gecko-dev/ef/gc/hash/dynahash.c

650 строки
18 KiB
C

/* -*- Mode: C; tab-width: 8; indent-tabs-mode: nil; -*-
*
* The contents of this file are subject to the Netscape Public
* License Version 1.1 (the "License"); you may not use this file
* except in compliance with the License. You may obtain a copy of
* the License at http://www.mozilla.org/NPL/
*
* Software distributed under the License is distributed on an "AS
* IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or
* implied. See the License for the specific language governing
* rights and limitations under the License.
*
* The Original Code is mozilla.org code.
*
* The Initial Developer of the Original Code is Netscape
* Communications Corporation. Portions created by Netscape are
* Copyright (C) 1998 Netscape Communications Corporation. All
* Rights Reserved.
*
* Contributor(s):
*/
/*******************************************************************************
* Hash Tables: Based on Larson's in-memory adaption of linear hashing
* [CACM, vol 31, #4, 1988 pp 446-457]
******************************************************************************/
#include "dynahash.h"
#include "prlog.h"
#define MUL(x, y) ((x) << (y##_SHIFT))
#define DIV(x, y) ((x) >> (y##_SHIFT))
#define MOD(x, y) ((x) & ((y) - 1))
PR_IMPLEMENT(PLDynahash*)
PL_NewDynahash(PRUint32 initialSize,
PLHashFun hash, PLEqualFun equal,
PLDestroyFun del, void* context)
{
PRStatus status;
PLDynahash* self = (PLDynahash*)malloc(sizeof(PLDynahash));
if (self == NULL)
return self;
status = PL_DynahashInit(self, initialSize, hash, equal, del, context);
if (status != PR_SUCCESS) {
PL_DynahashDestroy(self);
return NULL;
}
return self;
}
PR_IMPLEMENT(void)
PL_DynahashDestroy(PLDynahash* self)
{
PL_DynahashEmpty(self);
free(self);
}
PR_IMPLEMENT(PRStatus)
PL_DynahashInit(PLDynahash* self, PRUint32 initialSize, PLHashFun hash,
PLEqualFun equal, PLDestroyFun del, void* context)
{
PRUint32 count = DIV(initialSize, PL_DYNAHASH_BUCKETSIZE) + 1;
self->directory = (PLHashDir*)count;
self->hash = hash;
self->equal = equal;
self->destroy = del;
self->keyCount = 0;
self->nextBucketToSplit = 0;
self->maxSize = (PRUint16)MUL(count, PL_DYNAHASH_BUCKETSIZE);
self->bucketCount = 0;
self->context = context;
#ifdef PL_DYNAHASH_STATISTICS
self->accesses = 0;
self->collisions = 0;
self->expansions = 0;
#endif /* PL_DYNAHASH_STATISTICS */
return PR_SUCCESS;
}
static PRStatus
pl_DynahashAllocBucket(PLDynahash* self)
{
PLHashDir* newDir;
PRUint32 i, size;
PRUint16 count = 1;
if (self->bucketCount == 0) {
count = (PRUint16)(PRInt32)self->directory;
}
/* realloc directory */
size = self->bucketCount + count;
newDir = (PLHashDir*)calloc(sizeof(PLHashDir*), size);
if (newDir == NULL)
return PR_FAILURE;
// memset((void*)newDir, 0, size * sizeof(PLHashBucket*));
if (self->bucketCount) {
memcpy((void*)newDir, self->directory, self->bucketCount
* sizeof(PLHashBucket*));
}
/* add new bucket(s) */
for (i = self->bucketCount; i < size; i++) {
PLHashBucket* newBucket;
newBucket = (PLHashBucket*)calloc(sizeof(PLHashBucket*), PL_DYNAHASH_BUCKETSIZE);
if (newBucket == NULL) return PR_FAILURE;
// memset((void*)newBucket, 0, sizeof(PLDynahashElement*)
// * PL_DYNAHASH_BUCKETSIZE);
newDir[i] = newBucket;
}
if (self->bucketCount != 0) {
free(self->directory);
}
self->directory = newDir;
self->bucketCount += count;
return PR_SUCCESS;
}
PR_IMPLEMENT(void)
PL_DynahashCleanup(PLDynahash* self)
{
PL_DynahashEmpty(self);
}
#define PL_DYNAHASH_HASH(resultVar, self, key) \
{ \
PRUword _h_ = self->hash(self->context, key); \
PRUword _a_ = MOD(_h_, self->maxSize); \
if (_a_ < self->nextBucketToSplit) { \
_a_ = MOD(_h_, self->maxSize << 1); /* h % (2 * maxSize) */ \
} \
resultVar = _a_; \
} \
PR_IMPLEMENT(PLDynahashElement*)
PL_DynahashLookup(PLDynahash* self, PLDynahashElement* element)
{
PRUword h;
PRUint32 bucketIndex, binIndex;
PLHashBucket* bin;
PLDynahashElement* result = NULL;
PLHashDir* dir;
PLDynahashElement* e;
if (self->bucketCount == 0) return NULL;
PL_DYNAHASH_HASH(h, self, element);
bucketIndex = DIV(h, PL_DYNAHASH_BUCKETSIZE);
binIndex = MOD(h, PL_DYNAHASH_BUCKETSIZE);
/* hash ensures valid segment */
dir = self->directory;
bin = dir[bucketIndex];
PR_ASSERT(bin);
for (e = bin[binIndex]; e != NULL; e = e->next) {
if (self->equal(self->context, e, element)) {
return e;
}
}
return result;
}
static PRStatus
pl_DynahashExpand(PLDynahash* self)
{
PRStatus err;
if (self->maxSize + self->nextBucketToSplit
< MUL(PL_DYNAHASH_MAX_DIR_SIZE, PL_DYNAHASH_BUCKETSIZE)) {
PRUint32 oldDirIndex, oldBinIndex, newDirIndex, newBinIndex;
PLHashBucket* oldSegment;
PLHashBucket* newSegment;
PRUword newAddress;
PLDynahashElement** previous;
PLDynahashElement* current;
PLDynahashElement** lastOfNew;
/* locate the bucket to split */
oldDirIndex = DIV(self->nextBucketToSplit, PL_DYNAHASH_BUCKETSIZE);
oldSegment = self->directory[oldDirIndex];
oldBinIndex = MOD(self->nextBucketToSplit, PL_DYNAHASH_BUCKETSIZE);
/* expand address space */
newAddress = self->maxSize + self->nextBucketToSplit;
newDirIndex = DIV(newAddress, PL_DYNAHASH_BUCKETSIZE);
newBinIndex = MOD(newAddress, PL_DYNAHASH_BUCKETSIZE);
/* we need to unlock the directory handle here in case AllocBucket grows
the directory */
if (newBinIndex == 0) {
err = pl_DynahashAllocBucket(self);
if (err != PR_SUCCESS) return err;
}
PR_ASSERT(newDirIndex < self->bucketCount);
newSegment = self->directory[newDirIndex];
/* adjust state */
self->nextBucketToSplit++;
if (self->nextBucketToSplit == self->maxSize) {
self->maxSize <<= 1; /* maxSize *= 2 */
self->nextBucketToSplit = 0;
}
/* relocate records to the new bucket */
previous = &oldSegment[oldBinIndex];
current = *previous;
lastOfNew = &newSegment[newBinIndex];
*lastOfNew = NULL;
while (current != NULL) {
PRUword hash;
PL_DYNAHASH_HASH(hash, self, current);
if (hash == newAddress) {
/* attach to the end of the new chain */
*lastOfNew = current;
/* remove from old chain */
*previous = current->next;
lastOfNew = &current->next;
current = current->next;
*lastOfNew = NULL;
}
else {
/* leave it on the old chain */
previous = &current->next;
current = current->next;
}
}
}
return PR_SUCCESS;
}
PR_IMPLEMENT(PRStatus)
PL_DynahashAdd(PLDynahash* self, PLDynahashElement* element, PRBool replace,
PLDynahashElement* *elementToDestroy)
{
/* Returns a pointer to the unused thing (either the key or the replaced key). */
PRStatus err = PR_SUCCESS;
PRUword h;
PRUint32 bucketIndex, binIndex;
PLHashBucket* b;
PLDynahashElement** p;
PLDynahashElement* q;
PLDynahashElement* n;
*elementToDestroy = NULL;
if (self->bucketCount == 0) {
err = pl_DynahashAllocBucket(self);
if (err != PR_SUCCESS) return err;
}
PL_DYNAHASH_HASH(h, self, element);
bucketIndex = DIV(h, PL_DYNAHASH_BUCKETSIZE);
binIndex = MOD(h, PL_DYNAHASH_BUCKETSIZE);
/* hash ensures valid segment */
b = self->directory[bucketIndex];
PR_ASSERT(b);
/* follow collision chain */
p = &b[binIndex];
q = *p;
while (q && !self->equal(self->context, q, element)) {
p = &q->next;
q = *p;
}
if (q && !replace) {
*elementToDestroy = element;
goto done;
}
n = element;
*p = n;
if (q) {
/* replace q with n */
n->next = q->next;
}
else {
/* table over full? expand it: */
if (++self->keyCount / MUL(self->bucketCount, PL_DYNAHASH_BUCKETSIZE)
> PL_DYNAHASH_LOAD_FACTOR) {
err = pl_DynahashExpand(self);
if (err != PR_SUCCESS) return err;
}
}
*elementToDestroy = q;
done:;
return err;
}
PR_IMPLEMENT(PLDynahashElement*)
PL_DynahashRemove(PLDynahash* self, PLDynahashElement* element)
{
PRUword h;
PRUint32 bucketIndex, binIndex;
PLHashBucket* b;
PLDynahashElement** p;
PLDynahashElement* q;
if (self->bucketCount == 0)
return NULL;
PL_DYNAHASH_HASH(h, self, element);
bucketIndex = DIV(h, PL_DYNAHASH_BUCKETSIZE);
binIndex = MOD(h, PL_DYNAHASH_BUCKETSIZE);
b = self->directory[bucketIndex];
PR_ASSERT(b);
p = &b[binIndex];
q = *p;
while (q && !self->equal(self->context, q, element)) {
p = &q->next;
q = *p;
}
if (q) {
/* remove q */
*p = q->next;
--self->keyCount;
/* XXX table can be shrunk? shrink it */
}
return q;
}
static PLDynahashElement*
PL_DynahashRemove1(PLDynahash* self, PRUint32 bucketIndex, PRInt32 binIndex,
int elemIndex)
{
/* only used by PLDynahashIterator, below */
PLHashBucket* b;
PLDynahashElement** p;
PLDynahashElement* q;
if (self->bucketCount == 0)
return NULL;
b = self->directory[bucketIndex];
p = &b[binIndex];
q = *p;
while (elemIndex--) {
p = &q->next;
q = *p;
}
if (q) {
*p = q->next;
--self->keyCount;
}
return q;
}
PR_IMPLEMENT(void)
PL_DynahashEmpty(PLDynahash* self)
{
PRUint32 i, j;
for (i = 0; i < self->bucketCount; i++) {
PLHashBucket* b = self->directory[i];
for (j = 0; j < PL_DYNAHASH_BUCKETSIZE; j++) {
PLDynahashElement* q;
PLDynahashElement* p = b[j];
while (p) {
q = p->next;
self->destroy(self->context, p);
p = q;
}
}
free(b);
}
if (self->bucketCount != 0) {
free(self->directory);
}
self->keyCount = 0;
self->nextBucketToSplit = 0;
self->bucketCount = 0;
self->directory = (PLHashDir*)1;
self->maxSize = MUL(1, PL_DYNAHASH_BUCKETSIZE);
}
#ifdef DEBUG
PR_IMPLEMENT(PRBool)
PL_DynahashIsValid(PLDynahash* self)
{
PLDynahashElement* hep;
PLDynahashIterator iter;
PL_DynahashIteratorInit(&iter, self);
while ((hep = PL_DynahashIteratorNext(&iter)) != NULL) {
if (PL_DynahashLookup(self, hep) != hep) {
/* something else matches */
return PR_FALSE;
}
}
return PR_TRUE;
}
#endif /* DEBUG */
/*******************************************************************************
* Hash Table Iterator
******************************************************************************/
PR_IMPLEMENT(void)
PL_DynahashIteratorInit(PLDynahashIterator* self, PLDynahash* ht)
{
self->table = ht;
self->bucket = 0;
self->bin = -1;
self->curno = 0;
self->lbucket = 0;
self->lbin = 0;
self->lcurno = 0;
self->nextp = NULL;
self->curp = NULL;
if (ht->bucketCount == 0) {
self->done = PR_TRUE;
self->b = NULL;
}
else {
self->done = PR_FALSE;
self->b = ht->directory[self->bucket];
PL_DynahashIteratorNext(self); /* setup nextp */
}
}
static void
pl_DynahashIteratorNext1(PLDynahashIterator* self)
{
if (!self->done) {
/* save last position in case we have to reset */
self->lbucket = self->bucket;
self->lbin = self->bin;
self->lcurno = self->curno;
if (self->nextp) {
self->nextp = self->nextp->next;
self->curno++;
}
while (!self->nextp) {
if (++self->bin >= PL_DYNAHASH_BUCKETSIZE) {
self->bin = 0;
if (++self->bucket < self->table->bucketCount) {
self->b = self->table->directory[self->bucket];
}
else {
self->done = PR_TRUE;
break;
}
}
self->nextp = self->b[self->bin];
self->curno = 0;
}
}
}
PR_IMPLEMENT(PLDynahashElement*)
PL_DynahashIteratorNext(PLDynahashIterator* self)
{
/* This iterator precomputes one next element ahead so that we can detect
when we dropped the current element and not skip elements. */
if (self->curp != self->nextp) {
self->done = PR_TRUE;
}
else {
pl_DynahashIteratorNext1(self);
}
return self->curp;
}
PR_IMPLEMENT(PRUword)
PL_DynahashIteratorCurrentAddress(PLDynahashIterator* self)
{
return (PRUword)(MUL(self->lbucket, PL_DYNAHASH_BUCKETSIZE) + self->lbin);
}
static PLDynahashElement*
PL_DynahashIteratorRemoveCurrent(PLDynahashIterator* self)
{
PLDynahashElement* dp = PL_DynahashRemove1(self->table,
self->lbucket,
self->lbin,
(int) self->lcurno);
if (dp != NULL) {
if ((self->bin == self->lbin) && (self->bucket == self->lbucket)) {
self->curno = self->lcurno;
}
}
return dp;
}
PR_IMPLEMENT(void)
PL_DynahashIteratorReset(PLDynahashIterator* self, PLDynahash* ht)
{
PRInt32 i;
self->table = ht;
if (self->done) return;
if (self->lbin == -1) {
/* a single next has occurred, will emit first element twice */
self->lbin = 0;
}
/* recompute curp from lbucket/lbin/lcurno */
if (self->lbucket >= self->table->bucketCount) {
/* we're done */
self->curp = self->nextp = 0;
self->done = PR_TRUE;
return;
}
self->b = self->table->directory[self->lbucket];
self->curp = self->b[self->lbin];
for (i = 0; i < self->lcurno && self->curp; i++) {
self->curp = self->curp->next;
}
/* nextp also cannot be trusted */
self->bucket = self->lbucket;
self->bin = self->lbin;
self->curno = self->lcurno;
if (self->curp == self->nextp) {
/* special case where curp has been destroyed and nextp was the next one
in the bin */
return;
}
/* curp mibht be zero, it doesn't matter */
self->nextp = self->curp;
pl_DynahashIteratorNext1(self);
}
/*******************************************************************************
* Test
******************************************************************************/
#ifdef DEBUG
typedef struct MyHashElement {
PLDynahashElement hashElement; /* must be first */
PRUint32 key;
PRUint32 value;
} MyHashElement;
PRUint32
Dynahash_Test_Hash(void* context, MyHashElement* x)
{
#ifdef XP_MAC
#pragma unused(context)
#endif
return x->key;
}
PRBool
Dynahash_Test_equal(void* context, MyHashElement* x, MyHashElement* y)
{
#ifdef XP_MAC
#pragma unused(context)
#endif
return (PRBool)(x->key == y->key);
}
void
Dynahash_Test_destroy(void* context, MyHashElement* x)
{
#ifdef XP_MAC
#pragma unused(context)
#endif
free(x);
}
#define kDynahash_Test_Iterations 1000
#include <stdio.h>
PR_IMPLEMENT(PRBool)
PL_DynahashTest(void)
{
PRStatus err;
PRUint32 i;
PRBool result = PR_FALSE;
PLDynahash* ht = PL_NewDynahash(14,
(PLHashFun)Dynahash_Test_Hash,
(PLEqualFun)Dynahash_Test_equal,
(PLDestroyFun)Dynahash_Test_destroy,
NULL);
for (i = 0; i < kDynahash_Test_Iterations; i++) {
MyHashElement* oldElem;
MyHashElement* elem;
printf("seeding %ld\n", i);
elem = (MyHashElement*)malloc(sizeof(MyHashElement));
if (elem == NULL) goto done;
elem->hashElement.next = NULL;
elem->key = i;
elem->value = i;
err = PL_DynahashAdd(ht, (PLDynahashElement*)elem, PR_FALSE,
(PLDynahashElement**)&oldElem);
if (err != PR_SUCCESS) goto done;
free(oldElem);
}
for (i = 0; i < kDynahash_Test_Iterations; i += 2) {
MyHashElement elem;
MyHashElement* found;
PRBool wasFound = PR_FALSE;
printf("deleting %ld\n", i);
elem.hashElement.next = NULL;
elem.key = i;
found = (MyHashElement*)PL_DynahashRemove(ht, (PLDynahashElement*)&elem);
if (found == NULL)
goto done;
wasFound = (PRBool)(found->value == i);
if (!wasFound)
goto done;
}
for (i = 0; i < kDynahash_Test_Iterations; i++) {
MyHashElement elem;
MyHashElement* found;
printf("getting %ld\n", i);
elem.hashElement.next = NULL;
elem.key = i;
found = (MyHashElement*)PL_DynahashLookup(ht, (PLDynahashElement*)&elem);
if (found) {
if (i & 1 == 1) {
PRBool wasFound = (PRBool)(found->value == i);
if (!wasFound) goto done;
}
else {
goto done;
}
}
else {
if (i & 1 == 1) {
goto done;
}
/* else ok */
}
}
result = PR_TRUE;
done:;
PL_DynahashDestroy(ht);
PR_ASSERT(result);
return result;
}
#endif /* DEBUG */
/******************************************************************************/