/*
* heap_fill_tuple
* Load data portion of a tuple from values/isnull arrays
*
* We also fill the null bitmap (if any) and set the infomask bits
* that reflect the tuple's data contents.
*
* NOTE: it is now REQUIRED that the caller have pre-zeroed the data area.
*/
void
heap_fill_tuple(TupleDesc tupleDesc,
Datum *values, bool *isnull,
char *data, Size data_size,
uint16 *infomask, bits8 *bit)
{
bits8 *bitP;
int bitmask;
int i;
int numberOfAttributes = tupleDesc->natts;
Form_pg_attribute *att = tupleDesc->attrs;
#ifdef USE_ASSERT_CHECKING
char *start = data;
#endif
if (bit != NULL)
{
bitP = &bit[-1];
bitmask = HIGHBIT;
}
else
{
/* just to keep compiler quiet */
bitP = NULL;
bitmask = 0;
}
*infomask &= ~(HEAP_HASNULL | HEAP_HASVARWIDTH | HEAP_HASEXTERNAL);
for (i = 0; i < numberOfAttributes; i++)
{
Size data_length;
if (bit != NULL)
{
if (bitmask != HIGHBIT)
bitmask <<= 1;
else
{
bitP += 1;
*bitP = 0x0;
bitmask = 1;
}
if (isnull[i])
{
*infomask |= HEAP_HASNULL;
continue;
}
*bitP |= bitmask;
}
/*
* XXX we use the att_align macros on the pointer value itself, not on
* an offset. This is a bit of a hack.
*/
if (att[i]->attbyval)
{
/* pass-by-value */
data = (char *) att_align_nominal(data, att[i]->attalign);
store_att_byval(data, values[i], att[i]->attlen);
data_length = att[i]->attlen;
}
else if (att[i]->attlen == -1)
{
/* varlena */
Pointer val = DatumGetPointer(values[i]);
*infomask |= HEAP_HASVARWIDTH;
if (VARATT_IS_EXTERNAL(val))
{
*infomask |= HEAP_HASEXTERNAL;
/* no alignment, since it's short by definition */
data_length = VARSIZE_EXTERNAL(val);
memcpy(data, val, data_length);
}
else if (VARATT_IS_SHORT(val))
{
/* no alignment for short varlenas */
data_length = VARSIZE_SHORT(val);
memcpy(data, val, data_length);
}
else if (VARLENA_ATT_IS_PACKABLE(att[i]) &&
VARATT_CAN_MAKE_SHORT(val))
{
/* convert to short varlena -- no alignment */
data_length = VARATT_CONVERTED_SHORT_SIZE(val);
SET_VARSIZE_SHORT(data, data_length);
memcpy(data + 1, VARDATA(val), data_length - 1);
}
else
{
/* full 4-byte header varlena */
data = (char *) att_align_nominal(data,
att[i]->attalign);
data_length = VARSIZE(val);
memcpy(data, val, data_length);
}
}
else if (att[i]->attlen == -2)
{
/* cstring ... never needs alignment */
*infomask |= HEAP_HASVARWIDTH;
Assert(att[i]->attalign == 'c');
data_length = strlen(DatumGetCString(values[i])) + 1;
memcpy(data, DatumGetPointer(values[i]), data_length);
}
else
{
/* fixed-length pass-by-reference */
data = (char *) att_align_nominal(data, att[i]->attalign);
Assert(att[i]->attlen > 0);
data_length = att[i]->attlen;
memcpy(data, DatumGetPointer(values[i]), data_length);
}
data += data_length;
}
Assert((data - start) == data_size);
}
static void
compileTheSubstitute(DictThesaurus *d)
{
int i;
for (i = 0; i < d->nsubst; i++)
{
TSLexeme *rem = d->subst[i].res,
*outptr,
*inptr;
int n = 2;
outptr = d->subst[i].res = (TSLexeme *) palloc(sizeof(TSLexeme) * n);
outptr->lexeme = NULL;
inptr = rem;
while (inptr && inptr->lexeme)
{
TSLexeme *lexized,
tmplex[2];
if (inptr->flags & DT_USEASIS)
{ /* do not lexize */
tmplex[0] = *inptr;
tmplex[0].flags = 0;
tmplex[1].lexeme = NULL;
lexized = tmplex;
}
else
{
lexized = (TSLexeme *) DatumGetPointer(
FunctionCall4(
&(d->subdict->lexize),
PointerGetDatum(d->subdict->dictData),
PointerGetDatum(inptr->lexeme),
Int32GetDatum(strlen(inptr->lexeme)),
PointerGetDatum(NULL)
)
);
}
if (lexized && lexized->lexeme)
{
int toset = (lexized->lexeme && outptr != d->subst[i].res) ? (outptr - d->subst[i].res) : -1;
while (lexized->lexeme)
{
if (outptr - d->subst[i].res + 1 >= n)
{
int diff = outptr - d->subst[i].res;
n *= 2;
d->subst[i].res = (TSLexeme *) repalloc(d->subst[i].res, sizeof(TSLexeme) * n);
outptr = d->subst[i].res + diff;
}
*outptr = *lexized;
outptr->lexeme = pstrdup(lexized->lexeme);
outptr++;
lexized++;
}
if (toset > 0)
d->subst[i].res[toset].flags |= TSL_ADDPOS;
}
else if (lexized)
{
ereport(ERROR,
(errcode(ERRCODE_CONFIG_FILE_ERROR),
errmsg("thesaurus substitute word \"%s\" is a stop word (rule %d)",
inptr->lexeme, i + 1)));
}
else
{
ereport(ERROR,
(errcode(ERRCODE_CONFIG_FILE_ERROR),
errmsg("thesaurus substitute word \"%s\" isn't recognized by subdictionary (rule %d)",
inptr->lexeme, i + 1)));
}
if (inptr->lexeme)
pfree(inptr->lexeme);
inptr++;
}
if (outptr == d->subst[i].res)
ereport(ERROR,
(errcode(ERRCODE_CONFIG_FILE_ERROR),
errmsg("thesaurus substitute phrase is empty (rule %d)",
i + 1)));
d->subst[i].reslen = outptr - d->subst[i].res;
pfree(rem);
}
}
/*
* Compute the list of TIDs to be visited, by evaluating the expressions
* for them.
*
* (The result is actually an array, not a list.)
*/
static void
TidListCreate(TidScanState *tidstate)
{
List *evalList = tidstate->tss_tidquals;
ExprContext *econtext = tidstate->ss.ps.ps_ExprContext;
BlockNumber nblocks;
ItemPointerData *tidList;
int numAllocTids;
int numTids;
ListCell *l;
/*
* We silently discard any TIDs that are out of range at the time of scan
* start. (Since we hold at least AccessShareLock on the table, it won't
* be possible for someone to truncate away the blocks we intend to
* visit.)
*/
nblocks = RelationGetNumberOfBlocks(tidstate->ss.ss_currentRelation);
/*
* We initialize the array with enough slots for the case that all quals
* are simple OpExprs or CurrentOfExprs. If there are any
* ScalarArrayOpExprs, we may have to enlarge the array.
*/
numAllocTids = list_length(evalList);
tidList = (ItemPointerData *)
palloc(numAllocTids * sizeof(ItemPointerData));
numTids = 0;
tidstate->tss_isCurrentOf = false;
foreach(l, evalList)
{
ExprState *exstate = (ExprState *) lfirst(l);
Expr *expr = exstate->expr;
ItemPointer itemptr;
bool isNull;
if (is_opclause(expr))
{
FuncExprState *fexstate = (FuncExprState *) exstate;
Node *arg1;
Node *arg2;
arg1 = get_leftop(expr);
arg2 = get_rightop(expr);
if (IsCTIDVar(arg1))
exstate = (ExprState *) lsecond(fexstate->args);
else if (IsCTIDVar(arg2))
exstate = (ExprState *) linitial(fexstate->args);
else
elog(ERROR, "could not identify CTID variable");
itemptr = (ItemPointer)
DatumGetPointer(ExecEvalExprSwitchContext(exstate,
econtext,
&isNull,
NULL));
if (!isNull &&
ItemPointerIsValid(itemptr) &&
ItemPointerGetBlockNumber(itemptr) < nblocks)
{
if (numTids >= numAllocTids)
{
numAllocTids *= 2;
tidList = (ItemPointerData *)
repalloc(tidList,
numAllocTids * sizeof(ItemPointerData));
}
tidList[numTids++] = *itemptr;
}
}
else if (expr && IsA(expr, ScalarArrayOpExpr))
{
ScalarArrayOpExprState *saexstate = (ScalarArrayOpExprState *) exstate;
Datum arraydatum;
ArrayType *itemarray;
Datum *ipdatums;
bool *ipnulls;
int ndatums;
int i;
exstate = (ExprState *) lsecond(saexstate->fxprstate.args);
arraydatum = ExecEvalExprSwitchContext(exstate,
econtext,
&isNull,
NULL);
if (isNull)
continue;
itemarray = DatumGetArrayTypeP(arraydatum);
deconstruct_array(itemarray,
TIDOID, SizeOfIptrData, false, 's',
&ipdatums, &ipnulls, &ndatums);
if (numTids + ndatums > numAllocTids)
{
numAllocTids = numTids + ndatums;
tidList = (ItemPointerData *)
repalloc(tidList,
numAllocTids * sizeof(ItemPointerData));
}
for (i = 0; i < ndatums; i++)
{
if (!ipnulls[i])
{
itemptr = (ItemPointer) DatumGetPointer(ipdatums[i]);
if (ItemPointerIsValid(itemptr) &&
ItemPointerGetBlockNumber(itemptr) < nblocks)
tidList[numTids++] = *itemptr;
}
}
pfree(ipdatums);
pfree(ipnulls);
}
else if (expr && IsA(expr, CurrentOfExpr))
{
CurrentOfExpr *cexpr = (CurrentOfExpr *) expr;
ItemPointerData cursor_tid;
if (execCurrentOf(cexpr, econtext,
RelationGetRelid(tidstate->ss.ss_currentRelation),
&cursor_tid))
{
if (numTids >= numAllocTids)
{
numAllocTids *= 2;
tidList = (ItemPointerData *)
repalloc(tidList,
numAllocTids * sizeof(ItemPointerData));
}
tidList[numTids++] = cursor_tid;
tidstate->tss_isCurrentOf = true;
}
}
else
elog(ERROR, "could not identify CTID expression");
}
/*
* Convert a HeapTuple into a byte-sequence, and store it directly
* into a chunklist for transmission.
*
* This code is based on the printtup_internal_20() function in printtup.c.
*/
void
SerializeTupleIntoChunks(GenericTuple gtuple, SerTupInfo *pSerInfo, TupleChunkList tcList)
{
TupleChunkListItem tcItem = NULL;
MemoryContext oldCtxt;
TupleDesc tupdesc;
int i,
natts;
AssertArg(tcList != NULL);
AssertArg(gtuple != NULL);
AssertArg(pSerInfo != NULL);
tupdesc = pSerInfo->tupdesc;
natts = tupdesc->natts;
/* get ready to go */
tcList->p_first = NULL;
tcList->p_last = NULL;
tcList->num_chunks = 0;
tcList->serialized_data_length = 0;
tcList->max_chunk_length = Gp_max_tuple_chunk_size;
if (natts == 0)
{
tcItem = getChunkFromCache(&pSerInfo->chunkCache);
if (tcItem == NULL)
{
ereport(FATAL, (errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("Could not allocate space for first chunk item in new chunk list.")));
}
/* TC_EMTPY is just one chunk */
SetChunkType(tcItem->chunk_data, TC_EMPTY);
tcItem->chunk_length = TUPLE_CHUNK_HEADER_SIZE;
appendChunkToTCList(tcList, tcItem);
return;
}
tcItem = getChunkFromCache(&pSerInfo->chunkCache);
if (tcItem == NULL)
{
ereport(FATAL, (errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("Could not allocate space for first chunk item in new chunk list.")));
}
/* assume that we'll take a single chunk */
SetChunkType(tcItem->chunk_data, TC_WHOLE);
tcItem->chunk_length = TUPLE_CHUNK_HEADER_SIZE;
appendChunkToTCList(tcList, tcItem);
AssertState(s_tupSerMemCtxt != NULL);
if (is_memtuple(gtuple))
{
MemTuple mtuple = (MemTuple) gtuple;
addByteStringToChunkList(tcList, (char *) mtuple, memtuple_get_size(mtuple), &pSerInfo->chunkCache);
addPadding(tcList, &pSerInfo->chunkCache, memtuple_get_size(mtuple));
}
else
{
HeapTuple tuple = (HeapTuple) gtuple;
HeapTupleHeader t_data = tuple->t_data;
TupSerHeader tsh;
unsigned int datalen;
unsigned int nullslen;
datalen = tuple->t_len - t_data->t_hoff;
if (HeapTupleHasNulls(tuple))
nullslen = BITMAPLEN(HeapTupleHeaderGetNatts(t_data));
else
nullslen = 0;
tsh.tuplen = sizeof(TupSerHeader) + TYPEALIGN(TUPLE_CHUNK_ALIGN,nullslen) + datalen;
tsh.natts = HeapTupleHeaderGetNatts(t_data);
tsh.infomask = t_data->t_infomask;
addByteStringToChunkList(tcList, (char *)&tsh, sizeof(TupSerHeader), &pSerInfo->chunkCache);
/* If we don't have any attributes which have been toasted, we
* can be very very simple: just send the raw data. */
if ((tsh.infomask & HEAP_HASEXTERNAL) == 0)
{
if (nullslen)
{
addByteStringToChunkList(tcList, (char *)t_data->t_bits, nullslen, &pSerInfo->chunkCache);
addPadding(tcList,&pSerInfo->chunkCache,nullslen);
}
addByteStringToChunkList(tcList, (char *)t_data + t_data->t_hoff, datalen, &pSerInfo->chunkCache);
addPadding(tcList,&pSerInfo->chunkCache,datalen);
}
else
{
/* We have to be more careful when we have tuples that
* have been toasted. Ideally we'd like to send the
* untoasted attributes in as "raw" a format as possible
* but that makes rebuilding the tuple harder .
*/
oldCtxt = MemoryContextSwitchTo(s_tupSerMemCtxt);
/* deconstruct the tuple (faster than a heap_getattr loop) */
heap_deform_tuple(tuple, tupdesc, pSerInfo->values, pSerInfo->nulls);
MemoryContextSwitchTo(oldCtxt);
/* Send the nulls character-array. */
addByteStringToChunkList(tcList, pSerInfo->nulls, natts, &pSerInfo->chunkCache);
addPadding(tcList,&pSerInfo->chunkCache,natts);
/*
* send the attributes of this tuple: NOTE anything which allocates
* temporary space (e.g. could result in a PG_DETOAST_DATUM) should be
* executed with the memory context set to s_tupSerMemCtxt
*/
for (i = 0; i < natts; ++i)
{
SerAttrInfo *attrInfo = pSerInfo->myinfo + i;
Datum origattr = pSerInfo->values[i],
attr;
/* skip null attributes (already taken care of above) */
if (pSerInfo->nulls[i])
continue;
if (attrInfo->typlen == -1)
{
int32 sz;
char *data;
/*
* If we have a toasted datum, forcibly detoast it here to avoid
* memory leakage: we want to force the detoast allocation(s) to
* happen in our reset-able serialization context.
*/
oldCtxt = MemoryContextSwitchTo(s_tupSerMemCtxt);
attr = PointerGetDatum(PG_DETOAST_DATUM_PACKED(origattr));
MemoryContextSwitchTo(oldCtxt);
sz = VARSIZE_ANY_EXHDR(attr);
data = VARDATA_ANY(attr);
/* Send length first, then data */
addInt32ToChunkList(tcList, sz, &pSerInfo->chunkCache);
addByteStringToChunkList(tcList, data, sz, &pSerInfo->chunkCache);
addPadding(tcList, &pSerInfo->chunkCache, sz);
}
else if (attrInfo->typlen == -2)
{
int32 sz;
char *data;
/* CString, we would send the string with the terminating '\0' */
data = DatumGetCString(origattr);
sz = strlen(data) + 1;
/* Send length first, then data */
addInt32ToChunkList(tcList, sz, &pSerInfo->chunkCache);
addByteStringToChunkList(tcList, data, sz, &pSerInfo->chunkCache);
addPadding(tcList, &pSerInfo->chunkCache, sz);
}
else if (attrInfo->typbyval)
{
/*
* We send a full-width Datum for all pass-by-value types, regardless of
* the actual size.
*/
addByteStringToChunkList(tcList, (char *) &origattr, sizeof(Datum), &pSerInfo->chunkCache);
addPadding(tcList, &pSerInfo->chunkCache, sizeof(Datum));
}
else
{
addByteStringToChunkList(tcList, DatumGetPointer(origattr), attrInfo->typlen, &pSerInfo->chunkCache);
addPadding(tcList, &pSerInfo->chunkCache, attrInfo->typlen);
attr = origattr;
}
}
MemoryContextReset(s_tupSerMemCtxt);
}
}
/*
* if we have more than 1 chunk we have to set the chunk types on our
* first chunk and last chunk
*/
if (tcList->num_chunks > 1)
{
TupleChunkListItem first,
last;
first = tcList->p_first;
last = tcList->p_last;
Assert(first != NULL);
Assert(first != last);
Assert(last != NULL);
SetChunkType(first->chunk_data, TC_PARTIAL_START);
SetChunkType(last->chunk_data, TC_PARTIAL_END);
/*
* any intervening chunks are already set to TC_PARTIAL_MID when
* allocated
*/
}
return;
}
/*
* hstoreUpgrade: PG_DETOAST_DATUM plus support for conversion of old hstores
*/
HStore *
hstoreUpgrade(Datum orig)
{
HStore *hs = (HStore *) PG_DETOAST_DATUM(orig);
int valid_new;
int valid_old;
/* Return immediately if no conversion needed */
if (hs->size_ & HS_FLAG_NEWVERSION)
return hs;
/* Do we have a writable copy? If not, make one. */
if ((void *) hs == (void *) DatumGetPointer(orig))
hs = (HStore *) PG_DETOAST_DATUM_COPY(orig);
if (hs->size_ == 0 ||
(VARSIZE(hs) < 32768 && HSE_ISFIRST((ARRPTR(hs)[0]))))
{
HS_SETCOUNT(hs, HS_COUNT(hs));
HS_FIXSIZE(hs, HS_COUNT(hs));
return hs;
}
valid_new = hstoreValidNewFormat(hs);
valid_old = hstoreValidOldFormat(hs);
if (!valid_old || hs->size_ == 0)
{
if (valid_new)
{
/*
* force the "new version" flag and the correct varlena length.
*/
HS_SETCOUNT(hs, HS_COUNT(hs));
HS_FIXSIZE(hs, HS_COUNT(hs));
return hs;
}
else
{
elog(ERROR, "invalid hstore value found");
}
}
/*
* this is the tricky edge case. It is only possible in some quite extreme
* cases (the hstore must have had a lot of wasted padding space at the
* end). But the only way a "new" hstore value could get here is if we're
* upgrading in place from a pre-release version of hstore-new (NOT
* contrib/hstore), so we work off the following assumptions: 1. If you're
* moving from old contrib/hstore to hstore-new, you're required to fix up
* any potential conflicts first, e.g. by running ALTER TABLE ... USING
* col::text::hstore; on all hstore columns before upgrading. 2. If you're
* moving from old contrib/hstore to new contrib/hstore, then "new" values
* are impossible here 3. If you're moving from pre-release hstore-new to
* hstore-new, then "old" values are impossible here 4. If you're moving
* from pre-release hstore-new to new contrib/hstore, you're not doing so
* as an in-place upgrade, so there is no issue So the upshot of all this
* is that we can treat all the edge cases as "new" if we're being built
* as hstore-new, and "old" if we're being built as contrib/hstore.
*
* XXX the WARNING can probably be downgraded to DEBUG1 once this has been
* beta-tested. But for now, it would be very useful to know if anyone can
* actually reach this case in a non-contrived setting.
*/
if (valid_new)
{
#if HSTORE_IS_HSTORE_NEW
elog(WARNING, "ambiguous hstore value resolved as hstore-new");
/*
* force the "new version" flag and the correct varlena length.
*/
HS_SETCOUNT(hs, HS_COUNT(hs));
HS_FIXSIZE(hs, HS_COUNT(hs));
return hs;
#else
elog(WARNING, "ambiguous hstore value resolved as hstore-old");
#endif
}
/*
* must have an old-style value. Overwrite it in place as a new-style one.
*/
{
int count = hs->size_;
HEntry *new_entries = ARRPTR(hs);
HOldEntry *old_entries = (HOldEntry *) ARRPTR(hs);
int i;
for (i = 0; i < count; ++i)
{
uint32 pos = old_entries[i].pos;
uint32 keylen = old_entries[i].keylen;
uint32 vallen = old_entries[i].vallen;
bool isnull = old_entries[i].valisnull;
if (isnull)
vallen = 0;
new_entries[2 * i].entry = (pos + keylen) & HENTRY_POSMASK;
new_entries[2 * i + 1].entry = (((pos + keylen + vallen) & HENTRY_POSMASK)
| ((isnull) ? HENTRY_ISNULL : 0));
}
if (count)
new_entries[0].entry |= HENTRY_ISFIRST;
HS_SETCOUNT(hs, count);
HS_FIXSIZE(hs, count);
}
return hs;
}
Datum
_numeric_weighted_stddev_samp_intermediate(PG_FUNCTION_ARGS)
{
WeightedStddevSampInternalState *state;
Datum value,
weight,
old_s_0,
old_s_1,
old_s_2,
w_v,
w_v2;
MemoryContext aggcontext,
oldcontext;
if (!AggCheckCallContext(fcinfo, &aggcontext))
/* cannot be called directly because of internal-type argument */
elog(ERROR, "_weighted_stddev_samp_intermediate called in non-aggregate context");
if (PG_ARGISNULL(0))
{
oldcontext = MemoryContextSwitchTo(aggcontext);
state = (WeightedStddevSampInternalState *) palloc(sizeof(WeightedStddevSampInternalState));
state->s_2 = make_numeric(0);
state->s_1 = make_numeric(0);
state->s_0 = make_numeric(0);
state->zero = make_numeric(0);
state->n_prime = 0;
MemoryContextSwitchTo(oldcontext);
}
else
state = (WeightedStddevSampInternalState *) PG_GETARG_POINTER(0);
/*
* We're non-strict, so we MUST check args for nullity ourselves before
* using them. To preserve the behaviour of null inputs, we skip updating
* on them.
*/
if (PG_ARGISNULL(1) || PG_ARGISNULL(2))
PG_RETURN_POINTER(state);
/*
* We fetch and process the input in the shortlived calling context to
* avoid leaking memory in aggcontext per cycle. We force the input to be
* detoasted here, too, in the shortlived context. (PG_GETARG_DATUM does
* not detoast, but PG_GETARG_NUMERIC does.)
*/
value = NumericGetDatum(PG_GETARG_NUMERIC(1));
weight = NumericGetDatum(PG_GETARG_NUMERIC(2));
/*
* We also skip updating when the weight is zero.
*/
if (DatumGetBool(DirectFunctionCall2(numeric_eq, weight, state->zero)))
PG_RETURN_POINTER(state);
/*
* Compute intermediate values w*v and w*(v^2) in the short-lived context
*/
w_v = DirectFunctionCall2(numeric_mul, weight, value);
w_v2 = DirectFunctionCall2(numeric_mul, w_v, value);
/*
* The new running totals must be allocated in the long-lived context. We
* rely on the numeric_* functions to clean up after themselves (which they
* currently do, but only if the input is already detoasted); we could play
* safe and copy only the final results into aggcontext, but this turns out
* to have a measurable performance hit.
*/
oldcontext = MemoryContextSwitchTo(aggcontext);
old_s_2 = state->s_2;
old_s_1 = state->s_1;
old_s_0 = state->s_0;
state->s_0 = DirectFunctionCall2(numeric_add, old_s_0, weight);
state->s_1 = DirectFunctionCall2(numeric_add, old_s_1, w_v);
state->s_2 = DirectFunctionCall2(numeric_add, old_s_2, w_v2);
state->n_prime += 1;
pfree(DatumGetPointer(old_s_2));
pfree(DatumGetPointer(old_s_1));
pfree(DatumGetPointer(old_s_0));
MemoryContextSwitchTo(oldcontext);
PG_RETURN_POINTER(state);
}
Datum
tsquery_rewrite_query(PG_FUNCTION_ARGS)
{
TSQuery query = PG_GETARG_TSQUERY_COPY(0);
text *in = PG_GETARG_TEXT_P(1);
TSQuery rewrited = query;
MemoryContext outercontext = CurrentMemoryContext;
MemoryContext oldcontext;
QTNode *tree;
char *buf;
SPIPlanPtr plan;
Portal portal;
bool isnull;
int i;
if (query->size == 0)
{
PG_FREE_IF_COPY(in, 1);
PG_RETURN_POINTER(rewrited);
}
tree = QT2QTN(GETQUERY(query), GETOPERAND(query));
QTNTernary(tree);
QTNSort(tree);
buf = text_to_cstring(in);
SPI_connect();
if ((plan = SPI_prepare(buf, 0, NULL)) == NULL)
elog(ERROR, "SPI_prepare(\"%s\") failed", buf);
if ((portal = SPI_cursor_open(NULL, plan, NULL, NULL, true)) == NULL)
elog(ERROR, "SPI_cursor_open(\"%s\") failed", buf);
SPI_cursor_fetch(portal, true, 100);
if (SPI_tuptable == NULL ||
SPI_tuptable->tupdesc->natts != 2 ||
SPI_gettypeid(SPI_tuptable->tupdesc, 1) != TSQUERYOID ||
SPI_gettypeid(SPI_tuptable->tupdesc, 2) != TSQUERYOID)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("ts_rewrite query must return two tsquery columns")));
while (SPI_processed > 0 && tree)
{
for (i = 0; i < SPI_processed && tree; i++)
{
Datum qdata = SPI_getbinval(SPI_tuptable->vals[i], SPI_tuptable->tupdesc, 1, &isnull);
Datum sdata;
if (isnull)
continue;
sdata = SPI_getbinval(SPI_tuptable->vals[i], SPI_tuptable->tupdesc, 2, &isnull);
if (!isnull)
{
TSQuery qtex = DatumGetTSQuery(qdata);
TSQuery qtsubs = DatumGetTSQuery(sdata);
QTNode *qex,
*qsubs = NULL;
if (qtex->size == 0)
{
if (qtex != (TSQuery) DatumGetPointer(qdata))
pfree(qtex);
if (qtsubs != (TSQuery) DatumGetPointer(sdata))
pfree(qtsubs);
continue;
}
qex = QT2QTN(GETQUERY(qtex), GETOPERAND(qtex));
QTNTernary(qex);
QTNSort(qex);
if (qtsubs->size)
qsubs = QT2QTN(GETQUERY(qtsubs), GETOPERAND(qtsubs));
oldcontext = MemoryContextSwitchTo(outercontext);
tree = findsubquery(tree, qex, qsubs, NULL);
MemoryContextSwitchTo(oldcontext);
QTNFree(qex);
if (qtex != (TSQuery) DatumGetPointer(qdata))
pfree(qtex);
QTNFree(qsubs);
if (qtsubs != (TSQuery) DatumGetPointer(sdata))
pfree(qtsubs);
if (tree)
{
/* ready the tree for another pass */
QTNClearFlags(tree, QTN_NOCHANGE);
QTNSort(tree);
}
}
}
SPI_freetuptable(SPI_tuptable);
SPI_cursor_fetch(portal, true, 100);
}
SPI_freetuptable(SPI_tuptable);
SPI_cursor_close(portal);
SPI_freeplan(plan);
SPI_finish();
if (tree)
{
QTNBinary(tree);
rewrited = QTN2QT(tree);
QTNFree(tree);
PG_FREE_IF_COPY(query, 0);
}
else
{
SET_VARSIZE(rewrited, HDRSIZETQ);
rewrited->size = 0;
}
pfree(buf);
PG_FREE_IF_COPY(in, 1);
PG_RETURN_POINTER(rewrited);
}
/*
* pgstrom_create_param_buffer
*
* It construct a param-buffer on the shared memory segment, according to
* the supplied Const/Param list. Its initial reference counter is 1, so
* this buffer can be released using pgstrom_put_param_buffer().
*/
kern_parambuf *
pgstrom_create_kern_parambuf(List *used_params,
ExprContext *econtext)
{
StringInfoData str;
kern_parambuf *kpbuf;
char padding[STROMALIGN_LEN];
ListCell *cell;
Size offset;
int index = 0;
int nparams = list_length(used_params);
/* seek to the head of variable length field */
offset = STROMALIGN(offsetof(kern_parambuf, poffset[nparams]));
initStringInfo(&str);
enlargeStringInfo(&str, offset);
memset(str.data, 0, offset);
str.len = offset;
/* walks on the Para/Const list */
foreach (cell, used_params)
{
Node *node = lfirst(cell);
if (IsA(node, Const))
{
Const *con = (Const *) node;
kpbuf = (kern_parambuf *)str.data;
if (con->constisnull)
kpbuf->poffset[index] = 0; /* null */
else
{
kpbuf->poffset[index] = str.len;
if (con->constlen > 0)
appendBinaryStringInfo(&str,
(char *)&con->constvalue,
con->constlen);
else
appendBinaryStringInfo(&str,
DatumGetPointer(con->constvalue),
VARSIZE(con->constvalue));
}
}
else if (IsA(node, Param))
{
ParamListInfo param_info = econtext->ecxt_param_list_info;
Param *param = (Param *) node;
if (param_info &&
param->paramid > 0 && param->paramid <= param_info->numParams)
{
ParamExternData *prm = ¶m_info->params[param->paramid - 1];
/* give hook a chance in case parameter is dynamic */
if (!OidIsValid(prm->ptype) && param_info->paramFetch != NULL)
(*param_info->paramFetch) (param_info, param->paramid);
kpbuf = (kern_parambuf *)str.data;
if (!OidIsValid(prm->ptype))
{
elog(INFO, "debug: Param has no particular data type");
kpbuf->poffset[index++] = 0; /* null */
continue;
}
/* safety check in case hook did something unexpected */
if (prm->ptype != param->paramtype)
ereport(ERROR,
(errcode(ERRCODE_DATATYPE_MISMATCH),
errmsg("type of parameter %d (%s) does not match that when preparing the plan (%s)",
param->paramid,
format_type_be(prm->ptype),
format_type_be(param->paramtype))));
if (prm->isnull)
kpbuf->poffset[index] = 0; /* null */
else
{
int typlen = get_typlen(prm->ptype);
if (typlen == 0)
elog(ERROR, "cache lookup failed for type %u",
prm->ptype);
if (typlen > 0)
appendBinaryStringInfo(&str,
(char *)&prm->value,
typlen);
else
appendBinaryStringInfo(&str,
DatumGetPointer(prm->value),
VARSIZE(prm->value));
}
}
}
else
elog(ERROR, "unexpected node: %s", nodeToString(node));
/* alignment */
if (STROMALIGN(str.len) != str.len)
appendBinaryStringInfo(&str, padding,
STROMALIGN(str.len) - str.len);
index++;
}
Datum
gtsvector_compress(PG_FUNCTION_ARGS)
{
GISTENTRY *entry = (GISTENTRY *) PG_GETARG_POINTER(0);
GISTENTRY *retval = entry;
if (entry->leafkey)
{ /* tsvector */
SignTSVector *res;
TSVector val = DatumGetTSVector(entry->key);
int32 len;
int32 *arr;
WordEntry *ptr = ARRPTR(val);
char *words = STRPTR(val);
len = CALCGTSIZE(ARRKEY, val->size);
res = (SignTSVector *) palloc(len);
SET_VARSIZE(res, len);
res->flag = ARRKEY;
arr = GETARR(res);
len = val->size;
while (len--)
{
pg_crc32 c;
INIT_LEGACY_CRC32(c);
COMP_LEGACY_CRC32(c, words + ptr->pos, ptr->len);
FIN_LEGACY_CRC32(c);
*arr = *(int32 *) &c;
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 = (SignTSVector *) repalloc((void *) res, len);
SET_VARSIZE(res, len);
}
/* make signature, if array is too long */
if (VARSIZE(res) > TOAST_INDEX_TARGET)
{
SignTSVector *ressign;
len = CALCGTSIZE(SIGNKEY, 0);
ressign = (SignTSVector *) 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)))
{
int32 i,
len;
SignTSVector *res;
BITVECP sign = GETSIGN(DatumGetPointer(entry->key));
LOOPBYTE
{
if ((sign[i] & 0xff) != 0xff)
PG_RETURN_POINTER(retval);
}
len = CALCGTSIZE(SIGNKEY | ALLISTRUE, 0);
res = (SignTSVector *) palloc(len);
SET_VARSIZE(res, len);
res->flag = SIGNKEY | ALLISTRUE;
retval = (GISTENTRY *) palloc(sizeof(GISTENTRY));
gistentryinit(*retval, PointerGetDatum(res),
entry->rel, entry->page,
entry->offset, FALSE);
}
/*
* Turn a composite / record into JSON.
*/
static void
composite_to_json(Datum composite, StringInfo result, bool use_line_feeds)
{
HeapTupleHeader td;
Oid tupType;
int32 tupTypmod;
TupleDesc tupdesc;
HeapTupleData tmptup, *tuple;
int i;
bool needsep = false;
char *sep;
sep = use_line_feeds ? ",\n " : ",";
td = DatumGetHeapTupleHeader(composite);
/* Extract rowtype info and find a tupdesc */
tupType = HeapTupleHeaderGetTypeId(td);
tupTypmod = HeapTupleHeaderGetTypMod(td);
tupdesc = lookup_rowtype_tupdesc(tupType, tupTypmod);
/* Build a temporary HeapTuple control structure */
tmptup.t_len = HeapTupleHeaderGetDatumLength(td);
tmptup.t_data = td;
tuple = &tmptup;
appendStringInfoChar(result,'{');
for (i = 0; i < tupdesc->natts; i++)
{
Datum val, origval;
bool isnull;
char *attname;
TYPCATEGORY tcategory;
Oid typoutput;
bool typisvarlena;
if (tupdesc->attrs[i]->attisdropped)
continue;
if (needsep)
appendStringInfoString(result,sep);
needsep = true;
attname = NameStr(tupdesc->attrs[i]->attname);
escape_json(result,attname);
appendStringInfoChar(result,':');
origval = heap_getattr(tuple, i + 1, tupdesc, &isnull);
if (tupdesc->attrs[i]->atttypid == RECORDARRAYOID)
tcategory = TYPCATEGORY_ARRAY;
else if (tupdesc->attrs[i]->atttypid == RECORDOID)
tcategory = TYPCATEGORY_COMPOSITE;
else
tcategory = TypeCategory(tupdesc->attrs[i]->atttypid);
getTypeOutputInfo(tupdesc->attrs[i]->atttypid,
&typoutput, &typisvarlena);
/*
* If we have a toasted datum, forcibly detoast it here to avoid memory
* leakage inside the type's output routine.
*/
if (typisvarlena && ! isnull)
val = PointerGetDatum(PG_DETOAST_DATUM(origval));
else
val = origval;
datum_to_json(val, result, tcategory, typoutput);
/* Clean up detoasted copy, if any */
if (val != origval)
pfree(DatumGetPointer(val));
}
appendStringInfoChar(result,'}');
ReleaseTupleDesc(tupdesc);
}
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);
}
/*
** 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);
}
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);
);
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);
}
/*
* Fetch dictionary cache entry
*/
TSDictionaryCacheEntry *
lookup_ts_dictionary_cache(Oid dictId)
{
TSDictionaryCacheEntry *entry;
if (TSDictionaryCacheHash == NULL)
{
/* First time through: initialize the hash table */
HASHCTL ctl;
MemSet(&ctl, 0, sizeof(ctl));
ctl.keysize = sizeof(Oid);
ctl.entrysize = sizeof(TSDictionaryCacheEntry);
TSDictionaryCacheHash = hash_create("Tsearch dictionary cache", 8,
&ctl, HASH_ELEM | HASH_BLOBS);
/* Flush cache on pg_ts_dict and pg_ts_template changes */
CacheRegisterSyscacheCallback(TSDICTOID, InvalidateTSCacheCallBack,
PointerGetDatum(TSDictionaryCacheHash));
CacheRegisterSyscacheCallback(TSTEMPLATEOID, InvalidateTSCacheCallBack,
PointerGetDatum(TSDictionaryCacheHash));
/* Also make sure CacheMemoryContext exists */
if (!CacheMemoryContext)
CreateCacheMemoryContext();
}
/* Check single-entry cache */
if (lastUsedDictionary && lastUsedDictionary->dictId == dictId &&
lastUsedDictionary->isvalid)
return lastUsedDictionary;
/* Try to look up an existing entry */
entry = (TSDictionaryCacheEntry *) hash_search(TSDictionaryCacheHash,
(void *) &dictId,
HASH_FIND, NULL);
if (entry == NULL || !entry->isvalid)
{
/*
* If we didn't find one, we want to make one. But first look up the
* object to be sure the OID is real.
*/
HeapTuple tpdict,
tptmpl;
Form_pg_ts_dict dict;
Form_pg_ts_template ctemplate;
MemoryContext saveCtx;
tpdict = SearchSysCache1(TSDICTOID, ObjectIdGetDatum(dictId));
if (!HeapTupleIsValid(tpdict))
elog(ERROR, "cache lookup failed for text search dictionary %u",
dictId);
dict = (Form_pg_ts_dict) GETSTRUCT(tpdict);
/*
* Sanity checks
*/
if (!OidIsValid(dict->dicttemplate))
elog(ERROR, "text search dictionary %u has no ctemplate", dictId);
/*
* Retrieve dictionary's ctemplate
*/
tptmpl = SearchSysCache1(TSTEMPLATEOID,
ObjectIdGetDatum(dict->dicttemplate));
if (!HeapTupleIsValid(tptmpl))
elog(ERROR, "cache lookup failed for text search ctemplate %u",
dict->dicttemplate);
ctemplate = (Form_pg_ts_template) GETSTRUCT(tptmpl);
/*
* Sanity checks
*/
if (!OidIsValid(ctemplate->tmpllexize))
elog(ERROR, "text search ctemplate %u has no lexize method",
ctemplate->tmpllexize);
if (entry == NULL)
{
bool found;
/* Now make the cache entry */
entry = (TSDictionaryCacheEntry *)
hash_search(TSDictionaryCacheHash,
(void *) &dictId,
HASH_ENTER, &found);
Assert(!found); /* it wasn't there a moment ago */
/* Create private___ memory context the first time through */
saveCtx = AllocSetContextCreate(CacheMemoryContext,
NameStr(dict->dictname),
ALLOCSET_SMALL_MINSIZE,
ALLOCSET_SMALL_INITSIZE,
ALLOCSET_SMALL_MAXSIZE);
}
else
{
/* Clear the existing entry's private___ context */
saveCtx = entry->dictCtx;
MemoryContextResetAndDeleteChildren(saveCtx);
}
MemSet(entry, 0, sizeof(TSDictionaryCacheEntry));
entry->dictId = dictId;
entry->dictCtx = saveCtx;
entry->lexizeOid = ctemplate->tmpllexize;
if (OidIsValid(ctemplate->tmplinit))
{
List *dictoptions;
Datum opt;
bool isnull;
MemoryContext oldcontext;
/*
* Init method runs in dictionary's private___ memory context, and we
* make sure the options are stored there too
*/
oldcontext = MemoryContextSwitchTo(entry->dictCtx);
opt = SysCacheGetAttr(TSDICTOID, tpdict,
Anum_pg_ts_dict_dictinitoption,
&isnull);
if (isnull)
dictoptions = NIL;
else
dictoptions = deserialize_deflist(opt);
entry->dictData =
DatumGetPointer(OidFunctionCall1(ctemplate->tmplinit,
PointerGetDatum(dictoptions)));
MemoryContextSwitchTo(oldcontext);
}
ReleaseSysCache(tptmpl);
ReleaseSysCache(tpdict);
fmgr_info_cxt(entry->lexizeOid, &entry->lexize, entry->dictCtx);
entry->isvalid = true;
}
lastUsedDictionary = entry;
return entry;
}
/*
** The GiST PickSplit method for segments
** We use Guttman's poly time split algorithm
*/
GIST_SPLITVEC *
gseg_picksplit(GistEntryVector *entryvec,
GIST_SPLITVEC *v)
{
OffsetNumber i,
j;
SEG *datum_alpha,
*datum_beta;
SEG *datum_l,
*datum_r;
SEG *union_d,
*union_dl,
*union_dr;
SEG *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 = 1,
seed_2 = 2;
OffsetNumber *left,
*right;
OffsetNumber maxoff;
#ifdef GIST_DEBUG
fprintf(stderr, "picksplit\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 = (SEG *) DatumGetPointer(entryvec->vector[i].key);
for (j = OffsetNumberNext(i); j <= maxoff; j = OffsetNumberNext(j))
{
datum_beta = (SEG *) DatumGetPointer(entryvec->vector[j].key);
/* compute the wasted space by unioning these guys */
/* size_waste = size_union - size_inter; */
union_d = seg_union(datum_alpha, datum_beta);
rt_seg_size(union_d, &size_union);
inter_d = seg_inter(datum_alpha, datum_beta);
rt_seg_size(inter_d, &size_inter);
size_waste = size_union - size_inter;
/*
* 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;
datum_alpha = (SEG *) DatumGetPointer(entryvec->vector[seed_1].key);
datum_l = seg_union(datum_alpha, datum_alpha);
rt_seg_size(datum_l, &size_l);
datum_beta = (SEG *) DatumGetPointer(entryvec->vector[seed_2].key);
datum_r = seg_union(datum_beta, datum_beta);
rt_seg_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 = (SEG *) DatumGetPointer(entryvec->vector[i].key);
union_dl = seg_union(datum_l, datum_alpha);
union_dr = seg_union(datum_r, datum_alpha);
rt_seg_size(union_dl, &size_alpha);
rt_seg_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_alpha;
*right++ = i;
v->spl_nright++;
}
}
*left = *right = FirstOffsetNumber; /* sentinel value, see dosplit() */
v->spl_ldatum = PointerGetDatum(datum_l);
v->spl_rdatum = PointerGetDatum(datum_r);
return v;
}
/*
* compute_tsvector_stats() -- compute statistics for a tsvector column
*
* This functions computes statistics that are useful for determining @@
* operations' selectivity, along with the fraction of non-null rows and
* average width.
*
* Instead of finding the most common values, as we do for most datatypes,
* we're looking for the most common lexemes. This is more useful, because
* there most probably won't be any two rows with the same tsvector and thus
* the notion of a MCV is a bit bogus with this datatype. With a list of the
* most common lexemes we can do a better job at figuring out @@ selectivity.
*
* For the same reasons we assume that tsvector columns are unique when
* determining the number of distinct values.
*
* The algorithm used is Lossy Counting, as proposed in the paper "Approximate
* frequency counts over data streams" by G. S. Manku and R. Motwani, in
* Proceedings of the 28th International Conference on Very Large Data Bases,
* Hong Kong, China, August 2002, section 4.2. The paper is available at
* http://www.vldb.org/conf/2002/S10P03.pdf
*
* The Lossy Counting (aka LC) algorithm goes like this:
* Let D be a set of triples (e, f, d), where e is an element value, f is
* that element's frequency (occurrence count) and d is the maximum error in
* f. We start with D empty and process the elements in batches of size
* w. (The batch size is also known as "bucket size".) Let the current batch
* number be b_current, starting with 1. For each element e we either
* increment its f count, if it's already in D, or insert a new triple into D
* with values (e, 1, b_current - 1). After processing each batch we prune D,
* by removing from it all elements with f + d <= b_current. Finally, we
* gather elements with largest f. The LC paper proves error bounds on f
* dependent on the batch size w, and shows that the required table size
* is no more than a few times w.
*
* We use a hashtable for the D structure and a bucket width of
* statistics_target * 10, where 10 is an arbitrarily chosen constant,
* meant to approximate the number of lexemes in a single tsvector.
*/
static void
compute_tsvector_stats(VacAttrStats *stats,
AnalyzeAttrFetchFunc fetchfunc,
int samplerows,
double totalrows)
{
int num_mcelem;
int null_cnt = 0;
double total_width = 0;
/* This is D from the LC algorithm. */
HTAB *lexemes_tab;
HASHCTL hash_ctl;
HASH_SEQ_STATUS scan_status;
/* This is the current bucket number from the LC algorithm */
int b_current;
/* This is 'w' from the LC algorithm */
int bucket_width;
int vector_no,
lexeme_no;
LexemeHashKey hash_key;
TrackItem *item;
/* We want statistics_target * 10 lexemes in the MCELEM array */
num_mcelem = stats->attr->attstattarget * 10;
/*
* We set bucket width equal to the target number of result lexemes. This
* is probably about right but perhaps might need to be scaled up or down
* a bit?
*/
bucket_width = num_mcelem;
/*
* Create the hashtable. It will be in local memory, so we don't need to
* worry about initial size too much. Also we don't need to pay any
* attention to locking and memory management.
*/
MemSet(&hash_ctl, 0, sizeof(hash_ctl));
hash_ctl.keysize = sizeof(LexemeHashKey);
hash_ctl.entrysize = sizeof(TrackItem);
hash_ctl.hash = lexeme_hash;
hash_ctl.match = lexeme_match;
hash_ctl.hcxt = CurrentMemoryContext;
lexemes_tab = hash_create("Analyzed lexemes table",
bucket_width * 4,
&hash_ctl,
HASH_ELEM | HASH_FUNCTION | HASH_COMPARE | HASH_CONTEXT);
/* Initialize counters. */
b_current = 1;
lexeme_no = 1;
/* Loop over the tsvectors. */
for (vector_no = 0; vector_no < samplerows; vector_no++)
{
Datum value;
bool isnull;
TSVector vector;
WordEntry *curentryptr;
char *lexemesptr;
int j;
vacuum_delay_point();
value = fetchfunc(stats, vector_no, &isnull);
/*
* Check for null/nonnull.
*/
if (isnull)
{
null_cnt++;
continue;
}
/*
* Add up widths for average-width calculation. Since it's a
* tsvector, we know it's varlena. As in the regular
* compute_minimal_stats function, we use the toasted width for this
* calculation.
*/
total_width += VARSIZE_ANY(DatumGetPointer(value));
/*
* Now detoast the tsvector if needed.
*/
vector = DatumGetTSVector(value);
/*
* We loop through the lexemes in the tsvector and add them to our
* tracking hashtable. Note: the hashtable entries will point into
* the (detoasted) tsvector value, therefore we cannot free that
* storage until we're done.
*/
lexemesptr = STRPTR(vector);
curentryptr = ARRPTR(vector);
for (j = 0; j < vector->size; j++)
{
bool found;
/* Construct a hash key */
hash_key.lexeme = lexemesptr + curentryptr->pos;
hash_key.length = curentryptr->len;
/* Lookup current lexeme in hashtable, adding it if new */
item = (TrackItem *) hash_search(lexemes_tab,
(const void *) &hash_key,
HASH_ENTER, &found);
if (found)
{
/* The lexeme is already on the tracking list */
item->frequency++;
}
else
{
/* Initialize new tracking list element */
item->frequency = 1;
item->delta = b_current - 1;
}
/* We prune the D structure after processing each bucket */
if (lexeme_no % bucket_width == 0)
{
prune_lexemes_hashtable(lexemes_tab, b_current);
b_current++;
}
/* Advance to the next WordEntry in the tsvector */
lexeme_no++;
curentryptr++;
}
}
/* We can only compute real stats if we found some non-null values. */
if (null_cnt < samplerows)
{
int nonnull_cnt = samplerows - null_cnt;
int i;
TrackItem **sort_table;
int track_len;
int minfreq,
maxfreq;
stats->stats_valid = true;
/* Do the simple null-frac and average width stats */
stats->stanullfrac = (double) null_cnt / (double) samplerows;
stats->stawidth = total_width / (double) nonnull_cnt;
/* Assume it's a unique column (see notes above) */
stats->stadistinct = -1.0;
/*
* Determine the top-N lexemes by simply copying pointers from the
* hashtable into an array and applying qsort()
*/
track_len = hash_get_num_entries(lexemes_tab);
sort_table = (TrackItem **) palloc(sizeof(TrackItem *) * track_len);
hash_seq_init(&scan_status, lexemes_tab);
i = 0;
while ((item = (TrackItem *) hash_seq_search(&scan_status)) != NULL)
{
sort_table[i++] = item;
}
Assert(i == track_len);
qsort(sort_table, track_len, sizeof(TrackItem *),
trackitem_compare_frequencies_desc);
/* Suppress any single-occurrence items */
while (track_len > 0)
{
if (sort_table[track_len - 1]->frequency > 1)
break;
track_len--;
}
/* Determine the number of most common lexemes to be stored */
if (num_mcelem > track_len)
num_mcelem = track_len;
/* Generate MCELEM slot entry */
if (num_mcelem > 0)
{
MemoryContext old_context;
Datum *mcelem_values;
float4 *mcelem_freqs;
/* Grab the minimal and maximal frequencies that will get stored */
minfreq = sort_table[num_mcelem - 1]->frequency;
maxfreq = sort_table[0]->frequency;
/*
* We want to store statistics sorted on the lexeme value using
* first length, then byte-for-byte comparison. The reason for
* doing length comparison first is that we don't care about the
* ordering so long as it's consistent, and comparing lengths
* first gives us a chance to avoid a strncmp() call.
*
* This is different from what we do with scalar statistics --
* they get sorted on frequencies. The rationale is that we
* usually search through most common elements looking for a
* specific value, so we can grab its frequency. When values are
* presorted we can employ binary search for that. See
* ts_selfuncs.c for a real usage scenario.
*/
qsort(sort_table, num_mcelem, sizeof(TrackItem *),
trackitem_compare_lexemes);
/* Must copy the target values into anl_context */
old_context = MemoryContextSwitchTo(stats->anl_context);
/*
* We sorted statistics on the lexeme value, but we want to be
* able to find out the minimal and maximal frequency without
* going through all the values. We keep those two extra
* frequencies in two extra cells in mcelem_freqs.
*/
mcelem_values = (Datum *) palloc(num_mcelem * sizeof(Datum));
mcelem_freqs = (float4 *) palloc((num_mcelem + 2) * sizeof(float4));
for (i = 0; i < num_mcelem; i++)
{
TrackItem *item = sort_table[i];
mcelem_values[i] =
PointerGetDatum(cstring_to_text_with_len(item->key.lexeme,
item->key.length));
mcelem_freqs[i] = (double) item->frequency / (double) nonnull_cnt;
}
mcelem_freqs[i++] = (double) minfreq / (double) nonnull_cnt;
mcelem_freqs[i] = (double) maxfreq / (double) nonnull_cnt;
MemoryContextSwitchTo(old_context);
stats->stakind[0] = STATISTIC_KIND_MCELEM;
stats->staop[0] = TextEqualOperator;
stats->stanumbers[0] = mcelem_freqs;
/* See above comment about two extra frequency fields */
stats->numnumbers[0] = num_mcelem + 2;
stats->stavalues[0] = mcelem_values;
stats->numvalues[0] = num_mcelem;
/* We are storing text values */
stats->statypid[0] = TEXTOID;
stats->statyplen[0] = -1; /* typlen, -1 for varlena */
stats->statypbyval[0] = false;
stats->statypalign[0] = 'i';
}
}
else
{
/* We found only nulls; assume the column is entirely null */
stats->stats_valid = true;
stats->stanullfrac = 1.0;
stats->stawidth = 0; /* "unknown" */
stats->stadistinct = 0.0; /* "unknown" */
}
/*
* We don't need to bother cleaning up any of our temporary palloc's. The
* hashtable should also go away, as it used a child memory context.
*/
}
static TSLexeme *
LexizeExec(LexizeData *ld, ParsedLex **correspondLexem)
{
int i;
ListDictionary *map;
TSDictionaryCacheEntry *dict;
TSLexeme *res;
if (ld->curDictId == InvalidOid)
{
/*
* usial mode: dictionary wants only one word, but we should keep in
* mind that we should go through all stack
*/
while (ld->towork.head)
{
ParsedLex *curVal = ld->towork.head;
char *curValLemm = curVal->lemm;
int curValLenLemm = curVal->lenlemm;
map = ld->cfg->map + curVal->type;
if (curVal->type == 0 || curVal->type >= ld->cfg->lenmap || map->len == 0)
{
/* skip this type of lexeme */
RemoveHead(ld);
continue;
}
for (i = ld->posDict; i < map->len; i++)
{
dict = lookup_ts_dictionary_cache(map->dictIds[i]);
ld->dictState.isend = ld->dictState.getnext = false;
ld->dictState.private_state = NULL;
res = (TSLexeme *) DatumGetPointer(FunctionCall4(
&(dict->lexize),
PointerGetDatum(dict->dictData),
PointerGetDatum(curValLemm),
Int32GetDatum(curValLenLemm),
PointerGetDatum(&ld->dictState)
));
if (ld->dictState.getnext)
{
/*
* dictionary wants next word, so setup and store current
* position and go to multiword mode
*/
ld->curDictId = DatumGetObjectId(map->dictIds[i]);
ld->posDict = i + 1;
ld->curSub = curVal->next;
if (res)
setNewTmpRes(ld, curVal, res);
return LexizeExec(ld, correspondLexem);
}
if (!res) /* dictionary doesn't know this lexeme */
continue;
if (res->flags & TSL_FILTER)
{
curValLemm = res->lexeme;
curValLenLemm = strlen(res->lexeme);
continue;
}
RemoveHead(ld);
setCorrLex(ld, correspondLexem);
return res;
}
RemoveHead(ld);
}
}
else
{ /* curDictId is valid */
dict = lookup_ts_dictionary_cache(ld->curDictId);
/*
* Dictionary ld->curDictId asks us about following words
*/
while (ld->curSub)
{
ParsedLex *curVal = ld->curSub;
map = ld->cfg->map + curVal->type;
if (curVal->type != 0)
{
bool dictExists = false;
if (curVal->type >= ld->cfg->lenmap || map->len == 0)
{
/* skip this type of lexeme */
ld->curSub = curVal->next;
continue;
}
/*
* We should be sure that current type of lexeme is recognized
* by our dictinonary: we just check is it exist in list of
* dictionaries ?
*/
for (i = 0; i < map->len && !dictExists; i++)
if (ld->curDictId == DatumGetObjectId(map->dictIds[i]))
dictExists = true;
if (!dictExists)
{
/*
* Dictionary can't work with current tpe of lexeme,
* return to basic mode and redo all stored lexemes
*/
ld->curDictId = InvalidOid;
return LexizeExec(ld, correspondLexem);
}
}
ld->dictState.isend = (curVal->type == 0) ? true : false;
ld->dictState.getnext = false;
res = (TSLexeme *) DatumGetPointer(FunctionCall4(
&(dict->lexize),
PointerGetDatum(dict->dictData),
PointerGetDatum(curVal->lemm),
Int32GetDatum(curVal->lenlemm),
PointerGetDatum(&ld->dictState)
));
if (ld->dictState.getnext)
{
/* Dictionary wants one more */
ld->curSub = curVal->next;
if (res)
setNewTmpRes(ld, curVal, res);
continue;
}
if (res || ld->tmpRes)
{
/*
* Dictionary normalizes lexemes, so we remove from stack all
* used lexemes, return to basic mode and redo end of stack
* (if it exists)
*/
if (res)
{
moveToWaste(ld, ld->curSub);
}
else
{
res = ld->tmpRes;
moveToWaste(ld, ld->lastRes);
}
/* reset to initial state */
ld->curDictId = InvalidOid;
ld->posDict = 0;
ld->lastRes = NULL;
ld->tmpRes = NULL;
setCorrLex(ld, correspondLexem);
return res;
}
/*
* Dict don't want next lexem and didn't recognize anything, redo
* from ld->towork.head
*/
ld->curDictId = InvalidOid;
return LexizeExec(ld, correspondLexem);
}
}
setCorrLex(ld, correspondLexem);
return NULL;
}
Datum
_numeric_weighted_mean_intermediate(PG_FUNCTION_ARGS)
{
WeightedMeanInternalState *state;
Datum value,
weight,
temp_total,
old_sum,
old_weight;
MemoryContext aggcontext,
oldcontext;
if (!AggCheckCallContext(fcinfo, &aggcontext))
/* cannot be called directly because of internal-type argument */
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("_numeric_weighted_mean_intermediate called in non-aggregate context")));
if (PG_ARGISNULL(0))
{
oldcontext = MemoryContextSwitchTo(aggcontext);
state = (WeightedMeanInternalState *) palloc(sizeof(WeightedMeanInternalState));
state->running_sum = make_numeric(0);
state->running_weight = make_numeric(0);
MemoryContextSwitchTo(oldcontext);
}
else
state = (WeightedMeanInternalState *) PG_GETARG_POINTER(0);
/*
* We're non-strict, so we MUST check args for nullity ourselves before
* using them. To preserve the behaviour of null inputs, we skip updating
* on them.
*/
if (PG_ARGISNULL(1) || PG_ARGISNULL(2))
PG_RETURN_POINTER(state);
/*
* We fetch and process the input in the shortlived calling context to
* avoid leaking memory in aggcontext per cycle. We force the input to be
* detoasted here, too, in the shortlived context. (PG_GETARG_DATUM does
* not detoast, but PG_GETARG_NUMERIC does.)
*/
value = NumericGetDatum(PG_GETARG_NUMERIC(1));
weight = NumericGetDatum(PG_GETARG_NUMERIC(2));
temp_total = DirectFunctionCall2(numeric_mul, value, weight);
/*
* The new running totals must be allocated in the long-lived context. We
* rely on the numeric_* functions to clean up after themselves (which they
* currently do, but only if the input is already detoasted); we could play
* safe and copy only the final results into aggcontext, but this turns out
* to have a measurable performance hit.
*/
oldcontext = MemoryContextSwitchTo(aggcontext);
old_sum = state->running_sum;
old_weight = state->running_weight;
state->running_sum = DirectFunctionCall2(numeric_add,
state->running_sum, temp_total);
state->running_weight = DirectFunctionCall2(numeric_add,
state->running_weight, weight);
pfree(DatumGetPointer(old_sum));
pfree(DatumGetPointer(old_weight));
MemoryContextSwitchTo(oldcontext);
PG_RETURN_POINTER(state);
}
/*
* Parse string and lexize words.
*
* prs will be filled in.
*/
void
parsetext(Oid cfgId, ParsedText *prs, char *buf, int buflen)
{
int type,
lenlemm;
char *lemm = NULL;
LexizeData ldata;
TSLexeme *norms;
TSConfigCacheEntry *cfg;
TSParserCacheEntry *prsobj;
void *prsdata;
cfg = lookup_ts_config_cache(cfgId);
prsobj = lookup_ts_parser_cache(cfg->prsId);
prsdata = (void *) DatumGetPointer(FunctionCall2(&prsobj->prsstart,
PointerGetDatum(buf),
Int32GetDatum(buflen)));
LexizeInit(&ldata, cfg);
do
{
type = DatumGetInt32(FunctionCall3(&(prsobj->prstoken),
PointerGetDatum(prsdata),
PointerGetDatum(&lemm),
PointerGetDatum(&lenlemm)));
if (type > 0 && lenlemm >= MAXSTRLEN)
{
#ifdef IGNORE_LONGLEXEME
ereport(NOTICE,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("word is too long to be indexed"),
errdetail("Words longer than %d characters are ignored.",
MAXSTRLEN)));
continue;
#else
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("word is too long to be indexed"),
errdetail("Words longer than %d characters are ignored.",
MAXSTRLEN)));
#endif
}
LexizeAddLemm(&ldata, type, lemm, lenlemm);
while ((norms = LexizeExec(&ldata, NULL)) != NULL)
{
TSLexeme *ptr = norms;
prs->pos++; /* set pos */
while (ptr->lexeme)
{
if (prs->curwords == prs->lenwords)
{
prs->lenwords *= 2;
prs->words = (ParsedWord *) repalloc((void *) prs->words, prs->lenwords * sizeof(ParsedWord));
}
if (ptr->flags & TSL_ADDPOS)
prs->pos++;
prs->words[prs->curwords].len = strlen(ptr->lexeme);
prs->words[prs->curwords].word = ptr->lexeme;
prs->words[prs->curwords].nvariant = ptr->nvariant;
prs->words[prs->curwords].flags = ptr->flags & TSL_PREFIX;
prs->words[prs->curwords].alen = 0;
prs->words[prs->curwords].pos.pos = LIMITPOS(prs->pos);
ptr++;
prs->curwords++;
}
pfree(norms);
}
} while (type > 0);
FunctionCall1(&(prsobj->prsend), PointerGetDatum(prsdata));
}
void
ginNewScanKey(IndexScanDesc scan)
{
ScanKey scankey = scan->keyData;
GinScanOpaque so = (GinScanOpaque) scan->opaque;
int i;
bool hasNullQuery = false;
/* if no scan keys provided, allocate extra EVERYTHING GinScanKey */
so->keys = (GinScanKey)
palloc(Max(scan->numberOfKeys, 1) * sizeof(GinScanKeyData));
so->nkeys = 0;
/* initialize expansible array of GinScanEntry pointers */
so->totalentries = 0;
so->allocentries = 32;
so->entries = (GinScanEntry *)
palloc0(so->allocentries * sizeof(GinScanEntry));
so->isVoidRes = false;
for (i = 0; i < scan->numberOfKeys; i++)
{
ScanKey skey = &scankey[i];
Datum *queryValues;
int32 nQueryValues = 0;
bool *partial_matches = NULL;
Pointer *extra_data = NULL;
bool *nullFlags = NULL;
int32 searchMode = GIN_SEARCH_MODE_DEFAULT;
/*
* We assume that GIN-indexable operators are strict, so a null query
* argument means an unsatisfiable query.
*/
if (skey->sk_flags & SK_ISNULL)
{
so->isVoidRes = true;
break;
}
/* OK to call the extractQueryFn */
queryValues = (Datum *)
DatumGetPointer(FunctionCall7(&so->ginstate.extractQueryFn[skey->sk_attno - 1],
skey->sk_argument,
PointerGetDatum(&nQueryValues),
UInt16GetDatum(skey->sk_strategy),
PointerGetDatum(&partial_matches),
PointerGetDatum(&extra_data),
PointerGetDatum(&nullFlags),
PointerGetDatum(&searchMode)));
/*
* If bogus searchMode is returned, treat as GIN_SEARCH_MODE_ALL; note
* in particular we don't allow extractQueryFn to select
* GIN_SEARCH_MODE_EVERYTHING.
*/
if (searchMode < GIN_SEARCH_MODE_DEFAULT ||
searchMode > GIN_SEARCH_MODE_ALL)
searchMode = GIN_SEARCH_MODE_ALL;
/* Non-default modes require the index to have placeholders */
if (searchMode != GIN_SEARCH_MODE_DEFAULT)
hasNullQuery = true;
/*
* In default mode, no keys means an unsatisfiable query.
*/
if (queryValues == NULL || nQueryValues <= 0)
{
if (searchMode == GIN_SEARCH_MODE_DEFAULT)
{
so->isVoidRes = true;
break;
}
nQueryValues = 0; /* ensure sane value */
}
/*
* If the extractQueryFn didn't create a nullFlags array, create one,
* assuming that everything's non-null. Otherwise, run through the
* array and make sure each value is exactly 0 or 1; this ensures
* binary compatibility with the GinNullCategory representation. While
* at it, detect whether any null keys are present.
*/
if (nullFlags == NULL)
nullFlags = (bool *) palloc0(nQueryValues * sizeof(bool));
else
{
int32 j;
for (j = 0; j < nQueryValues; j++)
{
if (nullFlags[j])
{
nullFlags[j] = true; /* not any other nonzero value */
hasNullQuery = true;
}
}
}
/* now we can use the nullFlags as category codes */
ginFillScanKey(so, skey->sk_attno,
skey->sk_strategy, searchMode,
skey->sk_argument, nQueryValues,
queryValues, (GinNullCategory *) nullFlags,
partial_matches, extra_data);
}
/*
* If there are no regular scan keys, generate an EVERYTHING scankey to
* drive a full-index scan.
*/
if (so->nkeys == 0 && !so->isVoidRes)
{
hasNullQuery = true;
ginFillScanKey(so, FirstOffsetNumber,
InvalidStrategy, GIN_SEARCH_MODE_EVERYTHING,
(Datum) 0, 0,
NULL, NULL, NULL, NULL);
}
/*
* If the index is version 0, it may be missing null and placeholder
* entries, which would render searches for nulls and full-index scans
* unreliable. Throw an error if so.
*/
if (hasNullQuery && !so->isVoidRes)
{
GinStatsData ginStats;
ginGetStats(scan->indexRelation, &ginStats);
if (ginStats.ginVersion < 1)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("old GIN indexes do not support whole-index scans nor searches for nulls"),
errhint("To fix this, do REINDEX INDEX \"%s\".",
RelationGetRelationName(scan->indexRelation))));
}
pgstat_count_index_scan(scan->indexRelation);
}
void
hlparsetext(Oid cfgId, HeadlineParsedText *prs, TSQuery query, char *buf, int buflen)
{
int type,
lenlemm;
char *lemm = NULL;
LexizeData ldata;
TSLexeme *norms;
ParsedLex *lexs;
TSConfigCacheEntry *cfg;
TSParserCacheEntry *prsobj;
void *prsdata;
cfg = lookup_ts_config_cache(cfgId);
prsobj = lookup_ts_parser_cache(cfg->prsId);
prsdata = (void *) DatumGetPointer(FunctionCall2(&(prsobj->prsstart),
PointerGetDatum(buf),
Int32GetDatum(buflen)));
LexizeInit(&ldata, cfg);
do
{
type = DatumGetInt32(FunctionCall3(&(prsobj->prstoken),
PointerGetDatum(prsdata),
PointerGetDatum(&lemm),
PointerGetDatum(&lenlemm)));
if (type > 0 && lenlemm >= MAXSTRLEN)
{
#ifdef IGNORE_LONGLEXEME
ereport(NOTICE,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("word is too long to be indexed"),
errdetail("Words longer than %d characters are ignored.",
MAXSTRLEN)));
continue;
#else
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("word is too long to be indexed"),
errdetail("Words longer than %d characters are ignored.",
MAXSTRLEN)));
#endif
}
LexizeAddLemm(&ldata, type, lemm, lenlemm);
do
{
if ((norms = LexizeExec(&ldata, &lexs)) != NULL)
addHLParsedLex(prs, query, lexs, norms);
else
addHLParsedLex(prs, query, lexs, NULL);
} while (norms);
} while (type > 0);
FunctionCall1(&(prsobj->prsend), PointerGetDatum(prsdata));
}
/* ----------------------------------------------------------------
* ProcedureCreate
*
* Note: allParameterTypes, parameterModes, parameterNames, and proconfig
* are either arrays of the proper types or NULL. We declare them Datum,
* not "ArrayType *", to avoid importing array.h into pg_proc_fn.h.
* ----------------------------------------------------------------
*/
Oid
ProcedureCreate(const char *procedureName,
Oid procNamespace,
bool replace,
bool returnsSet,
Oid returnType,
Oid languageObjectId,
Oid languageValidator,
const char *prosrc,
const char *probin,
bool isAgg,
bool isWindowFunc,
bool security_definer,
bool isStrict,
char volatility,
oidvector *parameterTypes,
Datum allParameterTypes,
Datum parameterModes,
Datum parameterNames,
List *parameterDefaults,
Datum proconfig,
float4 procost,
float4 prorows)
{
Oid retval;
int parameterCount;
int allParamCount;
Oid *allParams;
bool genericInParam = false;
bool genericOutParam = false;
bool internalInParam = false;
bool internalOutParam = false;
Oid variadicType = InvalidOid;
Oid proowner = GetUserId();
Acl *proacl = NULL;
Relation rel;
HeapTuple tup;
HeapTuple oldtup;
bool nulls[Natts_pg_proc];
Datum values[Natts_pg_proc];
bool replaces[Natts_pg_proc];
Oid relid;
NameData procname;
TupleDesc tupDesc;
bool is_update;
ObjectAddress myself,
referenced;
int i;
/*
* sanity checks
*/
Assert(PointerIsValid(prosrc));
parameterCount = parameterTypes->dim1;
if (parameterCount < 0 || parameterCount > FUNC_MAX_ARGS)
ereport(ERROR,
(errcode(ERRCODE_TOO_MANY_ARGUMENTS),
errmsg_plural("functions cannot have more than %d argument",
"functions cannot have more than %d arguments",
FUNC_MAX_ARGS,
FUNC_MAX_ARGS)));
/* note: the above is correct, we do NOT count output arguments */
if (allParameterTypes != PointerGetDatum(NULL))
{
/*
* We expect the array to be a 1-D OID array; verify that. We don't
* need to use deconstruct_array() since the array data is just going
* to look like a C array of OID values.
*/
ArrayType *allParamArray = (ArrayType *) DatumGetPointer(allParameterTypes);
allParamCount = ARR_DIMS(allParamArray)[0];
if (ARR_NDIM(allParamArray) != 1 ||
allParamCount <= 0 ||
ARR_HASNULL(allParamArray) ||
ARR_ELEMTYPE(allParamArray) != OIDOID)
elog(ERROR, "allParameterTypes is not a 1-D Oid array");
allParams = (Oid *) ARR_DATA_PTR(allParamArray);
Assert(allParamCount >= parameterCount);
/* we assume caller got the contents right */
}
else
{
allParamCount = parameterCount;
allParams = parameterTypes->values;
}
/*
* Do not allow polymorphic return type unless at least one input argument
* is polymorphic. Also, do not allow return type INTERNAL unless at
* least one input argument is INTERNAL.
*/
for (i = 0; i < parameterCount; i++)
{
switch (parameterTypes->values[i])
{
case ANYARRAYOID:
case ANYELEMENTOID:
case ANYNONARRAYOID:
case ANYENUMOID:
genericInParam = true;
break;
case INTERNALOID:
internalInParam = true;
break;
}
}
if (allParameterTypes != PointerGetDatum(NULL))
{
for (i = 0; i < allParamCount; i++)
{
/*
* We don't bother to distinguish input and output params here, so
* if there is, say, just an input INTERNAL param then we will
* still set internalOutParam. This is OK since we don't really
* care.
*/
switch (allParams[i])
{
case ANYARRAYOID:
case ANYELEMENTOID:
case ANYNONARRAYOID:
case ANYENUMOID:
genericOutParam = true;
break;
case INTERNALOID:
internalOutParam = true;
break;
}
}
}
if ((IsPolymorphicType(returnType) || genericOutParam)
&& !genericInParam)
ereport(ERROR,
(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
errmsg("cannot determine result data type"),
errdetail("A function returning a polymorphic type must have at least one polymorphic argument.")));
if ((returnType == INTERNALOID || internalOutParam) && !internalInParam)
ereport(ERROR,
(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
errmsg("unsafe use of pseudo-type \"internal\""),
errdetail("A function returning \"internal\" must have at least one \"internal\" argument.")));
/*
* don't allow functions of complex types that have the same name as
* existing attributes of the type
*/
if (parameterCount == 1 &&
OidIsValid(parameterTypes->values[0]) &&
(relid = typeidTypeRelid(parameterTypes->values[0])) != InvalidOid &&
get_attnum(relid, procedureName) != InvalidAttrNumber)
ereport(ERROR,
(errcode(ERRCODE_DUPLICATE_COLUMN),
errmsg("\"%s\" is already an attribute of type %s",
procedureName,
format_type_be(parameterTypes->values[0]))));
if (parameterModes != PointerGetDatum(NULL))
{
/*
* We expect the array to be a 1-D CHAR array; verify that. We don't
* need to use deconstruct_array() since the array data is just going
* to look like a C array of char values.
*/
ArrayType *modesArray = (ArrayType *) DatumGetPointer(parameterModes);
char *modes;
if (ARR_NDIM(modesArray) != 1 ||
ARR_DIMS(modesArray)[0] != allParamCount ||
ARR_HASNULL(modesArray) ||
ARR_ELEMTYPE(modesArray) != CHAROID)
elog(ERROR, "parameterModes is not a 1-D char array");
modes = (char *) ARR_DATA_PTR(modesArray);
/*
* Only the last input parameter can be variadic; if it is, save its
* element type. Errors here are just elog since caller should have
* checked this already.
*/
for (i = 0; i < allParamCount; i++)
{
switch (modes[i])
{
case PROARGMODE_IN:
case PROARGMODE_INOUT:
if (OidIsValid(variadicType))
elog(ERROR, "variadic parameter must be last");
break;
case PROARGMODE_OUT:
case PROARGMODE_TABLE:
/* okay */
break;
case PROARGMODE_VARIADIC:
if (OidIsValid(variadicType))
elog(ERROR, "variadic parameter must be last");
switch (allParams[i])
{
case ANYOID:
variadicType = ANYOID;
break;
case ANYARRAYOID:
variadicType = ANYELEMENTOID;
break;
default:
variadicType = get_element_type(allParams[i]);
if (!OidIsValid(variadicType))
elog(ERROR, "variadic parameter is not an array");
break;
}
break;
default:
elog(ERROR, "invalid parameter mode '%c'", modes[i]);
break;
}
}
}
/*
* All seems OK; prepare the data to be inserted into pg_proc.
*/
for (i = 0; i < Natts_pg_proc; ++i)
{
nulls[i] = false;
values[i] = (Datum) 0;
replaces[i] = true;
}
namestrcpy(&procname, procedureName);
values[Anum_pg_proc_proname - 1] = NameGetDatum(&procname);
values[Anum_pg_proc_pronamespace - 1] = ObjectIdGetDatum(procNamespace);
values[Anum_pg_proc_proowner - 1] = ObjectIdGetDatum(proowner);
values[Anum_pg_proc_prolang - 1] = ObjectIdGetDatum(languageObjectId);
values[Anum_pg_proc_procost - 1] = Float4GetDatum(procost);
values[Anum_pg_proc_prorows - 1] = Float4GetDatum(prorows);
values[Anum_pg_proc_provariadic - 1] = ObjectIdGetDatum(variadicType);
values[Anum_pg_proc_proisagg - 1] = BoolGetDatum(isAgg);
values[Anum_pg_proc_proiswindow - 1] = BoolGetDatum(isWindowFunc);
values[Anum_pg_proc_prosecdef - 1] = BoolGetDatum(security_definer);
values[Anum_pg_proc_proisstrict - 1] = BoolGetDatum(isStrict);
values[Anum_pg_proc_proretset - 1] = BoolGetDatum(returnsSet);
values[Anum_pg_proc_provolatile - 1] = CharGetDatum(volatility);
values[Anum_pg_proc_pronargs - 1] = UInt16GetDatum(parameterCount);
values[Anum_pg_proc_pronargdefaults - 1] = UInt16GetDatum(list_length(parameterDefaults));
values[Anum_pg_proc_prorettype - 1] = ObjectIdGetDatum(returnType);
values[Anum_pg_proc_proargtypes - 1] = PointerGetDatum(parameterTypes);
if (allParameterTypes != PointerGetDatum(NULL))
values[Anum_pg_proc_proallargtypes - 1] = allParameterTypes;
else
nulls[Anum_pg_proc_proallargtypes - 1] = true;
if (parameterModes != PointerGetDatum(NULL))
values[Anum_pg_proc_proargmodes - 1] = parameterModes;
else
nulls[Anum_pg_proc_proargmodes - 1] = true;
if (parameterNames != PointerGetDatum(NULL))
values[Anum_pg_proc_proargnames - 1] = parameterNames;
else
nulls[Anum_pg_proc_proargnames - 1] = true;
if (parameterDefaults != NIL)
values[Anum_pg_proc_proargdefaults - 1] = CStringGetTextDatum(nodeToString(parameterDefaults));
else
nulls[Anum_pg_proc_proargdefaults - 1] = true;
values[Anum_pg_proc_prosrc - 1] = CStringGetTextDatum(prosrc);
if (probin)
values[Anum_pg_proc_probin - 1] = CStringGetTextDatum(probin);
else
nulls[Anum_pg_proc_probin - 1] = true;
if (proconfig != PointerGetDatum(NULL))
values[Anum_pg_proc_proconfig - 1] = proconfig;
else
nulls[Anum_pg_proc_proconfig - 1] = true;
/* proacl will be determined later */
rel = heap_open(ProcedureRelationId, RowExclusiveLock);
tupDesc = RelationGetDescr(rel);
/* Check for pre-existing definition */
oldtup = SearchSysCache3(PROCNAMEARGSNSP,
PointerGetDatum(procedureName),
PointerGetDatum(parameterTypes),
ObjectIdGetDatum(procNamespace));
if (HeapTupleIsValid(oldtup))
{
/* There is one; okay to replace it? */
Form_pg_proc oldproc = (Form_pg_proc) GETSTRUCT(oldtup);
Datum proargnames;
bool isnull;
if (!replace)
ereport(ERROR,
(errcode(ERRCODE_DUPLICATE_FUNCTION),
errmsg("function \"%s\" already exists with same argument types",
procedureName)));
if (!pg_proc_ownercheck(HeapTupleGetOid(oldtup), proowner))
aclcheck_error(ACLCHECK_NOT_OWNER, ACL_KIND_PROC,
procedureName);
/*
* Not okay to change the return type of the existing proc, since
* existing rules, views, etc may depend on the return type.
*/
if (returnType != oldproc->prorettype ||
returnsSet != oldproc->proretset)
ereport(ERROR,
(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
errmsg("cannot change return type of existing function"),
errhint("Use DROP FUNCTION first.")));
/*
* If it returns RECORD, check for possible change of record type
* implied by OUT parameters
*/
if (returnType == RECORDOID)
{
TupleDesc olddesc;
TupleDesc newdesc;
olddesc = build_function_result_tupdesc_t(oldtup);
newdesc = build_function_result_tupdesc_d(allParameterTypes,
parameterModes,
parameterNames);
if (olddesc == NULL && newdesc == NULL)
/* ok, both are runtime-defined RECORDs */ ;
else if (olddesc == NULL || newdesc == NULL ||
!equalTupleDescs(olddesc, newdesc))
ereport(ERROR,
(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
errmsg("cannot change return type of existing function"),
errdetail("Row type defined by OUT parameters is different."),
errhint("Use DROP FUNCTION first.")));
}
/*
* If there were any named input parameters, check to make sure the
* names have not been changed, as this could break existing calls. We
* allow adding names to formerly unnamed parameters, though.
*/
proargnames = SysCacheGetAttr(PROCNAMEARGSNSP, oldtup,
Anum_pg_proc_proargnames,
&isnull);
if (!isnull)
{
Datum proargmodes;
char **old_arg_names;
char **new_arg_names;
int n_old_arg_names;
int n_new_arg_names;
int j;
proargmodes = SysCacheGetAttr(PROCNAMEARGSNSP, oldtup,
Anum_pg_proc_proargmodes,
&isnull);
if (isnull)
proargmodes = PointerGetDatum(NULL); /* just to be sure */
n_old_arg_names = get_func_input_arg_names(proargnames,
proargmodes,
&old_arg_names);
n_new_arg_names = get_func_input_arg_names(parameterNames,
parameterModes,
&new_arg_names);
for (j = 0; j < n_old_arg_names; j++)
{
if (old_arg_names[j] == NULL)
continue;
if (j >= n_new_arg_names || new_arg_names[j] == NULL ||
strcmp(old_arg_names[j], new_arg_names[j]) != 0)
ereport(ERROR,
(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
errmsg("cannot change name of input parameter \"%s\"",
old_arg_names[j]),
errhint("Use DROP FUNCTION first.")));
}
}
/*
* If there are existing defaults, check compatibility: redefinition
* must not remove any defaults nor change their types. (Removing a
* default might cause a function to fail to satisfy an existing call.
* Changing type would only be possible if the associated parameter is
* polymorphic, and in such cases a change of default type might alter
* the resolved output type of existing calls.)
*/
if (oldproc->pronargdefaults != 0)
{
Datum proargdefaults;
List *oldDefaults;
ListCell *oldlc;
ListCell *newlc;
if (list_length(parameterDefaults) < oldproc->pronargdefaults)
ereport(ERROR,
(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
errmsg("cannot remove parameter defaults from existing function"),
errhint("Use DROP FUNCTION first.")));
proargdefaults = SysCacheGetAttr(PROCNAMEARGSNSP, oldtup,
Anum_pg_proc_proargdefaults,
&isnull);
Assert(!isnull);
oldDefaults = (List *) stringToNode(TextDatumGetCString(proargdefaults));
Assert(IsA(oldDefaults, List));
Assert(list_length(oldDefaults) == oldproc->pronargdefaults);
/* new list can have more defaults than old, advance over 'em */
newlc = list_head(parameterDefaults);
for (i = list_length(parameterDefaults) - oldproc->pronargdefaults;
i > 0;
i--)
newlc = lnext(newlc);
foreach(oldlc, oldDefaults)
{
Node *oldDef = (Node *) lfirst(oldlc);
Node *newDef = (Node *) lfirst(newlc);
if (exprType(oldDef) != exprType(newDef))
ereport(ERROR,
(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
errmsg("cannot change data type of existing parameter default value"),
errhint("Use DROP FUNCTION first.")));
newlc = lnext(newlc);
}
}
static jvalue _TupleDesc_coerceDatum(Type self, Datum arg)
{
jvalue result;
result.l = TupleDesc_create((TupleDesc)DatumGetPointer(arg));
return result;
}
static void
compileTheLexeme(DictThesaurus *d)
{
int i,
nnw = 0,
tnm = 16;
TheLexeme *newwrds = (TheLexeme *) palloc(sizeof(TheLexeme) * tnm),
*ptrwrds;
for (i = 0; i < d->nwrds; i++)
{
TSLexeme *ptr;
if (strcmp(d->wrds[i].lexeme, "?") == 0) /* Is stop word marker? */
newwrds = addCompiledLexeme(newwrds, &nnw, &tnm, NULL, d->wrds[i].entries, 0);
else
{
ptr = (TSLexeme *) DatumGetPointer(FunctionCall4(&(d->subdict->lexize),
PointerGetDatum(d->subdict->dictData),
PointerGetDatum(d->wrds[i].lexeme),
Int32GetDatum(strlen(d->wrds[i].lexeme)),
PointerGetDatum(NULL)));
if (!ptr)
ereport(ERROR,
(errcode(ERRCODE_CONFIG_FILE_ERROR),
errmsg("thesaurus sample word \"%s\" isn't recognized by subdictionary (rule %d)",
d->wrds[i].lexeme,
d->wrds[i].entries->idsubst + 1)));
else if (!(ptr->lexeme))
ereport(ERROR,
(errcode(ERRCODE_CONFIG_FILE_ERROR),
errmsg("thesaurus sample word \"%s\" is a stop word (rule %d)",
d->wrds[i].lexeme,
d->wrds[i].entries->idsubst + 1),
errhint("Use \"?\" to represent a stop word within a sample phrase.")));
else
{
while (ptr->lexeme)
{
TSLexeme *remptr = ptr + 1;
int tnvar = 1;
int curvar = ptr->nvariant;
/* compute n words in one variant */
while (remptr->lexeme)
{
if (remptr->nvariant != (remptr - 1)->nvariant)
break;
tnvar++;
remptr++;
}
remptr = ptr;
while (remptr->lexeme && remptr->nvariant == curvar)
{
newwrds = addCompiledLexeme(newwrds, &nnw, &tnm, remptr, d->wrds[i].entries, tnvar);
remptr++;
}
ptr = remptr;
}
}
}
pfree(d->wrds[i].lexeme);
pfree(d->wrds[i].entries);
}
if (d->wrds)
pfree(d->wrds);
d->wrds = newwrds;
d->nwrds = nnw;
d->ntwrds = tnm;
if (d->nwrds > 1)
{
qsort(d->wrds, d->nwrds, sizeof(TheLexeme), cmpTheLexeme);
/* uniq */
newwrds = d->wrds;
ptrwrds = d->wrds + 1;
while (ptrwrds - d->wrds < d->nwrds)
{
if (cmpLexeme(ptrwrds, newwrds) == 0)
{
if (cmpLexemeInfo(ptrwrds->entries, newwrds->entries))
{
ptrwrds->entries->nextentry = newwrds->entries;
newwrds->entries = ptrwrds->entries;
}
else
pfree(ptrwrds->entries);
if (ptrwrds->lexeme)
pfree(ptrwrds->lexeme);
}
else
{
newwrds++;
*newwrds = *ptrwrds;
}
ptrwrds++;
}
d->nwrds = newwrds - d->wrds + 1;
d->wrds = (TheLexeme *) repalloc(d->wrds, sizeof(TheLexeme) * d->nwrds);
}
}
/* ----------------
* printtup --- print a tuple in protocol 3.0
* ----------------
*/
static void
printtup(TupleTableSlot *slot, DestReceiver *self)
{
TupleDesc typeinfo = slot->tts_tupleDescriptor;
DR_printtup *myState = (DR_printtup *) self;
StringInfoData buf;
int natts = typeinfo->natts;
int i;
/* Set or update my derived attribute info, if needed */
if (myState->attrinfo != typeinfo || myState->nattrs != natts)
printtup_prepare_info(myState, typeinfo, natts);
/* Make sure the tuple is fully deconstructed */
slot_getallattrs(slot);
/*
* Prepare a DataRow message
*/
pq_beginmessage(&buf, 'D');
pq_sendint(&buf, natts, 2);
/*
* send the attributes of this tuple
*/
for (i = 0; i < natts; ++i)
{
PrinttupAttrInfo *thisState = myState->myinfo + i;
Datum origattr = slot->tts_values[i],
attr;
if (slot->tts_isnull[i])
{
pq_sendint(&buf, -1, 4);
continue;
}
/*
* If we have a toasted datum, forcibly detoast it here to avoid
* memory leakage inside the type's output routine.
*/
if (thisState->typisvarlena)
attr = PointerGetDatum(PG_DETOAST_DATUM(origattr));
else
attr = origattr;
if (thisState->format == 0)
{
/* Text output */
char *outputstr;
outputstr = OutputFunctionCall(&thisState->finfo, attr);
pq_sendcountedtext(&buf, outputstr, strlen(outputstr), false);
pfree(outputstr);
}
else
{
/* Binary output */
bytea *outputbytes;
outputbytes = SendFunctionCall(&thisState->finfo, attr);
pq_sendint(&buf, VARSIZE(outputbytes) - VARHDRSZ, 4);
pq_sendbytes(&buf, VARDATA(outputbytes),
VARSIZE(outputbytes) - VARHDRSZ);
pfree(outputbytes);
}
/* Clean up detoasted copy, if any */
if (attr != origattr)
pfree(DatumGetPointer(attr));
}
pq_endmessage(&buf);
}
GISTENTRY *
gbt_num_compress(GISTENTRY *retval, GISTENTRY *entry, const gbtree_ninfo *tinfo)
{
if (entry->leafkey)
{
union
{
int16 i2;
int32 i4;
int64 i8;
float4 f4;
float8 f8;
DateADT dt;
TimeADT tm;
Timestamp ts;
Cash ch;
} v;
GBT_NUMKEY *r = (GBT_NUMKEY *) palloc0(tinfo->indexsize);
void *leaf = NULL;
switch (tinfo->t)
{
case gbt_t_int2:
v.i2 = DatumGetInt16(entry->key);
leaf = &v.i2;
break;
case gbt_t_int4:
v.i4 = DatumGetInt32(entry->key);
leaf = &v.i4;
break;
case gbt_t_int8:
v.i8 = DatumGetInt64(entry->key);
leaf = &v.i8;
break;
case gbt_t_oid:
v.i4 = DatumGetObjectId(entry->key);
leaf = &v.i4;
break;
case gbt_t_float4:
v.f4 = DatumGetFloat4(entry->key);
leaf = &v.f4;
break;
case gbt_t_float8:
v.f8 = DatumGetFloat8(entry->key);
leaf = &v.f8;
break;
case gbt_t_date:
v.dt = DatumGetDateADT(entry->key);
leaf = &v.dt;
break;
case gbt_t_time:
v.tm = DatumGetTimeADT(entry->key);
leaf = &v.tm;
break;
case gbt_t_ts:
v.ts = DatumGetTimestamp(entry->key);
leaf = &v.ts;
break;
case gbt_t_cash:
v.ch = DatumGetCash(entry->key);
leaf = &v.ch;
break;
default:
leaf = DatumGetPointer(entry->key);
}
Assert(tinfo->indexsize >= 2 * tinfo->size);
memcpy((void *) &r[0], leaf, tinfo->size);
memcpy((void *) &r[tinfo->size], leaf, tinfo->size);
retval = palloc(sizeof(GISTENTRY));
gistentryinit(*retval, PointerGetDatum(r), entry->rel, entry->page,
entry->offset, FALSE);
}
else
retval = entry;
return retval;
}
/* ----------------
* printtup_internal_20 --- print a binary tuple in protocol 2.0
*
* We use a different message type, i.e. 'B' instead of 'D' to
* indicate a tuple in internal (binary) form.
*
* This is largely same as printtup_20, except we use binary formatting.
* ----------------
*/
static void
printtup_internal_20(TupleTableSlot *slot, DestReceiver *self)
{
TupleDesc typeinfo = slot->tts_tupleDescriptor;
DR_printtup *myState = (DR_printtup *) self;
StringInfoData buf;
int natts = typeinfo->natts;
int i,
j,
k;
/* Set or update my derived attribute info, if needed */
if (myState->attrinfo != typeinfo || myState->nattrs != natts)
printtup_prepare_info(myState, typeinfo, natts);
/* Make sure the tuple is fully deconstructed */
slot_getallattrs(slot);
/*
* tell the frontend to expect new tuple data (in binary style)
*/
pq_beginmessage(&buf, 'B');
/*
* send a bitmap of which attributes are not null
*/
j = 0;
k = 1 << 7;
for (i = 0; i < natts; ++i)
{
if (!slot->tts_isnull[i])
j |= k; /* set bit if not null */
k >>= 1;
if (k == 0) /* end of byte? */
{
pq_sendint(&buf, j, 1);
j = 0;
k = 1 << 7;
}
}
if (k != (1 << 7)) /* flush last partial byte */
pq_sendint(&buf, j, 1);
/*
* send the attributes of this tuple
*/
for (i = 0; i < natts; ++i)
{
PrinttupAttrInfo *thisState = myState->myinfo + i;
Datum origattr = slot->tts_values[i],
attr;
bytea *outputbytes;
if (slot->tts_isnull[i])
continue;
Assert(thisState->format == 1);
/*
* If we have a toasted datum, forcibly detoast it here to avoid
* memory leakage inside the type's output routine.
*/
if (thisState->typisvarlena)
attr = PointerGetDatum(PG_DETOAST_DATUM(origattr));
else
attr = origattr;
outputbytes = SendFunctionCall(&thisState->finfo, attr);
/* We assume the result will not have been toasted */
pq_sendint(&buf, VARSIZE(outputbytes) - VARHDRSZ, 4);
pq_sendbytes(&buf, VARDATA(outputbytes),
VARSIZE(outputbytes) - VARHDRSZ);
pfree(outputbytes);
/* Clean up detoasted copy, if any */
if (attr != origattr)
pfree(DatumGetPointer(attr));
}
pq_endmessage(&buf);
}
/*
* Add an attribute to the hash calculation.
* **IMPORTANT: any new hard coded support for a data type in here
* must be added to isGreenplumDbHashable() below!
*
* Note that the caller should provide the base type if the datum is
* of a domain type. It is quite expensive to call get_typtype() and
* getBaseType() here since this function gets called a lot for the
* same set of Datums.
*
* @param hashFn called to update the hash value.
* @param clientData passed to hashFn.
*/
void
hashDatum(Datum datum, Oid type, datumHashFunction hashFn, void *clientData)
{
void *buf = NULL; /* pointer to the data */
size_t len = 0; /* length for the data buffer */
int64 intbuf; /* an 8 byte buffer for all integer sizes */
float4 buf_f4;
float8 buf_f8;
Timestamp tsbuf; /* timestamp data dype is either a double or
* int8 (determined in compile time) */
TimestampTz tstzbuf;
DateADT datebuf;
TimeADT timebuf;
TimeTzADT *timetzptr;
Interval *intervalptr;
AbsoluteTime abstime_buf;
RelativeTime reltime_buf;
TimeInterval tinterval;
AbsoluteTime tinterval_len;
Numeric num;
bool bool_buf;
char char_buf;
Name namebuf;
ArrayType *arrbuf;
inet *inetptr; /* inet/cidr */
unsigned char inet_hkey[sizeof(inet_struct)];
macaddr *macptr; /* MAC address */
VarBit *vbitptr;
int2vector *i2vec_buf;
oidvector *oidvec_buf;
Cash cash_buf;
AclItem *aclitem_ptr;
uint32 aclitem_buf;
/*
* special case buffers
*/
uint32 nanbuf;
uint32 invalidbuf;
void *tofree = NULL;
/*
* Select the hash to be performed according to the field type we are adding to the
* hash.
*/
switch (type)
{
/*
* ======= NUMERIC TYPES ========
*/
case INT2OID: /* -32 thousand to 32 thousand, 2-byte storage */
intbuf = (int64) DatumGetInt16(datum); /* cast to 8 byte before
* hashing */
buf = &intbuf;
len = sizeof(intbuf);
break;
case INT4OID: /* -2 billion to 2 billion integer, 4-byte
* storage */
intbuf = (int64) DatumGetInt32(datum); /* cast to 8 byte before
* hashing */
buf = &intbuf;
len = sizeof(intbuf);
break;
case INT8OID: /* ~18 digit integer, 8-byte storage */
intbuf = DatumGetInt64(datum); /* cast to 8 byte before
* hashing */
buf = &intbuf;
len = sizeof(intbuf);
break;
case FLOAT4OID: /* single-precision floating point number,
* 4-byte storage */
buf_f4 = DatumGetFloat4(datum);
/*
* On IEEE-float machines, minus zero and zero have different bit
* patterns but should compare as equal. We must ensure that they
* have the same hash value, which is most easily done this way:
*/
if (buf_f4 == (float4) 0)
buf_f4 = 0.0;
buf = &buf_f4;
len = sizeof(buf_f4);
break;
case FLOAT8OID: /* double-precision floating point number,
* 8-byte storage */
buf_f8 = DatumGetFloat8(datum);
/*
* On IEEE-float machines, minus zero and zero have different bit
* patterns but should compare as equal. We must ensure that they
* have the same hash value, which is most easily done this way:
*/
if (buf_f8 == (float8) 0)
buf_f8 = 0.0;
buf = &buf_f8;
len = sizeof(buf_f8);
break;
case NUMERICOID:
num = DatumGetNumeric(datum);
if (NUMERIC_IS_NAN(num))
{
nanbuf = NAN_VAL;
buf = &nanbuf;
len = sizeof(nanbuf);
}
else
/* not a nan */
{
buf = num->n_data;
len = (VARSIZE(num) - NUMERIC_HDRSZ);
}
/*
* If we did a pg_detoast_datum, we need to remember to pfree,
* or we will leak memory. Because of the 1-byte varlena header stuff.
*/
if (num != DatumGetPointer(datum))
tofree = num;
break;
/*
* ====== CHARACTER TYPES =======
*/
case CHAROID: /* char(1), single character */
char_buf = DatumGetChar(datum);
buf = &char_buf;
len = 1;
break;
case BPCHAROID: /* char(n), blank-padded string, fixed storage */
case TEXTOID: /* text */
case VARCHAROID: /* varchar */
case BYTEAOID: /* bytea */
{
int tmplen;
varattrib_untoast_ptr_len(datum, (char **) &buf, &tmplen, &tofree);
/* adjust length to not include trailing blanks */
if (type != BYTEAOID && tmplen > 1)
tmplen = ignoreblanks((char *) buf, tmplen);
len = tmplen;
break;
}
case NAMEOID:
namebuf = DatumGetName(datum);
len = NAMEDATALEN;
buf = NameStr(*namebuf);
/* adjust length to not include trailing blanks */
if (len > 1)
len = ignoreblanks((char *) buf, len);
break;
/*
* ====== OBJECT IDENTIFIER TYPES ======
*/
case OIDOID: /* object identifier(oid), maximum 4 billion */
case REGPROCOID: /* function name */
case REGPROCEDUREOID: /* function name with argument types */
case REGOPEROID: /* operator name */
case REGOPERATOROID: /* operator with argument types */
case REGCLASSOID: /* relation name */
case REGTYPEOID: /* data type name */
intbuf = (int64) DatumGetUInt32(datum); /* cast to 8 byte before hashing */
buf = &intbuf;
len = sizeof(intbuf);
break;
case TIDOID: /* tuple id (6 bytes) */
buf = DatumGetPointer(datum);
len = SizeOfIptrData;
break;
/*
* ====== DATE/TIME TYPES ======
*/
case TIMESTAMPOID: /* date and time */
tsbuf = DatumGetTimestamp(datum);
buf = &tsbuf;
len = sizeof(tsbuf);
break;
case TIMESTAMPTZOID: /* date and time with time zone */
tstzbuf = DatumGetTimestampTz(datum);
buf = &tstzbuf;
len = sizeof(tstzbuf);
break;
case DATEOID: /* ANSI SQL date */
datebuf = DatumGetDateADT(datum);
buf = &datebuf;
len = sizeof(datebuf);
break;
case TIMEOID: /* hh:mm:ss, ANSI SQL time */
timebuf = DatumGetTimeADT(datum);
buf = &timebuf;
len = sizeof(timebuf);
break;
case TIMETZOID: /* time with time zone */
/*
* will not compare to TIMEOID on equal values.
* Postgres never attempts to compare the two as well.
*/
timetzptr = DatumGetTimeTzADTP(datum);
buf = (unsigned char *) timetzptr;
/*
* Specify hash length as sizeof(double) + sizeof(int4), not as
* sizeof(TimeTzADT), so that any garbage pad bytes in the structure
* won't be included in the hash!
*/
len = sizeof(timetzptr->time) + sizeof(timetzptr->zone);
break;
case INTERVALOID: /* @ <number> <units>, time interval */
intervalptr = DatumGetIntervalP(datum);
buf = (unsigned char *) intervalptr;
/*
* Specify hash length as sizeof(double) + sizeof(int4), not as
* sizeof(Interval), so that any garbage pad bytes in the structure
* won't be included in the hash!
*/
len = sizeof(intervalptr->time) + sizeof(intervalptr->month);
break;
case ABSTIMEOID:
abstime_buf = DatumGetAbsoluteTime(datum);
if (abstime_buf == INVALID_ABSTIME)
{
/* hash to a constant value */
invalidbuf = INVALID_VAL;
len = sizeof(invalidbuf);
buf = &invalidbuf;
}
else
{
len = sizeof(abstime_buf);
buf = &abstime_buf;
}
break;
case RELTIMEOID:
reltime_buf = DatumGetRelativeTime(datum);
if (reltime_buf == INVALID_RELTIME)
{
/* hash to a constant value */
invalidbuf = INVALID_VAL;
len = sizeof(invalidbuf);
buf = &invalidbuf;
}
else
{
len = sizeof(reltime_buf);
buf = &reltime_buf;
}
break;
case TINTERVALOID:
tinterval = DatumGetTimeInterval(datum);
/*
* check if a valid interval. the '0' status code
* stands for T_INTERVAL_INVAL which is defined in
* nabstime.c. We use the actual value instead
* of defining it again here.
*/
if(tinterval->status == 0 ||
tinterval->data[0] == INVALID_ABSTIME ||
tinterval->data[1] == INVALID_ABSTIME)
{
/* hash to a constant value */
invalidbuf = INVALID_VAL;
len = sizeof(invalidbuf);
buf = &invalidbuf;
}
else
{
/* normalize on length of the time interval */
tinterval_len = tinterval->data[1] - tinterval->data[0];
len = sizeof(tinterval_len);
buf = &tinterval_len;
}
break;
/*
* ======= NETWORK TYPES ========
*/
case INETOID:
case CIDROID:
inetptr = DatumGetInetP(datum);
len = inet_getkey(inetptr, inet_hkey, sizeof(inet_hkey)); /* fill-in inet_key & get len */
buf = inet_hkey;
break;
case MACADDROID:
macptr = DatumGetMacaddrP(datum);
len = sizeof(macaddr);
buf = (unsigned char *) macptr;
break;
/*
* ======== BIT STRINGS ========
*/
case BITOID:
case VARBITOID:
/*
* Note that these are essentially strings.
* we don't need to worry about '10' and '010'
* to compare, b/c they will not, by design.
* (see SQL standard, and varbit.c)
*/
vbitptr = DatumGetVarBitP(datum);
len = VARBITBYTES(vbitptr);
buf = (char *) VARBITS(vbitptr);
break;
/*
* ======= other types =======
*/
case BOOLOID: /* boolean, 'true'/'false' */
bool_buf = DatumGetBool(datum);
buf = &bool_buf;
len = sizeof(bool_buf);
break;
/*
* We prepare the hash key for aclitems just like postgresql does.
* (see code and comment in acl.c: hash_aclitem() ).
*/
case ACLITEMOID:
aclitem_ptr = DatumGetAclItemP(datum);
aclitem_buf = (uint32) (aclitem_ptr->ai_privs + aclitem_ptr->ai_grantee + aclitem_ptr->ai_grantor);
buf = &aclitem_buf;
len = sizeof(aclitem_buf);
break;
/*
* ANYARRAY is a pseudo-type. We use it to include
* any of the array types (OIDs 1007-1033 in pg_type.h).
* caller needs to be sure the type is ANYARRAYOID
* before calling cdbhash on an array (INSERT and COPY do so).
*/
case ANYARRAYOID:
arrbuf = DatumGetArrayTypeP(datum);
len = VARSIZE(arrbuf) - VARHDRSZ;
buf = VARDATA(arrbuf);
break;
case INT2VECTOROID:
i2vec_buf = (int2vector *) DatumGetPointer(datum);
len = i2vec_buf->dim1 * sizeof(int2);
buf = (void *)i2vec_buf->values;
break;
case OIDVECTOROID:
oidvec_buf = (oidvector *) DatumGetPointer(datum);
len = oidvec_buf->dim1 * sizeof(Oid);
buf = oidvec_buf->values;
break;
case CASHOID: /* cash is stored in int32 internally */
cash_buf = (* (Cash *)DatumGetPointer(datum));
len = sizeof(Cash);
buf = &cash_buf;
break;
default:
ereport(ERROR,
(errcode(ERRCODE_CDB_FEATURE_NOT_YET),
errmsg("Type %u is not hashable.", type)));
} /* switch(type) */
/* do the hash using the selected algorithm */
hashFn(clientData, buf, len);
if(tofree)
pfree(tofree);
}
/*
* heap_form_tuple
* construct a tuple from the given values[] and isnull[] arrays,
* which are of the length indicated by tupleDescriptor->natts
*
* The result is allocated in the current memory context.
*/
HeapTuple
heaptuple_form_to(TupleDesc tupleDescriptor, Datum *values, bool *isnull, HeapTuple dst, uint32 *dstlen)
{
HeapTuple tuple; /* return tuple */
HeapTupleHeader td; /* tuple data */
Size actual_len;
Size len,
data_len;
int hoff;
bool hasnull = false;
Form_pg_attribute *att = tupleDescriptor->attrs;
int numberOfAttributes = tupleDescriptor->natts;
int i;
if (numberOfAttributes > MaxTupleAttributeNumber)
ereport(ERROR,
(errcode(ERRCODE_TOO_MANY_COLUMNS),
errmsg("number of columns (%d) exceeds limit (%d)",
numberOfAttributes, MaxTupleAttributeNumber)));
/*
* Check for nulls and embedded tuples; expand any toasted attributes in
* embedded tuples. This preserves the invariant that toasting can only
* go one level deep.
*
* We can skip calling toast_flatten_tuple_attribute() if the attribute
* couldn't possibly be of composite type. All composite datums are
* varlena and have alignment 'd'; furthermore they aren't arrays. Also,
* if an attribute is already toasted, it must have been sent to disk
* already and so cannot contain toasted attributes.
*/
for (i = 0; i < numberOfAttributes; i++)
{
if (isnull[i])
hasnull = true;
else if (att[i]->attlen == -1 &&
att[i]->attalign == 'd' &&
att[i]->attndims == 0 &&
!VARATT_IS_EXTENDED(DatumGetPointer(values[i])))
{
values[i] = toast_flatten_tuple_attribute(values[i],
att[i]->atttypid,
att[i]->atttypmod);
}
}
/*
* Determine total space needed
*/
len = offsetof(HeapTupleHeaderData, t_bits);
if (hasnull)
len += BITMAPLEN(numberOfAttributes);
if (tupleDescriptor->tdhasoid)
len += sizeof(Oid);
hoff = len = MAXALIGN(len); /* align user data safely */
data_len = heap_compute_data_size(tupleDescriptor, values, isnull);
len += data_len;
if (dstlen && (*dstlen) < (HEAPTUPLESIZE + len))
{
*dstlen = HEAPTUPLESIZE + len;
return NULL;
}
if (dstlen)
{
*dstlen = HEAPTUPLESIZE + len;
tuple = dst;
memset(tuple, 0, HEAPTUPLESIZE + len);
}
else
tuple = (HeapTuple) palloc0(HEAPTUPLESIZE + len);
/*
* Allocate and zero the space needed. Note that the tuple body and
* HeapTupleData management structure are allocated in one chunk.
*/
tuple->t_data = td = (HeapTupleHeader) ((char *) tuple + HEAPTUPLESIZE);
/*
* And fill in the information. Note we fill the Datum fields even though
* this tuple may never become a Datum.
*/
tuple->t_len = len;
ItemPointerSetInvalid(&(tuple->t_self));
HeapTupleHeaderSetDatumLength(td, len);
HeapTupleHeaderSetTypeId(td, tupleDescriptor->tdtypeid);
HeapTupleHeaderSetTypMod(td, tupleDescriptor->tdtypmod);
HeapTupleHeaderSetNatts(td, numberOfAttributes);
td->t_hoff = hoff;
if (tupleDescriptor->tdhasoid) /* else leave infomask = 0 */
td->t_infomask = HEAP_HASOID;
actual_len = heap_fill_tuple(tupleDescriptor,
values,
isnull,
(char *) td + hoff,
data_len,
&td->t_infomask,
(hasnull ? td->t_bits : NULL));
Assert(data_len == actual_len);
Assert(!is_heaptuple_memtuple(tuple));
return tuple;
}