postgresql/contrib/pg_trgm/trgm_gin.c
Tom Lane 6e2f3ae884 Support LIKE and ILIKE index searches via contrib/pg_trgm indexes.
Unlike Btree-based LIKE optimization, this works for non-left-anchored
search patterns.  The effectiveness of the search depends on how many
trigrams can be extracted from the pattern.  (The worst case, with no
trigrams, degrades to a full-table scan, so this isn't a panacea.  But
it can be very useful.)

Alexander Korotkov, reviewed by Jan Urbanski
2011-01-31 21:34:49 -05:00

199 lines
4.5 KiB
C

/*
* contrib/pg_trgm/trgm_gin.c
*/
#include "postgres.h"
#include "trgm.h"
#include "access/gin.h"
#include "access/itup.h"
#include "access/skey.h"
#include "access/tuptoaster.h"
#include "storage/bufpage.h"
#include "utils/array.h"
#include "utils/builtins.h"
PG_FUNCTION_INFO_V1(gin_extract_trgm);
Datum gin_extract_trgm(PG_FUNCTION_ARGS);
PG_FUNCTION_INFO_V1(gin_extract_value_trgm);
Datum gin_extract_value_trgm(PG_FUNCTION_ARGS);
PG_FUNCTION_INFO_V1(gin_extract_query_trgm);
Datum gin_extract_query_trgm(PG_FUNCTION_ARGS);
PG_FUNCTION_INFO_V1(gin_trgm_consistent);
Datum gin_trgm_consistent(PG_FUNCTION_ARGS);
/*
* This function can only be called if a pre-9.1 version of the GIN operator
* class definition is present in the catalogs (probably as a consequence
* of upgrade-in-place). Complain.
*/
Datum
gin_extract_trgm(PG_FUNCTION_ARGS)
{
ereport(ERROR,
(errmsg("GIN operator class for pg_trgm is out of date"),
errhint("Please drop and re-create the pg_trgm catalog entries.")));
PG_RETURN_NULL();
}
Datum
gin_extract_value_trgm(PG_FUNCTION_ARGS)
{
text *val = (text *) PG_GETARG_TEXT_P(0);
int32 *nentries = (int32 *) PG_GETARG_POINTER(1);
Datum *entries = NULL;
TRGM *trg;
int32 trglen;
*nentries = 0;
trg = generate_trgm(VARDATA(val), VARSIZE(val) - VARHDRSZ);
trglen = ARRNELEM(trg);
if (trglen > 0)
{
trgm *ptr;
int32 i;
*nentries = trglen;
entries = (Datum *) palloc(sizeof(Datum) * trglen);
ptr = GETARR(trg);
for (i = 0; i < trglen; i++)
{
int32 item = trgm2int(ptr);
entries[i] = Int32GetDatum(item);
ptr++;
}
}
PG_RETURN_POINTER(entries);
}
Datum
gin_extract_query_trgm(PG_FUNCTION_ARGS)
{
text *val = (text *) PG_GETARG_TEXT_P(0);
int32 *nentries = (int32 *) PG_GETARG_POINTER(1);
StrategyNumber strategy = PG_GETARG_UINT16(2);
/* bool **pmatch = (bool **) PG_GETARG_POINTER(3); */
/* Pointer *extra_data = (Pointer *) PG_GETARG_POINTER(4); */
/* bool **nullFlags = (bool **) PG_GETARG_POINTER(5); */
int32 *searchMode = (int32 *) PG_GETARG_POINTER(6);
Datum *entries = NULL;
TRGM *trg;
int32 trglen;
trgm *ptr;
int32 i;
switch (strategy)
{
case SimilarityStrategyNumber:
trg = generate_trgm(VARDATA(val), VARSIZE(val) - VARHDRSZ);
break;
case ILikeStrategyNumber:
#ifndef IGNORECASE
elog(ERROR, "cannot handle ~~* with case-sensitive trigrams");
#endif
/* FALL THRU */
case LikeStrategyNumber:
/*
* For wildcard search we extract all the trigrams that every
* potentially-matching string must include.
*/
trg = generate_wildcard_trgm(VARDATA(val), VARSIZE(val) - VARHDRSZ);
break;
default:
elog(ERROR, "unrecognized strategy number: %d", strategy);
trg = NULL; /* keep compiler quiet */
break;
}
trglen = ARRNELEM(trg);
*nentries = trglen;
if (trglen > 0)
{
entries = (Datum *) palloc(sizeof(Datum) * trglen);
ptr = GETARR(trg);
for (i = 0; i < trglen; i++)
{
int32 item = trgm2int(ptr);
entries[i] = Int32GetDatum(item);
ptr++;
}
}
/*
* If no trigram was extracted then we have to scan all the index.
*/
if (trglen == 0)
*searchMode = GIN_SEARCH_MODE_ALL;
PG_RETURN_POINTER(entries);
}
Datum
gin_trgm_consistent(PG_FUNCTION_ARGS)
{
bool *check = (bool *) PG_GETARG_POINTER(0);
StrategyNumber strategy = PG_GETARG_UINT16(1);
/* text *query = PG_GETARG_TEXT_P(2); */
int32 nkeys = PG_GETARG_INT32(3);
/* Pointer *extra_data = (Pointer *) PG_GETARG_POINTER(4); */
bool *recheck = (bool *) PG_GETARG_POINTER(5);
bool res;
int32 i,
ntrue;
/* All cases served by this function are inexact */
*recheck = true;
switch (strategy)
{
case SimilarityStrategyNumber:
/* Count the matches */
ntrue = 0;
for (i = 0; i < nkeys; i++)
{
if (check[i])
ntrue++;
}
#ifdef DIVUNION
res = (nkeys == ntrue) ? true : ((((((float4) ntrue) / ((float4) (nkeys - ntrue)))) >= trgm_limit) ? true : false);
#else
res = (nkeys == 0) ? false : ((((((float4) ntrue) / ((float4) nkeys))) >= trgm_limit) ? true : false);
#endif
break;
case ILikeStrategyNumber:
#ifndef IGNORECASE
elog(ERROR, "cannot handle ~~* with case-sensitive trigrams");
#endif
/* FALL THRU */
case LikeStrategyNumber:
/* Check if all extracted trigrams are presented. */
res = true;
for (i = 0; i < nkeys; i++)
{
if (!check[i])
{
res = false;
break;
}
}
break;
default:
elog(ERROR, "unrecognized strategy number: %d", strategy);
res = false; /* keep compiler quiet */
break;
}
PG_RETURN_BOOL(res);
}