/*
* Returns whether trg2 contains all trigrams in trg1.
* This relies on the trigram arrays being sorted.
*/
bool
trgm_contained_by(TRGM *trg1, TRGM *trg2)
{
trgm *ptr1,
*ptr2;
int len1,
len2;
ptr1 = GETARR(trg1);
ptr2 = GETARR(trg2);
len1 = ARRNELEM(trg1);
len2 = ARRNELEM(trg2);
while (ptr1 - GETARR(trg1) < len1 && ptr2 - GETARR(trg2) < len2)
{
int res = CMPTRGM(ptr1, ptr2);
if (res < 0)
return false;
else if (res > 0)
ptr2++;
else
{
ptr1++;
ptr2++;
}
}
if (ptr1 - GETARR(trg1) < len1)
return false;
else
return true;
}
Datum
show_trgm(PG_FUNCTION_ARGS)
{
text *in = PG_GETARG_TEXT_P(0);
TRGM *trg;
Datum *d;
ArrayType *a;
trgm *ptr;
int i;
trg = generate_trgm(VARDATA(in), VARSIZE(in) - VARHDRSZ);
d = (Datum *) palloc(sizeof(Datum) * (1 + ARRNELEM(trg)));
for (i = 0, ptr = GETARR(trg); i < ARRNELEM(trg); i++, ptr++)
{
text *item = (text *) palloc(VARHDRSZ + Max(12, pg_database_encoding_max_length() * 3));
if (pg_database_encoding_max_length() > 1 && !ISPRINTABLETRGM(ptr))
{
snprintf(VARDATA(item), 12, "0x%06x", trgm2int(ptr));
SET_VARSIZE(item, VARHDRSZ + strlen(VARDATA(item)));
}
else
{
SET_VARSIZE(item, VARHDRSZ + 3);
CPTRGM(VARDATA(item), ptr);
}
d[i] = PointerGetDatum(item);
}
a = construct_array(
d,
ARRNELEM(trg),
TEXTOID,
-1,
false,
'i'
);
for (i = 0; i < ARRNELEM(trg); i++)
pfree(DatumGetPointer(d[i]));
pfree(d);
pfree(trg);
PG_FREE_IF_COPY(in, 0);
PG_RETURN_POINTER(a);
}
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_trgm(PG_FUNCTION_ARGS)
{
text *val = (text *) PG_GETARG_TEXT_P(0);
int32 *nentries = (int32 *) PG_GETARG_POINTER(1);
Datum *entries = NULL;
TRGM *trg;
int4 trglen;
*nentries = 0;
trg = generate_trgm(VARDATA(val), VARSIZE(val) - VARHDRSZ);
trglen = ARRNELEM(trg);
if (trglen > 0)
{
trgm *ptr;
int4 i = 0,
item;
*nentries = (int32) trglen;
entries = (Datum *) palloc(sizeof(Datum) * trglen);
ptr = GETARR(trg);
while (ptr - GETARR(trg) < ARRNELEM(trg))
{
item = trgm2int(ptr);
entries[i++] = Int32GetDatum(item);
ptr++;
}
if (PG_NARGS() > 4)
{
/*
* Function called from query extracting
*/
Pointer **extra_data = (Pointer **) PG_GETARG_POINTER(4);
*extra_data = (Pointer *) palloc0(sizeof(Pointer) * (*nentries));
*(int32 *) (*extra_data) = trglen;
}
}
PG_RETURN_POINTER(entries);
}
float4
cnt_sml(TRGM *trg1, TRGM *trg2)
{
trgm *ptr1,
*ptr2;
int count = 0;
int len1,
len2;
ptr1 = GETARR(trg1);
ptr2 = GETARR(trg2);
len1 = ARRNELEM(trg1);
len2 = ARRNELEM(trg2);
/* explicit test is needed to avoid 0/0 division when both lengths are 0 */
if (len1 <= 0 || len2 <= 0)
return (float4) 0.0;
while (ptr1 - GETARR(trg1) < len1 && ptr2 - GETARR(trg2) < len2)
{
int res = CMPTRGM(ptr1, ptr2);
if (res < 0)
ptr1++;
else if (res > 0)
ptr2++;
else
{
ptr1++;
ptr2++;
count++;
}
}
#ifdef DIVUNION
return ((float4) count) / ((float4) (len1 + len2 - count));
#else
return ((float4) count) / ((float4) ((len1 > len2) ? len1 : len2));
#endif
}
float4
cnt_sml(TRGM *trg1, TRGM *trg2, bool inexact)
{
trgm *ptr1,
*ptr2;
int count = 0;
int len1,
len2;
ptr1 = GETARR(trg1);
ptr2 = GETARR(trg2);
len1 = ARRNELEM(trg1);
len2 = ARRNELEM(trg2);
/* explicit test is needed to avoid 0/0 division when both lengths are 0 */
if (len1 <= 0 || len2 <= 0)
return (float4) 0.0;
while (ptr1 - GETARR(trg1) < len1 && ptr2 - GETARR(trg2) < len2)
{
int res = CMPTRGM(ptr1, ptr2);
if (res < 0)
ptr1++;
else if (res > 0)
ptr2++;
else
{
ptr1++;
ptr2++;
count++;
}
}
/*
* If inexact then len2 is equal to count, because we don't know actual
* length of second string in inexact search and we can assume that count
* is a lower bound of len2.
*/
return CALCSML(count, len1, inexact ? count : len2);
}
static void
makesign(BITVECP sign, SignTSVector *a)
{
int32 k,
len = ARRNELEM(a);
int32 *ptr = GETARR(a);
MemSet((void *) sign, 0, sizeof(BITVEC));
for (k = 0; k < len; k++)
HASH(sign, ptr[k]);
}
static void
makesign(BITVECP sign, GISTTYPE * a)
{
int4 k,
len = ARRNELEM(a);
int4 *ptr = GETARR(a);
MemSet((void *) sign, 0, sizeof(BITVEC));
for (k = 0; k < len; k++)
HASH(sign, ptr[k]);
}
float4
cnt_sml(TRGM *trg1, TRGM *trg2)
{
trgm *ptr1,
*ptr2;
int count = 0;
int len1,
len2;
ptr1 = GETARR(trg1);
ptr2 = GETARR(trg2);
len1 = ARRNELEM(trg1);
len2 = ARRNELEM(trg2);
while (ptr1 - GETARR(trg1) < len1 && ptr2 - GETARR(trg2) < len2)
{
int res = CMPTRGM(ptr1, ptr2);
if (res < 0)
ptr1++;
else if (res > 0)
ptr2++;
else
{
ptr1++;
ptr2++;
count++;
}
}
#ifdef DIVUNION
return ((((float4) count) / ((float4) (len1 + len2 - count))));
#else
return (((float) count) / ((float) ((len1 > len2) ? len1 : len2)));
#endif
}
/*
* Return a palloc'd boolean array showing, for each trigram in "query",
* whether it is present in the trigram array "key".
* This relies on the "key" array being sorted, but "query" need not be.
*/
bool *
trgm_presence_map(TRGM *query, TRGM *key)
{
bool *result;
trgm *ptrq = GETARR(query),
*ptrk = GETARR(key);
int lenq = ARRNELEM(query),
lenk = ARRNELEM(key),
i;
result = (bool *) palloc0(lenq * sizeof(bool));
/* for each query trigram, do a binary search in the key array */
for (i = 0; i < lenq; i++)
{
int lo = 0;
int hi = lenk;
while (lo < hi)
{
int mid = (lo + hi) / 2;
int res = CMPTRGM(ptrq, ptrk + mid);
if (res < 0)
hi = mid;
else if (res > 0)
lo = mid + 1;
else
{
result[i] = true;
break;
}
}
ptrq++;
}
return result;
}
static void
makesign(BITVECP sign, TRGM *a)
{
int32 k,
len = ARRNELEM(a);
trgm *ptr = GETARR(a);
int32 tmp = 0;
MemSet((void *) sign, 0, sizeof(BITVEC));
SETBIT(sign, SIGLENBIT); /* set last unused bit */
for (k = 0; k < len; k++)
{
CPTRGM(((char *) &tmp), ptr + k);
HASH(sign, tmp);
}
}
Datum
gtsvectorout(PG_FUNCTION_ARGS)
{
SignTSVector *key = (SignTSVector *) DatumGetPointer(PG_DETOAST_DATUM(PG_GETARG_POINTER(0)));
char *outbuf;
if (outbuf_maxlen == 0)
outbuf_maxlen = 2 * EXTRALEN + Max(strlen(SINGOUTSTR), strlen(ARROUTSTR)) + 1;
outbuf = palloc(outbuf_maxlen);
if (ISARRKEY(key))
sprintf(outbuf, ARROUTSTR, (int) ARRNELEM(key));
else
{
int cnttrue = (ISALLTRUE(key)) ? SIGLENBIT : sizebitvec(GETSIGN(key));
sprintf(outbuf, SINGOUTSTR, cnttrue, (int) SIGLENBIT - cnttrue);
}
PG_FREE_IF_COPY(key, 0);
PG_RETURN_POINTER(outbuf);
}
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;
TrgmPackedGraph *graph;
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;
case RegExpICaseStrategyNumber:
#ifndef IGNORECASE
elog(ERROR, "cannot handle ~* with case-sensitive trigrams");
#endif
/* FALL THRU */
case RegExpStrategyNumber:
trg = createTrgmNFA(val, PG_GET_COLLATION(),
&graph, CurrentMemoryContext);
if (trg && ARRNELEM(trg) > 0)
{
/*
* Successful regex processing: store NFA-like graph as
* extra_data. GIN API requires an array of nentries
* Pointers, but we just put the same value in each element.
*/
trglen = ARRNELEM(trg);
*extra_data = (Pointer *) palloc(sizeof(Pointer) * trglen);
for (i = 0; i < trglen; i++)
(*extra_data)[i] = (Pointer) graph;
}
else
{
/* No result: have to do full index scan. */
*nentries = 0;
*searchMode = GIN_SEARCH_MODE_ALL;
PG_RETURN_POINTER(entries);
}
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_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);
}
