1/*
2******************************************************************************
3*   Copyright (C) 1997-2011, International Business Machines
4*   Corporation and others.  All Rights Reserved.
5******************************************************************************
6*   Date        Name        Description
7*   03/22/00    aliu        Adapted from original C++ ICU Hashtable.
8*   07/06/01    aliu        Modified to support int32_t keys on
9*                           platforms with sizeof(void*) < 32.
10******************************************************************************
11*/
12
13#include "uhash.h"
14#include "unicode/ustring.h"
15#include "cstring.h"
16#include "cmemory.h"
17#include "uassert.h"
18#include "ustr_imp.h"
19
20/* This hashtable is implemented as a double hash.  All elements are
21 * stored in a single array with no secondary storage for collision
22 * resolution (no linked list, etc.).  When there is a hash collision
23 * (when two unequal keys have the same hashcode) we resolve this by
24 * using a secondary hash.  The secondary hash is an increment
25 * computed as a hash function (a different one) of the primary
26 * hashcode.  This increment is added to the initial hash value to
27 * obtain further slots assigned to the same hash code.  For this to
28 * work, the length of the array and the increment must be relatively
29 * prime.  The easiest way to achieve this is to have the length of
30 * the array be prime, and the increment be any value from
31 * 1..length-1.
32 *
33 * Hashcodes are 32-bit integers.  We make sure all hashcodes are
34 * non-negative by masking off the top bit.  This has two effects: (1)
35 * modulo arithmetic is simplified.  If we allowed negative hashcodes,
36 * then when we computed hashcode % length, we could get a negative
37 * result, which we would then have to adjust back into range.  It's
38 * simpler to just make hashcodes non-negative. (2) It makes it easy
39 * to check for empty vs. occupied slots in the table.  We just mark
40 * empty or deleted slots with a negative hashcode.
41 *
42 * The central function is _uhash_find().  This function looks for a
43 * slot matching the given key and hashcode.  If one is found, it
44 * returns a pointer to that slot.  If the table is full, and no match
45 * is found, it returns NULL -- in theory.  This would make the code
46 * more complicated, since all callers of _uhash_find() would then
47 * have to check for a NULL result.  To keep this from happening, we
48 * don't allow the table to fill.  When there is only one
49 * empty/deleted slot left, uhash_put() will refuse to increase the
50 * count, and fail.  This simplifies the code.  In practice, one will
51 * seldom encounter this using default UHashtables.  However, if a
52 * hashtable is set to a U_FIXED resize policy, or if memory is
53 * exhausted, then the table may fill.
54 *
55 * High and low water ratios control rehashing.  They establish levels
56 * of fullness (from 0 to 1) outside of which the data array is
57 * reallocated and repopulated.  Setting the low water ratio to zero
58 * means the table will never shrink.  Setting the high water ratio to
59 * one means the table will never grow.  The ratios should be
60 * coordinated with the ratio between successive elements of the
61 * PRIMES table, so that when the primeIndex is incremented or
62 * decremented during rehashing, it brings the ratio of count / length
63 * back into the desired range (between low and high water ratios).
64 */
65
66/********************************************************************
67 * PRIVATE Constants, Macros
68 ********************************************************************/
69
70/* This is a list of non-consecutive primes chosen such that
71 * PRIMES[i+1] ~ 2*PRIMES[i].  (Currently, the ratio ranges from 1.81
72 * to 2.18; the inverse ratio ranges from 0.459 to 0.552.)  If this
73 * ratio is changed, the low and high water ratios should also be
74 * adjusted to suit.
75 *
76 * These prime numbers were also chosen so that they are the largest
77 * prime number while being less than a power of two.
78 */
79static const int32_t PRIMES[] = {
80    13, 31, 61, 127, 251, 509, 1021, 2039, 4093, 8191, 16381, 32749,
81    65521, 131071, 262139, 524287, 1048573, 2097143, 4194301, 8388593,
82    16777213, 33554393, 67108859, 134217689, 268435399, 536870909,
83    1073741789, 2147483647 /*, 4294967291 */
84};
85
86#define PRIMES_LENGTH (sizeof(PRIMES) / sizeof(PRIMES[0]))
87#define DEFAULT_PRIME_INDEX 3
88
89/* These ratios are tuned to the PRIMES array such that a resize
90 * places the table back into the zone of non-resizing.  That is,
91 * after a call to _uhash_rehash(), a subsequent call to
92 * _uhash_rehash() should do nothing (should not churn).  This is only
93 * a potential problem with U_GROW_AND_SHRINK.
94 */
95static const float RESIZE_POLICY_RATIO_TABLE[6] = {
96    /* low, high water ratio */
97    0.0F, 0.5F, /* U_GROW: Grow on demand, do not shrink */
98    0.1F, 0.5F, /* U_GROW_AND_SHRINK: Grow and shrink on demand */
99    0.0F, 1.0F  /* U_FIXED: Never change size */
100};
101
102/*
103  Invariants for hashcode values:
104
105  * DELETED < 0
106  * EMPTY < 0
107  * Real hashes >= 0
108
109  Hashcodes may not start out this way, but internally they are
110  adjusted so that they are always positive.  We assume 32-bit
111  hashcodes; adjust these constants for other hashcode sizes.
112*/
113#define HASH_DELETED    ((int32_t) 0x80000000)
114#define HASH_EMPTY      ((int32_t) HASH_DELETED + 1)
115
116#define IS_EMPTY_OR_DELETED(x) ((x) < 0)
117
118/* This macro expects a UHashTok.pointer as its keypointer and
119   valuepointer parameters */
120#define HASH_DELETE_KEY_VALUE(hash, keypointer, valuepointer) \
121            if (hash->keyDeleter != NULL && keypointer != NULL) { \
122                (*hash->keyDeleter)(keypointer); \
123            } \
124            if (hash->valueDeleter != NULL && valuepointer != NULL) { \
125                (*hash->valueDeleter)(valuepointer); \
126            }
127
128/*
129 * Constants for hinting whether a key or value is an integer
130 * or a pointer.  If a hint bit is zero, then the associated
131 * token is assumed to be an integer.
132 */
133#define HINT_KEY_POINTER   (1)
134#define HINT_VALUE_POINTER (2)
135
136/********************************************************************
137 * PRIVATE Implementation
138 ********************************************************************/
139
140static UHashTok
141_uhash_setElement(UHashtable *hash, UHashElement* e,
142                  int32_t hashcode,
143                  UHashTok key, UHashTok value, int8_t hint) {
144
145    UHashTok oldValue = e->value;
146    if (hash->keyDeleter != NULL && e->key.pointer != NULL &&
147        e->key.pointer != key.pointer) { /* Avoid double deletion */
148        (*hash->keyDeleter)(e->key.pointer);
149    }
150    if (hash->valueDeleter != NULL) {
151        if (oldValue.pointer != NULL &&
152            oldValue.pointer != value.pointer) { /* Avoid double deletion */
153            (*hash->valueDeleter)(oldValue.pointer);
154        }
155        oldValue.pointer = NULL;
156    }
157    /* Compilers should copy the UHashTok union correctly, but even if
158     * they do, memory heap tools (e.g. BoundsChecker) can get
159     * confused when a pointer is cloaked in a union and then copied.
160     * TO ALLEVIATE THIS, we use hints (based on what API the user is
161     * calling) to copy pointers when we know the user thinks
162     * something is a pointer. */
163    if (hint & HINT_KEY_POINTER) {
164        e->key.pointer = key.pointer;
165    } else {
166        e->key = key;
167    }
168    if (hint & HINT_VALUE_POINTER) {
169        e->value.pointer = value.pointer;
170    } else {
171        e->value = value;
172    }
173    e->hashcode = hashcode;
174    return oldValue;
175}
176
177/**
178 * Assumes that the given element is not empty or deleted.
179 */
180static UHashTok
181_uhash_internalRemoveElement(UHashtable *hash, UHashElement* e) {
182    UHashTok empty;
183    U_ASSERT(!IS_EMPTY_OR_DELETED(e->hashcode));
184    --hash->count;
185    empty.pointer = NULL; empty.integer = 0;
186    return _uhash_setElement(hash, e, HASH_DELETED, empty, empty, 0);
187}
188
189static void
190_uhash_internalSetResizePolicy(UHashtable *hash, enum UHashResizePolicy policy) {
191    U_ASSERT(hash != NULL);
192    U_ASSERT(((int32_t)policy) >= 0);
193    U_ASSERT(((int32_t)policy) < 3);
194    hash->lowWaterRatio  = RESIZE_POLICY_RATIO_TABLE[policy * 2];
195    hash->highWaterRatio = RESIZE_POLICY_RATIO_TABLE[policy * 2 + 1];
196}
197
198/**
199 * Allocate internal data array of a size determined by the given
200 * prime index.  If the index is out of range it is pinned into range.
201 * If the allocation fails the status is set to
202 * U_MEMORY_ALLOCATION_ERROR and all array storage is freed.  In
203 * either case the previous array pointer is overwritten.
204 *
205 * Caller must ensure primeIndex is in range 0..PRIME_LENGTH-1.
206 */
207static void
208_uhash_allocate(UHashtable *hash,
209                int32_t primeIndex,
210                UErrorCode *status) {
211
212    UHashElement *p, *limit;
213    UHashTok emptytok;
214
215    if (U_FAILURE(*status)) return;
216
217    U_ASSERT(primeIndex >= 0 && primeIndex < PRIMES_LENGTH);
218
219    hash->primeIndex = primeIndex;
220    hash->length = PRIMES[primeIndex];
221
222    p = hash->elements = (UHashElement*)
223        uprv_malloc(sizeof(UHashElement) * hash->length);
224
225    if (hash->elements == NULL) {
226        *status = U_MEMORY_ALLOCATION_ERROR;
227        return;
228    }
229
230    emptytok.pointer = NULL; /* Only one of these two is needed */
231    emptytok.integer = 0;    /* but we don't know which one. */
232
233    limit = p + hash->length;
234    while (p < limit) {
235        p->key = emptytok;
236        p->value = emptytok;
237        p->hashcode = HASH_EMPTY;
238        ++p;
239    }
240
241    hash->count = 0;
242    hash->lowWaterMark = (int32_t)(hash->length * hash->lowWaterRatio);
243    hash->highWaterMark = (int32_t)(hash->length * hash->highWaterRatio);
244}
245
246static UHashtable*
247_uhash_init(UHashtable *result,
248              UHashFunction *keyHash,
249              UKeyComparator *keyComp,
250              UValueComparator *valueComp,
251              int32_t primeIndex,
252              UErrorCode *status)
253{
254    if (U_FAILURE(*status)) return NULL;
255    U_ASSERT(keyHash != NULL);
256    U_ASSERT(keyComp != NULL);
257
258    result->keyHasher       = keyHash;
259    result->keyComparator   = keyComp;
260    result->valueComparator = valueComp;
261    result->keyDeleter      = NULL;
262    result->valueDeleter    = NULL;
263    result->allocated       = FALSE;
264    _uhash_internalSetResizePolicy(result, U_GROW);
265
266    _uhash_allocate(result, primeIndex, status);
267
268    if (U_FAILURE(*status)) {
269        return NULL;
270    }
271
272    return result;
273}
274
275static UHashtable*
276_uhash_create(UHashFunction *keyHash,
277              UKeyComparator *keyComp,
278              UValueComparator *valueComp,
279              int32_t primeIndex,
280              UErrorCode *status) {
281    UHashtable *result;
282
283    if (U_FAILURE(*status)) return NULL;
284
285    result = (UHashtable*) uprv_malloc(sizeof(UHashtable));
286    if (result == NULL) {
287        *status = U_MEMORY_ALLOCATION_ERROR;
288        return NULL;
289    }
290
291    _uhash_init(result, keyHash, keyComp, valueComp, primeIndex, status);
292    result->allocated       = TRUE;
293
294    if (U_FAILURE(*status)) {
295        uprv_free(result);
296        return NULL;
297    }
298
299    return result;
300}
301
302/**
303 * Look for a key in the table, or if no such key exists, the first
304 * empty slot matching the given hashcode.  Keys are compared using
305 * the keyComparator function.
306 *
307 * First find the start position, which is the hashcode modulo
308 * the length.  Test it to see if it is:
309 *
310 * a. identical:  First check the hash values for a quick check,
311 *    then compare keys for equality using keyComparator.
312 * b. deleted
313 * c. empty
314 *
315 * Stop if it is identical or empty, otherwise continue by adding a
316 * "jump" value (moduloing by the length again to keep it within
317 * range) and retesting.  For efficiency, there need enough empty
318 * values so that the searchs stop within a reasonable amount of time.
319 * This can be changed by changing the high/low water marks.
320 *
321 * In theory, this function can return NULL, if it is full (no empty
322 * or deleted slots) and if no matching key is found.  In practice, we
323 * prevent this elsewhere (in uhash_put) by making sure the last slot
324 * in the table is never filled.
325 *
326 * The size of the table should be prime for this algorithm to work;
327 * otherwise we are not guaranteed that the jump value (the secondary
328 * hash) is relatively prime to the table length.
329 */
330static UHashElement*
331_uhash_find(const UHashtable *hash, UHashTok key,
332            int32_t hashcode) {
333
334    int32_t firstDeleted = -1;  /* assume invalid index */
335    int32_t theIndex, startIndex;
336    int32_t jump = 0; /* lazy evaluate */
337    int32_t tableHash;
338    UHashElement *elements = hash->elements;
339
340    hashcode &= 0x7FFFFFFF; /* must be positive */
341    startIndex = theIndex = (hashcode ^ 0x4000000) % hash->length;
342
343    do {
344        tableHash = elements[theIndex].hashcode;
345        if (tableHash == hashcode) {          /* quick check */
346            if ((*hash->keyComparator)(key, elements[theIndex].key)) {
347                return &(elements[theIndex]);
348            }
349        } else if (!IS_EMPTY_OR_DELETED(tableHash)) {
350            /* We have hit a slot which contains a key-value pair,
351             * but for which the hash code does not match.  Keep
352             * looking.
353             */
354        } else if (tableHash == HASH_EMPTY) { /* empty, end o' the line */
355            break;
356        } else if (firstDeleted < 0) { /* remember first deleted */
357            firstDeleted = theIndex;
358        }
359        if (jump == 0) { /* lazy compute jump */
360            /* The jump value must be relatively prime to the table
361             * length.  As long as the length is prime, then any value
362             * 1..length-1 will be relatively prime to it.
363             */
364            jump = (hashcode % (hash->length - 1)) + 1;
365        }
366        theIndex = (theIndex + jump) % hash->length;
367    } while (theIndex != startIndex);
368
369    if (firstDeleted >= 0) {
370        theIndex = firstDeleted; /* reset if had deleted slot */
371    } else if (tableHash != HASH_EMPTY) {
372        /* We get to this point if the hashtable is full (no empty or
373         * deleted slots), and we've failed to find a match.  THIS
374         * WILL NEVER HAPPEN as long as uhash_put() makes sure that
375         * count is always < length.
376         */
377        U_ASSERT(FALSE);
378        return NULL; /* Never happens if uhash_put() behaves */
379    }
380    return &(elements[theIndex]);
381}
382
383/**
384 * Attempt to grow or shrink the data arrays in order to make the
385 * count fit between the high and low water marks.  hash_put() and
386 * hash_remove() call this method when the count exceeds the high or
387 * low water marks.  This method may do nothing, if memory allocation
388 * fails, or if the count is already in range, or if the length is
389 * already at the low or high limit.  In any case, upon return the
390 * arrays will be valid.
391 */
392static void
393_uhash_rehash(UHashtable *hash, UErrorCode *status) {
394
395    UHashElement *old = hash->elements;
396    int32_t oldLength = hash->length;
397    int32_t newPrimeIndex = hash->primeIndex;
398    int32_t i;
399
400    if (hash->count > hash->highWaterMark) {
401        if (++newPrimeIndex >= PRIMES_LENGTH) {
402            return;
403        }
404    } else if (hash->count < hash->lowWaterMark) {
405        if (--newPrimeIndex < 0) {
406            return;
407        }
408    } else {
409        return;
410    }
411
412    _uhash_allocate(hash, newPrimeIndex, status);
413
414    if (U_FAILURE(*status)) {
415        hash->elements = old;
416        hash->length = oldLength;
417        return;
418    }
419
420    for (i = oldLength - 1; i >= 0; --i) {
421        if (!IS_EMPTY_OR_DELETED(old[i].hashcode)) {
422            UHashElement *e = _uhash_find(hash, old[i].key, old[i].hashcode);
423            U_ASSERT(e != NULL);
424            U_ASSERT(e->hashcode == HASH_EMPTY);
425            e->key = old[i].key;
426            e->value = old[i].value;
427            e->hashcode = old[i].hashcode;
428            ++hash->count;
429        }
430    }
431
432    uprv_free(old);
433}
434
435static UHashTok
436_uhash_remove(UHashtable *hash,
437              UHashTok key) {
438    /* First find the position of the key in the table.  If the object
439     * has not been removed already, remove it.  If the user wanted
440     * keys deleted, then delete it also.  We have to put a special
441     * hashcode in that position that means that something has been
442     * deleted, since when we do a find, we have to continue PAST any
443     * deleted values.
444     */
445    UHashTok result;
446    UHashElement* e = _uhash_find(hash, key, hash->keyHasher(key));
447    U_ASSERT(e != NULL);
448    result.pointer = NULL;
449    result.integer = 0;
450    if (!IS_EMPTY_OR_DELETED(e->hashcode)) {
451        result = _uhash_internalRemoveElement(hash, e);
452        if (hash->count < hash->lowWaterMark) {
453            UErrorCode status = U_ZERO_ERROR;
454            _uhash_rehash(hash, &status);
455        }
456    }
457    return result;
458}
459
460static UHashTok
461_uhash_put(UHashtable *hash,
462           UHashTok key,
463           UHashTok value,
464           int8_t hint,
465           UErrorCode *status) {
466
467    /* Put finds the position in the table for the new value.  If the
468     * key is already in the table, it is deleted, if there is a
469     * non-NULL keyDeleter.  Then the key, the hash and the value are
470     * all put at the position in their respective arrays.
471     */
472    int32_t hashcode;
473    UHashElement* e;
474    UHashTok emptytok;
475
476    if (U_FAILURE(*status)) {
477        goto err;
478    }
479    U_ASSERT(hash != NULL);
480    /* Cannot always check pointer here or iSeries sees NULL every time. */
481    if ((hint & HINT_VALUE_POINTER) && value.pointer == NULL) {
482        /* Disallow storage of NULL values, since NULL is returned by
483         * get() to indicate an absent key.  Storing NULL == removing.
484         */
485        return _uhash_remove(hash, key);
486    }
487    if (hash->count > hash->highWaterMark) {
488        _uhash_rehash(hash, status);
489        if (U_FAILURE(*status)) {
490            goto err;
491        }
492    }
493
494    hashcode = (*hash->keyHasher)(key);
495    e = _uhash_find(hash, key, hashcode);
496    U_ASSERT(e != NULL);
497
498    if (IS_EMPTY_OR_DELETED(e->hashcode)) {
499        /* Important: We must never actually fill the table up.  If we
500         * do so, then _uhash_find() will return NULL, and we'll have
501         * to check for NULL after every call to _uhash_find().  To
502         * avoid this we make sure there is always at least one empty
503         * or deleted slot in the table.  This only is a problem if we
504         * are out of memory and rehash isn't working.
505         */
506        ++hash->count;
507        if (hash->count == hash->length) {
508            /* Don't allow count to reach length */
509            --hash->count;
510            *status = U_MEMORY_ALLOCATION_ERROR;
511            goto err;
512        }
513    }
514
515    /* We must in all cases handle storage properly.  If there was an
516     * old key, then it must be deleted (if the deleter != NULL).
517     * Make hashcodes stored in table positive.
518     */
519    return _uhash_setElement(hash, e, hashcode & 0x7FFFFFFF, key, value, hint);
520
521 err:
522    /* If the deleters are non-NULL, this method adopts its key and/or
523     * value arguments, and we must be sure to delete the key and/or
524     * value in all cases, even upon failure.
525     */
526    HASH_DELETE_KEY_VALUE(hash, key.pointer, value.pointer);
527    emptytok.pointer = NULL; emptytok.integer = 0;
528    return emptytok;
529}
530
531
532/********************************************************************
533 * PUBLIC API
534 ********************************************************************/
535
536U_CAPI UHashtable* U_EXPORT2
537uhash_open(UHashFunction *keyHash,
538           UKeyComparator *keyComp,
539           UValueComparator *valueComp,
540           UErrorCode *status) {
541
542    return _uhash_create(keyHash, keyComp, valueComp, DEFAULT_PRIME_INDEX, status);
543}
544
545U_CAPI UHashtable* U_EXPORT2
546uhash_openSize(UHashFunction *keyHash,
547               UKeyComparator *keyComp,
548               UValueComparator *valueComp,
549               int32_t size,
550               UErrorCode *status) {
551
552    /* Find the smallest index i for which PRIMES[i] >= size. */
553    int32_t i = 0;
554    while (i<(PRIMES_LENGTH-1) && PRIMES[i]<size) {
555        ++i;
556    }
557
558    return _uhash_create(keyHash, keyComp, valueComp, i, status);
559}
560
561U_CAPI UHashtable* U_EXPORT2
562uhash_init(UHashtable *fillinResult,
563           UHashFunction *keyHash,
564           UKeyComparator *keyComp,
565           UValueComparator *valueComp,
566           UErrorCode *status) {
567
568    return _uhash_init(fillinResult, keyHash, keyComp, valueComp, DEFAULT_PRIME_INDEX, status);
569}
570
571U_CAPI void U_EXPORT2
572uhash_close(UHashtable *hash) {
573    if (hash == NULL) {
574        return;
575    }
576    if (hash->elements != NULL) {
577        if (hash->keyDeleter != NULL || hash->valueDeleter != NULL) {
578            int32_t pos=-1;
579            UHashElement *e;
580            while ((e = (UHashElement*) uhash_nextElement(hash, &pos)) != NULL) {
581                HASH_DELETE_KEY_VALUE(hash, e->key.pointer, e->value.pointer);
582            }
583        }
584        uprv_free(hash->elements);
585        hash->elements = NULL;
586    }
587    if (hash->allocated) {
588        uprv_free(hash);
589    }
590}
591
592U_CAPI UHashFunction *U_EXPORT2
593uhash_setKeyHasher(UHashtable *hash, UHashFunction *fn) {
594    UHashFunction *result = hash->keyHasher;
595    hash->keyHasher = fn;
596    return result;
597}
598
599U_CAPI UKeyComparator *U_EXPORT2
600uhash_setKeyComparator(UHashtable *hash, UKeyComparator *fn) {
601    UKeyComparator *result = hash->keyComparator;
602    hash->keyComparator = fn;
603    return result;
604}
605U_CAPI UValueComparator *U_EXPORT2
606uhash_setValueComparator(UHashtable *hash, UValueComparator *fn){
607    UValueComparator *result = hash->valueComparator;
608    hash->valueComparator = fn;
609    return result;
610}
611
612U_CAPI UObjectDeleter *U_EXPORT2
613uhash_setKeyDeleter(UHashtable *hash, UObjectDeleter *fn) {
614    UObjectDeleter *result = hash->keyDeleter;
615    hash->keyDeleter = fn;
616    return result;
617}
618
619U_CAPI UObjectDeleter *U_EXPORT2
620uhash_setValueDeleter(UHashtable *hash, UObjectDeleter *fn) {
621    UObjectDeleter *result = hash->valueDeleter;
622    hash->valueDeleter = fn;
623    return result;
624}
625
626U_CAPI void U_EXPORT2
627uhash_setResizePolicy(UHashtable *hash, enum UHashResizePolicy policy) {
628    UErrorCode status = U_ZERO_ERROR;
629    _uhash_internalSetResizePolicy(hash, policy);
630    hash->lowWaterMark  = (int32_t)(hash->length * hash->lowWaterRatio);
631    hash->highWaterMark = (int32_t)(hash->length * hash->highWaterRatio);
632    _uhash_rehash(hash, &status);
633}
634
635U_CAPI int32_t U_EXPORT2
636uhash_count(const UHashtable *hash) {
637    return hash->count;
638}
639
640U_CAPI void* U_EXPORT2
641uhash_get(const UHashtable *hash,
642          const void* key) {
643    UHashTok keyholder;
644    keyholder.pointer = (void*) key;
645    return _uhash_find(hash, keyholder, hash->keyHasher(keyholder))->value.pointer;
646}
647
648U_CAPI void* U_EXPORT2
649uhash_iget(const UHashtable *hash,
650           int32_t key) {
651    UHashTok keyholder;
652    keyholder.integer = key;
653    return _uhash_find(hash, keyholder, hash->keyHasher(keyholder))->value.pointer;
654}
655
656U_CAPI int32_t U_EXPORT2
657uhash_geti(const UHashtable *hash,
658           const void* key) {
659    UHashTok keyholder;
660    keyholder.pointer = (void*) key;
661    return _uhash_find(hash, keyholder, hash->keyHasher(keyholder))->value.integer;
662}
663
664U_CAPI int32_t U_EXPORT2
665uhash_igeti(const UHashtable *hash,
666           int32_t key) {
667    UHashTok keyholder;
668    keyholder.integer = key;
669    return _uhash_find(hash, keyholder, hash->keyHasher(keyholder))->value.integer;
670}
671
672U_CAPI void* U_EXPORT2
673uhash_put(UHashtable *hash,
674          void* key,
675          void* value,
676          UErrorCode *status) {
677    UHashTok keyholder, valueholder;
678    keyholder.pointer = key;
679    valueholder.pointer = value;
680    return _uhash_put(hash, keyholder, valueholder,
681                      HINT_KEY_POINTER | HINT_VALUE_POINTER,
682                      status).pointer;
683}
684
685U_CAPI void* U_EXPORT2
686uhash_iput(UHashtable *hash,
687           int32_t key,
688           void* value,
689           UErrorCode *status) {
690    UHashTok keyholder, valueholder;
691    keyholder.integer = key;
692    valueholder.pointer = value;
693    return _uhash_put(hash, keyholder, valueholder,
694                      HINT_VALUE_POINTER,
695                      status).pointer;
696}
697
698U_CAPI int32_t U_EXPORT2
699uhash_puti(UHashtable *hash,
700           void* key,
701           int32_t value,
702           UErrorCode *status) {
703    UHashTok keyholder, valueholder;
704    keyholder.pointer = key;
705    valueholder.integer = value;
706    return _uhash_put(hash, keyholder, valueholder,
707                      HINT_KEY_POINTER,
708                      status).integer;
709}
710
711
712U_CAPI int32_t U_EXPORT2
713uhash_iputi(UHashtable *hash,
714           int32_t key,
715           int32_t value,
716           UErrorCode *status) {
717    UHashTok keyholder, valueholder;
718    keyholder.integer = key;
719    valueholder.integer = value;
720    return _uhash_put(hash, keyholder, valueholder,
721                      0, /* neither is a ptr */
722                      status).integer;
723}
724
725U_CAPI void* U_EXPORT2
726uhash_remove(UHashtable *hash,
727             const void* key) {
728    UHashTok keyholder;
729    keyholder.pointer = (void*) key;
730    return _uhash_remove(hash, keyholder).pointer;
731}
732
733U_CAPI void* U_EXPORT2
734uhash_iremove(UHashtable *hash,
735              int32_t key) {
736    UHashTok keyholder;
737    keyholder.integer = key;
738    return _uhash_remove(hash, keyholder).pointer;
739}
740
741U_CAPI int32_t U_EXPORT2
742uhash_removei(UHashtable *hash,
743              const void* key) {
744    UHashTok keyholder;
745    keyholder.pointer = (void*) key;
746    return _uhash_remove(hash, keyholder).integer;
747}
748
749U_CAPI int32_t U_EXPORT2
750uhash_iremovei(UHashtable *hash,
751               int32_t key) {
752    UHashTok keyholder;
753    keyholder.integer = key;
754    return _uhash_remove(hash, keyholder).integer;
755}
756
757U_CAPI void U_EXPORT2
758uhash_removeAll(UHashtable *hash) {
759    int32_t pos = -1;
760    const UHashElement *e;
761    U_ASSERT(hash != NULL);
762    if (hash->count != 0) {
763        while ((e = uhash_nextElement(hash, &pos)) != NULL) {
764            uhash_removeElement(hash, e);
765        }
766    }
767    U_ASSERT(hash->count == 0);
768}
769
770U_CAPI const UHashElement* U_EXPORT2
771uhash_find(const UHashtable *hash, const void* key) {
772    UHashTok keyholder;
773    const UHashElement *e;
774    keyholder.pointer = (void*) key;
775    e = _uhash_find(hash, keyholder, hash->keyHasher(keyholder));
776    return IS_EMPTY_OR_DELETED(e->hashcode) ? NULL : e;
777}
778
779U_CAPI const UHashElement* U_EXPORT2
780uhash_nextElement(const UHashtable *hash, int32_t *pos) {
781    /* Walk through the array until we find an element that is not
782     * EMPTY and not DELETED.
783     */
784    int32_t i;
785    U_ASSERT(hash != NULL);
786    for (i = *pos + 1; i < hash->length; ++i) {
787        if (!IS_EMPTY_OR_DELETED(hash->elements[i].hashcode)) {
788            *pos = i;
789            return &(hash->elements[i]);
790        }
791    }
792
793    /* No more elements */
794    return NULL;
795}
796
797U_CAPI void* U_EXPORT2
798uhash_removeElement(UHashtable *hash, const UHashElement* e) {
799    U_ASSERT(hash != NULL);
800    U_ASSERT(e != NULL);
801    if (!IS_EMPTY_OR_DELETED(e->hashcode)) {
802        UHashElement *nce = (UHashElement *)e;
803        return _uhash_internalRemoveElement(hash, nce).pointer;
804    }
805    return NULL;
806}
807
808/********************************************************************
809 * UHashTok convenience
810 ********************************************************************/
811
812/**
813 * Return a UHashTok for an integer.
814 */
815/*U_CAPI UHashTok U_EXPORT2
816uhash_toki(int32_t i) {
817    UHashTok tok;
818    tok.integer = i;
819    return tok;
820}*/
821
822/**
823 * Return a UHashTok for a pointer.
824 */
825/*U_CAPI UHashTok U_EXPORT2
826uhash_tokp(void* p) {
827    UHashTok tok;
828    tok.pointer = p;
829    return tok;
830}*/
831
832/********************************************************************
833 * PUBLIC Key Hash Functions
834 ********************************************************************/
835
836U_CAPI int32_t U_EXPORT2
837uhash_hashUChars(const UHashTok key) {
838    const UChar *s = (const UChar *)key.pointer;
839    return s == NULL ? 0 : ustr_hashUCharsN(s, u_strlen(s));
840}
841
842U_CAPI int32_t U_EXPORT2
843uhash_hashChars(const UHashTok key) {
844    const char *s = (const char *)key.pointer;
845    return s == NULL ? 0 : ustr_hashCharsN(s, uprv_strlen(s));
846}
847
848U_CAPI int32_t U_EXPORT2
849uhash_hashIChars(const UHashTok key) {
850    const char *s = (const char *)key.pointer;
851    return s == NULL ? 0 : ustr_hashICharsN(s, uprv_strlen(s));
852}
853
854U_CAPI UBool U_EXPORT2
855uhash_equals(const UHashtable* hash1, const UHashtable* hash2){
856    int32_t count1, count2, pos, i;
857
858    if(hash1==hash2){
859        return TRUE;
860    }
861
862    /*
863     * Make sure that we are comparing 2 valid hashes of the same type
864     * with valid comparison functions.
865     * Without valid comparison functions, a binary comparison
866     * of the hash values will yield random results on machines
867     * with 64-bit pointers and 32-bit integer hashes.
868     * A valueComparator is normally optional.
869     */
870    if (hash1==NULL || hash2==NULL ||
871        hash1->keyComparator != hash2->keyComparator ||
872        hash1->valueComparator != hash2->valueComparator ||
873        hash1->valueComparator == NULL)
874    {
875        /*
876        Normally we would return an error here about incompatible hash tables,
877        but we return FALSE instead.
878        */
879        return FALSE;
880    }
881
882    count1 = uhash_count(hash1);
883    count2 = uhash_count(hash2);
884    if(count1!=count2){
885        return FALSE;
886    }
887
888    pos=-1;
889    for(i=0; i<count1; i++){
890        const UHashElement* elem1 = uhash_nextElement(hash1, &pos);
891        const UHashTok key1 = elem1->key;
892        const UHashTok val1 = elem1->value;
893        /* here the keys are not compared, instead the key form hash1 is used to fetch
894         * value from hash2. If the hashes are equal then then both hashes should
895         * contain equal values for the same key!
896         */
897        const UHashElement* elem2 = _uhash_find(hash2, key1, hash2->keyHasher(key1));
898        const UHashTok val2 = elem2->value;
899        if(hash1->valueComparator(val1, val2)==FALSE){
900            return FALSE;
901        }
902    }
903    return TRUE;
904}
905
906/********************************************************************
907 * PUBLIC Comparator Functions
908 ********************************************************************/
909
910U_CAPI UBool U_EXPORT2
911uhash_compareUChars(const UHashTok key1, const UHashTok key2) {
912    const UChar *p1 = (const UChar*) key1.pointer;
913    const UChar *p2 = (const UChar*) key2.pointer;
914    if (p1 == p2) {
915        return TRUE;
916    }
917    if (p1 == NULL || p2 == NULL) {
918        return FALSE;
919    }
920    while (*p1 != 0 && *p1 == *p2) {
921        ++p1;
922        ++p2;
923    }
924    return (UBool)(*p1 == *p2);
925}
926
927U_CAPI UBool U_EXPORT2
928uhash_compareChars(const UHashTok key1, const UHashTok key2) {
929    const char *p1 = (const char*) key1.pointer;
930    const char *p2 = (const char*) key2.pointer;
931    if (p1 == p2) {
932        return TRUE;
933    }
934    if (p1 == NULL || p2 == NULL) {
935        return FALSE;
936    }
937    while (*p1 != 0 && *p1 == *p2) {
938        ++p1;
939        ++p2;
940    }
941    return (UBool)(*p1 == *p2);
942}
943
944U_CAPI UBool U_EXPORT2
945uhash_compareIChars(const UHashTok key1, const UHashTok key2) {
946    const char *p1 = (const char*) key1.pointer;
947    const char *p2 = (const char*) key2.pointer;
948    if (p1 == p2) {
949        return TRUE;
950    }
951    if (p1 == NULL || p2 == NULL) {
952        return FALSE;
953    }
954    while (*p1 != 0 && uprv_tolower(*p1) == uprv_tolower(*p2)) {
955        ++p1;
956        ++p2;
957    }
958    return (UBool)(*p1 == *p2);
959}
960
961/********************************************************************
962 * PUBLIC int32_t Support Functions
963 ********************************************************************/
964
965U_CAPI int32_t U_EXPORT2
966uhash_hashLong(const UHashTok key) {
967    return key.integer;
968}
969
970U_CAPI UBool U_EXPORT2
971uhash_compareLong(const UHashTok key1, const UHashTok key2) {
972    return (UBool)(key1.integer == key2.integer);
973}
974