dynahash.c

来自「PostgreSQL7.4.6 for Linux」· C语言 代码 · 共 953 行 · 第 1/2 页

/*-------------------------------------------------------------------------
 *
 * dynahash.c
 *	  dynamic hash tables
 *
 *
 * Portions Copyright (c) 1996-2003, PostgreSQL Global Development Group
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 *
 * IDENTIFICATION
 *	  $Header: /cvsroot/pgsql/src/backend/utils/hash/dynahash.c,v 1.48 2003/08/19 01:13:41 tgl Exp $
 *
 *-------------------------------------------------------------------------
 */
/*
 *
 * Dynamic hashing, after CACM April 1988 pp 446-457, by Per-Ake Larson.
 * Coded into C, with minor code improvements, and with hsearch(3) interface,
 * by [e-mail], Jul 26, 1988: 13:16;
 * also, hcreate/hdestroy routines added to simulate hsearch(3).
 *
 * These routines simulate hsearch(3) and family, with the important
 * difference that the hash table is dynamic - can grow indefinitely
 * beyond its original size (as supplied to hcreate()).
 *
 * Performance appears to be comparable to that of hsearch(3).
 * The 'source-code' options referred to in hsearch(3)'s 'man' page
 * are not implemented; otherwise functionality is identical.
 *
 * Compilation controls:
 * DEBUG controls some informative traces, mainly for debugging.
 * HASH_STATISTICS causes HashAccesses and HashCollisions to be maintained;
 * when combined with HASH_DEBUG, these are displayed by hdestroy().
 *
 * Problems & fixes to [e-mail]. WARNING: relies on pre-processor
 * concatenation property, in probably unnecessary code 'optimisation'.
 *
 * Modified [e-mail] February 1990
 *		added multiple table interface
 * Modified by [e-mail] April 1990
 *		changed ctl structure for shared memory
 */

#include "postgres.h"

#include "utils/dynahash.h"
#include "utils/hsearch.h"
#include "utils/memutils.h"

/*
 * Key (also entry) part of a HASHELEMENT
 */
#define ELEMENTKEY(helem)  (((char *)(helem)) + MAXALIGN(sizeof(HASHELEMENT)))

/*
 * Fast MOD arithmetic, assuming that y is a power of 2 !
 */
#define MOD(x,y)			   ((x) & ((y)-1))

/*
 * Private function prototypes
 */
static void *DynaHashAlloc(Size size);
static HASHSEGMENT seg_alloc(HTAB *hashp);
static bool element_alloc(HTAB *hashp);
static bool dir_realloc(HTAB *hashp);
static bool expand_table(HTAB *hashp);
static bool hdefault(HTAB *hashp);
static bool init_htab(HTAB *hashp, long nelem);
static void hash_corrupted(HTAB *hashp);


/*
 * memory allocation routines
 */
static MemoryContext DynaHashCxt = NULL;
static MemoryContext CurrentDynaHashCxt = NULL;

static void *
DynaHashAlloc(Size size)
{
	Assert(MemoryContextIsValid(CurrentDynaHashCxt));
	return MemoryContextAlloc(CurrentDynaHashCxt, size);
}

#define MEM_ALLOC		DynaHashAlloc
#define MEM_FREE		pfree


#if HASH_STATISTICS
static long hash_accesses,
			hash_collisions,
			hash_expansions;
#endif


/************************** CREATE ROUTINES **********************/

HTAB *
hash_create(const char *tabname, long nelem, HASHCTL *info, int flags)
{
	HTAB	   *hashp;
	HASHHDR    *hctl;

	/* First time through, create a memory context for hash tables */
	if (!DynaHashCxt)
		DynaHashCxt = AllocSetContextCreate(TopMemoryContext,
											"DynaHash",
											ALLOCSET_DEFAULT_MINSIZE,
											ALLOCSET_DEFAULT_INITSIZE,
											ALLOCSET_DEFAULT_MAXSIZE);

	/* Select allocation context for this hash table */
	if (flags & HASH_CONTEXT)
		CurrentDynaHashCxt = info->hcxt;
	else
		CurrentDynaHashCxt = DynaHashCxt;

	/* Initialize the hash header */
	hashp = (HTAB *) MEM_ALLOC(sizeof(HTAB));
	if (!hashp)
		return NULL;
	MemSet(hashp, 0, sizeof(HTAB));

	hashp->tabname = (char *) MEM_ALLOC(strlen(tabname) + 1);
	strcpy(hashp->tabname, tabname);

	if (flags & HASH_FUNCTION)
		hashp->hash = info->hash;
	else
		hashp->hash = string_hash;		/* default hash function */

	/*
	 * If you don't specify a match function, it defaults to strncmp() if
	 * you used string_hash (either explicitly or by default) and to
	 * memcmp() otherwise.  (Prior to PostgreSQL 7.4, memcmp() was always
	 * used.)
	 */
	if (flags & HASH_COMPARE)
		hashp->match = info->match;
	else if (hashp->hash == string_hash)
		hashp->match = (HashCompareFunc) strncmp;
	else
		hashp->match = memcmp;

	if (flags & HASH_SHARED_MEM)
	{
		/*
		 * ctl structure is preallocated for shared memory tables. Note
		 * that HASH_DIRSIZE had better be set as well.
		 */
		hashp->hctl = info->hctl;
		hashp->dir = info->dir;
		hashp->alloc = info->alloc;
		hashp->hcxt = NULL;
		hashp->isshared = true;

		/* hash table already exists, we're just attaching to it */
		if (flags & HASH_ATTACH)
			return hashp;
	}
	else
	{
		/* setup hash table defaults */
		hashp->hctl = NULL;
		hashp->dir = NULL;
		hashp->alloc = MEM_ALLOC;
		hashp->hcxt = CurrentDynaHashCxt;
		hashp->isshared = false;
	}

	if (!hashp->hctl)
	{
		hashp->hctl = (HASHHDR *) hashp->alloc(sizeof(HASHHDR));
		if (!hashp->hctl)
			return NULL;
	}

	if (!hdefault(hashp))
		return NULL;

	hctl = hashp->hctl;
#ifdef HASH_STATISTICS
	hctl->accesses = hctl->collisions = 0;
#endif

	if (flags & HASH_SEGMENT)
	{
		hctl->ssize = info->ssize;
		hctl->sshift = my_log2(info->ssize);
		/* ssize had better be a power of 2 */
		Assert(hctl->ssize == (1L << hctl->sshift));
	}
	if (flags & HASH_FFACTOR)
		hctl->ffactor = info->ffactor;

	/*
	 * SHM hash tables have fixed directory size passed by the caller.
	 */
	if (flags & HASH_DIRSIZE)
	{
		hctl->max_dsize = info->max_dsize;
		hctl->dsize = info->dsize;
	}

	/*
	 * hash table now allocates space for key and data but you have to say
	 * how much space to allocate
	 */
	if (flags & HASH_ELEM)
	{
		hctl->keysize = info->keysize;
		hctl->entrysize = info->entrysize;
	}

	if (flags & HASH_ALLOC)
		hashp->alloc = info->alloc;
	else
	{
		/* remaining hash table structures live in child of given context */
		hashp->hcxt = AllocSetContextCreate(CurrentDynaHashCxt,
											"DynaHashTable",
											ALLOCSET_DEFAULT_MINSIZE,
											ALLOCSET_DEFAULT_INITSIZE,
											ALLOCSET_DEFAULT_MAXSIZE);
		CurrentDynaHashCxt = hashp->hcxt;
	}

	if (!init_htab(hashp, nelem))
	{
		hash_destroy(hashp);
		return NULL;
	}
	return hashp;
}

/*
 * Set default HASHHDR parameters.
 */
static bool
hdefault(HTAB *hashp)
{
	HASHHDR    *hctl = hashp->hctl;

	MemSet(hctl, 0, sizeof(HASHHDR));

	hctl->ssize = DEF_SEGSIZE;
	hctl->sshift = DEF_SEGSIZE_SHIFT;
	hctl->dsize = DEF_DIRSIZE;
	hctl->ffactor = DEF_FFACTOR;
	hctl->nentries = 0;
	hctl->nsegs = 0;

	/* I added these MS. */

	/* rather pointless defaults for key & entry size */
	hctl->keysize = sizeof(char *);
	hctl->entrysize = 2 * sizeof(char *);

	/* table has no fixed maximum size */
	hctl->max_dsize = NO_MAX_DSIZE;

	/* garbage collection for HASH_REMOVE */
	hctl->freeList = NULL;

	return true;
}


static bool
init_htab(HTAB *hashp, long nelem)
{
	HASHHDR    *hctl = hashp->hctl;
	HASHSEGMENT *segp;
	int			nbuckets;
	int			nsegs;

	/*
	 * Divide number of elements by the fill factor to determine a desired
	 * number of buckets.  Allocate space for the next greater power of
	 * two number of buckets
	 */
	nelem = (nelem - 1) / hctl->ffactor + 1;

	nbuckets = 1 << my_log2(nelem);

	hctl->max_bucket = hctl->low_mask = nbuckets - 1;
	hctl->high_mask = (nbuckets << 1) - 1;

	/*
	 * Figure number of directory segments needed, round up to a power of
	 * 2
	 */
	nsegs = (nbuckets - 1) / hctl->ssize + 1;
	nsegs = 1 << my_log2(nsegs);

	/*
	 * Make sure directory is big enough. If pre-allocated directory is
	 * too small, choke (caller screwed up).
	 */
	if (nsegs > hctl->dsize)
	{
		if (!(hashp->dir))
			hctl->dsize = nsegs;
		else
			return false;
	}

	/* Allocate a directory */
	if (!(hashp->dir))
	{
		CurrentDynaHashCxt = hashp->hcxt;
		hashp->dir = (HASHSEGMENT *)
			hashp->alloc(hctl->dsize * sizeof(HASHSEGMENT));
		if (!hashp->dir)
			return false;
	}

	/* Allocate initial segments */
	for (segp = hashp->dir; hctl->nsegs < nsegs; hctl->nsegs++, segp++)
	{
		*segp = seg_alloc(hashp);
		if (*segp == NULL)
			return false;
	}

#if HASH_DEBUG
	fprintf(stderr, "init_htab:\n%s%p\n%s%ld\n%s%ld\n%s%d\n%s%ld\n%s%u\n%s%x\n%s%x\n%s%ld\n%s%ld\n",
			"TABLE POINTER   ", hashp,
			"DIRECTORY SIZE  ", hctl->dsize,
			"SEGMENT SIZE    ", hctl->ssize,
			"SEGMENT SHIFT   ", hctl->sshift,
			"FILL FACTOR     ", hctl->ffactor,
			"MAX BUCKET      ", hctl->max_bucket,
			"HIGH MASK       ", hctl->high_mask,
			"LOW  MASK       ", hctl->low_mask,
			"NSEGS           ", hctl->nsegs,
			"NENTRIES        ", hctl->nentries);
#endif
	return true;
}

/*
 * Estimate the space needed for a hashtable containing the given number
 * of entries of given size.
 * NOTE: this is used to estimate the footprint of hashtables in shared
 * memory; therefore it does not count HTAB which is in local memory.
 * NB: assumes that all hash structure parameters have default values!
 */
long
hash_estimate_size(long num_entries, Size entrysize)
{
	long		size = 0;
	long		nBuckets,
				nSegments,
				nDirEntries,
				nElementAllocs,
				elementSize;

	/* estimate number of buckets wanted */
	nBuckets = 1L << my_log2((num_entries - 1) / DEF_FFACTOR + 1);
	/* # of segments needed for nBuckets */
	nSegments = 1L << my_log2((nBuckets - 1) / DEF_SEGSIZE + 1);
	/* directory entries */
	nDirEntries = DEF_DIRSIZE;
	while (nDirEntries < nSegments)
		nDirEntries <<= 1;		/* dir_alloc doubles dsize at each call */

	/* fixed control info */
	size += MAXALIGN(sizeof(HASHHDR));	/* but not HTAB, per above */
	/* directory */
	size += MAXALIGN(nDirEntries * sizeof(HASHSEGMENT));
	/* segments */
	size += nSegments * MAXALIGN(DEF_SEGSIZE * sizeof(HASHBUCKET));
	/* elements --- allocated in groups of HASHELEMENT_ALLOC_INCR */
	elementSize = MAXALIGN(sizeof(HASHELEMENT)) + MAXALIGN(entrysize);
	nElementAllocs = (num_entries - 1) / HASHELEMENT_ALLOC_INCR + 1;
	size += nElementAllocs * HASHELEMENT_ALLOC_INCR * elementSize;

	return size;
}

/*
 * Select an appropriate directory size for a hashtable with the given
 * maximum number of entries.
 * This is only needed for hashtables in shared memory, whose directories
 * cannot be expanded dynamically.
 * NB: assumes that all hash structure parameters have default values!
 *
 * XXX this had better agree with the behavior of init_htab()...
 */
long
hash_select_dirsize(long num_entries)
{
	long		nBuckets,
				nSegments,
				nDirEntries;

	/* estimate number of buckets wanted */
	nBuckets = 1L << my_log2((num_entries - 1) / DEF_FFACTOR + 1);
	/* # of segments needed for nBuckets */
	nSegments = 1L << my_log2((nBuckets - 1) / DEF_SEGSIZE + 1);
	/* directory entries */
	nDirEntries = DEF_DIRSIZE;
	while (nDirEntries < nSegments)
		nDirEntries <<= 1;		/* dir_alloc doubles dsize at each call */

	return nDirEntries;
}


/********************** DESTROY ROUTINES ************************/

void
hash_destroy(HTAB *hashp)
{
	if (hashp != NULL)
	{
		/* allocation method must be one we know how to free, too */
		Assert(hashp->alloc == MEM_ALLOC);
		/* so this hashtable must have it's own context */
		Assert(hashp->hcxt != NULL);

		hash_stats("destroy", hashp);

		/*
		 * Free buckets, dir etc. by destroying the hash table's memory
		 * context.
		 */
		MemoryContextDelete(hashp->hcxt);

		/*
		 * Free the HTAB and control structure, which are allocated in the
		 * parent context (DynaHashCxt or the context given by the caller
		 * of hash_create()).
		 */
		MEM_FREE(hashp->hctl);
		MEM_FREE(hashp->tabname);
		MEM_FREE(hashp);
	}
}

void
hash_stats(const char *where, HTAB *hashp)
{
#if HASH_STATISTICS
	fprintf(stderr, "%s: this HTAB -- accesses %ld collisions %ld\n",
			where, hashp->hctl->accesses, hashp->hctl->collisions);

	fprintf(stderr, "hash_stats: entries %ld keysize %ld maxp %u segmentcount %ld\n",
			hashp->hctl->nentries, hashp->hctl->keysize,
			hashp->hctl->max_bucket, hashp->hctl->nsegs);
	fprintf(stderr, "%s: total accesses %ld total collisions %ld\n",
			where, hash_accesses, hash_collisions);
	fprintf(stderr, "hash_stats: total expansions %ld\n",
			hash_expansions);
#endif
}

/*******************************SEARCH ROUTINES *****************************/


/* Convert a hash value to a bucket number */
static inline uint32
calc_bucket(HASHHDR *hctl, uint32 hash_val)
{
	uint32		bucket;

	bucket = hash_val & hctl->high_mask;
	if (bucket > hctl->max_bucket)
		bucket = bucket & hctl->low_mask;

	return bucket;
}

/*----------