
2794     }
2795     if (hi != null) {
2796     if (toKey == null) {
2797     toKey = hi;
2798     toInclusive = hiInclusive;
2799     }
2800     else {
2801     int c = m.compare(toKey, hi);
2802     if (c > 0 || (c == 0 && !hiInclusive && toInclusive))
2803     throw new IllegalArgumentException("key out of range");
2804     }
2805 dl 1.1 }
2806 dl 1.47 return new SubMap<K,V>(m, fromKey, fromInclusive,
2807 dl 1.46 toKey, toInclusive, isDescending);
2808 dl 1.1 }
2809    
2810 dl 1.47 public SubMap<K,V> subMap(K fromKey,
2811     boolean fromInclusive,
2812     K toKey,
2813     boolean toInclusive) {
2814 dl 1.1 if (fromKey == null || toKey == null)
2815     throw new NullPointerException();
2816 dl 1.46 return newSubMap(fromKey, fromInclusive, toKey, toInclusive);
2817 dl 1.1 }
2818 dl 1.47
2819     public SubMap<K,V> headMap(K toKey,
2820     boolean inclusive) {
2821 dl 1.1 if (toKey == null)
2822     throw new NullPointerException();
2823 dl 1.46 return newSubMap(null, false, toKey, inclusive);
2824 dl 1.1 }
2825 dl 1.47
2826     public SubMap<K,V> tailMap(K fromKey,
2827     boolean inclusive) {
2828 dl 1.1 if (fromKey == null)
2829     throw new NullPointerException();
2830 dl 1.46 return newSubMap(fromKey, inclusive, null, false);
2831     }
2832    
2833     public SubMap<K,V> subMap(K fromKey, K toKey) {
2834 dl 1.47 return subMap(fromKey, true, toKey, false);
2835 dl 1.1 }
2836    
2837 dl 1.46 public SubMap<K,V> headMap(K toKey) {
2838 dl 1.47 return headMap(toKey, false);
2839 dl 1.6 }
2840    
2841 dl 1.46 public SubMap<K,V> tailMap(K fromKey) {
2842 dl 1.47 return tailMap(fromKey, true);
2843 dl 1.6 }
2844    
2845 dl 1.46 public SubMap<K,V> descendingMap() {
2846 dl 1.47 return new SubMap<K,V>(m, lo, loInclusive,
2847 dl 1.46 hi, hiInclusive, !isDescending);
2848 dl 1.6 }
2849    
2850 dl 1.1 /* ---------------- Relational methods -------------- */
2851    
2852     public Map.Entry<K,V> ceilingEntry(K key) {
2853 dl 1.46 return getNearEntry(key, (m.GT|m.EQ));
2854 dl 1.1 }
2855    
2856     public K ceilingKey(K key) {
2857 dl 1.46 return getNearKey(key, (m.GT|m.EQ));
2858 dl 1.1 }
2859    
2860     public Map.Entry<K,V> lowerEntry(K key) {
2861 dl 1.46 return getNearEntry(key, (m.LT));
2862 dl 1.1 }
2863    
2864     public K lowerKey(K key) {
2865 dl 1.46 return getNearKey(key, (m.LT));
2866 dl 1.1 }
2867    
2868     public Map.Entry<K,V> floorEntry(K key) {
2869 dl 1.46 return getNearEntry(key, (m.LT|m.EQ));
2870 dl 1.1 }
2871    
2872     public K floorKey(K key) {
2873 dl 1.46 return getNearKey(key, (m.LT|m.EQ));
2874 dl 1.1 }
2875    
2876     public Map.Entry<K,V> higherEntry(K key) {
2877 dl 1.46 return getNearEntry(key, (m.GT));
2878 dl 1.1 }
2879    
2880     public K higherKey(K key) {
2881 dl 1.46 return getNearKey(key, (m.GT));
2882     }
2883    
2884     public K firstKey() {
2885 jsr166 1.61 return isDescending ? highestKey() : lowestKey();
2886 dl 1.46 }
2887    
2888     public K lastKey() {
2889 jsr166 1.61 return isDescending ? lowestKey() : highestKey();
2890 dl 1.1 }
2891    
2892     public Map.Entry<K,V> firstEntry() {
2893 jsr166 1.61 return isDescending ? highestEntry() : lowestEntry();
2894 dl 1.1 }
2895    
2896     public Map.Entry<K,V> lastEntry() {
2897 jsr166 1.61 return isDescending ? lowestEntry() : highestEntry();
2898 dl 1.1 }
2899    
2900     public Map.Entry<K,V> pollFirstEntry() {
2901 jsr166 1.61 return isDescending ? removeHighest() : removeLowest();
2902 dl 1.1 }
2903    
2904     public Map.Entry<K,V> pollLastEntry() {
2905 jsr166 1.61 return isDescending ? removeLowest() : removeHighest();
2906 dl 1.1 }
2907    
2908     /* ---------------- Submap Views -------------- */
2909    
2910 jsr166 1.51 public NavigableSet<K> keySet() {
2911 dl 1.46 KeySet<K> ks = keySetView;
2912     return (ks != null) ? ks : (keySetView = new KeySet(this));
2913 dl 1.1 }
2914    
2915 dl 1.46 public NavigableSet<K> navigableKeySet() {
2916     KeySet<K> ks = keySetView;
2917     return (ks != null) ? ks : (keySetView = new KeySet(this));
2918     }
2919 dl 1.45
2920 dl 1.46 public Collection<V> values() {
2921     Collection<V> vs = valuesView;
2922     return (vs != null) ? vs : (valuesView = new Values(this));
2923 dl 1.1 }
2924    
2925 dl 1.46 public Set<Map.Entry<K,V>> entrySet() {
2926     Set<Map.Entry<K,V>> es = entrySetView;
2927     return (es != null) ? es : (entrySetView = new EntrySet(this));
2928 dl 1.1 }
2929    
2930 dl 1.46 public NavigableSet<K> descendingKeySet() {
2931     return descendingMap().navigableKeySet();
2932 dl 1.1 }
2933    
2934 dl 1.46 Iterator<K> keyIterator() {
2935     return new SubMapKeyIterator();
2936 dl 1.1 }
2937    
2938 dl 1.46 Iterator<V> valueIterator() {
2939     return new SubMapValueIterator();
2940 dl 1.1 }
2941    
2942 dl 1.46 Iterator<Map.Entry<K,V>> entryIterator() {
2943     return new SubMapEntryIterator();
2944 dl 1.1 }
2945    
2946 dl 1.46 /**
2947     * Variant of main Iter class to traverse through submaps.
2948     */
2949     abstract class SubMapIter<T> implements Iterator<T> {
2950     /** the last node returned by next() */
2951 jsr166 1.52 Node<K,V> lastReturned;
2952 dl 1.46 /** the next node to return from next(); */
2953     Node<K,V> next;
2954     /** Cache of next value field to maintain weak consistency */
2955 jsr166 1.52 V nextValue;
2956 dl 1.46
2957 dl 1.47 SubMapIter() {
2958 dl 1.46 for (;;) {
2959 jsr166 1.52 next = isDescending ? hiNode() : loNode();
2960 dl 1.46 if (next == null)
2961     break;
2962 jsr166 1.55 Object x = next.value;
2963 jsr166 1.52 if (x != null && x != next) {
2964 jsr166 1.55 if (! inBounds(next.key))
2965 dl 1.46 next = null;
2966 jsr166 1.55 else
2967     nextValue = (V) x;
2968 dl 1.46 break;
2969     }
2970     }
2971 dl 1.1 }
2972 dl 1.46
2973     public final boolean hasNext() {
2974     return next != null;
2975 dl 1.1 }
2976 dl 1.46
2977     final void advance() {
2978 jsr166 1.54 if (next == null)
2979 dl 1.46 throw new NoSuchElementException();
2980 jsr166 1.55 lastReturned = next;
2981 dl 1.46 if (isDescending)
2982     descend();
2983     else
2984     ascend();
2985 dl 1.1 }
2986 dl 1.46
2987     private void ascend() {
2988     for (;;) {
2989     next = next.next;
2990     if (next == null)
2991     break;
2992 jsr166 1.55 Object x = next.value;
2993 jsr166 1.52 if (x != null && x != next) {
2994     if (tooHigh(next.key))
2995 dl 1.46 next = null;
2996 jsr166 1.52 else
2997 jsr166 1.55 nextValue = (V) x;
2998 dl 1.46 break;
2999     }
3000     }
3001     }
3002    
3003     private void descend() {
3004     for (;;) {
3005 jsr166 1.52 next = m.findNear(lastReturned.key, LT);
3006 dl 1.46 if (next == null)
3007     break;
3008 jsr166 1.55 Object x = next.value;
3009 jsr166 1.52 if (x != null && x != next) {
3010     if (tooLow(next.key))
3011 dl 1.46 next = null;
3012 jsr166 1.55 else
3013 jsr166 1.52 nextValue = (V) x;
3014 dl 1.46 break;
3015     }
3016     }
3017 dl 1.1 }
3018 dl 1.46
3019     public void remove() {
3020 jsr166 1.52 Node<K,V> l = lastReturned;
3021 dl 1.46 if (l == null)
3022     throw new IllegalStateException();
3023     m.remove(l.key);
3024 jsr166 1.55 lastReturned = null;
3025 dl 1.1 }
3026 dl 1.46
3027     }
3028    
3029     final class SubMapValueIterator extends SubMapIter<V> {
3030     public V next() {
3031 jsr166 1.52 V v = nextValue;
3032 dl 1.46 advance();
3033 jsr166 1.52 return v;
3034 dl 1.45 }
3035 dl 1.1 }
3036    
3037 dl 1.46 final class SubMapKeyIterator extends SubMapIter<K> {
3038     public K next() {
3039     Node<K,V> n = next;
3040     advance();
3041     return n.key;
3042     }
3043 dl 1.1 }
3044    
3045 dl 1.46 final class SubMapEntryIterator extends SubMapIter<Map.Entry<K,V>> {
3046     public Map.Entry<K,V> next() {
3047     Node<K,V> n = next;
3048 jsr166 1.52 V v = nextValue;
3049 dl 1.46 advance();
3050     return new AbstractMap.SimpleImmutableEntry<K,V>(n.key, v);
3051 dl 1.1 }
3052     }
3053     }
3054 dl 1.59
3055     // Unsafe mechanics
3056     private static final sun.misc.Unsafe UNSAFE = sun.misc.Unsafe.getUnsafe();
3057     private static final long headOffset =
3058     objectFieldOffset(UNSAFE, "head", ConcurrentSkipListMap.class);
3059 jsr166 1.60
3060 dl 1.59 static long objectFieldOffset(sun.misc.Unsafe UNSAFE,
3061     String field, Class<?> klazz) {
3062     try {
3063     return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
3064     } catch (NoSuchFieldException e) {
3065     // Convert Exception to corresponding Error
3066     NoSuchFieldError error = new NoSuchFieldError(field);
3067     error.initCause(e);
3068     throw error;
3069     }
3070     }
3071    
3072 dl 1.1 }