Upstream version 8.37.180.0
[platform/framework/web/crosswalk.git] / src / third_party / WebKit / Source / wtf / HashTable.h
1 /*
2  * Copyright (C) 2005, 2006, 2007, 2008, 2011, 2012 Apple Inc. All rights reserved.
3  * Copyright (C) 2008 David Levin <levin@chromium.org>
4  *
5  * This library is free software; you can redistribute it and/or
6  * modify it under the terms of the GNU Library General Public
7  * License as published by the Free Software Foundation; either
8  * version 2 of the License, or (at your option) any later version.
9  *
10  * This library is distributed in the hope that it will be useful,
11  * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU * Library General Public License for more details.
12  *
13  * You should have received a copy of the GNU Library General Public License
14  * along with this library; see the file COPYING.LIB.  If not, write to
15  * the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
16  * Boston, MA 02110-1301, USA.
17  *
18  */
19
20 #ifndef WTF_HashTable_h
21 #define WTF_HashTable_h
22
23 #include "wtf/Alignment.h"
24 #include "wtf/Assertions.h"
25 #include "wtf/DefaultAllocator.h"
26 #include "wtf/HashTraits.h"
27 #include "wtf/WTF.h"
28
29 #define DUMP_HASHTABLE_STATS 0
30 #define DUMP_HASHTABLE_STATS_PER_TABLE 0
31
32 #if DUMP_HASHTABLE_STATS_PER_TABLE
33 #include "wtf/DataLog.h"
34 #endif
35
36 #if DUMP_HASHTABLE_STATS
37 #if DUMP_HASHTABLE_STATS_PER_TABLE
38 #define UPDATE_PROBE_COUNTS()                            \
39     ++probeCount;                                        \
40     HashTableStats::recordCollisionAtCount(probeCount);  \
41     ++perTableProbeCount;                                \
42     m_stats->recordCollisionAtCount(perTableProbeCount)
43 #define UPDATE_ACCESS_COUNTS()                           \
44     atomicIncrement(&HashTableStats::numAccesses);       \
45     int probeCount = 0;                                  \
46     ++m_stats->numAccesses;                              \
47     int perTableProbeCount = 0
48 #else
49 #define UPDATE_PROBE_COUNTS()                            \
50     ++probeCount;                                        \
51     HashTableStats::recordCollisionAtCount(probeCount)
52 #define UPDATE_ACCESS_COUNTS()                           \
53     atomicIncrement(&HashTableStats::numAccesses);       \
54     int probeCount = 0
55 #endif
56 #else
57 #if DUMP_HASHTABLE_STATS_PER_TABLE
58 #define UPDATE_PROBE_COUNTS()                            \
59     ++perTableProbeCount;                                \
60     m_stats->recordCollisionAtCount(perTableProbeCount)
61 #define UPDATE_ACCESS_COUNTS()                           \
62     ++m_stats->numAccesses;                              \
63     int perTableProbeCount = 0
64 #else
65 #define UPDATE_PROBE_COUNTS() do { } while (0)
66 #define UPDATE_ACCESS_COUNTS() do { } while (0)
67 #endif
68 #endif
69
70 namespace WTF {
71
72 #if DUMP_HASHTABLE_STATS
73
74     struct HashTableStats {
75         // The following variables are all atomically incremented when modified.
76         static int numAccesses;
77         static int numRehashes;
78         static int numRemoves;
79         static int numReinserts;
80
81         // The following variables are only modified in the recordCollisionAtCount method within a mutex.
82         static int maxCollisions;
83         static int numCollisions;
84         static int collisionGraph[4096];
85
86         static void recordCollisionAtCount(int count);
87         static void dumpStats();
88     };
89
90 #endif
91
92     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
93     class HashTable;
94     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
95     class HashTableIterator;
96     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
97     class HashTableConstIterator;
98     template<typename Value, typename HashFunctions, typename HashTraits, typename Allocator>
99     class LinkedHashSet;
100     template<WeakHandlingFlag x, typename T, typename U, typename V, typename W, typename X, typename Y, typename Z>
101     struct WeakProcessingHashTableHelper;
102
103     typedef enum { HashItemKnownGood } HashItemKnownGoodTag;
104
105     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
106     class HashTableConstIterator {
107     private:
108         typedef HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator> HashTableType;
109         typedef HashTableIterator<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator> iterator;
110         typedef HashTableConstIterator<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator> const_iterator;
111         typedef Value ValueType;
112         typedef typename Traits::IteratorConstGetType GetType;
113         typedef const ValueType* PointerType;
114
115         friend class HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>;
116         friend class HashTableIterator<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>;
117
118         void skipEmptyBuckets()
119         {
120             while (m_position != m_endPosition && HashTableType::isEmptyOrDeletedBucket(*m_position))
121                 ++m_position;
122         }
123
124         HashTableConstIterator(PointerType position, PointerType endPosition, const HashTableType* container)
125             : m_position(position)
126             , m_endPosition(endPosition)
127 #ifdef ASSERT_ENABLED
128             , m_container(container)
129             , m_containerModifications(container->modifications())
130 #endif
131         {
132             skipEmptyBuckets();
133         }
134
135         HashTableConstIterator(PointerType position, PointerType endPosition, const HashTableType* container, HashItemKnownGoodTag)
136             : m_position(position)
137             , m_endPosition(endPosition)
138 #ifdef ASSERT_ENABLED
139             , m_container(container)
140             , m_containerModifications(container->modifications())
141 #endif
142         {
143             ASSERT(m_containerModifications == m_container->modifications());
144         }
145
146         void checkModifications() const
147         {
148             // HashTable and collections that build on it do not support
149             // modifications while there is an iterator in use. The exception
150             // is ListHashSet, which has its own iterators that tolerate
151             // modification of the underlying set.
152             ASSERT(m_containerModifications == m_container->modifications());
153         }
154
155     public:
156         HashTableConstIterator()
157         {
158         }
159
160         GetType get() const
161         {
162             checkModifications();
163             return m_position;
164         }
165         typename Traits::IteratorConstReferenceType operator*() const { return Traits::getToReferenceConstConversion(get()); }
166         GetType operator->() const { return get(); }
167
168         const_iterator& operator++()
169         {
170             ASSERT(m_position != m_endPosition);
171             checkModifications();
172             ++m_position;
173             skipEmptyBuckets();
174             return *this;
175         }
176
177         // postfix ++ intentionally omitted
178
179         // Comparison.
180         bool operator==(const const_iterator& other) const
181         {
182             return m_position == other.m_position;
183         }
184         bool operator!=(const const_iterator& other) const
185         {
186             return m_position != other.m_position;
187         }
188         bool operator==(const iterator& other) const
189         {
190             return *this == static_cast<const_iterator>(other);
191         }
192         bool operator!=(const iterator& other) const
193         {
194             return *this != static_cast<const_iterator>(other);
195         }
196
197     private:
198         PointerType m_position;
199         PointerType m_endPosition;
200 #ifdef ASSERT_ENABLED
201         const HashTableType* m_container;
202         int64_t m_containerModifications;
203 #endif
204     };
205
206     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
207     class HashTableIterator {
208     private:
209         typedef HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator> HashTableType;
210         typedef HashTableIterator<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator> iterator;
211         typedef HashTableConstIterator<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator> const_iterator;
212         typedef Value ValueType;
213         typedef typename Traits::IteratorGetType GetType;
214         typedef ValueType* PointerType;
215
216         friend class HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>;
217
218         HashTableIterator(PointerType pos, PointerType end, const HashTableType* container) : m_iterator(pos, end, container) { }
219         HashTableIterator(PointerType pos, PointerType end, const HashTableType* container, HashItemKnownGoodTag tag) : m_iterator(pos, end, container, tag) { }
220
221     public:
222         HashTableIterator() { }
223
224         // default copy, assignment and destructor are OK
225
226         GetType get() const { return const_cast<GetType>(m_iterator.get()); }
227         typename Traits::IteratorReferenceType operator*() const { return Traits::getToReferenceConversion(get()); }
228         GetType operator->() const { return get(); }
229
230         iterator& operator++() { ++m_iterator; return *this; }
231
232         // postfix ++ intentionally omitted
233
234         // Comparison.
235         bool operator==(const iterator& other) const { return m_iterator == other.m_iterator; }
236         bool operator!=(const iterator& other) const { return m_iterator != other.m_iterator; }
237         bool operator==(const const_iterator& other) const { return m_iterator == other; }
238         bool operator!=(const const_iterator& other) const { return m_iterator != other; }
239
240         operator const_iterator() const { return m_iterator; }
241
242     private:
243         const_iterator m_iterator;
244     };
245
246     using std::swap;
247
248     // Work around MSVC's standard library, whose swap for pairs does not swap by component.
249     template<typename T> inline void hashTableSwap(T& a, T& b)
250     {
251         swap(a, b);
252     }
253
254     template<typename T, typename U> inline void hashTableSwap(KeyValuePair<T, U>& a, KeyValuePair<T, U>& b)
255     {
256         swap(a.key, b.key);
257         swap(a.value, b.value);
258     }
259
260     template<typename T, bool useSwap> struct Mover;
261     template<typename T> struct Mover<T, true> { static void move(T& from, T& to) { hashTableSwap(from, to); } };
262     template<typename T> struct Mover<T, false> { static void move(T& from, T& to) { to = from; } };
263
264     template<typename HashFunctions> class IdentityHashTranslator {
265     public:
266         template<typename T> static unsigned hash(const T& key) { return HashFunctions::hash(key); }
267         template<typename T, typename U> static bool equal(const T& a, const U& b) { return HashFunctions::equal(a, b); }
268         template<typename T, typename U, typename V> static void translate(T& location, const U&, const V& value) { location = value; }
269     };
270
271     template<typename HashTableType, typename ValueType> struct HashTableAddResult {
272         HashTableAddResult(const HashTableType* container, ValueType* storedValue, bool isNewEntry)
273             : storedValue(storedValue)
274             , isNewEntry(isNewEntry)
275 #if SECURITY_ASSERT_ENABLED
276             , m_container(container)
277             , m_containerModifications(container->modifications())
278 #endif
279         {
280             ASSERT_UNUSED(container, container);
281         }
282
283         ~HashTableAddResult()
284         {
285             // If rehash happened before accessing storedValue, it's
286             // use-after-free. Any modification may cause a rehash, so we check
287             // for modifications here.
288             // Rehash after accessing storedValue is harmless but will assert if
289             // the AddResult destructor takes place after a modification. You
290             // may need to limit the scope of the AddResult.
291             ASSERT_WITH_SECURITY_IMPLICATION(m_containerModifications == m_container->modifications());
292         }
293
294         ValueType* storedValue;
295         bool isNewEntry;
296
297 #if SECURITY_ASSERT_ENABLED
298     private:
299         const HashTableType* m_container;
300         const int64_t m_containerModifications;
301 #endif
302     };
303
304     template<typename Value, typename Extractor, typename KeyTraits>
305     struct HashTableHelper {
306         static bool isEmptyBucket(const Value& value) { return isHashTraitsEmptyValue<KeyTraits>(Extractor::extract(value)); }
307         static bool isDeletedBucket(const Value& value) { return KeyTraits::isDeletedValue(Extractor::extract(value)); }
308         static bool isEmptyOrDeletedBucket(const Value& value) { return isEmptyBucket(value) || isDeletedBucket(value); }
309     };
310
311     template<typename HashTranslator, typename KeyTraits, bool safeToCompareToEmptyOrDeleted>
312     struct HashTableKeyChecker {
313         // There's no simple generic way to make this check if safeToCompareToEmptyOrDeleted is false,
314         // so the check always passes.
315         template <typename T>
316         static bool checkKey(const T&) { return true; }
317     };
318
319     template<typename HashTranslator, typename KeyTraits>
320     struct HashTableKeyChecker<HashTranslator, KeyTraits, true> {
321         template <typename T>
322         static bool checkKey(const T& key)
323         {
324             // FIXME : Check also equality to the deleted value.
325             return !HashTranslator::equal(KeyTraits::emptyValue(), key);
326         }
327     };
328
329     // Don't declare a destructor for HeapAllocated hash tables.
330     template<typename Derived, bool isGarbageCollected>
331     class HashTableDestructorBase;
332
333     template<typename Derived>
334     class HashTableDestructorBase<Derived, true> { };
335
336     template<typename Derived>
337     class HashTableDestructorBase<Derived, false> {
338     public:
339         ~HashTableDestructorBase() { static_cast<Derived*>(this)->finalize(); }
340     };
341
342     // Note: empty or deleted key values are not allowed, using them may lead to undefined behavior.
343     // For pointer keys this means that null pointers are not allowed unless you supply custom key traits.
344     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
345     class HashTable : public HashTableDestructorBase<HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>, Allocator::isGarbageCollected> {
346     public:
347         typedef HashTableIterator<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator> iterator;
348         typedef HashTableConstIterator<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator> const_iterator;
349         typedef Traits ValueTraits;
350         typedef Key KeyType;
351         typedef typename KeyTraits::PeekInType KeyPeekInType;
352         typedef typename KeyTraits::PassInType KeyPassInType;
353         typedef Value ValueType;
354         typedef Extractor ExtractorType;
355         typedef KeyTraits KeyTraitsType;
356         typedef typename Traits::PassInType ValuePassInType;
357         typedef IdentityHashTranslator<HashFunctions> IdentityTranslatorType;
358         typedef HashTableAddResult<HashTable, ValueType> AddResult;
359
360 #if DUMP_HASHTABLE_STATS_PER_TABLE
361         struct Stats {
362             Stats()
363                 : numAccesses(0)
364                 , numRehashes(0)
365                 , numRemoves(0)
366                 , numReinserts(0)
367                 , maxCollisions(0)
368                 , numCollisions(0)
369                 , collisionGraph()
370             {
371             }
372
373             int numAccesses;
374             int numRehashes;
375             int numRemoves;
376             int numReinserts;
377
378             int maxCollisions;
379             int numCollisions;
380             int collisionGraph[4096];
381
382             void recordCollisionAtCount(int count)
383             {
384                 if (count > maxCollisions)
385                     maxCollisions = count;
386                 numCollisions++;
387                 collisionGraph[count]++;
388             }
389
390             void dumpStats()
391             {
392                 dataLogF("\nWTF::HashTable::Stats dump\n\n");
393                 dataLogF("%d accesses\n", numAccesses);
394                 dataLogF("%d total collisions, average %.2f probes per access\n", numCollisions, 1.0 * (numAccesses + numCollisions) / numAccesses);
395                 dataLogF("longest collision chain: %d\n", maxCollisions);
396                 for (int i = 1; i <= maxCollisions; i++) {
397                     dataLogF("  %d lookups with exactly %d collisions (%.2f%% , %.2f%% with this many or more)\n", collisionGraph[i], i, 100.0 * (collisionGraph[i] - collisionGraph[i+1]) / numAccesses, 100.0 * collisionGraph[i] / numAccesses);
398                 }
399                 dataLogF("%d rehashes\n", numRehashes);
400                 dataLogF("%d reinserts\n", numReinserts);
401             }
402         };
403 #endif
404
405         HashTable();
406         void finalize()
407         {
408             ASSERT(!Allocator::isGarbageCollected);
409             if (LIKELY(!m_table))
410                 return;
411             deleteAllBucketsAndDeallocate(m_table, m_tableSize);
412             m_table = 0;
413         }
414
415         HashTable(const HashTable&);
416         void swap(HashTable&);
417         HashTable& operator=(const HashTable&);
418
419         // When the hash table is empty, just return the same iterator for end as for begin.
420         // This is more efficient because we don't have to skip all the empty and deleted
421         // buckets, and iterating an empty table is a common case that's worth optimizing.
422         iterator begin() { return isEmpty() ? end() : makeIterator(m_table); }
423         iterator end() { return makeKnownGoodIterator(m_table + m_tableSize); }
424         const_iterator begin() const { return isEmpty() ? end() : makeConstIterator(m_table); }
425         const_iterator end() const { return makeKnownGoodConstIterator(m_table + m_tableSize); }
426
427         unsigned size() const { return m_keyCount; }
428         unsigned capacity() const { return m_tableSize; }
429         bool isEmpty() const { return !m_keyCount; }
430
431         AddResult add(ValuePassInType value)
432         {
433             return add<IdentityTranslatorType>(Extractor::extract(value), value);
434         }
435
436         // A special version of add() that finds the object by hashing and comparing
437         // with some other type, to avoid the cost of type conversion if the object is already
438         // in the table.
439         template<typename HashTranslator, typename T, typename Extra> AddResult add(const T& key, const Extra&);
440         template<typename HashTranslator, typename T, typename Extra> AddResult addPassingHashCode(const T& key, const Extra&);
441
442         iterator find(KeyPeekInType key) { return find<IdentityTranslatorType>(key); }
443         const_iterator find(KeyPeekInType key) const { return find<IdentityTranslatorType>(key); }
444         bool contains(KeyPeekInType key) const { return contains<IdentityTranslatorType>(key); }
445
446         template<typename HashTranslator, typename T> iterator find(const T&);
447         template<typename HashTranslator, typename T> const_iterator find(const T&) const;
448         template<typename HashTranslator, typename T> bool contains(const T&) const;
449
450         void remove(KeyPeekInType);
451         void remove(iterator);
452         void remove(const_iterator);
453         void clear();
454
455         static bool isEmptyBucket(const ValueType& value) { return isHashTraitsEmptyValue<KeyTraits>(Extractor::extract(value)); }
456         static bool isDeletedBucket(const ValueType& value) { return KeyTraits::isDeletedValue(Extractor::extract(value)); }
457         static bool isEmptyOrDeletedBucket(const ValueType& value) { return HashTableHelper<ValueType, Extractor, KeyTraits>:: isEmptyOrDeletedBucket(value); }
458
459         ValueType* lookup(KeyPeekInType key) { return lookup<IdentityTranslatorType, KeyPeekInType>(key); }
460         template<typename HashTranslator, typename T> ValueType* lookup(T);
461         template<typename HashTranslator, typename T> const ValueType* lookup(T) const;
462
463         void trace(typename Allocator::Visitor*);
464
465 #ifdef ASSERT_ENABLED
466         int64_t modifications() const { return m_modifications; }
467         void registerModification() { m_modifications++; }
468         // HashTable and collections that build on it do not support
469         // modifications while there is an iterator in use. The exception is
470         // ListHashSet, which has its own iterators that tolerate modification
471         // of the underlying set.
472         void checkModifications(int64_t mods) const { ASSERT(mods == m_modifications); }
473 #else
474         int64_t modifications() const { return 0; }
475         void registerModification() { }
476         void checkModifications(int64_t mods) const { }
477 #endif
478
479     private:
480         static ValueType* allocateTable(unsigned size);
481         static void deleteAllBucketsAndDeallocate(ValueType* table, unsigned size);
482
483         typedef std::pair<ValueType*, bool> LookupType;
484         typedef std::pair<LookupType, unsigned> FullLookupType;
485
486         LookupType lookupForWriting(const Key& key) { return lookupForWriting<IdentityTranslatorType>(key); };
487         template<typename HashTranslator, typename T> FullLookupType fullLookupForWriting(const T&);
488         template<typename HashTranslator, typename T> LookupType lookupForWriting(const T&);
489
490         void remove(ValueType*);
491
492         bool shouldExpand() const { return (m_keyCount + m_deletedCount) * m_maxLoad >= m_tableSize; }
493         bool mustRehashInPlace() const { return m_keyCount * m_minLoad < m_tableSize * 2; }
494         bool shouldShrink() const { return m_keyCount * m_minLoad < m_tableSize && m_tableSize > KeyTraits::minimumTableSize; }
495         ValueType* expand(ValueType* entry = 0);
496         void shrink() { rehash(m_tableSize / 2, 0); }
497
498         ValueType* rehash(unsigned newTableSize, ValueType* entry);
499         ValueType* reinsert(ValueType&);
500
501         static void initializeBucket(ValueType& bucket);
502         static void deleteBucket(ValueType& bucket) { bucket.~ValueType(); Traits::constructDeletedValue(bucket); }
503
504         FullLookupType makeLookupResult(ValueType* position, bool found, unsigned hash)
505             { return FullLookupType(LookupType(position, found), hash); }
506
507         iterator makeIterator(ValueType* pos) { return iterator(pos, m_table + m_tableSize, this); }
508         const_iterator makeConstIterator(ValueType* pos) const { return const_iterator(pos, m_table + m_tableSize, this); }
509         iterator makeKnownGoodIterator(ValueType* pos) { return iterator(pos, m_table + m_tableSize, this, HashItemKnownGood); }
510         const_iterator makeKnownGoodConstIterator(ValueType* pos) const { return const_iterator(pos, m_table + m_tableSize, this, HashItemKnownGood); }
511
512         static const unsigned m_maxLoad = 2;
513         static const unsigned m_minLoad = 6;
514
515         unsigned tableSizeMask() const
516         {
517             size_t mask = m_tableSize - 1;
518             ASSERT((mask & m_tableSize) == 0);
519             return mask;
520         }
521
522         ValueType* m_table;
523         unsigned m_tableSize;
524         unsigned m_keyCount;
525         unsigned m_deletedCount;
526 #ifdef ASSERT_ENABLED
527         unsigned m_modifications;
528 #endif
529
530 #if DUMP_HASHTABLE_STATS_PER_TABLE
531     public:
532         mutable OwnPtr<Stats> m_stats;
533 #endif
534
535         template<WeakHandlingFlag x, typename T, typename U, typename V, typename W, typename X, typename Y, typename Z> friend struct WeakProcessingHashTableHelper;
536         template<typename T, typename U, typename V, typename W> friend class LinkedHashSet;
537     };
538
539     // Set all the bits to one after the most significant bit: 00110101010 -> 00111111111.
540     template<unsigned size> struct OneifyLowBits;
541     template<>
542     struct OneifyLowBits<0> {
543         static const unsigned value = 0;
544     };
545     template<unsigned number>
546     struct OneifyLowBits {
547         static const unsigned value = number | OneifyLowBits<(number >> 1)>::value;
548     };
549     // Compute the first power of two integer that is an upper bound of the parameter 'number'.
550     template<unsigned number>
551     struct UpperPowerOfTwoBound {
552         static const unsigned value = (OneifyLowBits<number - 1>::value + 1) * 2;
553     };
554
555     // Because power of two numbers are the limit of maxLoad, their capacity is twice the
556     // UpperPowerOfTwoBound, or 4 times their values.
557     template<unsigned size, bool isPowerOfTwo> struct HashTableCapacityForSizeSplitter;
558     template<unsigned size>
559     struct HashTableCapacityForSizeSplitter<size, true> {
560         static const unsigned value = size * 4;
561     };
562     template<unsigned size>
563     struct HashTableCapacityForSizeSplitter<size, false> {
564         static const unsigned value = UpperPowerOfTwoBound<size>::value;
565     };
566
567     // HashTableCapacityForSize computes the upper power of two capacity to hold the size parameter.
568     // This is done at compile time to initialize the HashTraits.
569     template<unsigned size>
570     struct HashTableCapacityForSize {
571         static const unsigned value = HashTableCapacityForSizeSplitter<size, !(size & (size - 1))>::value;
572         COMPILE_ASSERT(size > 0, HashTableNonZeroMinimumCapacity);
573         COMPILE_ASSERT(!static_cast<int>(value >> 31), HashTableNoCapacityOverflow);
574         COMPILE_ASSERT(value > (2 * size), HashTableCapacityHoldsContentSize);
575     };
576
577     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
578     inline HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::HashTable()
579         : m_table(0)
580         , m_tableSize(0)
581         , m_keyCount(0)
582         , m_deletedCount(0)
583 #ifdef ASSERT_ENABLED
584         , m_modifications(0)
585 #endif
586 #if DUMP_HASHTABLE_STATS_PER_TABLE
587         , m_stats(adoptPtr(new Stats))
588 #endif
589     {
590     }
591
592     inline unsigned doubleHash(unsigned key)
593     {
594         key = ~key + (key >> 23);
595         key ^= (key << 12);
596         key ^= (key >> 7);
597         key ^= (key << 2);
598         key ^= (key >> 20);
599         return key;
600     }
601
602     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
603     template<typename HashTranslator, typename T>
604     inline Value* HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::lookup(T key)
605     {
606         return const_cast<Value*>(const_cast<const HashTable*>(this)->lookup<HashTranslator, T>(key));
607     }
608
609     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
610     template<typename HashTranslator, typename T>
611     inline const Value* HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::lookup(T key) const
612     {
613         ASSERT((HashTableKeyChecker<HashTranslator, KeyTraits, HashFunctions::safeToCompareToEmptyOrDeleted>::checkKey(key)));
614         const ValueType* table = m_table;
615         if (!table)
616             return 0;
617
618         size_t k = 0;
619         size_t sizeMask = tableSizeMask();
620         unsigned h = HashTranslator::hash(key);
621         size_t i = h & sizeMask;
622
623         UPDATE_ACCESS_COUNTS();
624
625         while (1) {
626             const ValueType* entry = table + i;
627
628             if (HashFunctions::safeToCompareToEmptyOrDeleted) {
629                 if (HashTranslator::equal(Extractor::extract(*entry), key))
630                     return entry;
631
632                 if (isEmptyBucket(*entry))
633                     return 0;
634             } else {
635                 if (isEmptyBucket(*entry))
636                     return 0;
637
638                 if (!isDeletedBucket(*entry) && HashTranslator::equal(Extractor::extract(*entry), key))
639                     return entry;
640             }
641             UPDATE_PROBE_COUNTS();
642             if (!k)
643                 k = 1 | doubleHash(h);
644             i = (i + k) & sizeMask;
645         }
646     }
647
648     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
649     template<typename HashTranslator, typename T>
650     inline typename HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::LookupType HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::lookupForWriting(const T& key)
651     {
652         ASSERT(m_table);
653         registerModification();
654
655         ValueType* table = m_table;
656         size_t k = 0;
657         size_t sizeMask = tableSizeMask();
658         unsigned h = HashTranslator::hash(key);
659         size_t i = h & sizeMask;
660
661         UPDATE_ACCESS_COUNTS();
662
663         ValueType* deletedEntry = 0;
664
665         while (1) {
666             ValueType* entry = table + i;
667
668             if (isEmptyBucket(*entry))
669                 return LookupType(deletedEntry ? deletedEntry : entry, false);
670
671             if (HashFunctions::safeToCompareToEmptyOrDeleted) {
672                 if (HashTranslator::equal(Extractor::extract(*entry), key))
673                     return LookupType(entry, true);
674
675                 if (isDeletedBucket(*entry))
676                     deletedEntry = entry;
677             } else {
678                 if (isDeletedBucket(*entry))
679                     deletedEntry = entry;
680                 else if (HashTranslator::equal(Extractor::extract(*entry), key))
681                     return LookupType(entry, true);
682             }
683             UPDATE_PROBE_COUNTS();
684             if (!k)
685                 k = 1 | doubleHash(h);
686             i = (i + k) & sizeMask;
687         }
688     }
689
690     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
691     template<typename HashTranslator, typename T>
692     inline typename HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::FullLookupType HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::fullLookupForWriting(const T& key)
693     {
694         ASSERT(m_table);
695         registerModification();
696
697         ValueType* table = m_table;
698         size_t k = 0;
699         size_t sizeMask = tableSizeMask();
700         unsigned h = HashTranslator::hash(key);
701         size_t i = h & sizeMask;
702
703         UPDATE_ACCESS_COUNTS();
704
705         ValueType* deletedEntry = 0;
706
707         while (1) {
708             ValueType* entry = table + i;
709
710             if (isEmptyBucket(*entry))
711                 return makeLookupResult(deletedEntry ? deletedEntry : entry, false, h);
712
713             if (HashFunctions::safeToCompareToEmptyOrDeleted) {
714                 if (HashTranslator::equal(Extractor::extract(*entry), key))
715                     return makeLookupResult(entry, true, h);
716
717                 if (isDeletedBucket(*entry))
718                     deletedEntry = entry;
719             } else {
720                 if (isDeletedBucket(*entry))
721                     deletedEntry = entry;
722                 else if (HashTranslator::equal(Extractor::extract(*entry), key))
723                     return makeLookupResult(entry, true, h);
724             }
725             UPDATE_PROBE_COUNTS();
726             if (!k)
727                 k = 1 | doubleHash(h);
728             i = (i + k) & sizeMask;
729         }
730     }
731
732     template<bool emptyValueIsZero> struct HashTableBucketInitializer;
733
734     template<> struct HashTableBucketInitializer<false> {
735         template<typename Traits, typename Value> static void initialize(Value& bucket)
736         {
737             new (NotNull, &bucket) Value(Traits::emptyValue());
738         }
739     };
740
741     template<> struct HashTableBucketInitializer<true> {
742         template<typename Traits, typename Value> static void initialize(Value& bucket)
743         {
744             // This initializes the bucket without copying the empty value.
745             // That makes it possible to use this with types that don't support copying.
746             // The memset to 0 looks like a slow operation but is optimized by the compilers.
747             memset(&bucket, 0, sizeof(bucket));
748         }
749     };
750
751     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
752     inline void HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::initializeBucket(ValueType& bucket)
753     {
754         HashTableBucketInitializer<Traits::emptyValueIsZero>::template initialize<Traits>(bucket);
755     }
756
757     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
758     template<typename HashTranslator, typename T, typename Extra>
759     typename HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::AddResult HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::add(const T& key, const Extra& extra)
760     {
761         if (!m_table)
762             expand();
763
764         ASSERT(m_table);
765
766         ValueType* table = m_table;
767         size_t k = 0;
768         size_t sizeMask = tableSizeMask();
769         unsigned h = HashTranslator::hash(key);
770         size_t i = h & sizeMask;
771
772         UPDATE_ACCESS_COUNTS();
773
774         ValueType* deletedEntry = 0;
775         ValueType* entry;
776         while (1) {
777             entry = table + i;
778
779             if (isEmptyBucket(*entry))
780                 break;
781
782             if (HashFunctions::safeToCompareToEmptyOrDeleted) {
783                 if (HashTranslator::equal(Extractor::extract(*entry), key))
784                     return AddResult(this, entry, false);
785
786                 if (isDeletedBucket(*entry))
787                     deletedEntry = entry;
788             } else {
789                 if (isDeletedBucket(*entry))
790                     deletedEntry = entry;
791                 else if (HashTranslator::equal(Extractor::extract(*entry), key))
792                     return AddResult(this, entry, false);
793             }
794             UPDATE_PROBE_COUNTS();
795             if (!k)
796                 k = 1 | doubleHash(h);
797             i = (i + k) & sizeMask;
798         }
799
800         registerModification();
801
802         if (deletedEntry) {
803             initializeBucket(*deletedEntry);
804             entry = deletedEntry;
805             --m_deletedCount;
806         }
807
808         HashTranslator::translate(*entry, key, extra);
809         ASSERT(!isEmptyOrDeletedBucket(*entry));
810
811         ++m_keyCount;
812
813         if (shouldExpand())
814             entry = expand(entry);
815
816         return AddResult(this, entry, true);
817     }
818
819     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
820     template<typename HashTranslator, typename T, typename Extra>
821     typename HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::AddResult HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::addPassingHashCode(const T& key, const Extra& extra)
822     {
823         if (!m_table)
824             expand();
825
826         FullLookupType lookupResult = fullLookupForWriting<HashTranslator>(key);
827
828         ValueType* entry = lookupResult.first.first;
829         bool found = lookupResult.first.second;
830         unsigned h = lookupResult.second;
831
832         if (found)
833             return AddResult(this, entry, false);
834
835         registerModification();
836
837         if (isDeletedBucket(*entry)) {
838             initializeBucket(*entry);
839             --m_deletedCount;
840         }
841
842         HashTranslator::translate(*entry, key, extra, h);
843         ASSERT(!isEmptyOrDeletedBucket(*entry));
844
845         ++m_keyCount;
846         if (shouldExpand())
847             entry = expand(entry);
848
849         return AddResult(this, entry, true);
850     }
851
852     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
853     Value* HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::reinsert(ValueType& entry)
854     {
855         ASSERT(m_table);
856         registerModification();
857         ASSERT(!lookupForWriting(Extractor::extract(entry)).second);
858         ASSERT(!isDeletedBucket(*(lookupForWriting(Extractor::extract(entry)).first)));
859 #if DUMP_HASHTABLE_STATS
860         atomicIncrement(&HashTableStats::numReinserts);
861 #endif
862 #if DUMP_HASHTABLE_STATS_PER_TABLE
863         ++m_stats->numReinserts;
864 #endif
865         Value* newEntry = lookupForWriting(Extractor::extract(entry)).first;
866         Mover<ValueType, Traits::needsDestruction>::move(entry, *newEntry);
867
868         return newEntry;
869     }
870
871     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
872     template <typename HashTranslator, typename T>
873     inline typename HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::iterator HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::find(const T& key)
874     {
875         ValueType* entry = lookup<HashTranslator>(key);
876         if (!entry)
877             return end();
878
879         return makeKnownGoodIterator(entry);
880     }
881
882     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
883     template <typename HashTranslator, typename T>
884     inline typename HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::const_iterator HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::find(const T& key) const
885     {
886         ValueType* entry = const_cast<HashTable*>(this)->lookup<HashTranslator>(key);
887         if (!entry)
888             return end();
889
890         return makeKnownGoodConstIterator(entry);
891     }
892
893     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
894     template <typename HashTranslator, typename T>
895     bool HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::contains(const T& key) const
896     {
897         return const_cast<HashTable*>(this)->lookup<HashTranslator>(key);
898     }
899
900     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
901     void HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::remove(ValueType* pos)
902     {
903         registerModification();
904 #if DUMP_HASHTABLE_STATS
905         atomicIncrement(&HashTableStats::numRemoves);
906 #endif
907 #if DUMP_HASHTABLE_STATS_PER_TABLE
908         ++m_stats->numRemoves;
909 #endif
910
911         deleteBucket(*pos);
912         ++m_deletedCount;
913         --m_keyCount;
914
915         if (shouldShrink())
916             shrink();
917     }
918
919     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
920     inline void HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::remove(iterator it)
921     {
922         if (it == end())
923             return;
924
925         remove(const_cast<ValueType*>(it.m_iterator.m_position));
926     }
927
928     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
929     inline void HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::remove(const_iterator it)
930     {
931         if (it == end())
932             return;
933
934         remove(const_cast<ValueType*>(it.m_position));
935     }
936
937     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
938     inline void HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::remove(KeyPeekInType key)
939     {
940         remove(find(key));
941     }
942
943     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
944     Value* HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::allocateTable(unsigned size)
945     {
946         typedef typename Allocator::template HashTableBackingHelper<HashTable>::Type HashTableBacking;
947
948         size_t allocSize = size * sizeof(ValueType);
949         ValueType* result;
950         COMPILE_ASSERT(!Traits::emptyValueIsZero || !IsPolymorphic<ValueType>::value, EmptyValueCannotBeZeroForThingsWithAVtable);
951         if (Traits::emptyValueIsZero) {
952             result = Allocator::template zeroedBackingMalloc<ValueType*, HashTableBacking>(allocSize);
953         } else {
954             result = Allocator::template backingMalloc<ValueType*, HashTableBacking>(allocSize);
955             for (unsigned i = 0; i < size; i++)
956                 initializeBucket(result[i]);
957         }
958         return result;
959     }
960
961     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
962     void HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::deleteAllBucketsAndDeallocate(ValueType* table, unsigned size)
963     {
964         if (Traits::needsDestruction) {
965             for (unsigned i = 0; i < size; ++i) {
966                 // This code is called when the hash table is cleared or
967                 // resized. We have allocated a new backing store and we need
968                 // to run the destructors on the old backing store, as it is
969                 // being freed. If we are GCing we need to both call the
970                 // destructor and mark the bucket as deleted, otherwise the
971                 // destructor gets called again when the GC finds the backing
972                 // store. With the default allocator it's enough to call the
973                 // destructor, since we will free the memory explicitly and
974                 // we won't see the memory with the bucket again.
975                 if (!isEmptyOrDeletedBucket(table[i])) {
976                     if (Allocator::isGarbageCollected)
977                         deleteBucket(table[i]);
978                     else
979                         table[i].~ValueType();
980                 }
981             }
982         }
983         Allocator::backingFree(table);
984     }
985
986     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
987     Value* HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::expand(Value* entry)
988     {
989         unsigned newSize;
990         if (!m_tableSize) {
991             newSize = KeyTraits::minimumTableSize;
992         } else if (mustRehashInPlace()) {
993             newSize = m_tableSize;
994         } else {
995             newSize = m_tableSize * 2;
996             RELEASE_ASSERT(newSize > m_tableSize);
997         }
998
999         return rehash(newSize, entry);
1000     }
1001
1002     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
1003     Value* HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::rehash(unsigned newTableSize, Value* entry)
1004     {
1005         unsigned oldTableSize = m_tableSize;
1006         ValueType* oldTable = m_table;
1007
1008 #if DUMP_HASHTABLE_STATS
1009         if (oldTableSize != 0)
1010             atomicIncrement(&HashTableStats::numRehashes);
1011 #endif
1012
1013 #if DUMP_HASHTABLE_STATS_PER_TABLE
1014         if (oldTableSize != 0)
1015             ++m_stats->numRehashes;
1016 #endif
1017
1018         m_table = allocateTable(newTableSize);
1019         m_tableSize = newTableSize;
1020
1021         Value* newEntry = 0;
1022         for (unsigned i = 0; i != oldTableSize; ++i) {
1023             if (isEmptyOrDeletedBucket(oldTable[i])) {
1024                 ASSERT(&oldTable[i] != entry);
1025                 continue;
1026             }
1027
1028             Value* reinsertedEntry = reinsert(oldTable[i]);
1029             if (&oldTable[i] == entry) {
1030                 ASSERT(!newEntry);
1031                 newEntry = reinsertedEntry;
1032             }
1033         }
1034
1035         m_deletedCount = 0;
1036
1037         deleteAllBucketsAndDeallocate(oldTable, oldTableSize);
1038
1039         return newEntry;
1040     }
1041
1042     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
1043     void HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::clear()
1044     {
1045         registerModification();
1046         if (!m_table)
1047             return;
1048
1049         deleteAllBucketsAndDeallocate(m_table, m_tableSize);
1050         m_table = 0;
1051         m_tableSize = 0;
1052         m_keyCount = 0;
1053     }
1054
1055     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
1056     HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::HashTable(const HashTable& other)
1057         : m_table(0)
1058         , m_tableSize(0)
1059         , m_keyCount(0)
1060         , m_deletedCount(0)
1061 #ifdef ASSERT_ENABLED
1062         , m_modifications(0)
1063 #endif
1064 #if DUMP_HASHTABLE_STATS_PER_TABLE
1065         , m_stats(adoptPtr(new Stats(*other.m_stats)))
1066 #endif
1067     {
1068         // Copy the hash table the dumb way, by adding each element to the new table.
1069         // It might be more efficient to copy the table slots, but it's not clear that efficiency is needed.
1070         const_iterator end = other.end();
1071         for (const_iterator it = other.begin(); it != end; ++it)
1072             add(*it);
1073     }
1074
1075     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
1076     void HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::swap(HashTable& other)
1077     {
1078         std::swap(m_table, other.m_table);
1079         std::swap(m_tableSize, other.m_tableSize);
1080         std::swap(m_keyCount, other.m_keyCount);
1081         std::swap(m_deletedCount, other.m_deletedCount);
1082
1083 #ifdef ASSERT_ENABLED
1084         std::swap(m_modifications, other.m_modifications);
1085 #endif
1086
1087 #if DUMP_HASHTABLE_STATS_PER_TABLE
1088         m_stats.swap(other.m_stats);
1089 #endif
1090     }
1091
1092     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
1093     HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>& HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::operator=(const HashTable& other)
1094     {
1095         HashTable tmp(other);
1096         swap(tmp);
1097         return *this;
1098     }
1099
1100     template<WeakHandlingFlag weakHandlingFlag, typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
1101     struct WeakProcessingHashTableHelper;
1102
1103     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
1104     struct WeakProcessingHashTableHelper<NoWeakHandlingInCollections, Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator> {
1105         static void process(typename Allocator::Visitor* visitor, void* closure) { }
1106     };
1107
1108     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
1109     struct WeakProcessingHashTableHelper<WeakHandlingInCollections, Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator> {
1110         static void process(typename Allocator::Visitor* visitor, void* closure)
1111         {
1112             typedef HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator> HashTableType;
1113             HashTableType* table = reinterpret_cast<HashTableType*>(closure);
1114             if (table->m_table) {
1115                 // This just marks it live and does not push anything onto the
1116                 // marking stack.
1117                 Allocator::markNoTracing(visitor, table->m_table);
1118                 // Now perform weak processing (this is a no-op if the backing
1119                 // was accessible through an iterator and was already marked
1120                 // strongly).
1121                 for (typename HashTableType::ValueType* element = table->m_table + table->m_tableSize - 1; element >= table->m_table; element--) {
1122                     if (!HashTableType::isEmptyOrDeletedBucket(*element)) {
1123                         if (Traits::shouldRemoveFromCollection(visitor, *element)) {
1124                             table->registerModification();
1125                             HashTableType::deleteBucket(*element); // Also calls the destructor.
1126                             table->m_deletedCount++;
1127                             table->m_keyCount--;
1128                             // We don't rehash the backing until the next add
1129                             // or delete, because that would cause allocation
1130                             // during GC.
1131                         }
1132                     }
1133                 }
1134             }
1135         }
1136     };
1137
1138     template<typename Key, typename Value, typename Extractor, typename HashFunctions, typename Traits, typename KeyTraits, typename Allocator>
1139     void HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::trace(typename Allocator::Visitor* visitor)
1140     {
1141         // If someone else already marked the backing and queued up the trace
1142         // and/or weak callback then we are done. This optimization does not
1143         // happen for ListHashSet since its iterator does not point at the
1144         // backing.
1145         if (!m_table || visitor->isAlive(m_table))
1146             return;
1147         // Normally, we mark the backing store without performing trace. This
1148         // means it is marked live, but the pointers inside it are not marked.
1149         // Instead we will mark the pointers below. However, for backing
1150         // stores that contain weak pointers the handling is rather different.
1151         // We don't mark the backing store here, so the marking GC will leave
1152         // the backing unmarked. If the backing is found in any other way than
1153         // through its HashTable (ie from an iterator) then the mark bit will
1154         // be set and the pointers will be marked strongly, avoiding problems
1155         // with iterating over things that disappear due to weak processing
1156         // while we are iterating over them. The weakProcessing callback will
1157         // mark the backing as a void pointer, and will perform weak processing
1158         // if needed.
1159         if (Traits::weakHandlingFlag == NoWeakHandlingInCollections)
1160             Allocator::markNoTracing(visitor, m_table);
1161         else
1162             Allocator::registerWeakMembers(visitor, this, m_table, WeakProcessingHashTableHelper<Traits::weakHandlingFlag, Key, Value, Extractor, HashFunctions, Traits, KeyTraits, Allocator>::process);
1163         if (ShouldBeTraced<Traits>::value) {
1164             for (ValueType* element = m_table + m_tableSize - 1; element >= m_table; element--) {
1165                 if (!isEmptyOrDeletedBucket(*element))
1166                     Allocator::template trace<ValueType, Traits>(visitor, *element);
1167             }
1168         }
1169     }
1170
1171     // iterator adapters
1172
1173     template<typename HashTableType, typename Traits> struct HashTableConstIteratorAdapter {
1174         HashTableConstIteratorAdapter() {}
1175         HashTableConstIteratorAdapter(const typename HashTableType::const_iterator& impl) : m_impl(impl) {}
1176         typedef typename Traits::IteratorConstGetType GetType;
1177         typedef typename HashTableType::ValueTraits::IteratorConstGetType SourceGetType;
1178
1179         GetType get() const { return const_cast<GetType>(SourceGetType(m_impl.get())); }
1180         typename Traits::IteratorConstReferenceType operator*() const { return Traits::getToReferenceConstConversion(get()); }
1181         GetType operator->() const { return get(); }
1182
1183         HashTableConstIteratorAdapter& operator++() { ++m_impl; return *this; }
1184         // postfix ++ intentionally omitted
1185
1186         typename HashTableType::const_iterator m_impl;
1187     };
1188
1189     template<typename HashTableType, typename Traits> struct HashTableIteratorAdapter {
1190         typedef typename Traits::IteratorGetType GetType;
1191         typedef typename HashTableType::ValueTraits::IteratorGetType SourceGetType;
1192
1193         HashTableIteratorAdapter() {}
1194         HashTableIteratorAdapter(const typename HashTableType::iterator& impl) : m_impl(impl) {}
1195
1196         GetType get() const { return const_cast<GetType>(SourceGetType(m_impl.get())); }
1197         typename Traits::IteratorReferenceType operator*() const { return Traits::getToReferenceConversion(get()); }
1198         GetType operator->() const { return get(); }
1199
1200         HashTableIteratorAdapter& operator++() { ++m_impl; return *this; }
1201         // postfix ++ intentionally omitted
1202
1203         operator HashTableConstIteratorAdapter<HashTableType, Traits>()
1204         {
1205             typename HashTableType::const_iterator i = m_impl;
1206             return i;
1207         }
1208
1209         typename HashTableType::iterator m_impl;
1210     };
1211
1212     template<typename T, typename U>
1213     inline bool operator==(const HashTableConstIteratorAdapter<T, U>& a, const HashTableConstIteratorAdapter<T, U>& b)
1214     {
1215         return a.m_impl == b.m_impl;
1216     }
1217
1218     template<typename T, typename U>
1219     inline bool operator!=(const HashTableConstIteratorAdapter<T, U>& a, const HashTableConstIteratorAdapter<T, U>& b)
1220     {
1221         return a.m_impl != b.m_impl;
1222     }
1223
1224     template<typename T, typename U>
1225     inline bool operator==(const HashTableIteratorAdapter<T, U>& a, const HashTableIteratorAdapter<T, U>& b)
1226     {
1227         return a.m_impl == b.m_impl;
1228     }
1229
1230     template<typename T, typename U>
1231     inline bool operator!=(const HashTableIteratorAdapter<T, U>& a, const HashTableIteratorAdapter<T, U>& b)
1232     {
1233         return a.m_impl != b.m_impl;
1234     }
1235
1236     // All 4 combinations of ==, != and Const,non const.
1237     template<typename T, typename U>
1238     inline bool operator==(const HashTableConstIteratorAdapter<T, U>& a, const HashTableIteratorAdapter<T, U>& b)
1239     {
1240         return a.m_impl == b.m_impl;
1241     }
1242
1243     template<typename T, typename U>
1244     inline bool operator!=(const HashTableConstIteratorAdapter<T, U>& a, const HashTableIteratorAdapter<T, U>& b)
1245     {
1246         return a.m_impl != b.m_impl;
1247     }
1248
1249     template<typename T, typename U>
1250     inline bool operator==(const HashTableIteratorAdapter<T, U>& a, const HashTableConstIteratorAdapter<T, U>& b)
1251     {
1252         return a.m_impl == b.m_impl;
1253     }
1254
1255     template<typename T, typename U>
1256     inline bool operator!=(const HashTableIteratorAdapter<T, U>& a, const HashTableConstIteratorAdapter<T, U>& b)
1257     {
1258         return a.m_impl != b.m_impl;
1259     }
1260
1261     template<typename Collection1, typename Collection2>
1262     inline void removeAll(Collection1& collection, const Collection2& toBeRemoved)
1263     {
1264         if (collection.isEmpty() || toBeRemoved.isEmpty())
1265             return;
1266         typedef typename Collection2::const_iterator CollectionIterator;
1267         CollectionIterator end(toBeRemoved.end());
1268         for (CollectionIterator it(toBeRemoved.begin()); it != end; ++it)
1269             collection.remove(*it);
1270     }
1271
1272 } // namespace WTF
1273
1274 #include "wtf/HashIterators.h"
1275
1276 #endif // WTF_HashTable_h