/*
* Main entry point to GiST index build. Initially calls insert over and over,
* but switches to more efficient buffering build algorithm after a certain
* number of tuples (unless buffering mode is disabled).
*/
Datum
gistbuild(PG_FUNCTION_ARGS)
{
Relation heap = (Relation) PG_GETARG_POINTER(0);
Relation index = (Relation) PG_GETARG_POINTER(1);
IndexInfo *indexInfo = (IndexInfo *) PG_GETARG_POINTER(2);
IndexBuildResult *result;
double reltuples;
GISTBuildState buildstate;
Buffer buffer;
Page page;
MemoryContext oldcxt = CurrentMemoryContext;
int fillfactor;
buildstate.indexrel = index;
if (index->rd_options)
{
/* Get buffering mode from the options string */
GiSTOptions *options = (GiSTOptions *) index->rd_options;
char *bufferingMode = (char *) options + options->bufferingModeOffset;
if (strcmp(bufferingMode, "on") == 0)
buildstate.bufferingMode = GIST_BUFFERING_STATS;
else if (strcmp(bufferingMode, "off") == 0)
buildstate.bufferingMode = GIST_BUFFERING_DISABLED;
else
buildstate.bufferingMode = GIST_BUFFERING_AUTO;
fillfactor = options->fillfactor;
}
else
{
/*
* By default, switch to buffering mode when the index grows too large
* to fit in cache.
*/
buildstate.bufferingMode = GIST_BUFFERING_AUTO;
fillfactor = GIST_DEFAULT_FILLFACTOR;
}
/* Calculate target amount of free space to leave on pages */
buildstate.freespace = BLCKSZ * (100 - fillfactor) / 100;
/*
* We expect to be called exactly once for any index relation. If that's
* not the case, big trouble's what we have.
*/
if (RelationGetNumberOfBlocks(index) != 0)
elog(ERROR, "index \"%s\" already contains data",
RelationGetRelationName(index));
/* no locking is needed */
buildstate.giststate = initGISTstate(index);
/*
* Create a temporary memory context that is reset once for each tuple
* processed. (Note: we don't bother to make this a child of the
* giststate's scanCxt, so we have to delete it separately at the end.)
*/
buildstate.giststate->tempCxt = createTempGistContext();
/* initialize the root page */
buffer = gistNewBuffer(index);
Assert(BufferGetBlockNumber(buffer) == GIST_ROOT_BLKNO);
page = BufferGetPage(buffer);
START_CRIT_SECTION();
GISTInitBuffer(buffer, F_LEAF);
MarkBufferDirty(buffer);
if (RelationNeedsWAL(index))
{
XLogRecPtr recptr;
XLogRecData rdata;
rdata.data = (char *) &(index->rd_node);
rdata.len = sizeof(RelFileNode);
rdata.buffer = InvalidBuffer;
rdata.next = NULL;
recptr = XLogInsert(RM_GIST_ID, XLOG_GIST_CREATE_INDEX, &rdata);
PageSetLSN(page, recptr);
}
else
PageSetLSN(page, gistGetFakeLSN(heap));
UnlockReleaseBuffer(buffer);
END_CRIT_SECTION();
/* build the index */
buildstate.indtuples = 0;
buildstate.indtuplesSize = 0;
/*
* Do the heap scan.
*/
reltuples = IndexBuildHeapScan(heap, index, indexInfo, true,
gistBuildCallback, (void *) &buildstate);
/*
* If buffering was used, flush out all the tuples that are still in the
* buffers.
*/
if (buildstate.bufferingMode == GIST_BUFFERING_ACTIVE)
{
elog(DEBUG1, "all tuples processed, emptying buffers");
gistEmptyAllBuffers(&buildstate);
gistFreeBuildBuffers(buildstate.gfbb);
}
/* okay, all heap tuples are indexed */
MemoryContextSwitchTo(oldcxt);
MemoryContextDelete(buildstate.giststate->tempCxt);
freeGISTstate(buildstate.giststate);
/*
* Return statistics
*/
result = (IndexBuildResult *) palloc(sizeof(IndexBuildResult));
result->heap_tuples = reltuples;
result->index_tuples = (double) buildstate.indtuples;
PG_RETURN_POINTER(result);
}
/*
** GiST Compress and Decompress methods
*/
Datum
g_int_compress(PG_FUNCTION_ARGS)
{
GISTENTRY *entry = (GISTENTRY *) PG_GETARG_POINTER(0);
GISTENTRY *retval;
ArrayType *r;
int len;
int *dr;
int i,
min,
cand;
if (entry->leafkey)
{
r = (ArrayType *) PG_DETOAST_DATUM_COPY(entry->key);
PREPAREARR(r);
r->flags |= LEAFKEY;
retval = palloc(sizeof(GISTENTRY));
gistentryinit(*retval, PointerGetDatum(r),
entry->rel, entry->page, entry->offset, VARSIZE(r), FALSE);
PG_RETURN_POINTER(retval);
}
r = (ArrayType *) PG_DETOAST_DATUM(entry->key);
if (ISLEAFKEY(r) || ARRISVOID(r))
{
if (r != (ArrayType *) DatumGetPointer(entry->key))
pfree(r);
PG_RETURN_POINTER(entry);
}
if ((len = ARRNELEMS(r)) >= 2 * MAXNUMRANGE)
{ /* compress */
if (r == (ArrayType *) DatumGetPointer(entry->key))
r = (ArrayType *) PG_DETOAST_DATUM_COPY(entry->key);
r = resize_intArrayType(r, 2 * (len));
dr = ARRPTR(r);
for (i = len - 1; i >= 0; i--)
dr[2 * i] = dr[2 * i + 1] = dr[i];
len *= 2;
cand = 1;
while (len > MAXNUMRANGE * 2)
{
min = 0x7fffffff;
for (i = 2; i < len; i += 2)
if (min > (dr[i] - dr[i - 1]))
{
min = (dr[i] - dr[i - 1]);
cand = i;
}
memmove((void *) &dr[cand - 1], (void *) &dr[cand + 1], (len - cand - 1) * sizeof(int));
len -= 2;
}
r = resize_intArrayType(r, len);
retval = palloc(sizeof(GISTENTRY));
gistentryinit(*retval, PointerGetDatum(r),
entry->rel, entry->page, entry->offset, VARSIZE(r), FALSE);
PG_RETURN_POINTER(retval);
}
else
PG_RETURN_POINTER(entry);
PG_RETURN_POINTER(entry);
}
/*
* in without morphology
*/
Datum
ltxtq_in(PG_FUNCTION_ARGS)
{
PG_RETURN_POINTER(queryin((char *) PG_GETARG_POINTER(0)));
}
/*
* record_recv - binary input routine for any composite type.
*/
Datum
record_recv(PG_FUNCTION_ARGS)
{
StringInfo buf = (StringInfo) PG_GETARG_POINTER(0);
Oid tupType = PG_GETARG_OID(1);
#ifdef NOT_USED
int32 typmod = PG_GETARG_INT32(2);
#endif
HeapTupleHeader result;
int32 tupTypmod;
TupleDesc tupdesc;
HeapTuple tuple;
RecordIOData *my_extra;
int ncolumns;
int usercols;
int validcols;
int i;
Datum *values;
bool *nulls;
/*
* Use the passed type unless it's RECORD; we can't support input of
* anonymous types, mainly because there's no good way to figure out which
* anonymous type is wanted. Note that for RECORD, what we'll probably
* actually get is RECORD's typelem, ie, zero.
*/
if (tupType == InvalidOid || tupType == RECORDOID)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("input of anonymous composite types is not implemented")));
tupTypmod = -1; /* for all non-anonymous types */
tupdesc = lookup_rowtype_tupdesc(tupType, tupTypmod);
ncolumns = tupdesc->natts;
/*
* We arrange to look up the needed I/O info just once per series of
* calls, assuming the record type doesn't change underneath us.
*/
my_extra = (RecordIOData *) fcinfo->flinfo->fn_extra;
if (my_extra == NULL ||
my_extra->ncolumns != ncolumns)
{
fcinfo->flinfo->fn_extra =
MemoryContextAlloc(fcinfo->flinfo->fn_mcxt,
sizeof(RecordIOData) - sizeof(ColumnIOData)
+ ncolumns * sizeof(ColumnIOData));
my_extra = (RecordIOData *) fcinfo->flinfo->fn_extra;
my_extra->record_type = InvalidOid;
my_extra->record_typmod = 0;
}
if (my_extra->record_type != tupType ||
my_extra->record_typmod != tupTypmod)
{
MemSet(my_extra, 0,
sizeof(RecordIOData) - sizeof(ColumnIOData)
+ ncolumns * sizeof(ColumnIOData));
my_extra->record_type = tupType;
my_extra->record_typmod = tupTypmod;
my_extra->ncolumns = ncolumns;
}
values = (Datum *) palloc(ncolumns * sizeof(Datum));
nulls = (bool *) palloc(ncolumns * sizeof(bool));
/* Fetch number of columns user thinks it has */
usercols = pq_getmsgint(buf, 4);
/* Need to scan to count nondeleted columns */
validcols = 0;
for (i = 0; i < ncolumns; i++)
{
if (!tupdesc->attrs[i]->attisdropped)
validcols++;
}
if (usercols != validcols)
ereport(ERROR,
(errcode(ERRCODE_DATATYPE_MISMATCH),
errmsg("wrong number of columns: %d, expected %d",
usercols, validcols)));
/* Process each column */
for (i = 0; i < ncolumns; i++)
{
ColumnIOData *column_info = &my_extra->columns[i];
Oid column_type = tupdesc->attrs[i]->atttypid;
Oid coltypoid;
int itemlen;
StringInfoData item_buf;
StringInfo bufptr;
char csave;
/* Ignore dropped columns in datatype, but fill with nulls */
if (tupdesc->attrs[i]->attisdropped)
{
values[i] = (Datum) 0;
nulls[i] = true;
continue;
}
/* Verify column datatype */
coltypoid = pq_getmsgint(buf, sizeof(Oid));
if (coltypoid != column_type)
ereport(ERROR,
(errcode(ERRCODE_DATATYPE_MISMATCH),
errmsg("wrong data type: %u, expected %u",
coltypoid, column_type)));
/* Get and check the item length */
itemlen = pq_getmsgint(buf, 4);
if (itemlen < -1 || itemlen > (buf->len - buf->cursor))
ereport(ERROR,
(errcode(ERRCODE_INVALID_BINARY_REPRESENTATION),
errmsg("insufficient data left in message")));
if (itemlen == -1)
{
/* -1 length means NULL */
bufptr = NULL;
nulls[i] = true;
csave = 0; /* keep compiler quiet */
}
else
{
/*
* Rather than copying data around, we just set up a phony
* StringInfo pointing to the correct portion of the input buffer.
* We assume we can scribble on the input buffer so as to maintain
* the convention that StringInfos have a trailing null.
*/
item_buf.data = &buf->data[buf->cursor];
item_buf.maxlen = itemlen + 1;
item_buf.len = itemlen;
item_buf.cursor = 0;
buf->cursor += itemlen;
csave = buf->data[buf->cursor];
buf->data[buf->cursor] = '\0';
bufptr = &item_buf;
nulls[i] = false;
}
/* Now call the column's receiveproc */
if (column_info->column_type != column_type)
{
getTypeBinaryInputInfo(column_type,
&column_info->typiofunc,
&column_info->typioparam);
fmgr_info_cxt(column_info->typiofunc, &column_info->proc,
fcinfo->flinfo->fn_mcxt);
column_info->column_type = column_type;
}
values[i] = ReceiveFunctionCall(&column_info->proc,
bufptr,
column_info->typioparam,
tupdesc->attrs[i]->atttypmod);
if (bufptr)
{
/* Trouble if it didn't eat the whole buffer */
if (item_buf.cursor != itemlen)
ereport(ERROR,
(errcode(ERRCODE_INVALID_BINARY_REPRESENTATION),
errmsg("improper binary format in record column %d",
i + 1)));
buf->data[buf->cursor] = csave;
}
}
tuple = heap_form_tuple(tupdesc, values, nulls);
/*
* We cannot return tuple->t_data because heap_form_tuple allocates it as
* part of a larger chunk, and our caller may expect to be able to pfree
* our result. So must copy the info into a new palloc chunk.
*/
result = (HeapTupleHeader) palloc(tuple->t_len);
memcpy(result, tuple->t_data, tuple->t_len);
heap_freetuple(tuple);
pfree(values);
pfree(nulls);
ReleaseTupleDesc(tupdesc);
PG_RETURN_HEAPTUPLEHEADER(result);
}
Datum
ghstore_compress(PG_FUNCTION_ARGS)
{
GISTENTRY *entry = (GISTENTRY *) PG_GETARG_POINTER(0);
GISTENTRY *retval = entry;
if (entry->leafkey)
{
GISTTYPE *res = (GISTTYPE *) palloc0(CALCGTSIZE(0));
HStore *toastedval = (HStore *) DatumGetPointer(entry->key);
HStore *val = (HStore *) DatumGetPointer(PG_DETOAST_DATUM(entry->key));
HEntry *ptr = ARRPTR(val);
char *words = STRPTR(val);
SET_VARSIZE(res, CALCGTSIZE(0));
while (ptr - ARRPTR(val) < val->size)
{
int h;
h = crc32_sz((char *) (words + ptr->pos), ptr->keylen);
HASH(GETSIGN(res), h);
if (!ptr->valisnull)
{
h = crc32_sz((char *) (words + ptr->pos + ptr->keylen), ptr->vallen);
HASH(GETSIGN(res), h);
}
ptr++;
}
if (val != toastedval)
pfree(val);
retval = (GISTENTRY *) palloc(sizeof(GISTENTRY));
gistentryinit(*retval, PointerGetDatum(res),
entry->rel, entry->page,
entry->offset,
FALSE);
}
else if (!ISALLTRUE(DatumGetPointer(entry->key)))
{
int4 i;
GISTTYPE *res;
BITVECP sign = GETSIGN(DatumGetPointer(entry->key));
LOOPBYTE
{
if ((sign[i] & 0xff) != 0xff)
PG_RETURN_POINTER(retval);
}
res = (GISTTYPE *) palloc(CALCGTSIZE(ALLISTRUE));
SET_VARSIZE(res, CALCGTSIZE(ALLISTRUE));
res->flag = ALLISTRUE;
retval = (GISTENTRY *) palloc(sizeof(GISTENTRY));
gistentryinit(*retval, PointerGetDatum(res),
entry->rel, entry->page,
entry->offset,
FALSE);
}
/*
* gistgettuple() -- Get the next tuple in the scan
*/
Datum
gistgettuple(PG_FUNCTION_ARGS)
{
IndexScanDesc scan = (IndexScanDesc) PG_GETARG_POINTER(0);
ScanDirection dir = (ScanDirection) PG_GETARG_INT32(1);
GISTScanOpaque so = (GISTScanOpaque) scan->opaque;
if (dir != ForwardScanDirection)
elog(ERROR, "GiST only supports forward scan direction");
if (!so->qual_ok)
PG_RETURN_BOOL(false);
if (so->firstCall)
{
/* Begin the scan by processing the root page */
GISTSearchItem fakeItem;
pgstat_count_index_scan(scan->indexRelation);
so->firstCall = false;
so->curTreeItem = NULL;
so->curPageData = so->nPageData = 0;
fakeItem.blkno = GIST_ROOT_BLKNO;
memset(&fakeItem.data.parentlsn, 0, sizeof(GistNSN));
gistScanPage(scan, &fakeItem, NULL, NULL, NULL);
}
if (scan->numberOfOrderBys > 0)
{
/* Must fetch tuples in strict distance order */
PG_RETURN_BOOL(getNextNearest(scan));
}
else
{
/* Fetch tuples index-page-at-a-time */
for (;;)
{
if (so->curPageData < so->nPageData)
{
/* continuing to return tuples from a leaf page */
scan->xs_ctup.t_self = so->pageData[so->curPageData].heapPtr;
scan->xs_recheck = so->pageData[so->curPageData].recheck;
so->curPageData++;
PG_RETURN_BOOL(true);
}
/* find and process the next index page */
do
{
GISTSearchItem *item = getNextGISTSearchItem(so);
if (!item)
PG_RETURN_BOOL(false);
CHECK_FOR_INTERRUPTS();
/*
* While scanning a leaf page, ItemPointers of matching heap
* tuples are stored in so->pageData. If there are any on
* this page, we fall out of the inner "do" and loop around to
* return them.
*/
gistScanPage(scan, item, so->curTreeItem->distances, NULL, NULL);
pfree(item);
} while (so->nPageData == 0);
}
}
PG_RETURN_BOOL(false); /* keep compiler quiet */
}
Datum
spheretrans_out(PG_FUNCTION_ARGS)
{
SEuler *se = (SEuler *) PG_GETARG_POINTER(0);
char *buffer = (char *) palloc(255);
char buf[100];
char etype[4];
SPoint val[3];
unsigned char i,
t = 0;
unsigned int rdeg,
rmin;
double rsec;
val[0].lat = val[1].lat = val[2].lat = 0.0;
val[0].lng = se->phi;
val[1].lng = se->theta;
val[2].lng = se->psi;
spoint_check(&val[0]);
spoint_check(&val[1]);
spoint_check(&val[2]);
buffer[0] = '\0';
for (i = 0; i < 3; i++)
{
rdeg = rmin = 0;
rsec = 0.0;
switch (sphere_output)
{
case OUTPUT_DEG:
sprintf(&buf[0],
"%.*gd",
sphere_output_precision, RADIANS * val[i].lng);
break;
case OUTPUT_HMS:
case OUTPUT_DMS:
rad_to_dms(val[i].lng, &rdeg, &rmin, &rsec);
sprintf(&buf[0],
"%2ud %2um %.*gs",
rdeg, rmin, sphere_output_precision, rsec);
break;
default:
sprintf(&buf[0], "%.*g", sphere_output_precision, val[i].lng);
break;
}
strcat(&buf[0], ", ");
strcat(buffer, &buf[0]);
}
for (i = 0; i < 3; i++)
{
switch (i)
{
case 0:
t = se->phi_a;
break;
case 1:
t = se->theta_a;
break;
case 2:
t = se->psi_a;
break;
}
switch (t)
{
case EULER_AXIS_X:
etype[i] = 'X';
break;
case EULER_AXIS_Y:
etype[i] = 'Y';
break;
case EULER_AXIS_Z:
etype[i] = 'Z';
break;
}
}
etype[3] = '\0';
strcat(buffer, etype);
PG_RETURN_CSTRING(buffer);
}
Datum
g_intbig_compress(PG_FUNCTION_ARGS)
{
GISTENTRY *entry = (GISTENTRY *) PG_GETARG_POINTER(0);
if (entry->leafkey)
{
GISTENTRY *retval;
ArrayType *in = DatumGetArrayTypeP(entry->key);
int4 *ptr;
int num;
GISTTYPE *res = (GISTTYPE *) palloc0(CALCGTSIZE(0));
CHECKARRVALID(in);
if (ARRISEMPTY(in))
{
ptr = NULL;
num = 0;
}
else
{
ptr = ARRPTR(in);
num = ARRNELEMS(in);
}
SET_VARSIZE(res, CALCGTSIZE(0));
while (num--)
{
HASH(GETSIGN(res), *ptr);
ptr++;
}
retval = (GISTENTRY *) palloc(sizeof(GISTENTRY));
gistentryinit(*retval, PointerGetDatum(res),
entry->rel, entry->page,
entry->offset, FALSE);
if (in != DatumGetArrayTypeP(entry->key))
pfree(in);
PG_RETURN_POINTER(retval);
}
else if (!ISALLTRUE(DatumGetPointer(entry->key)))
{
GISTENTRY *retval;
int i;
BITVECP sign = GETSIGN(DatumGetPointer(entry->key));
GISTTYPE *res;
LOOPBYTE
{
if ((sign[i] & 0xff) != 0xff)
PG_RETURN_POINTER(entry);
}
res = (GISTTYPE *) palloc(CALCGTSIZE(ALLISTRUE));
SET_VARSIZE(res, CALCGTSIZE(ALLISTRUE));
res->flag = ALLISTRUE;
retval = (GISTENTRY *) palloc(sizeof(GISTENTRY));
gistentryinit(*retval, PointerGetDatum(res),
entry->rel, entry->page,
entry->offset, FALSE);
PG_RETURN_POINTER(retval);
}
/*
* extract_variadic_args
*
* Extract a set of argument values, types and NULL markers for a given
* input function which makes use of a VARIADIC input whose argument list
* depends on the caller context. When doing a VARIADIC call, the caller
* has provided one argument made of an array of values, so deconstruct the
* array data before using it for the next processing. If no VARIADIC call
* is used, just fill in the status data based on all the arguments given
* by the caller.
*
* This function returns the number of arguments generated, or -1 in the
* case of "VARIADIC NULL".
*/
int
extract_variadic_args(FunctionCallInfo fcinfo, int variadic_start,
bool convert_unknown, Datum **args, Oid **types,
bool **nulls)
{
bool variadic = get_fn_expr_variadic(fcinfo->flinfo);
Datum *args_res;
bool *nulls_res;
Oid *types_res;
int nargs,
i;
*args = NULL;
*types = NULL;
*nulls = NULL;
if (variadic)
{
ArrayType *array_in;
Oid element_type;
bool typbyval;
char typalign;
int16 typlen;
Assert(PG_NARGS() == variadic_start + 1);
if (PG_ARGISNULL(variadic_start))
return -1;
array_in = PG_GETARG_ARRAYTYPE_P(variadic_start);
element_type = ARR_ELEMTYPE(array_in);
get_typlenbyvalalign(element_type,
&typlen, &typbyval, &typalign);
deconstruct_array(array_in, element_type, typlen, typbyval,
typalign, &args_res, &nulls_res,
&nargs);
/* All the elements of the array have the same type */
types_res = (Oid *) palloc0(nargs * sizeof(Oid));
for (i = 0; i < nargs; i++)
types_res[i] = element_type;
}
else
{
nargs = PG_NARGS() - variadic_start;
Assert(nargs > 0);
nulls_res = (bool *) palloc0(nargs * sizeof(bool));
args_res = (Datum *) palloc0(nargs * sizeof(Datum));
types_res = (Oid *) palloc0(nargs * sizeof(Oid));
for (i = 0; i < nargs; i++)
{
nulls_res[i] = PG_ARGISNULL(i + variadic_start);
types_res[i] = get_fn_expr_argtype(fcinfo->flinfo,
i + variadic_start);
/*
* Turn a constant (more or less literal) value that's of unknown
* type into text if required. Unknowns come in as a cstring
* pointer. Note: for functions declared as taking type "any", the
* parser will not do any type conversion on unknown-type literals
* (that is, undecorated strings or NULLs).
*/
if (convert_unknown &&
types_res[i] == UNKNOWNOID &&
get_fn_expr_arg_stable(fcinfo->flinfo, i + variadic_start))
{
types_res[i] = TEXTOID;
if (PG_ARGISNULL(i + variadic_start))
args_res[i] = (Datum) 0;
else
args_res[i] =
CStringGetTextDatum(PG_GETARG_POINTER(i + variadic_start));
}
else
{
/* no conversion needed, just take the datum as given */
args_res[i] = PG_GETARG_DATUM(i + variadic_start);
}
if (!OidIsValid(types_res[i]) ||
(convert_unknown && types_res[i] == UNKNOWNOID))
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("could not determine data type for argument %d",
i + 1)));
}
}
/* Fill in results */
*args = args_res;
*nulls = nulls_res;
*types = types_res;
return nargs;
}
/*
* jsonb_object_agg aggregate function
*/
Datum
jsonb_object_agg_transfn(PG_FUNCTION_ARGS)
{
MemoryContext oldcontext,
aggcontext;
JsonbInState elem;
JsonbAggState *state;
Datum val;
JsonbInState *result;
bool single_scalar;
JsonbIterator *it;
Jsonb *jbkey,
*jbval;
JsonbValue v;
JsonbIteratorToken type;
if (!AggCheckCallContext(fcinfo, &aggcontext))
{
/* cannot be called directly because of internal-type argument */
elog(ERROR, "jsonb_object_agg_transfn called in non-aggregate context");
}
/* set up the accumulator on the first go round */
if (PG_ARGISNULL(0))
{
Oid arg_type;
oldcontext = MemoryContextSwitchTo(aggcontext);
state = palloc(sizeof(JsonbAggState));
result = palloc0(sizeof(JsonbInState));
state->res = result;
result->res = pushJsonbValue(&result->parseState,
WJB_BEGIN_OBJECT, NULL);
MemoryContextSwitchTo(oldcontext);
arg_type = get_fn_expr_argtype(fcinfo->flinfo, 1);
if (arg_type == InvalidOid)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("could not determine input data type")));
jsonb_categorize_type(arg_type, &state->key_category,
&state->key_output_func);
arg_type = get_fn_expr_argtype(fcinfo->flinfo, 2);
if (arg_type == InvalidOid)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("could not determine input data type")));
jsonb_categorize_type(arg_type, &state->val_category,
&state->val_output_func);
}
else
{
state = (JsonbAggState *) PG_GETARG_POINTER(0);
result = state->res;
}
/* turn the argument into jsonb in the normal function context */
if (PG_ARGISNULL(1))
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("field name must not be null")));
val = PG_GETARG_DATUM(1);
memset(&elem, 0, sizeof(JsonbInState));
datum_to_jsonb(val, false, &elem, state->key_category,
state->key_output_func, true);
jbkey = JsonbValueToJsonb(elem.res);
val = PG_ARGISNULL(2) ? (Datum) 0 : PG_GETARG_DATUM(2);
memset(&elem, 0, sizeof(JsonbInState));
datum_to_jsonb(val, PG_ARGISNULL(2), &elem, state->val_category,
state->val_output_func, false);
jbval = JsonbValueToJsonb(elem.res);
it = JsonbIteratorInit(&jbkey->root);
/* switch to the aggregate context for accumulation operations */
oldcontext = MemoryContextSwitchTo(aggcontext);
/*
* keys should be scalar, and we should have already checked for that
* above when calling datum_to_jsonb, so we only need to look for these
* things.
*/
while ((type = JsonbIteratorNext(&it, &v, false)) != WJB_DONE)
{
switch (type)
{
case WJB_BEGIN_ARRAY:
if (!v.val.array.rawScalar)
elog(ERROR, "unexpected structure for key");
break;
case WJB_ELEM:
if (v.type == jbvString)
{
/* copy string values in the aggregate context */
char *buf = palloc(v.val.string.len + 1);
snprintf(buf, v.val.string.len + 1, "%s", v.val.string.val);
v.val.string.val = buf;
}
else
{
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("object keys must be strings")));
}
result->res = pushJsonbValue(&result->parseState,
WJB_KEY, &v);
break;
case WJB_END_ARRAY:
break;
default:
elog(ERROR, "unexpected structure for key");
break;
}
}
it = JsonbIteratorInit(&jbval->root);
single_scalar = false;
/*
* values can be anything, including structured and null, so we treate
* them as in json_agg_transfn, except that single scalars are always
* pushed as WJB_VALUE items.
*/
while ((type = JsonbIteratorNext(&it, &v, false)) != WJB_DONE)
{
switch (type)
{
case WJB_BEGIN_ARRAY:
if (v.val.array.rawScalar)
single_scalar = true;
else
result->res = pushJsonbValue(&result->parseState,
type, NULL);
break;
case WJB_END_ARRAY:
if (!single_scalar)
result->res = pushJsonbValue(&result->parseState,
type, NULL);
break;
case WJB_BEGIN_OBJECT:
case WJB_END_OBJECT:
result->res = pushJsonbValue(&result->parseState,
type, NULL);
break;
case WJB_ELEM:
case WJB_KEY:
case WJB_VALUE:
if (v.type == jbvString)
{
/* copy string values in the aggregate context */
char *buf = palloc(v.val.string.len + 1);
snprintf(buf, v.val.string.len + 1, "%s", v.val.string.val);
v.val.string.val = buf;
}
else if (v.type == jbvNumeric)
{
/* same for numeric */
v.val.numeric =
DatumGetNumeric(DirectFunctionCall1(numeric_uplus,
NumericGetDatum(v.val.numeric)));
}
result->res = pushJsonbValue(&result->parseState,
single_scalar ? WJB_VALUE : type,
&v);
break;
default:
elog(ERROR, "unknown jsonb iterator token type");
}
}
MemoryContextSwitchTo(oldcontext);
PG_RETURN_POINTER(state);
}
Datum
tsa_rewrite_accum(PG_FUNCTION_ARGS)
{
TSQuery acc;
ArrayType *qa;
TSQuery q;
QTNode *qex = NULL,
*subs = NULL,
*acctree = NULL;
bool isfind = false;
Datum *elemsp;
int nelemsp;
MemoryContext aggcontext;
MemoryContext oldcontext;
if (!AggCheckCallContext(fcinfo, &aggcontext))
elog(ERROR, "tsa_rewrite_accum called in non-aggregate context");
if (PG_ARGISNULL(0) || PG_GETARG_POINTER(0) == NULL)
{
acc = (TSQuery) MemoryContextAlloc(aggcontext, HDRSIZETQ);
SET_VARSIZE(acc, HDRSIZETQ);
acc->size = 0;
}
else
acc = PG_GETARG_TSQUERY(0);
if (PG_ARGISNULL(1) || PG_GETARG_POINTER(1) == NULL)
PG_RETURN_TSQUERY(acc);
else
qa = PG_GETARG_ARRAYTYPE_P_COPY(1);
if (ARR_NDIM(qa) != 1)
elog(ERROR, "array must be one-dimensional, not %d dimensions",
ARR_NDIM(qa));
if (ArrayGetNItems(ARR_NDIM(qa), ARR_DIMS(qa)) != 3)
elog(ERROR, "array must have three elements");
if (ARR_ELEMTYPE(qa) != TSQUERYOID)
elog(ERROR, "array must contain tsquery elements");
deconstruct_array(qa, TSQUERYOID, -1, false, 'i', &elemsp, NULL, &nelemsp);
q = DatumGetTSQuery(elemsp[0]);
if (q->size == 0)
{
pfree(elemsp);
PG_RETURN_POINTER(acc);
}
if (!acc->size)
{
if (VARSIZE(acc) > HDRSIZETQ)
{
pfree(elemsp);
PG_RETURN_POINTER(acc);
}
else
acctree = QT2QTN(GETQUERY(q), GETOPERAND(q));
}
else
acctree = QT2QTN(GETQUERY(acc), GETOPERAND(acc));
QTNTernary(acctree);
QTNSort(acctree);
q = DatumGetTSQuery(elemsp[1]);
if (q->size == 0)
{
pfree(elemsp);
PG_RETURN_POINTER(acc);
}
qex = QT2QTN(GETQUERY(q), GETOPERAND(q));
QTNTernary(qex);
QTNSort(qex);
q = DatumGetTSQuery(elemsp[2]);
if (q->size)
subs = QT2QTN(GETQUERY(q), GETOPERAND(q));
acctree = findsubquery(acctree, qex, subs, &isfind);
if (isfind || !acc->size)
{
/* pfree( acc ); do not pfree(p), because nodeAgg.c will */
if (acctree)
{
QTNBinary(acctree);
oldcontext = MemoryContextSwitchTo(aggcontext);
acc = QTN2QT(acctree);
MemoryContextSwitchTo(oldcontext);
}
else
{
acc = (TSQuery) MemoryContextAlloc(aggcontext, HDRSIZETQ);
SET_VARSIZE(acc, HDRSIZETQ);
acc->size = 0;
}
}
pfree(elemsp);
QTNFree(qex);
QTNFree(subs);
QTNFree(acctree);
PG_RETURN_TSQUERY(acc);
}
/*
* jsonb_agg aggregate function
*/
Datum
jsonb_agg_transfn(PG_FUNCTION_ARGS)
{
MemoryContext oldcontext,
aggcontext;
JsonbAggState *state;
JsonbInState elem;
Datum val;
JsonbInState *result;
bool single_scalar = false;
JsonbIterator *it;
Jsonb *jbelem;
JsonbValue v;
JsonbIteratorToken type;
if (!AggCheckCallContext(fcinfo, &aggcontext))
{
/* cannot be called directly because of internal-type argument */
elog(ERROR, "jsonb_agg_transfn called in non-aggregate context");
}
/* set up the accumulator on the first go round */
if (PG_ARGISNULL(0))
{
Oid arg_type = get_fn_expr_argtype(fcinfo->flinfo, 1);
if (arg_type == InvalidOid)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("could not determine input data type")));
oldcontext = MemoryContextSwitchTo(aggcontext);
state = palloc(sizeof(JsonbAggState));
result = palloc0(sizeof(JsonbInState));
state->res = result;
result->res = pushJsonbValue(&result->parseState,
WJB_BEGIN_ARRAY, NULL);
MemoryContextSwitchTo(oldcontext);
jsonb_categorize_type(arg_type, &state->val_category,
&state->val_output_func);
}
else
{
state = (JsonbAggState *) PG_GETARG_POINTER(0);
result = state->res;
}
/* turn the argument into jsonb in the normal function context */
val = PG_ARGISNULL(1) ? (Datum) 0 : PG_GETARG_DATUM(1);
memset(&elem, 0, sizeof(JsonbInState));
datum_to_jsonb(val, PG_ARGISNULL(1), &elem, state->val_category,
state->val_output_func, false);
jbelem = JsonbValueToJsonb(elem.res);
/* switch to the aggregate context for accumulation operations */
oldcontext = MemoryContextSwitchTo(aggcontext);
it = JsonbIteratorInit(&jbelem->root);
while ((type = JsonbIteratorNext(&it, &v, false)) != WJB_DONE)
{
switch (type)
{
case WJB_BEGIN_ARRAY:
if (v.val.array.rawScalar)
single_scalar = true;
else
result->res = pushJsonbValue(&result->parseState,
type, NULL);
break;
case WJB_END_ARRAY:
if (!single_scalar)
result->res = pushJsonbValue(&result->parseState,
type, NULL);
break;
case WJB_BEGIN_OBJECT:
case WJB_END_OBJECT:
result->res = pushJsonbValue(&result->parseState,
type, NULL);
break;
case WJB_ELEM:
case WJB_KEY:
case WJB_VALUE:
if (v.type == jbvString)
{
/* copy string values in the aggregate context */
char *buf = palloc(v.val.string.len + 1);
snprintf(buf, v.val.string.len + 1, "%s", v.val.string.val);
v.val.string.val = buf;
}
else if (v.type == jbvNumeric)
{
/* same for numeric */
v.val.numeric =
DatumGetNumeric(DirectFunctionCall1(numeric_uplus,
NumericGetDatum(v.val.numeric)));
}
result->res = pushJsonbValue(&result->parseState,
type, &v);
break;
default:
elog(ERROR, "unknown jsonb iterator token type");
}
}
MemoryContextSwitchTo(oldcontext);
PG_RETURN_POINTER(state);
}
/*
* SQL function jsonb_build_object(variadic "any")
*/
Datum
jsonb_build_object(PG_FUNCTION_ARGS)
{
int nargs = PG_NARGS();
int i;
Datum arg;
Oid val_type;
JsonbInState result;
if (nargs % 2 != 0)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("invalid number or arguments: object must be matched key value pairs")));
memset(&result, 0, sizeof(JsonbInState));
result.res = pushJsonbValue(&result.parseState, WJB_BEGIN_OBJECT, NULL);
for (i = 0; i < nargs; i += 2)
{
/* process key */
if (PG_ARGISNULL(i))
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("arg %d: key cannot be null", i + 1)));
val_type = get_fn_expr_argtype(fcinfo->flinfo, i);
/*
* turn a constant (more or less literal) value that's of unknown type
* into text. Unknowns come in as a cstring pointer.
*/
if (val_type == UNKNOWNOID && get_fn_expr_arg_stable(fcinfo->flinfo, i))
{
val_type = TEXTOID;
if (PG_ARGISNULL(i))
arg = (Datum) 0;
else
arg = CStringGetTextDatum(PG_GETARG_POINTER(i));
}
else
{
arg = PG_GETARG_DATUM(i);
}
if (val_type == InvalidOid || val_type == UNKNOWNOID)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("arg %d: could not determine data type", i + 1)));
add_jsonb(arg, false, &result, val_type, true);
/* process value */
val_type = get_fn_expr_argtype(fcinfo->flinfo, i + 1);
/* see comments above */
if (val_type == UNKNOWNOID && get_fn_expr_arg_stable(fcinfo->flinfo, i + 1))
{
val_type = TEXTOID;
if (PG_ARGISNULL(i + 1))
arg = (Datum) 0;
else
arg = CStringGetTextDatum(PG_GETARG_POINTER(i + 1));
}
else
{
arg = PG_GETARG_DATUM(i + 1);
}
if (val_type == InvalidOid || val_type == UNKNOWNOID)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("arg %d: could not determine data type", i + 2)));
add_jsonb(arg, PG_ARGISNULL(i + 1), &result, val_type, false);
}
result.res = pushJsonbValue(&result.parseState, WJB_END_OBJECT, NULL);
PG_RETURN_POINTER(JsonbValueToJsonb(result.res));
}
/*
* tsmatchsel -- Selectivity of "@@"
*
* restriction selectivity function for tsvector @@ tsquery and
* tsquery @@ tsvector
*/
Datum
tsmatchsel(PG_FUNCTION_ARGS)
{
PlannerInfo *root = (PlannerInfo *) PG_GETARG_POINTER(0);
#ifdef NOT_USED
Oid operator = PG_GETARG_OID(1);
#endif
List *args = (List *) PG_GETARG_POINTER(2);
int varRelid = PG_GETARG_INT32(3);
VariableStatData vardata;
Node *other;
bool varonleft;
Selectivity selec;
/*
* If expression is not variable = something or something = variable, then
* punt and return a default estimate.
*/
if (!get_restriction_variable(root, args, varRelid,
&vardata, &other, &varonleft))
PG_RETURN_FLOAT8(DEFAULT_TS_MATCH_SEL);
/*
* Can't do anything useful if the something is not a constant, either.
*/
if (!IsA(other, Const))
{
ReleaseVariableStats(vardata);
PG_RETURN_FLOAT8(DEFAULT_TS_MATCH_SEL);
}
/*
* The "@@" operator is strict, so we can cope with NULL right away
*/
if (((Const *) other)->constisnull)
{
ReleaseVariableStats(vardata);
PG_RETURN_FLOAT8(0.0);
}
/*
* OK, there's a Var and a Const we're dealing with here. We need the
* Const to be a TSQuery, else we can't do anything useful. We have to
* check this because the Var might be the TSQuery not the TSVector.
*/
if (((Const *) other)->consttype == TSQUERYOID)
{
/* tsvector @@ tsquery or the other way around */
Assert(vardata.vartype == TSVECTOROID);
selec = tsquerysel(&vardata, ((Const *) other)->constvalue);
}
else
{
/* If we can't see the query structure, must punt */
selec = DEFAULT_TS_MATCH_SEL;
}
ReleaseVariableStats(vardata);
CLAMP_PROBABILITY(selec);
PG_RETURN_FLOAT8((float8) selec);
}
Datum
gin_consistent_jsonb(PG_FUNCTION_ARGS)
{
bool *check = (bool *) PG_GETARG_POINTER(0);
StrategyNumber strategy = PG_GETARG_UINT16(1);
/* Jsonb *query = PG_GETARG_JSONB(2); */
int32 nkeys = PG_GETARG_INT32(3);
/* Pointer *extra_data = (Pointer *) PG_GETARG_POINTER(4); */
bool *recheck = (bool *) PG_GETARG_POINTER(5);
bool res = true;
int32 i;
if (strategy == JsonbContainsStrategyNumber)
{
/*
* We must always recheck, since we can't tell from the index whether
* the positions of the matched items match the structure of the query
* object. (Even if we could, we'd also have to worry about hashed
* keys and the index's failure to distinguish keys from string array
* elements.) However, the tuple certainly doesn't match unless it
* contains all the query keys.
*/
*recheck = true;
for (i = 0; i < nkeys; i++)
{
if (!check[i])
{
res = false;
break;
}
}
}
else if (strategy == JsonbExistsStrategyNumber)
{
/*
* Although the key is certainly present in the index, we must recheck
* because (1) the key might be hashed, and (2) the index match might
* be for a key that's not at top level of the JSON object. For (1),
* we could look at the query key to see if it's hashed and not
* recheck if not, but the index lacks enough info to tell about (2).
*/
*recheck = true;
res = true;
}
else if (strategy == JsonbExistsAnyStrategyNumber)
{
/* As for plain exists, we must recheck */
*recheck = true;
res = true;
}
else if (strategy == JsonbExistsAllStrategyNumber)
{
/* As for plain exists, we must recheck */
*recheck = true;
/* ... but unless all the keys are present, we can say "false" */
for (i = 0; i < nkeys; i++)
{
if (!check[i])
{
res = false;
break;
}
}
}
else
elog(ERROR, "unrecognized strategy number: %d", strategy);
PG_RETURN_BOOL(res);
}
Datum
spg_text_choose(PG_FUNCTION_ARGS)
{
spgChooseIn *in = (spgChooseIn *) PG_GETARG_POINTER(0);
spgChooseOut *out = (spgChooseOut *) PG_GETARG_POINTER(1);
text *inText = DatumGetTextPP(in->datum);
char *inStr = VARDATA_ANY(inText);
int inSize = VARSIZE_ANY_EXHDR(inText);
uint8 nodeChar = '\0';
int i = 0;
int commonLen = 0;
/* Check for prefix match, set nodeChar to first byte after prefix */
if (in->hasPrefix)
{
text *prefixText = DatumGetTextPP(in->prefixDatum);
char *prefixStr = VARDATA_ANY(prefixText);
int prefixSize = VARSIZE_ANY_EXHDR(prefixText);
commonLen = commonPrefix(inStr + in->level,
prefixStr,
inSize - in->level,
prefixSize);
if (commonLen == prefixSize)
{
if (inSize - in->level > commonLen)
nodeChar = *(uint8 *) (inStr + in->level + commonLen);
else
nodeChar = '\0';
}
else
{
/* Must split tuple because incoming value doesn't match prefix */
out->resultType = spgSplitTuple;
if (commonLen == 0)
{
out->result.splitTuple.prefixHasPrefix = false;
}
else
{
out->result.splitTuple.prefixHasPrefix = true;
out->result.splitTuple.prefixPrefixDatum =
formTextDatum(prefixStr, commonLen);
}
out->result.splitTuple.nodeLabel =
UInt8GetDatum(*(prefixStr + commonLen));
if (prefixSize - commonLen == 1)
{
out->result.splitTuple.postfixHasPrefix = false;
}
else
{
out->result.splitTuple.postfixHasPrefix = true;
out->result.splitTuple.postfixPrefixDatum =
formTextDatum(prefixStr + commonLen + 1,
prefixSize - commonLen - 1);
}
PG_RETURN_VOID();
}
}
else if (inSize > in->level)
{
nodeChar = *(uint8 *) (inStr + in->level);
}
else
{
nodeChar = '\0';
}
/* Look up nodeChar in the node label array */
if (searchChar(in->nodeLabels, in->nNodes, nodeChar, &i))
{
/*
* Descend to existing node. (If in->allTheSame, the core code will
* ignore our nodeN specification here, but that's OK. We still have
* to provide the correct levelAdd and restDatum values, and those are
* the same regardless of which node gets chosen by core.)
*/
out->resultType = spgMatchNode;
out->result.matchNode.nodeN = i;
out->result.matchNode.levelAdd = commonLen + 1;
if (inSize - in->level - commonLen - 1 > 0)
out->result.matchNode.restDatum =
formTextDatum(inStr + in->level + commonLen + 1,
inSize - in->level - commonLen - 1);
else
out->result.matchNode.restDatum =
formTextDatum(NULL, 0);
}
else if (in->allTheSame)
{
/*
* Can't use AddNode action, so split the tuple. The upper tuple has
* the same prefix as before and uses an empty node label for the
* lower tuple. The lower tuple has no prefix and the same node
* labels as the original tuple.
*/
out->resultType = spgSplitTuple;
out->result.splitTuple.prefixHasPrefix = in->hasPrefix;
out->result.splitTuple.prefixPrefixDatum = in->prefixDatum;
out->result.splitTuple.nodeLabel = UInt8GetDatum('\0');
out->result.splitTuple.postfixHasPrefix = false;
}
else
{
/* Add a node for the not-previously-seen nodeChar value */
out->resultType = spgAddNode;
out->result.addNode.nodeLabel = UInt8GetDatum(nodeChar);
out->result.addNode.nodeN = i;
}
PG_RETURN_VOID();
}
Datum
gin_extract_jsonb_hash(PG_FUNCTION_ARGS)
{
Jsonb *jb = PG_GETARG_JSONB(0);
int32 *nentries = (int32 *) PG_GETARG_POINTER(1);
int total = 2 * JB_ROOT_COUNT(jb);
JsonbIterator *it;
JsonbValue v;
PathHashStack tail;
PathHashStack *stack;
int i = 0,
r;
Datum *entries;
/* If the root level is empty, we certainly have no keys */
if (total == 0)
{
*nentries = 0;
PG_RETURN_POINTER(NULL);
}
/* Otherwise, use 2 * root count as initial estimate of result size */
entries = (Datum *) palloc(sizeof(Datum) * total);
/* We keep a stack of hashes corresponding to parent key levels */
tail.parent = NULL;
tail.hash = 0;
stack = &tail;
it = JsonbIteratorInit(&jb->root);
while ((r = JsonbIteratorNext(&it, &v, false)) != WJB_DONE)
{
PathHashStack *parent;
/* Since we recurse into the object, we might need more space */
if (i >= total)
{
total *= 2;
entries = (Datum *) repalloc(entries, sizeof(Datum) * total);
}
switch (r)
{
case WJB_BEGIN_ARRAY:
case WJB_BEGIN_OBJECT:
/* Push a stack level for this object */
parent = stack;
stack = (PathHashStack *) palloc(sizeof(PathHashStack));
if (parent->parent)
{
/*
* We pass forward hashes from previous container nesting
* levels so that nested arrays with an outermost nested
* object will have element hashes mixed with the
* outermost key. It's also somewhat useful to have
* nested objects' innermost values have hashes that are a
* function of not just their own key, but outer keys too.
*
* Nesting an array within another array will not alter
* innermost scalar element hash values, but that seems
* inconsequential.
*/
stack->hash = parent->hash;
}
else
{
/*
* At the outermost level, initialize hash with container
* type proxy value. Note that this makes JB_FARRAY and
* JB_FOBJECT part of the on-disk representation, but they
* are that in the base jsonb object storage already.
*/
stack->hash = (r == WJB_BEGIN_ARRAY) ? JB_FARRAY : JB_FOBJECT;
}
stack->parent = parent;
break;
case WJB_KEY:
/* initialize hash from parent */
stack->hash = stack->parent->hash;
/* and mix in this key */
JsonbHashScalarValue(&v, &stack->hash);
/* hash is now ready to incorporate the value */
break;
case WJB_ELEM:
/* array elements use parent hash mixed with element's hash */
stack->hash = stack->parent->hash;
/* FALL THRU */
case WJB_VALUE:
/* mix the element or value's hash into the prepared hash */
JsonbHashScalarValue(&v, &stack->hash);
/* and emit an index entry */
entries[i++] = UInt32GetDatum(stack->hash);
/* Note: we assume we'll see KEY before another VALUE */
break;
case WJB_END_ARRAY:
case WJB_END_OBJECT:
/* Pop the stack */
parent = stack->parent;
pfree(stack);
stack = parent;
break;
default:
elog(ERROR, "invalid JsonbIteratorNext rc: %d", r);
}
}
*nentries = i;
PG_RETURN_POINTER(entries);
}
Datum
spg_text_picksplit(PG_FUNCTION_ARGS)
{
spgPickSplitIn *in = (spgPickSplitIn *) PG_GETARG_POINTER(0);
spgPickSplitOut *out = (spgPickSplitOut *) PG_GETARG_POINTER(1);
text *text0 = DatumGetTextPP(in->datums[0]);
int i,
commonLen;
spgNodePtr *nodes;
/* Identify longest common prefix, if any */
commonLen = VARSIZE_ANY_EXHDR(text0);
for (i = 1; i < in->nTuples && commonLen > 0; i++)
{
text *texti = DatumGetTextPP(in->datums[i]);
int tmp = commonPrefix(VARDATA_ANY(text0),
VARDATA_ANY(texti),
VARSIZE_ANY_EXHDR(text0),
VARSIZE_ANY_EXHDR(texti));
if (tmp < commonLen)
commonLen = tmp;
}
/*
* Limit the prefix length, if necessary, to ensure that the resulting
* inner tuple will fit on a page.
*/
commonLen = Min(commonLen, SPGIST_MAX_PREFIX_LENGTH);
/* Set node prefix to be that string, if it's not empty */
if (commonLen == 0)
{
out->hasPrefix = false;
}
else
{
out->hasPrefix = true;
out->prefixDatum = formTextDatum(VARDATA_ANY(text0), commonLen);
}
/* Extract the node label (first non-common byte) from each value */
nodes = (spgNodePtr *) palloc(sizeof(spgNodePtr) * in->nTuples);
for (i = 0; i < in->nTuples; i++)
{
text *texti = DatumGetTextPP(in->datums[i]);
if (commonLen < VARSIZE_ANY_EXHDR(texti))
nodes[i].c = *(uint8 *) (VARDATA_ANY(texti) + commonLen);
else
nodes[i].c = '\0'; /* use \0 if string is all common */
nodes[i].i = i;
nodes[i].d = in->datums[i];
}
/*
* Sort by label bytes so that we can group the values into nodes. This
* also ensures that the nodes are ordered by label value, allowing the
* use of binary search in searchChar.
*/
qsort(nodes, in->nTuples, sizeof(*nodes), cmpNodePtr);
/* And emit results */
out->nNodes = 0;
out->nodeLabels = (Datum *) palloc(sizeof(Datum) * in->nTuples);
out->mapTuplesToNodes = (int *) palloc(sizeof(int) * in->nTuples);
out->leafTupleDatums = (Datum *) palloc(sizeof(Datum) * in->nTuples);
for (i = 0; i < in->nTuples; i++)
{
text *texti = DatumGetTextPP(nodes[i].d);
Datum leafD;
if (i == 0 || nodes[i].c != nodes[i - 1].c)
{
out->nodeLabels[out->nNodes] = UInt8GetDatum(nodes[i].c);
out->nNodes++;
}
if (commonLen < VARSIZE_ANY_EXHDR(texti))
leafD = formTextDatum(VARDATA_ANY(texti) + commonLen + 1,
VARSIZE_ANY_EXHDR(texti) - commonLen - 1);
else
leafD = formTextDatum(NULL, 0);
out->leafTupleDatums[nodes[i].i] = leafD;
out->mapTuplesToNodes[nodes[i].i] = out->nNodes - 1;
}
PG_RETURN_VOID();
}
/*
* GiST DeCompress method for boxes (also used for points, polygons
* and circles)
*
* do not do anything --- we just use the stored box as is.
*/
Datum
gist_box_decompress(PG_FUNCTION_ARGS)
{
PG_RETURN_POINTER(PG_GETARG_POINTER(0));
}
Datum
spg_text_inner_consistent(PG_FUNCTION_ARGS)
{
spgInnerConsistentIn *in = (spgInnerConsistentIn *) PG_GETARG_POINTER(0);
spgInnerConsistentOut *out = (spgInnerConsistentOut *) PG_GETARG_POINTER(1);
bool collate_is_c = lc_collate_is_c(PG_GET_COLLATION());
text *reconstrText = NULL;
int maxReconstrLen = 0;
text *prefixText = NULL;
int prefixSize = 0;
int i;
/*
* Reconstruct values represented at this tuple, including parent data,
* prefix of this tuple if any, and the node label if any. in->level
* should be the length of the previously reconstructed value, and the
* number of bytes added here is prefixSize or prefixSize + 1.
*
* Note: we assume that in->reconstructedValue isn't toasted and doesn't
* have a short varlena header. This is okay because it must have been
* created by a previous invocation of this routine, and we always emit
* long-format reconstructed values.
*/
Assert(in->level == 0 ? DatumGetPointer(in->reconstructedValue) == NULL :
VARSIZE_ANY_EXHDR(DatumGetPointer(in->reconstructedValue)) == in->level);
maxReconstrLen = in->level + 1;
if (in->hasPrefix)
{
prefixText = DatumGetTextPP(in->prefixDatum);
prefixSize = VARSIZE_ANY_EXHDR(prefixText);
maxReconstrLen += prefixSize;
}
reconstrText = palloc(VARHDRSZ + maxReconstrLen);
SET_VARSIZE(reconstrText, VARHDRSZ + maxReconstrLen);
if (in->level)
memcpy(VARDATA(reconstrText),
VARDATA(DatumGetPointer(in->reconstructedValue)),
in->level);
if (prefixSize)
memcpy(((char *) VARDATA(reconstrText)) + in->level,
VARDATA_ANY(prefixText),
prefixSize);
/* last byte of reconstrText will be filled in below */
/*
* Scan the child nodes. For each one, complete the reconstructed value
* and see if it's consistent with the query. If so, emit an entry into
* the output arrays.
*/
out->nodeNumbers = (int *) palloc(sizeof(int) * in->nNodes);
out->levelAdds = (int *) palloc(sizeof(int) * in->nNodes);
out->reconstructedValues = (Datum *) palloc(sizeof(Datum) * in->nNodes);
out->nNodes = 0;
for (i = 0; i < in->nNodes; i++)
{
uint8 nodeChar = DatumGetUInt8(in->nodeLabels[i]);
int thisLen;
bool res = true;
int j;
/* If nodeChar is zero, don't include it in data */
if (nodeChar == '\0')
thisLen = maxReconstrLen - 1;
else
{
((char *) VARDATA(reconstrText))[maxReconstrLen - 1] = nodeChar;
thisLen = maxReconstrLen;
}
for (j = 0; j < in->nkeys; j++)
{
StrategyNumber strategy = in->scankeys[j].sk_strategy;
text *inText;
int inSize;
int r;
/*
* If it's a collation-aware operator, but the collation is C, we
* can treat it as non-collation-aware. With non-C collation we
* need to traverse whole tree :-( so there's no point in making
* any check here.
*/
if (strategy > 10)
{
if (collate_is_c)
strategy -= 10;
else
continue;
}
inText = DatumGetTextPP(in->scankeys[j].sk_argument);
inSize = VARSIZE_ANY_EXHDR(inText);
r = memcmp(VARDATA(reconstrText), VARDATA_ANY(inText),
Min(inSize, thisLen));
switch (strategy)
{
case BTLessStrategyNumber:
case BTLessEqualStrategyNumber:
if (r > 0)
res = false;
break;
case BTEqualStrategyNumber:
if (r != 0 || inSize < thisLen)
res = false;
break;
case BTGreaterEqualStrategyNumber:
case BTGreaterStrategyNumber:
if (r < 0)
res = false;
break;
default:
elog(ERROR, "unrecognized strategy number: %d",
in->scankeys[j].sk_strategy);
break;
}
if (!res)
break; /* no need to consider remaining conditions */
}
if (res)
{
out->nodeNumbers[out->nNodes] = i;
out->levelAdds[out->nNodes] = thisLen - in->level;
SET_VARSIZE(reconstrText, VARHDRSZ + thisLen);
out->reconstructedValues[out->nNodes] =
datumCopy(PointerGetDatum(reconstrText), false, -1);
out->nNodes++;
}
}
PG_RETURN_VOID();
}
/*
* consistent support function
*/
Datum
ginarrayconsistent(PG_FUNCTION_ARGS)
{
bool *check = (bool *) PG_GETARG_POINTER(0);
StrategyNumber strategy = PG_GETARG_UINT16(1);
/* ArrayType *query = PG_GETARG_ARRAYTYPE_P(2); */
int32 nkeys = PG_GETARG_INT32(3);
/* Pointer *extra_data = (Pointer *) PG_GETARG_POINTER(4); */
bool *recheck = (bool *) PG_GETARG_POINTER(5);
/* Datum *queryKeys = (Datum *) PG_GETARG_POINTER(6); */
bool *nullFlags = (bool *) PG_GETARG_POINTER(7);
bool res;
int32 i;
switch (strategy)
{
case GinOverlapStrategy:
/* result is not lossy */
*recheck = false;
/* must have a match for at least one non-null element */
res = false;
for (i = 0; i < nkeys; i++)
{
if (check[i] && !nullFlags[i])
{
res = true;
break;
}
}
break;
case GinContainsStrategy:
/* result is not lossy */
*recheck = false;
/* must have all elements in check[] true, and no nulls */
res = true;
for (i = 0; i < nkeys; i++)
{
if (!check[i] || nullFlags[i])
{
res = false;
break;
}
}
break;
case GinContainedStrategy:
/* we will need recheck */
*recheck = true;
/* can't do anything else useful here */
res = true;
break;
case GinEqualStrategy:
/* we will need recheck */
*recheck = true;
/*
* Must have all elements in check[] true; no discrimination
* against nulls here. This is because array_contain_compare and
* array_eq handle nulls differently ...
*/
res = true;
for (i = 0; i < nkeys; i++)
{
if (!check[i])
{
res = false;
break;
}
}
break;
default:
elog(ERROR, "ginarrayconsistent: unknown strategy number: %d",
strategy);
res = false;
}
PG_RETURN_BOOL(res);
}
Datum
spg_text_leaf_consistent(PG_FUNCTION_ARGS)
{
spgLeafConsistentIn *in = (spgLeafConsistentIn *) PG_GETARG_POINTER(0);
spgLeafConsistentOut *out = (spgLeafConsistentOut *) PG_GETARG_POINTER(1);
int level = in->level;
text *leafValue,
*reconstrValue = NULL;
char *fullValue;
int fullLen;
bool res;
int j;
/* all tests are exact */
out->recheck = false;
leafValue = DatumGetTextPP(in->leafDatum);
if (DatumGetPointer(in->reconstructedValue))
reconstrValue = DatumGetTextP(in->reconstructedValue);
Assert(level == 0 ? reconstrValue == NULL :
VARSIZE_ANY_EXHDR(reconstrValue) == level);
/* Reconstruct the full string represented by this leaf tuple */
fullLen = level + VARSIZE_ANY_EXHDR(leafValue);
if (VARSIZE_ANY_EXHDR(leafValue) == 0 && level > 0)
{
fullValue = VARDATA(reconstrValue);
out->leafValue = PointerGetDatum(reconstrValue);
}
else
{
text *fullText = palloc(VARHDRSZ + fullLen);
SET_VARSIZE(fullText, VARHDRSZ + fullLen);
fullValue = VARDATA(fullText);
if (level)
memcpy(fullValue, VARDATA(reconstrValue), level);
if (VARSIZE_ANY_EXHDR(leafValue) > 0)
memcpy(fullValue + level, VARDATA_ANY(leafValue),
VARSIZE_ANY_EXHDR(leafValue));
out->leafValue = PointerGetDatum(fullText);
}
/* Perform the required comparison(s) */
res = true;
for (j = 0; j < in->nkeys; j++)
{
StrategyNumber strategy = in->scankeys[j].sk_strategy;
text *query = DatumGetTextPP(in->scankeys[j].sk_argument);
int queryLen = VARSIZE_ANY_EXHDR(query);
int r;
if (strategy > 10)
{
/* Collation-aware comparison */
strategy -= 10;
/* If asserts enabled, verify encoding of reconstructed string */
Assert(pg_verifymbstr(fullValue, fullLen, false));
r = varstr_cmp(fullValue, Min(queryLen, fullLen),
VARDATA_ANY(query), Min(queryLen, fullLen),
PG_GET_COLLATION());
}
else
{
/* Non-collation-aware comparison */
r = memcmp(fullValue, VARDATA_ANY(query), Min(queryLen, fullLen));
}
if (r == 0)
{
if (queryLen > fullLen)
r = -1;
else if (queryLen < fullLen)
r = 1;
}
switch (strategy)
{
case BTLessStrategyNumber:
res = (r < 0);
break;
case BTLessEqualStrategyNumber:
res = (r <= 0);
break;
case BTEqualStrategyNumber:
res = (r == 0);
break;
case BTGreaterEqualStrategyNumber:
res = (r >= 0);
break;
case BTGreaterStrategyNumber:
res = (r > 0);
break;
default:
elog(ERROR, "unrecognized strategy number: %d",
in->scankeys[j].sk_strategy);
res = false;
break;
}
if (!res)
break; /* no need to consider remaining conditions */
}
PG_RETURN_BOOL(res);
}
Datum
gtsquery_picksplit(PG_FUNCTION_ARGS)
{
GistEntryVector *entryvec = (GistEntryVector *) PG_GETARG_POINTER(0);
GIST_SPLITVEC *v = (GIST_SPLITVEC *) PG_GETARG_POINTER(1);
OffsetNumber maxoff = entryvec->n - 2;
OffsetNumber k,
j;
TSQuerySign datum_l,
datum_r;
int4 size_alpha,
size_beta;
int4 size_waste,
waste = -1;
int4 nbytes;
OffsetNumber seed_1 = 0,
seed_2 = 0;
OffsetNumber *left,
*right;
SPLITCOST *costvector;
nbytes = (maxoff + 2) * sizeof(OffsetNumber);
left = v->spl_left = (OffsetNumber *) palloc(nbytes);
right = v->spl_right = (OffsetNumber *) palloc(nbytes);
v->spl_nleft = v->spl_nright = 0;
for (k = FirstOffsetNumber; k < maxoff; k = OffsetNumberNext(k))
for (j = OffsetNumberNext(k); j <= maxoff; j = OffsetNumberNext(j))
{
size_waste = hemdist(GETENTRY(entryvec, j), GETENTRY(entryvec, k));
if (size_waste > waste)
{
waste = size_waste;
seed_1 = k;
seed_2 = j;
}
}
if (seed_1 == 0 || seed_2 == 0)
{
seed_1 = 1;
seed_2 = 2;
}
datum_l = GETENTRY(entryvec, seed_1);
datum_r = GETENTRY(entryvec, seed_2);
maxoff = OffsetNumberNext(maxoff);
costvector = (SPLITCOST *) palloc(sizeof(SPLITCOST) * maxoff);
for (j = FirstOffsetNumber; j <= maxoff; j = OffsetNumberNext(j))
{
costvector[j - 1].pos = j;
size_alpha = hemdist(GETENTRY(entryvec, seed_1), GETENTRY(entryvec, j));
size_beta = hemdist(GETENTRY(entryvec, seed_2), GETENTRY(entryvec, j));
costvector[j - 1].cost = abs(size_alpha - size_beta);
}
qsort((void *) costvector, maxoff, sizeof(SPLITCOST), comparecost);
for (k = 0; k < maxoff; k++)
{
j = costvector[k].pos;
if (j == seed_1)
{
*left++ = j;
v->spl_nleft++;
continue;
}
else if (j == seed_2)
{
*right++ = j;
v->spl_nright++;
continue;
}
size_alpha = hemdist(datum_l, GETENTRY(entryvec, j));
size_beta = hemdist(datum_r, GETENTRY(entryvec, j));
if (size_alpha < size_beta + WISH_F(v->spl_nleft, v->spl_nright, 0.05))
{
datum_l |= GETENTRY(entryvec, j);
*left++ = j;
v->spl_nleft++;
}
else
{
datum_r |= GETENTRY(entryvec, j);
*right++ = j;
v->spl_nright++;
}
}
*right = *left = FirstOffsetNumber;
v->spl_ldatum = TSQuerySignGetDatum(datum_l);
v->spl_rdatum = TSQuerySignGetDatum(datum_r);
PG_RETURN_POINTER(v);
}
/*
** The GiST PickSplit method for boxes
** We use Guttman's poly time split algorithm
*/
Datum
g_cube_picksplit(PG_FUNCTION_ARGS)
{
GistEntryVector *entryvec;
GIST_SPLITVEC *v;
OffsetNumber i,
j;
NDBOX *datum_alpha,
*datum_beta;
NDBOX *datum_l,
*datum_r;
NDBOX *union_d,
*union_dl,
*union_dr;
NDBOX *inter_d;
bool firsttime;
double size_alpha,
size_beta,
size_union,
size_inter;
double size_waste,
waste;
double size_l,
size_r;
int nbytes;
OffsetNumber seed_1 = 1,
seed_2 = 2;
OffsetNumber *left,
*right;
OffsetNumber maxoff;
entryvec = (GistEntryVector *) PG_GETARG_POINTER(0);
v = (GIST_SPLITVEC *) PG_GETARG_POINTER(1);
/*
* fprintf(stderr, "picksplit\n");
*/
maxoff = entryvec->n - 2;
nbytes = (maxoff + 2) * sizeof(OffsetNumber);
v->spl_left = (OffsetNumber *) palloc(nbytes);
v->spl_right = (OffsetNumber *) palloc(nbytes);
firsttime = true;
waste = 0.0;
for (i = FirstOffsetNumber; i < maxoff; i = OffsetNumberNext(i))
{
datum_alpha = DatumGetNDBOX(entryvec->vector[i].key);
for (j = OffsetNumberNext(i); j <= maxoff; j = OffsetNumberNext(j))
{
datum_beta = DatumGetNDBOX(entryvec->vector[j].key);
/* compute the wasted space by unioning these guys */
/* size_waste = size_union - size_inter; */
union_d = cube_union_v0(datum_alpha, datum_beta);
rt_cube_size(union_d, &size_union);
inter_d = DatumGetNDBOX(DirectFunctionCall2(cube_inter,
entryvec->vector[i].key, entryvec->vector[j].key));
rt_cube_size(inter_d, &size_inter);
size_waste = size_union - size_inter;
/*
* are these a more promising split than what we've already seen?
*/
if (size_waste > waste || firsttime)
{
waste = size_waste;
seed_1 = i;
seed_2 = j;
firsttime = false;
}
}
}
left = v->spl_left;
v->spl_nleft = 0;
right = v->spl_right;
v->spl_nright = 0;
datum_alpha = DatumGetNDBOX(entryvec->vector[seed_1].key);
datum_l = cube_union_v0(datum_alpha, datum_alpha);
rt_cube_size(datum_l, &size_l);
datum_beta = DatumGetNDBOX(entryvec->vector[seed_2].key);
datum_r = cube_union_v0(datum_beta, datum_beta);
rt_cube_size(datum_r, &size_r);
/*
* Now split up the regions between the two seeds. An important property
* of this split algorithm is that the split vector v has the indices of
* items to be split in order in its left and right vectors. We exploit
* this property by doing a merge in the code that actually splits the
* page.
*
* For efficiency, we also place the new index tuple in this loop. This is
* handled at the very end, when we have placed all the existing tuples
* and i == maxoff + 1.
*/
maxoff = OffsetNumberNext(maxoff);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
/*
* If we've already decided where to place this item, just put it on
* the right list. Otherwise, we need to figure out which page needs
* the least enlargement in order to store the item.
*/
if (i == seed_1)
{
*left++ = i;
v->spl_nleft++;
continue;
}
else if (i == seed_2)
{
*right++ = i;
v->spl_nright++;
continue;
}
/* okay, which page needs least enlargement? */
datum_alpha = DatumGetNDBOX(entryvec->vector[i].key);
union_dl = cube_union_v0(datum_l, datum_alpha);
union_dr = cube_union_v0(datum_r, datum_alpha);
rt_cube_size(union_dl, &size_alpha);
rt_cube_size(union_dr, &size_beta);
/* pick which page to add it to */
if (size_alpha - size_l < size_beta - size_r)
{
datum_l = union_dl;
size_l = size_alpha;
*left++ = i;
v->spl_nleft++;
}
else
{
datum_r = union_dr;
size_r = size_beta;
*right++ = i;
v->spl_nright++;
}
}
*left = *right = FirstOffsetNumber; /* sentinel value, see dosplit() */
v->spl_ldatum = PointerGetDatum(datum_l);
v->spl_rdatum = PointerGetDatum(datum_r);
PG_RETURN_POINTER(v);
}
Datum
gtsvector_compress(PG_FUNCTION_ARGS)
{
GISTENTRY *entry = (GISTENTRY *) PG_GETARG_POINTER(0);
GISTENTRY *retval = entry;
if (entry->leafkey)
{ /* tsvector */
GISTTYPE *res;
tsvector *val = (tsvector *) PG_DETOAST_DATUM(entry->key);
int4 len;
int4 *arr;
WordEntry *ptr = ARRPTR(val);
char *words = STRPTR(val);
len = CALCGTSIZE(ARRKEY, val->size);
res = (GISTTYPE *) palloc(len);
SET_VARSIZE(res, len);
res->flag = ARRKEY;
arr = GETARR(res);
len = val->size;
while (len--)
{
*arr = crc32_sz(&words[ptr->pos], ptr->len);
arr++;
ptr++;
}
len = uniqueint(GETARR(res), val->size);
if (len != val->size)
{
/*
* there is a collision of hash-function; len is always less than
* val->size
*/
len = CALCGTSIZE(ARRKEY, len);
res = (GISTTYPE *) repalloc((void *) res, len);
SET_VARSIZE(res, len);
}
/* make signature, if array is too long */
if (VARSIZE(res) > TOAST_INDEX_TARGET)
{
GISTTYPE *ressign;
len = CALCGTSIZE(SIGNKEY, 0);
ressign = (GISTTYPE *) palloc(len);
SET_VARSIZE(ressign, len);
ressign->flag = SIGNKEY;
makesign(GETSIGN(ressign), res);
res = ressign;
}
retval = (GISTENTRY *) palloc(sizeof(GISTENTRY));
gistentryinit(*retval, PointerGetDatum(res),
entry->rel, entry->page,
entry->offset, FALSE);
}
else if (ISSIGNKEY(DatumGetPointer(entry->key)) &&
!ISALLTRUE(DatumGetPointer(entry->key)))
{
int4 i,
len;
GISTTYPE *res;
BITVECP sign = GETSIGN(DatumGetPointer(entry->key));
LOOPBYTE(
if ((sign[i] & 0xff) != 0xff)
PG_RETURN_POINTER(retval);
);
static Datum
argm_transfn_universal(PG_FUNCTION_ARGS, int32 compareFunctionResultToAdvance)
{
Oid type;
ArgmState *state;
MemoryContext aggcontext,
oldcontext;
int i;
int32 comparison_result;
if (!AggCheckCallContext(fcinfo, &aggcontext))
{
/* cannot be called directly because of internal-type argument */
elog(ERROR, "argm**_transfn called in non-aggregate context");
}
if (PG_ARGISNULL(0))
{
/* First time through --- initialize */
/* Make a temporary context to hold all the junk */
oldcontext = MemoryContextSwitchTo(aggcontext);
/* Initialize the state variable*/
state = palloc(sizeof(ArgmState));
state->key_count = PG_NARGS() - 1;
state->keys = palloc(sizeof(ArgmDatumWithType) * (state->key_count));
for (i = 0; i < state->key_count; i++)
{
type = get_fn_expr_argtype(fcinfo->flinfo, i + 1);
if (type == InvalidOid)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("could not determine input data type")));
state->keys[i].type = type;
get_typlenbyvalalign(type,
&state->keys[i].typlen,
&state->keys[i].typbyval,
&state->keys[i].typalign);
/* We do not need a sorting proc for payload */
if (i != 0)
state->keys[i].cmp_proc =
lookup_type_cache(state->keys[i].type,
TYPECACHE_CMP_PROC)->cmp_proc;
/* Copy initial values */
argm_copy_datum(PG_ARGISNULL(i + 1),
PG_GETARG_DATUM(i + 1),
&(state->keys[i]), false);
}
}
else
{
state = (ArgmState *) PG_GETARG_POINTER(0);
oldcontext = MemoryContextSwitchTo(aggcontext);
/* compare keys (but not payload) lexicographically */
for (i = 1; i < state->key_count; i++)
{
/* nulls last */
if (PG_ARGISNULL(i + 1) && !state->keys[i].is_null)
break;
if (!PG_ARGISNULL(i + 1) && state->keys[i].is_null) {
/* nulls last */
comparison_result = 1;
} else {
comparison_result = DatumGetInt32(OidFunctionCall2Coll(
state->keys[i].cmp_proc,
PG_GET_COLLATION(),
PG_GETARG_DATUM(i + 1),
state->keys[i].value
)) * compareFunctionResultToAdvance;
if (comparison_result < 0)
break;
}
if (comparison_result > 0)
{
for (i = 0; i < state->key_count; i++)
argm_copy_datum(PG_ARGISNULL(i + 1),
PG_GETARG_DATUM(i + 1),
&(state->keys[i]), true);
break;
}
}
}
MemoryContextSwitchTo(oldcontext);
PG_RETURN_POINTER(state);
}
/*
** The GiST PickSplit method for _intments
** We use Guttman's poly time split algorithm
*/
Datum
g_int_picksplit(PG_FUNCTION_ARGS)
{
GistEntryVector *entryvec = (GistEntryVector *) PG_GETARG_POINTER(0);
GIST_SPLITVEC *v = (GIST_SPLITVEC *) PG_GETARG_POINTER(1);
OffsetNumber i,
j;
ArrayType *datum_alpha,
*datum_beta;
ArrayType *datum_l,
*datum_r;
ArrayType *union_d,
*union_dl,
*union_dr;
ArrayType *inter_d;
bool firsttime;
float size_alpha,
size_beta,
size_union,
size_inter;
float size_waste,
waste;
float size_l,
size_r;
int nbytes;
OffsetNumber seed_1 = 0,
seed_2 = 0;
OffsetNumber *left,
*right;
OffsetNumber maxoff;
SPLITCOST *costvector;
#ifdef GIST_DEBUG
elog(DEBUG3, "--------picksplit %d", entryvec->n);
#endif
maxoff = entryvec->n - 2;
nbytes = (maxoff + 2) * sizeof(OffsetNumber);
v->spl_left = (OffsetNumber *) palloc(nbytes);
v->spl_right = (OffsetNumber *) palloc(nbytes);
firsttime = true;
waste = 0.0;
for (i = FirstOffsetNumber; i < maxoff; i = OffsetNumberNext(i))
{
datum_alpha = GETENTRY(entryvec, i);
for (j = OffsetNumberNext(i); j <= maxoff; j = OffsetNumberNext(j))
{
datum_beta = GETENTRY(entryvec, j);
/* compute the wasted space by unioning these guys */
/* size_waste = size_union - size_inter; */
union_d = inner_int_union(datum_alpha, datum_beta);
rt__int_size(union_d, &size_union);
inter_d = inner_int_inter(datum_alpha, datum_beta);
rt__int_size(inter_d, &size_inter);
size_waste = size_union - size_inter;
pfree(union_d);
if (inter_d != (ArrayType *) NULL)
pfree(inter_d);
/*
* are these a more promising split that what we've already seen?
*/
if (size_waste > waste || firsttime)
{
waste = size_waste;
seed_1 = i;
seed_2 = j;
firsttime = false;
}
}
}
left = v->spl_left;
v->spl_nleft = 0;
right = v->spl_right;
v->spl_nright = 0;
if (seed_1 == 0 || seed_2 == 0)
{
seed_1 = 1;
seed_2 = 2;
}
datum_alpha = GETENTRY(entryvec, seed_1);
datum_l = copy_intArrayType(datum_alpha);
rt__int_size(datum_l, &size_l);
datum_beta = GETENTRY(entryvec, seed_2);
datum_r = copy_intArrayType(datum_beta);
rt__int_size(datum_r, &size_r);
maxoff = OffsetNumberNext(maxoff);
/*
* sort entries
*/
costvector = (SPLITCOST *) palloc(sizeof(SPLITCOST) * maxoff);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
costvector[i - 1].pos = i;
datum_alpha = GETENTRY(entryvec, i);
union_d = inner_int_union(datum_l, datum_alpha);
rt__int_size(union_d, &size_alpha);
pfree(union_d);
union_d = inner_int_union(datum_r, datum_alpha);
rt__int_size(union_d, &size_beta);
pfree(union_d);
costvector[i - 1].cost = Abs((size_alpha - size_l) - (size_beta - size_r));
}
qsort((void *) costvector, maxoff, sizeof(SPLITCOST), comparecost);
/*
* Now split up the regions between the two seeds. An important property
* of this split algorithm is that the split vector v has the indices of
* items to be split in order in its left and right vectors. We exploit
* this property by doing a merge in the code that actually splits the
* page.
*
* For efficiency, we also place the new index tuple in this loop. This is
* handled at the very end, when we have placed all the existing tuples
* and i == maxoff + 1.
*/
for (j = 0; j < maxoff; j++)
{
i = costvector[j].pos;
/*
* If we've already decided where to place this item, just put it on
* the right list. Otherwise, we need to figure out which page needs
* the least enlargement in order to store the item.
*/
if (i == seed_1)
{
*left++ = i;
v->spl_nleft++;
continue;
}
else if (i == seed_2)
{
*right++ = i;
v->spl_nright++;
continue;
}
/* okay, which page needs least enlargement? */
datum_alpha = GETENTRY(entryvec, i);
union_dl = inner_int_union(datum_l, datum_alpha);
union_dr = inner_int_union(datum_r, datum_alpha);
rt__int_size(union_dl, &size_alpha);
rt__int_size(union_dr, &size_beta);
/* pick which page to add it to */
if (size_alpha - size_l < size_beta - size_r + WISH_F(v->spl_nleft, v->spl_nright, 0.01))
{
if (datum_l)
pfree(datum_l);
if (union_dr)
pfree(union_dr);
datum_l = union_dl;
size_l = size_alpha;
*left++ = i;
v->spl_nleft++;
}
else
{
if (datum_r)
pfree(datum_r);
if (union_dl)
pfree(union_dl);
datum_r = union_dr;
size_r = size_beta;
*right++ = i;
v->spl_nright++;
}
}
pfree(costvector);
*right = *left = FirstOffsetNumber;
datum_l->flags &= ~LEAFKEY;
datum_r->flags &= ~LEAFKEY;
v->spl_ldatum = PointerGetDatum(datum_l);
v->spl_rdatum = PointerGetDatum(datum_r);
PG_RETURN_POINTER(v);
}
Datum
gin_extract_jsonb_query(PG_FUNCTION_ARGS)
{
int32 *nentries = (int32 *) PG_GETARG_POINTER(1);
StrategyNumber strategy = PG_GETARG_UINT16(2);
int32 *searchMode = (int32 *) PG_GETARG_POINTER(6);
Datum *entries;
if (strategy == JsonbContainsStrategyNumber)
{
/* Query is a jsonb, so just apply gin_extract_jsonb... */
entries = (Datum *)
DatumGetPointer(DirectFunctionCall2(gin_extract_jsonb,
PG_GETARG_DATUM(0),
PointerGetDatum(nentries)));
/* ...although "contains {}" requires a full index scan */
if (*nentries == 0)
*searchMode = GIN_SEARCH_MODE_ALL;
}
else if (strategy == JsonbExistsStrategyNumber)
{
/* Query is a text string, which we treat as a key */
text *query = PG_GETARG_TEXT_PP(0);
*nentries = 1;
entries = (Datum *) palloc(sizeof(Datum));
entries[0] = make_text_key(JGINFLAG_KEY,
VARDATA_ANY(query),
VARSIZE_ANY_EXHDR(query));
}
else if (strategy == JsonbExistsAnyStrategyNumber ||
strategy == JsonbExistsAllStrategyNumber)
{
/* Query is a text array; each element is treated as a key */
ArrayType *query = PG_GETARG_ARRAYTYPE_P(0);
Datum *key_datums;
bool *key_nulls;
int key_count;
int i,
j;
deconstruct_array(query,
TEXTOID, -1, false, 'i',
&key_datums, &key_nulls, &key_count);
entries = (Datum *) palloc(sizeof(Datum) * key_count);
for (i = 0, j = 0; i < key_count; i++)
{
/* Nulls in the array are ignored */
if (key_nulls[i])
continue;
entries[j++] = make_text_key(JGINFLAG_KEY,
VARDATA_ANY(key_datums[i]),
VARSIZE_ANY_EXHDR(key_datums[i]));
}
*nentries = j;
/* ExistsAll with no keys should match everything */
if (j == 0 && strategy == JsonbExistsAllStrategyNumber)
*searchMode = GIN_SEARCH_MODE_ALL;
}
else
{
elog(ERROR, "unrecognized strategy number: %d", strategy);
entries = NULL; /* keep compiler quiet */
}
PG_RETURN_POINTER(entries);
}
/*
* input
*/
Datum
bqarr_in(PG_FUNCTION_ARGS)
{
char *buf = (char *) PG_GETARG_POINTER(0);
WORKSTATE state;
int32 i;
QUERYTYPE *query;
int32 commonlen;
ITEM *ptr;
NODE *tmp;
int32 pos = 0;
#ifdef BS_DEBUG
StringInfoData pbuf;
#endif
state.buf = buf;
state.state = WAITOPERAND;
state.count = 0;
state.num = 0;
state.str = NULL;
/* make polish notation (postfix, but in reverse order) */
makepol(&state);
if (!state.num)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("empty query")));
if (state.num > QUERYTYPEMAXITEMS)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("number of query items (%d) exceeds the maximum allowed (%d)",
state.num, (int) QUERYTYPEMAXITEMS)));
commonlen = COMPUTESIZE(state.num);
query = (QUERYTYPE *) palloc(commonlen);
SET_VARSIZE(query, commonlen);
query->size = state.num;
ptr = GETQUERY(query);
for (i = state.num - 1; i >= 0; i--)
{
ptr[i].type = state.str->type;
ptr[i].val = state.str->val;
tmp = state.str->next;
pfree(state.str);
state.str = tmp;
}
pos = query->size - 1;
findoprnd(ptr, &pos);
#ifdef BS_DEBUG
initStringInfo(&pbuf);
for (i = 0; i < query->size; i++)
{
if (ptr[i].type == OPR)
appendStringInfo(&pbuf, "%c(%d) ", ptr[i].val, ptr[i].left);
else
appendStringInfo(&pbuf, "%d ", ptr[i].val);
}
elog(DEBUG3, "POR: %s", pbuf.data);
pfree(pbuf.data);
#endif
PG_RETURN_POINTER(query);
}
/*
** The GiST Consistent method for _intments
** Should return false if for all data items x below entry,
** the predicate x op query == FALSE, where op is the oper
** corresponding to strategy in the pg_amop table.
*/
Datum
g_int_consistent(PG_FUNCTION_ARGS)
{
GISTENTRY *entry = (GISTENTRY *) PG_GETARG_POINTER(0);
ArrayType *query = PG_GETARG_ARRAYTYPE_P_COPY(1);
StrategyNumber strategy = (StrategyNumber) PG_GETARG_UINT16(2);
/* Oid subtype = PG_GETARG_OID(3); */
bool *recheck = (bool *) PG_GETARG_POINTER(4);
bool retval;
/* this is exact except for RTSameStrategyNumber */
*recheck = (strategy == RTSameStrategyNumber);
if (strategy == BooleanSearchStrategy)
{
retval = execconsistent((QUERYTYPE *) query,
(ArrayType *) DatumGetPointer(entry->key),
GIST_LEAF(entry));
pfree(query);
PG_RETURN_BOOL(retval);
}
/* sort query for fast search, key is already sorted */
CHECKARRVALID(query);
PREPAREARR(query);
switch (strategy)
{
case RTOverlapStrategyNumber:
retval = inner_int_overlap((ArrayType *) DatumGetPointer(entry->key),
query);
break;
case RTSameStrategyNumber:
if (GIST_LEAF(entry))
DirectFunctionCall3(g_int_same,
entry->key,
PointerGetDatum(query),
PointerGetDatum(&retval));
else
retval = inner_int_contains((ArrayType *) DatumGetPointer(entry->key),
query);
break;
case RTContainsStrategyNumber:
case RTOldContainsStrategyNumber:
retval = inner_int_contains((ArrayType *) DatumGetPointer(entry->key),
query);
break;
case RTContainedByStrategyNumber:
case RTOldContainedByStrategyNumber:
if (GIST_LEAF(entry))
retval = inner_int_contains(query,
(ArrayType *) DatumGetPointer(entry->key));
else
retval = inner_int_overlap((ArrayType *) DatumGetPointer(entry->key),
query);
break;
default:
retval = FALSE;
}
pfree(query);
PG_RETURN_BOOL(retval);
}
