mirror of
https://github.com/mozilla/gecko-dev.git
synced 2024-11-07 04:05:49 +00:00
032597fa19
the hashtable is shrinked only once. r=brendan
484 lines
13 KiB
C
484 lines
13 KiB
C
/* -*- Mode: C; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*-
|
|
*
|
|
* ***** BEGIN LICENSE BLOCK *****
|
|
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
|
|
*
|
|
* The contents of this file are subject to the Mozilla 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/MPL/
|
|
*
|
|
* 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 Communicator client code, released
|
|
* March 31, 1998.
|
|
*
|
|
* The Initial Developer of the Original Code is
|
|
* Netscape Communications Corporation.
|
|
* Portions created by the Initial Developer are Copyright (C) 1998
|
|
* the Initial Developer. All Rights Reserved.
|
|
*
|
|
* Contributor(s):
|
|
*
|
|
* Alternatively, the contents of this file may be used under the terms of
|
|
* either of the GNU General Public License Version 2 or later (the "GPL"),
|
|
* or the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
|
|
* in which case the provisions of the GPL or the LGPL are applicable instead
|
|
* of those above. If you wish to allow use of your version of this file only
|
|
* under the terms of either the GPL or the LGPL, and not to allow others to
|
|
* use your version of this file under the terms of the MPL, indicate your
|
|
* decision by deleting the provisions above and replace them with the notice
|
|
* and other provisions required by the GPL or the LGPL. If you do not delete
|
|
* the provisions above, a recipient may use your version of this file under
|
|
* the terms of any one of the MPL, the GPL or the LGPL.
|
|
*
|
|
* ***** END LICENSE BLOCK ***** */
|
|
|
|
/*
|
|
* PR hash table package.
|
|
*/
|
|
#include "jsstddef.h"
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include "jstypes.h"
|
|
#include "jsbit.h"
|
|
#include "jsutil.h" /* Added by JSIFY */
|
|
#include "jshash.h" /* Added by JSIFY */
|
|
|
|
/* Compute the number of buckets in ht */
|
|
#define NBUCKETS(ht) JS_BIT(JS_HASH_BITS - (ht)->shift)
|
|
|
|
/* The smallest table has 16 buckets */
|
|
#define MINBUCKETSLOG2 4
|
|
#define MINBUCKETS JS_BIT(MINBUCKETSLOG2)
|
|
|
|
/* Compute the maximum entries given n buckets that we will tolerate, ~90% */
|
|
#define OVERLOADED(n) ((n) - ((n) >> 3))
|
|
|
|
/* Compute the number of entries below which we shrink the table by half */
|
|
#define UNDERLOADED(n) (((n) > MINBUCKETS) ? ((n) >> 2) : 0)
|
|
|
|
/*
|
|
** Stubs for default hash allocator ops.
|
|
*/
|
|
static void *
|
|
DefaultAllocTable(void *pool, size_t size)
|
|
{
|
|
return malloc(size);
|
|
}
|
|
|
|
static void
|
|
DefaultFreeTable(void *pool, void *item)
|
|
{
|
|
free(item);
|
|
}
|
|
|
|
static JSHashEntry *
|
|
DefaultAllocEntry(void *pool, const void *key)
|
|
{
|
|
return (JSHashEntry*) malloc(sizeof(JSHashEntry));
|
|
}
|
|
|
|
static void
|
|
DefaultFreeEntry(void *pool, JSHashEntry *he, uintN flag)
|
|
{
|
|
if (flag == HT_FREE_ENTRY)
|
|
free(he);
|
|
}
|
|
|
|
static JSHashAllocOps defaultHashAllocOps = {
|
|
DefaultAllocTable, DefaultFreeTable,
|
|
DefaultAllocEntry, DefaultFreeEntry
|
|
};
|
|
|
|
JS_PUBLIC_API(JSHashTable *)
|
|
JS_NewHashTable(uint32 n, JSHashFunction keyHash,
|
|
JSHashComparator keyCompare, JSHashComparator valueCompare,
|
|
JSHashAllocOps *allocOps, void *allocPriv)
|
|
{
|
|
JSHashTable *ht;
|
|
size_t nb;
|
|
|
|
if (n <= MINBUCKETS) {
|
|
n = MINBUCKETSLOG2;
|
|
} else {
|
|
n = JS_CeilingLog2(n);
|
|
if ((int32)n < 0)
|
|
return NULL;
|
|
}
|
|
|
|
if (!allocOps) allocOps = &defaultHashAllocOps;
|
|
|
|
ht = (JSHashTable*) allocOps->allocTable(allocPriv, sizeof *ht);
|
|
if (!ht)
|
|
return NULL;
|
|
memset(ht, 0, sizeof *ht);
|
|
ht->shift = JS_HASH_BITS - n;
|
|
n = JS_BIT(n);
|
|
nb = n * sizeof(JSHashEntry *);
|
|
ht->buckets = (JSHashEntry**) allocOps->allocTable(allocPriv, nb);
|
|
if (!ht->buckets) {
|
|
allocOps->freeTable(allocPriv, ht);
|
|
return NULL;
|
|
}
|
|
memset(ht->buckets, 0, nb);
|
|
|
|
ht->keyHash = keyHash;
|
|
ht->keyCompare = keyCompare;
|
|
ht->valueCompare = valueCompare;
|
|
ht->allocOps = allocOps;
|
|
ht->allocPriv = allocPriv;
|
|
return ht;
|
|
}
|
|
|
|
JS_PUBLIC_API(void)
|
|
JS_HashTableDestroy(JSHashTable *ht)
|
|
{
|
|
uint32 i, n;
|
|
JSHashEntry *he, **hep;
|
|
JSHashAllocOps *allocOps = ht->allocOps;
|
|
void *allocPriv = ht->allocPriv;
|
|
|
|
n = NBUCKETS(ht);
|
|
for (i = 0; i < n; i++) {
|
|
hep = &ht->buckets[i];
|
|
while ((he = *hep) != NULL) {
|
|
*hep = he->next;
|
|
allocOps->freeEntry(allocPriv, he, HT_FREE_ENTRY);
|
|
}
|
|
}
|
|
#ifdef DEBUG
|
|
memset(ht->buckets, 0xDB, n * sizeof ht->buckets[0]);
|
|
#endif
|
|
allocOps->freeTable(allocPriv, ht->buckets);
|
|
#ifdef DEBUG
|
|
memset(ht, 0xDB, sizeof *ht);
|
|
#endif
|
|
allocOps->freeTable(allocPriv, ht);
|
|
}
|
|
|
|
/*
|
|
* Multiplicative hash, from Knuth 6.4.
|
|
*/
|
|
#define BUCKET_HEAD(ht, keyHash) \
|
|
(&(ht)->buckets[((keyHash) * JS_GOLDEN_RATIO) >> (ht)->shift])
|
|
|
|
JS_PUBLIC_API(JSHashEntry **)
|
|
JS_HashTableRawLookup(JSHashTable *ht, JSHashNumber keyHash, const void *key)
|
|
{
|
|
JSHashEntry *he, **hep, **hep0;
|
|
|
|
#ifdef HASHMETER
|
|
ht->nlookups++;
|
|
#endif
|
|
hep = hep0 = BUCKET_HEAD(ht, keyHash);
|
|
while ((he = *hep) != NULL) {
|
|
if (he->keyHash == keyHash && ht->keyCompare(key, he->key)) {
|
|
/* Move to front of chain if not already there */
|
|
if (hep != hep0) {
|
|
*hep = he->next;
|
|
he->next = *hep0;
|
|
*hep0 = he;
|
|
}
|
|
return hep0;
|
|
}
|
|
hep = &he->next;
|
|
#ifdef HASHMETER
|
|
ht->nsteps++;
|
|
#endif
|
|
}
|
|
return hep;
|
|
}
|
|
|
|
static JSBool
|
|
Resize(JSHashTable *ht, uint32 newshift)
|
|
{
|
|
size_t nb, nentries, i;
|
|
JSHashEntry **oldbuckets, *he, *next, **hep;
|
|
#ifdef DEBUG
|
|
size_t nold = NBUCKETS(ht);
|
|
#endif
|
|
|
|
JS_ASSERT(newshift < JS_HASH_BITS);
|
|
|
|
nb = (size_t)1 << (JS_HASH_BITS - newshift);
|
|
|
|
/* Integer overflow protection. */
|
|
if (nb > (size_t)-1 / sizeof(JSHashEntry*))
|
|
return JS_FALSE;
|
|
nb *= sizeof(JSHashEntry*);
|
|
|
|
oldbuckets = ht->buckets;
|
|
ht->buckets = (JSHashEntry**)ht->allocOps->allocTable(ht->allocPriv, nb);
|
|
if (!ht->buckets) {
|
|
ht->buckets = oldbuckets;
|
|
return JS_FALSE;
|
|
}
|
|
memset(ht->buckets, 0, nb);
|
|
|
|
ht->shift = newshift;
|
|
nentries = ht->nentries;
|
|
|
|
for (i = 0; nentries != 0; i++) {
|
|
for (he = oldbuckets[i]; he; he = next) {
|
|
JS_ASSERT(nentries != 0);
|
|
--nentries;
|
|
next = he->next;
|
|
hep = BUCKET_HEAD(ht, he->keyHash);
|
|
|
|
/*
|
|
* Since he comes from the old table, it must be unique and we
|
|
* simply add it to the head of bucket chain without chain lookup.
|
|
*/
|
|
he->next = *hep;
|
|
*hep = he;
|
|
}
|
|
}
|
|
#ifdef DEBUG
|
|
memset(oldbuckets, 0xDB, nold * sizeof oldbuckets[0]);
|
|
#endif
|
|
ht->allocOps->freeTable(ht->allocPriv, oldbuckets);
|
|
return JS_TRUE;
|
|
}
|
|
|
|
JS_PUBLIC_API(JSHashEntry *)
|
|
JS_HashTableRawAdd(JSHashTable *ht, JSHashEntry **hep,
|
|
JSHashNumber keyHash, const void *key, void *value)
|
|
{
|
|
uint32 n;
|
|
JSHashEntry *he;
|
|
|
|
/* Grow the table if it is overloaded */
|
|
n = NBUCKETS(ht);
|
|
if (ht->nentries >= OVERLOADED(n)) {
|
|
if (!Resize(ht, ht->shift - 1))
|
|
return NULL;
|
|
#ifdef HASHMETER
|
|
ht->ngrows++;
|
|
#endif
|
|
hep = JS_HashTableRawLookup(ht, keyHash, key);
|
|
}
|
|
|
|
/* Make a new key value entry */
|
|
he = ht->allocOps->allocEntry(ht->allocPriv, key);
|
|
if (!he)
|
|
return NULL;
|
|
he->keyHash = keyHash;
|
|
he->key = key;
|
|
he->value = value;
|
|
he->next = *hep;
|
|
*hep = he;
|
|
ht->nentries++;
|
|
return he;
|
|
}
|
|
|
|
JS_PUBLIC_API(JSHashEntry *)
|
|
JS_HashTableAdd(JSHashTable *ht, const void *key, void *value)
|
|
{
|
|
JSHashNumber keyHash;
|
|
JSHashEntry *he, **hep;
|
|
|
|
keyHash = ht->keyHash(key);
|
|
hep = JS_HashTableRawLookup(ht, keyHash, key);
|
|
if ((he = *hep) != NULL) {
|
|
/* Hit; see if values match */
|
|
if (ht->valueCompare(he->value, value)) {
|
|
/* key,value pair is already present in table */
|
|
return he;
|
|
}
|
|
if (he->value)
|
|
ht->allocOps->freeEntry(ht->allocPriv, he, HT_FREE_VALUE);
|
|
he->value = value;
|
|
return he;
|
|
}
|
|
return JS_HashTableRawAdd(ht, hep, keyHash, key, value);
|
|
}
|
|
|
|
JS_PUBLIC_API(void)
|
|
JS_HashTableRawRemove(JSHashTable *ht, JSHashEntry **hep, JSHashEntry *he)
|
|
{
|
|
uint32 n;
|
|
|
|
*hep = he->next;
|
|
ht->allocOps->freeEntry(ht->allocPriv, he, HT_FREE_ENTRY);
|
|
|
|
/* Shrink table if it's underloaded */
|
|
n = NBUCKETS(ht);
|
|
if (--ht->nentries < UNDERLOADED(n)) {
|
|
Resize(ht, ht->shift + 1);
|
|
#ifdef HASHMETER
|
|
ht->nshrinks++;
|
|
#endif
|
|
}
|
|
}
|
|
|
|
JS_PUBLIC_API(JSBool)
|
|
JS_HashTableRemove(JSHashTable *ht, const void *key)
|
|
{
|
|
JSHashNumber keyHash;
|
|
JSHashEntry *he, **hep;
|
|
|
|
keyHash = ht->keyHash(key);
|
|
hep = JS_HashTableRawLookup(ht, keyHash, key);
|
|
if ((he = *hep) == NULL)
|
|
return JS_FALSE;
|
|
|
|
/* Hit; remove element */
|
|
JS_HashTableRawRemove(ht, hep, he);
|
|
return JS_TRUE;
|
|
}
|
|
|
|
JS_PUBLIC_API(void *)
|
|
JS_HashTableLookup(JSHashTable *ht, const void *key)
|
|
{
|
|
JSHashNumber keyHash;
|
|
JSHashEntry *he, **hep;
|
|
|
|
keyHash = ht->keyHash(key);
|
|
hep = JS_HashTableRawLookup(ht, keyHash, key);
|
|
if ((he = *hep) != NULL) {
|
|
return he->value;
|
|
}
|
|
return NULL;
|
|
}
|
|
|
|
/*
|
|
** Iterate over the entries in the hash table calling func for each
|
|
** entry found. Stop if "f" says to (return value & JS_ENUMERATE_STOP).
|
|
** Return a count of the number of elements scanned.
|
|
*/
|
|
JS_PUBLIC_API(int)
|
|
JS_HashTableEnumerateEntries(JSHashTable *ht, JSHashEnumerator f, void *arg)
|
|
{
|
|
JSHashEntry *he, **hep, **bucket;
|
|
uint32 nlimit, n, nbuckets, newlog2;
|
|
int rv;
|
|
|
|
nlimit = ht->nentries;
|
|
n = 0;
|
|
for (bucket = ht->buckets; n != nlimit; ++bucket) {
|
|
hep = bucket;
|
|
while ((he = *hep) != NULL) {
|
|
JS_ASSERT(n < nlimit);
|
|
rv = f(he, n, arg);
|
|
n++;
|
|
if (rv & HT_ENUMERATE_REMOVE) {
|
|
*hep = he->next;
|
|
ht->allocOps->freeEntry(ht->allocPriv, he, HT_FREE_ENTRY);
|
|
--ht->nentries;
|
|
} else {
|
|
hep = &he->next;
|
|
}
|
|
if (rv & HT_ENUMERATE_STOP) {
|
|
goto out;
|
|
}
|
|
}
|
|
}
|
|
|
|
out:
|
|
/* Shrink table if removal of entries made it underloaded */
|
|
if (ht->nentries != nlimit) {
|
|
JS_ASSERT(ht->nentries < nlimit);
|
|
nbuckets = NBUCKETS(ht);
|
|
if (MINBUCKETS < nbuckets && ht->nentries < UNDERLOADED(nbuckets)) {
|
|
newlog2 = JS_CeilingLog2(ht->nentries);
|
|
if (newlog2 < MINBUCKETSLOG2)
|
|
newlog2 = MINBUCKETSLOG2;
|
|
|
|
/* Check that we really shrink the table. */
|
|
JS_ASSERT(JS_HASH_BITS - ht->shift > newlog2);
|
|
Resize(ht, JS_HASH_BITS - newlog2);
|
|
}
|
|
}
|
|
return (int)n;
|
|
}
|
|
|
|
#ifdef HASHMETER
|
|
#include <math.h>
|
|
#include <stdio.h>
|
|
|
|
JS_PUBLIC_API(void)
|
|
JS_HashTableDumpMeter(JSHashTable *ht, JSHashEnumerator dump, FILE *fp)
|
|
{
|
|
double sqsum, mean, variance, sigma;
|
|
uint32 nchains, nbuckets, nentries;
|
|
uint32 i, n, maxChain, maxChainLen;
|
|
JSHashEntry *he;
|
|
|
|
sqsum = 0;
|
|
nchains = 0;
|
|
maxChainLen = 0;
|
|
nbuckets = NBUCKETS(ht);
|
|
for (i = 0; i < nbuckets; i++) {
|
|
he = ht->buckets[i];
|
|
if (!he)
|
|
continue;
|
|
nchains++;
|
|
for (n = 0; he; he = he->next)
|
|
n++;
|
|
sqsum += n * n;
|
|
if (n > maxChainLen) {
|
|
maxChainLen = n;
|
|
maxChain = i;
|
|
}
|
|
}
|
|
nentries = ht->nentries;
|
|
mean = (double)nentries / nchains;
|
|
variance = nchains * sqsum - nentries * nentries;
|
|
if (variance < 0 || nchains == 1)
|
|
variance = 0;
|
|
else
|
|
variance /= nchains * (nchains - 1);
|
|
sigma = sqrt(variance);
|
|
|
|
fprintf(fp, "\nHash table statistics:\n");
|
|
fprintf(fp, " number of lookups: %u\n", ht->nlookups);
|
|
fprintf(fp, " number of entries: %u\n", ht->nentries);
|
|
fprintf(fp, " number of grows: %u\n", ht->ngrows);
|
|
fprintf(fp, " number of shrinks: %u\n", ht->nshrinks);
|
|
fprintf(fp, " mean steps per hash: %g\n", (double)ht->nsteps
|
|
/ ht->nlookups);
|
|
fprintf(fp, "mean hash chain length: %g\n", mean);
|
|
fprintf(fp, " standard deviation: %g\n", sigma);
|
|
fprintf(fp, " max hash chain length: %u\n", maxChainLen);
|
|
fprintf(fp, " max hash chain: [%u]\n", maxChain);
|
|
|
|
for (he = ht->buckets[maxChain], i = 0; he; he = he->next, i++)
|
|
if (dump(he, i, fp) != HT_ENUMERATE_NEXT)
|
|
break;
|
|
}
|
|
#endif /* HASHMETER */
|
|
|
|
JS_PUBLIC_API(int)
|
|
JS_HashTableDump(JSHashTable *ht, JSHashEnumerator dump, FILE *fp)
|
|
{
|
|
int count;
|
|
|
|
count = JS_HashTableEnumerateEntries(ht, dump, fp);
|
|
#ifdef HASHMETER
|
|
JS_HashTableDumpMeter(ht, dump, fp);
|
|
#endif
|
|
return count;
|
|
}
|
|
|
|
JS_PUBLIC_API(JSHashNumber)
|
|
JS_HashString(const void *key)
|
|
{
|
|
JSHashNumber h;
|
|
const unsigned char *s;
|
|
|
|
h = 0;
|
|
for (s = (const unsigned char *)key; *s; s++)
|
|
h = (h >> (JS_HASH_BITS - 4)) ^ (h << 4) ^ *s;
|
|
return h;
|
|
}
|
|
|
|
JS_PUBLIC_API(int)
|
|
JS_CompareValues(const void *v1, const void *v2)
|
|
{
|
|
return v1 == v2;
|
|
}
|