Restructure SPGiST opclass interface API to support whole-index scans.

The original API definition was incapable of supporting whole-index scans
because there was no way to invoke leaf-value reconstruction without
checking any qual conditions.  Also, it was inefficient for
multiple-qual-condition scans because value reconstruction got done over
again for each qual condition, and because other internal work in the
consistent functions likewise had to be done for each qual.  To fix these
issues, pass the whole scankey array to the opclass consistent functions,
instead of only letting them see one item at a time.  (Essentially, the
loop over scankey entries is now inside the consistent functions not
outside them.  This makes the consistent functions a bit more complicated,
but not unreasonably so.)

In itself this commit does nothing except save a few cycles in
multiple-qual-condition index scans, since we can't support whole-index
scans on SPGiST indexes until nulls are included in the index.  However,
I consider this a must-fix for 9.2 because once we release it will get
very much harder to change the opclass API definition.
This commit is contained in:
Tom Lane 2012-03-10 18:36:49 -05:00
parent 39d74e346c
commit 03e56f798e
7 changed files with 523 additions and 532 deletions

View File

@ -439,8 +439,8 @@ CREATE FUNCTION my_inner_consistent(internal, internal) RETURNS void ...
<programlisting> <programlisting>
typedef struct spgInnerConsistentIn typedef struct spgInnerConsistentIn
{ {
StrategyNumber strategy; /* operator strategy number */ ScanKey scankeys; /* array of operators and comparison values */
Datum query; /* operator's RHS value */ int nkeys; /* length of array */
Datum reconstructedValue; /* value reconstructed at parent */ Datum reconstructedValue; /* value reconstructed at parent */
int level; /* current level (counting from zero) */ int level; /* current level (counting from zero) */
@ -463,8 +463,17 @@ typedef struct spgInnerConsistentOut
} spgInnerConsistentOut; } spgInnerConsistentOut;
</programlisting> </programlisting>
<structfield>strategy</> and The array <structfield>scankeys</>, of length <structfield>nkeys</>,
<structfield>query</> describe the index search condition. describes the index search condition(s). These conditions are
combined with AND &mdash; only index entries that satisfy all of
them are interesting. (Note that <structfield>nkeys</> = 0 implies
that all index entries satisfy the query.) Usually the consistent
function only cares about the <structfield>sk_strategy</> and
<structfield>sk_argument</> fields of each array entry, which
respectively give the indexable operator and comparison value.
In particular it is not necessary to check <structfield>sk_flags</> to
see if the comparison value is NULL, because the SP-GiST core code
will filter out such conditions.
<structfield>reconstructedValue</> is the value reconstructed for the <structfield>reconstructedValue</> is the value reconstructed for the
parent tuple; it is <literal>(Datum) 0</> at the root level or if the parent tuple; it is <literal>(Datum) 0</> at the root level or if the
<function>inner_consistent</> function did not provide a value at the <function>inner_consistent</> function did not provide a value at the
@ -527,8 +536,8 @@ CREATE FUNCTION my_leaf_consistent(internal, internal) RETURNS bool ...
<programlisting> <programlisting>
typedef struct spgLeafConsistentIn typedef struct spgLeafConsistentIn
{ {
StrategyNumber strategy; /* operator strategy number */ ScanKey scankeys; /* array of operators and comparison values */
Datum query; /* operator's RHS value */ int nkeys; /* length of array */
Datum reconstructedValue; /* value reconstructed at parent */ Datum reconstructedValue; /* value reconstructed at parent */
int level; /* current level (counting from zero) */ int level; /* current level (counting from zero) */
@ -544,8 +553,17 @@ typedef struct spgLeafConsistentOut
} spgLeafConsistentOut; } spgLeafConsistentOut;
</programlisting> </programlisting>
<structfield>strategy</> and The array <structfield>scankeys</>, of length <structfield>nkeys</>,
<structfield>query</> define the index search condition. describes the index search condition(s). These conditions are
combined with AND &mdash; only index entries that satisfy all of
them satisfy the query. (Note that <structfield>nkeys</> = 0 implies
that all index entries satisfy the query.) Usually the consistent
function only cares about the <structfield>sk_strategy</> and
<structfield>sk_argument</> fields of each array entry, which
respectively give the indexable operator and comparison value.
In particular it is not necessary to check <structfield>sk_flags</> to
see if the comparison value is NULL, because the SP-GiST core code
will filter out such conditions.
<structfield>reconstructedValue</> is the value reconstructed for the <structfield>reconstructedValue</> is the value reconstructed for the
parent tuple; it is <literal>(Datum) 0</> at the root level or if the parent tuple; it is <literal>(Datum) 0</> at the root level or if the
<function>inner_consistent</> function did not provide a value at the <function>inner_consistent</> function did not provide a value at the
@ -566,8 +584,8 @@ typedef struct spgLeafConsistentOut
<structfield>leafValue</> must be set to the value originally supplied <structfield>leafValue</> must be set to the value originally supplied
to be indexed for this leaf tuple. Also, to be indexed for this leaf tuple. Also,
<structfield>recheck</> may be set to <literal>true</> if the match <structfield>recheck</> may be set to <literal>true</> if the match
is uncertain and so the operator must be re-applied to the actual heap is uncertain and so the operator(s) must be re-applied to the actual
tuple to verify the match. heap tuple to verify the match.
</para> </para>
</listitem> </listitem>
</varlistentry> </varlistentry>

View File

@ -159,11 +159,10 @@ spg_kd_inner_consistent(PG_FUNCTION_ARGS)
{ {
spgInnerConsistentIn *in = (spgInnerConsistentIn *) PG_GETARG_POINTER(0); spgInnerConsistentIn *in = (spgInnerConsistentIn *) PG_GETARG_POINTER(0);
spgInnerConsistentOut *out = (spgInnerConsistentOut *) PG_GETARG_POINTER(1); spgInnerConsistentOut *out = (spgInnerConsistentOut *) PG_GETARG_POINTER(1);
Point *query;
BOX *boxQuery;
double coord; double coord;
int which;
int i;
query = DatumGetPointP(in->query);
Assert(in->hasPrefix); Assert(in->hasPrefix);
coord = DatumGetFloat8(in->prefixDatum); coord = DatumGetFloat8(in->prefixDatum);
@ -171,123 +170,96 @@ spg_kd_inner_consistent(PG_FUNCTION_ARGS)
elog(ERROR, "allTheSame should not occur for k-d trees"); elog(ERROR, "allTheSame should not occur for k-d trees");
Assert(in->nNodes == 2); Assert(in->nNodes == 2);
/* "which" is a bitmask of children that satisfy all constraints */
which = (1 << 1) | (1 << 2);
for (i = 0; i < in->nkeys; i++)
{
Point *query = DatumGetPointP(in->scankeys[i].sk_argument);
BOX *boxQuery;
switch (in->scankeys[i].sk_strategy)
{
case RTLeftStrategyNumber:
if ((in->level % 2) != 0 && FPlt(query->x, coord))
which &= (1 << 1);
break;
case RTRightStrategyNumber:
if ((in->level % 2) != 0 && FPgt(query->x, coord))
which &= (1 << 2);
break;
case RTSameStrategyNumber:
if ((in->level % 2) != 0)
{
if (FPlt(query->x, coord))
which &= (1 << 1);
else if (FPgt(query->x, coord))
which &= (1 << 2);
}
else
{
if (FPlt(query->y, coord))
which &= (1 << 1);
else if (FPgt(query->y, coord))
which &= (1 << 2);
}
break;
case RTBelowStrategyNumber:
if ((in->level % 2) == 0 && FPlt(query->y, coord))
which &= (1 << 1);
break;
case RTAboveStrategyNumber:
if ((in->level % 2) == 0 && FPgt(query->y, coord))
which &= (1 << 2);
break;
case RTContainedByStrategyNumber:
/*
* For this operator, the query is a box not a point. We
* cheat to the extent of assuming that DatumGetPointP won't
* do anything that would be bad for a pointer-to-box.
*/
boxQuery = DatumGetBoxP(in->scankeys[i].sk_argument);
if ((in->level % 2) != 0)
{
if (FPlt(boxQuery->high.x, coord))
which &= (1 << 1);
else if (FPgt(boxQuery->low.x, coord))
which &= (1 << 2);
}
else
{
if (FPlt(boxQuery->high.y, coord))
which &= (1 << 1);
else if (FPgt(boxQuery->low.y, coord))
which &= (1 << 2);
}
break;
default:
elog(ERROR, "unrecognized strategy number: %d",
in->scankeys[i].sk_strategy);
break;
}
if (which == 0)
break; /* no need to consider remaining conditions */
}
/* We must descend into the children identified by which */
out->nodeNumbers = (int *) palloc(sizeof(int) * 2); out->nodeNumbers = (int *) palloc(sizeof(int) * 2);
out->nNodes = 0;
for (i = 1; i <= 2; i++)
{
if (which & (1 << i))
out->nodeNumbers[out->nNodes++] = i - 1;
}
/* Set up level increments, too */
out->levelAdds = (int *) palloc(sizeof(int) * 2); out->levelAdds = (int *) palloc(sizeof(int) * 2);
out->levelAdds[0] = 1; out->levelAdds[0] = 1;
out->levelAdds[1] = 1; out->levelAdds[1] = 1;
out->nNodes = 0;
switch (in->strategy)
{
case RTLeftStrategyNumber:
out->nNodes = 1;
out->nodeNumbers[0] = 0;
if ((in->level % 2) == 0 || FPge(query->x, coord))
{
out->nodeNumbers[1] = 1;
out->nNodes++;
}
break;
case RTRightStrategyNumber:
out->nNodes = 1;
out->nodeNumbers[0] = 1;
if ((in->level % 2) == 0 || FPle(query->x, coord))
{
out->nodeNumbers[1] = 0;
out->nNodes++;
}
break;
case RTSameStrategyNumber:
if (in->level % 2)
{
if (FPle(query->x, coord))
{
out->nodeNumbers[out->nNodes] = 0;
out->nNodes++;
}
if (FPge(query->x, coord))
{
out->nodeNumbers[out->nNodes] = 1;
out->nNodes++;
}
}
else
{
if (FPle(query->y, coord))
{
out->nodeNumbers[out->nNodes] = 0;
out->nNodes++;
}
if (FPge(query->y, coord))
{
out->nodeNumbers[out->nNodes] = 1;
out->nNodes++;
}
}
break;
case RTBelowStrategyNumber:
out->nNodes = 1;
out->nodeNumbers[0] = 0;
if ((in->level % 2) == 1 || FPge(query->y, coord))
{
out->nodeNumbers[1] = 1;
out->nNodes++;
}
break;
case RTAboveStrategyNumber:
out->nNodes = 1;
out->nodeNumbers[0] = 1;
if ((in->level % 2) == 1 || FPle(query->y, coord))
{
out->nodeNumbers[1] = 0;
out->nNodes++;
}
break;
case RTContainedByStrategyNumber:
/*
* For this operator, the query is a box not a point. We cheat to
* the extent of assuming that DatumGetPointP won't do anything
* that would be bad for a pointer-to-box.
*/
boxQuery = DatumGetBoxP(in->query);
out->nNodes = 1;
if (in->level % 2)
{
if (FPlt(boxQuery->high.x, coord))
out->nodeNumbers[0] = 0;
else if (FPgt(boxQuery->low.x, coord))
out->nodeNumbers[0] = 1;
else
{
out->nodeNumbers[0] = 0;
out->nodeNumbers[1] = 1;
out->nNodes = 2;
}
}
else
{
if (FPlt(boxQuery->high.y, coord))
out->nodeNumbers[0] = 0;
else if (FPgt(boxQuery->low.y, coord))
out->nodeNumbers[0] = 1;
else
{
out->nodeNumbers[0] = 0;
out->nodeNumbers[1] = 1;
out->nNodes = 2;
}
}
break;
default:
elog(ERROR, "unrecognized strategy number: %d", in->strategy);
break;
}
PG_RETURN_VOID(); PG_RETURN_VOID();
} }

View File

@ -190,45 +190,21 @@ spg_quad_picksplit(PG_FUNCTION_ARGS)
} }
/* Subroutine to fill out->nodeNumbers[] for spg_quad_inner_consistent */
static void
setNodes(spgInnerConsistentOut *out, bool isAll, int first, int second)
{
if (isAll)
{
out->nNodes = 4;
out->nodeNumbers[0] = 0;
out->nodeNumbers[1] = 1;
out->nodeNumbers[2] = 2;
out->nodeNumbers[3] = 3;
}
else
{
out->nNodes = 2;
out->nodeNumbers[0] = first - 1;
out->nodeNumbers[1] = second - 1;
}
}
Datum Datum
spg_quad_inner_consistent(PG_FUNCTION_ARGS) spg_quad_inner_consistent(PG_FUNCTION_ARGS)
{ {
spgInnerConsistentIn *in = (spgInnerConsistentIn *) PG_GETARG_POINTER(0); spgInnerConsistentIn *in = (spgInnerConsistentIn *) PG_GETARG_POINTER(0);
spgInnerConsistentOut *out = (spgInnerConsistentOut *) PG_GETARG_POINTER(1); spgInnerConsistentOut *out = (spgInnerConsistentOut *) PG_GETARG_POINTER(1);
Point *query, Point *centroid;
*centroid; int which;
BOX *boxQuery; int i;
query = DatumGetPointP(in->query);
Assert(in->hasPrefix); Assert(in->hasPrefix);
centroid = DatumGetPointP(in->prefixDatum); centroid = DatumGetPointP(in->prefixDatum);
if (in->allTheSame) if (in->allTheSame)
{ {
/* Report that all nodes should be visited */ /* Report that all nodes should be visited */
int i;
out->nNodes = in->nNodes; out->nNodes = in->nNodes;
out->nodeNumbers = (int *) palloc(sizeof(int) * in->nNodes); out->nodeNumbers = (int *) palloc(sizeof(int) * in->nNodes);
for (i = 0; i < in->nNodes; i++) for (i = 0; i < in->nNodes; i++)
@ -237,76 +213,86 @@ spg_quad_inner_consistent(PG_FUNCTION_ARGS)
} }
Assert(in->nNodes == 4); Assert(in->nNodes == 4);
out->nodeNumbers = (int *) palloc(sizeof(int) * 4);
switch (in->strategy) /* "which" is a bitmask of quadrants that satisfy all constraints */
which = (1 << 1) | (1 << 2) | (1 << 3) | (1 << 4);
for (i = 0; i < in->nkeys; i++)
{ {
case RTLeftStrategyNumber: Point *query = DatumGetPointP(in->scankeys[i].sk_argument);
setNodes(out, SPTEST(point_left, centroid, query), 3, 4); BOX *boxQuery;
break;
case RTRightStrategyNumber:
setNodes(out, SPTEST(point_right, centroid, query), 1, 2);
break;
case RTSameStrategyNumber:
out->nNodes = 1;
out->nodeNumbers[0] = getQuadrant(centroid, query) - 1;
break;
case RTBelowStrategyNumber:
setNodes(out, SPTEST(point_below, centroid, query), 2, 3);
break;
case RTAboveStrategyNumber:
setNodes(out, SPTEST(point_above, centroid, query), 1, 4);
break;
case RTContainedByStrategyNumber:
/* switch (in->scankeys[i].sk_strategy)
* For this operator, the query is a box not a point. We cheat to {
* the extent of assuming that DatumGetPointP won't do anything case RTLeftStrategyNumber:
* that would be bad for a pointer-to-box. if (SPTEST(point_right, centroid, query))
*/ which &= (1 << 3) | (1 << 4);
boxQuery = DatumGetBoxP(in->query); break;
case RTRightStrategyNumber:
if (SPTEST(point_left, centroid, query))
which &= (1 << 1) | (1 << 2);
break;
case RTSameStrategyNumber:
which &= (1 << getQuadrant(centroid, query));
break;
case RTBelowStrategyNumber:
if (SPTEST(point_above, centroid, query))
which &= (1 << 2) | (1 << 3);
break;
case RTAboveStrategyNumber:
if (SPTEST(point_below, centroid, query))
which &= (1 << 1) | (1 << 4);
break;
case RTContainedByStrategyNumber:
if (DatumGetBool(DirectFunctionCall2(box_contain_pt, /*
PointerGetDatum(boxQuery), * For this operator, the query is a box not a point. We
PointerGetDatum(centroid)))) * cheat to the extent of assuming that DatumGetPointP won't
{ * do anything that would be bad for a pointer-to-box.
/* centroid is in box, so descend to all quadrants */ */
setNodes(out, true, 0, 0); boxQuery = DatumGetBoxP(in->scankeys[i].sk_argument);
}
else
{
/* identify quadrant(s) containing all corners of box */
Point p;
int i,
r = 0;
p = boxQuery->low; if (DatumGetBool(DirectFunctionCall2(box_contain_pt,
r |= 1 << (getQuadrant(centroid, &p) - 1); PointerGetDatum(boxQuery),
PointerGetDatum(centroid))))
p.y = boxQuery->high.y;
r |= 1 << (getQuadrant(centroid, &p) - 1);
p = boxQuery->high;
r |= 1 << (getQuadrant(centroid, &p) - 1);
p.x = boxQuery->low.x;
r |= 1 << (getQuadrant(centroid, &p) - 1);
/* we must descend into those quadrant(s) */
out->nNodes = 0;
for (i = 0; i < 4; i++)
{ {
if (r & (1 << i)) /* centroid is in box, so all quadrants are OK */
{
out->nodeNumbers[out->nNodes] = i;
out->nNodes++;
}
} }
} else
break; {
default: /* identify quadrant(s) containing all corners of box */
elog(ERROR, "unrecognized strategy number: %d", in->strategy); Point p;
break; int r = 0;
p = boxQuery->low;
r |= 1 << getQuadrant(centroid, &p);
p.y = boxQuery->high.y;
r |= 1 << getQuadrant(centroid, &p);
p = boxQuery->high;
r |= 1 << getQuadrant(centroid, &p);
p.x = boxQuery->low.x;
r |= 1 << getQuadrant(centroid, &p);
which &= r;
}
break;
default:
elog(ERROR, "unrecognized strategy number: %d",
in->scankeys[i].sk_strategy);
break;
}
if (which == 0)
break; /* no need to consider remaining conditions */
}
/* We must descend into the quadrant(s) identified by which */
out->nodeNumbers = (int *) palloc(sizeof(int) * 4);
out->nNodes = 0;
for (i = 1; i <= 4; i++)
{
if (which & (1 << i))
out->nodeNumbers[out->nNodes++] = i - 1;
} }
PG_RETURN_VOID(); PG_RETURN_VOID();
@ -318,9 +304,9 @@ spg_quad_leaf_consistent(PG_FUNCTION_ARGS)
{ {
spgLeafConsistentIn *in = (spgLeafConsistentIn *) PG_GETARG_POINTER(0); spgLeafConsistentIn *in = (spgLeafConsistentIn *) PG_GETARG_POINTER(0);
spgLeafConsistentOut *out = (spgLeafConsistentOut *) PG_GETARG_POINTER(1); spgLeafConsistentOut *out = (spgLeafConsistentOut *) PG_GETARG_POINTER(1);
Point *query = DatumGetPointP(in->query);
Point *datum = DatumGetPointP(in->leafDatum); Point *datum = DatumGetPointP(in->leafDatum);
bool res; bool res;
int i;
/* all tests are exact */ /* all tests are exact */
out->recheck = false; out->recheck = false;
@ -328,35 +314,45 @@ spg_quad_leaf_consistent(PG_FUNCTION_ARGS)
/* leafDatum is what it is... */ /* leafDatum is what it is... */
out->leafValue = in->leafDatum; out->leafValue = in->leafDatum;
switch (in->strategy) /* Perform the required comparison(s) */
res = true;
for (i = 0; i < in->nkeys; i++)
{ {
case RTLeftStrategyNumber: Point *query = DatumGetPointP(in->scankeys[i].sk_argument);
res = SPTEST(point_left, datum, query);
break;
case RTRightStrategyNumber:
res = SPTEST(point_right, datum, query);
break;
case RTSameStrategyNumber:
res = SPTEST(point_eq, datum, query);
break;
case RTBelowStrategyNumber:
res = SPTEST(point_below, datum, query);
break;
case RTAboveStrategyNumber:
res = SPTEST(point_above, datum, query);
break;
case RTContainedByStrategyNumber:
/* switch (in->scankeys[i].sk_strategy)
* For this operator, the query is a box not a point. We cheat to {
* the extent of assuming that DatumGetPointP won't do anything case RTLeftStrategyNumber:
* that would be bad for a pointer-to-box. res = SPTEST(point_left, datum, query);
*/ break;
res = SPTEST(box_contain_pt, query, datum); case RTRightStrategyNumber:
break; res = SPTEST(point_right, datum, query);
default: break;
elog(ERROR, "unrecognized strategy number: %d", in->strategy); case RTSameStrategyNumber:
res = false; res = SPTEST(point_eq, datum, query);
break;
case RTBelowStrategyNumber:
res = SPTEST(point_below, datum, query);
break;
case RTAboveStrategyNumber:
res = SPTEST(point_above, datum, query);
break;
case RTContainedByStrategyNumber:
/*
* For this operator, the query is a box not a point. We
* cheat to the extent of assuming that DatumGetPointP won't
* do anything that would be bad for a pointer-to-box.
*/
res = SPTEST(box_contain_pt, query, datum);
break;
default:
elog(ERROR, "unrecognized strategy number: %d",
in->scankeys[i].sk_strategy);
break;
}
if (!res)
break; break;
} }

View File

@ -56,18 +56,25 @@ freeScanStack(SpGistScanOpaque so)
} }
/* /*
* Initialize scanStack with a single entry for the root page, resetting * Initialize scanStack to search the root page, resetting
* any previously active scan * any previously active scan
*/ */
static void static void
resetSpGistScanOpaque(SpGistScanOpaque so) resetSpGistScanOpaque(SpGistScanOpaque so)
{ {
ScanStackEntry *startEntry = palloc0(sizeof(ScanStackEntry)); ScanStackEntry *startEntry;
ItemPointerSet(&startEntry->ptr, SPGIST_HEAD_BLKNO, FirstOffsetNumber);
freeScanStack(so); freeScanStack(so);
so->scanStack = list_make1(startEntry);
Assert(!so->searchNulls); /* XXX fixme */
if (so->searchNonNulls)
{
/* Stack a work item to scan the non-null index entries */
startEntry = (ScanStackEntry *) palloc0(sizeof(ScanStackEntry));
ItemPointerSet(&startEntry->ptr, SPGIST_HEAD_BLKNO, FirstOffsetNumber);
so->scanStack = list_make1(startEntry);
}
if (so->want_itup) if (so->want_itup)
{ {
@ -80,6 +87,82 @@ resetSpGistScanOpaque(SpGistScanOpaque so)
so->iPtr = so->nPtrs = 0; so->iPtr = so->nPtrs = 0;
} }
/*
* Prepare scan keys in SpGistScanOpaque from caller-given scan keys
*
* Sets searchNulls, searchNonNulls, numberOfKeys, keyData fields of *so.
*
* The point here is to eliminate null-related considerations from what the
* opclass consistent functions need to deal with. We assume all SPGiST-
* indexable operators are strict, so any null RHS value makes the scan
* condition unsatisfiable. We also pull out any IS NULL/IS NOT NULL
* conditions; their effect is reflected into searchNulls/searchNonNulls.
*/
static void
spgPrepareScanKeys(IndexScanDesc scan)
{
SpGistScanOpaque so = (SpGistScanOpaque) scan->opaque;
bool qual_ok;
bool haveIsNull;
bool haveNotNull;
int nkeys;
int i;
if (scan->numberOfKeys <= 0)
{
/* If no quals, whole-index scan is required */
so->searchNulls = true;
so->searchNonNulls = true;
so->numberOfKeys = 0;
return;
}
/* Examine the given quals */
qual_ok = true;
haveIsNull = haveNotNull = false;
nkeys = 0;
for (i = 0; i < scan->numberOfKeys; i++)
{
ScanKey skey = &scan->keyData[i];
if (skey->sk_flags & SK_SEARCHNULL)
haveIsNull = true;
else if (skey->sk_flags & SK_SEARCHNOTNULL)
haveNotNull = true;
else if (skey->sk_flags & SK_ISNULL)
{
/* ordinary qual with null argument - unsatisfiable */
qual_ok = false;
break;
}
else
{
/* ordinary qual, propagate into so->keyData */
so->keyData[nkeys++] = *skey;
/* this effectively creates a not-null requirement */
haveNotNull = true;
}
}
/* IS NULL in combination with something else is unsatisfiable */
if (haveIsNull && haveNotNull)
qual_ok = false;
/* Emit results */
if (qual_ok)
{
so->searchNulls = haveIsNull;
so->searchNonNulls = haveNotNull;
so->numberOfKeys = nkeys;
}
else
{
so->searchNulls = false;
so->searchNonNulls = false;
so->numberOfKeys = 0;
}
}
Datum Datum
spgbeginscan(PG_FUNCTION_ARGS) spgbeginscan(PG_FUNCTION_ARGS)
{ {
@ -92,13 +175,16 @@ spgbeginscan(PG_FUNCTION_ARGS)
scan = RelationGetIndexScan(rel, keysz, 0); scan = RelationGetIndexScan(rel, keysz, 0);
so = (SpGistScanOpaque) palloc0(sizeof(SpGistScanOpaqueData)); so = (SpGistScanOpaque) palloc0(sizeof(SpGistScanOpaqueData));
if (keysz > 0)
so->keyData = (ScanKey) palloc(sizeof(ScanKeyData) * keysz);
else
so->keyData = NULL;
initSpGistState(&so->state, scan->indexRelation); initSpGistState(&so->state, scan->indexRelation);
so->tempCxt = AllocSetContextCreate(CurrentMemoryContext, so->tempCxt = AllocSetContextCreate(CurrentMemoryContext,
"SP-GiST search temporary context", "SP-GiST search temporary context",
ALLOCSET_DEFAULT_MINSIZE, ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE, ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE); ALLOCSET_DEFAULT_MAXSIZE);
resetSpGistScanOpaque(so);
/* Set up indexTupDesc and xs_itupdesc in case it's an index-only scan */ /* Set up indexTupDesc and xs_itupdesc in case it's an index-only scan */
so->indexTupDesc = scan->xs_itupdesc = RelationGetDescr(rel); so->indexTupDesc = scan->xs_itupdesc = RelationGetDescr(rel);
@ -115,12 +201,17 @@ spgrescan(PG_FUNCTION_ARGS)
SpGistScanOpaque so = (SpGistScanOpaque) scan->opaque; SpGistScanOpaque so = (SpGistScanOpaque) scan->opaque;
ScanKey scankey = (ScanKey) PG_GETARG_POINTER(1); ScanKey scankey = (ScanKey) PG_GETARG_POINTER(1);
/* copy scankeys into local storage */
if (scankey && scan->numberOfKeys > 0) if (scankey && scan->numberOfKeys > 0)
{ {
memmove(scan->keyData, scankey, memmove(scan->keyData, scankey,
scan->numberOfKeys * sizeof(ScanKeyData)); scan->numberOfKeys * sizeof(ScanKeyData));
} }
/* preprocess scankeys, set up the representation in *so */
spgPrepareScanKeys(scan);
/* set up starting stack entries */
resetSpGistScanOpaque(so); resetSpGistScanOpaque(so);
PG_RETURN_VOID(); PG_RETURN_VOID();
@ -162,53 +253,34 @@ spgLeafTest(Relation index, SpGistScanOpaque so, Datum leafDatum,
int level, Datum reconstructedValue, int level, Datum reconstructedValue,
Datum *leafValue, bool *recheck) Datum *leafValue, bool *recheck)
{ {
bool result = true; bool result;
spgLeafConsistentIn in; spgLeafConsistentIn in;
spgLeafConsistentOut out; spgLeafConsistentOut out;
FmgrInfo *procinfo; FmgrInfo *procinfo;
MemoryContext oldCtx; MemoryContext oldCtx;
int i;
*leafValue = (Datum) 0; /* use temp context for calling leaf_consistent */
*recheck = false; oldCtx = MemoryContextSwitchTo(so->tempCxt);
/* set up values that are the same for all quals */ in.scankeys = so->keyData;
in.nkeys = so->numberOfKeys;
in.reconstructedValue = reconstructedValue; in.reconstructedValue = reconstructedValue;
in.level = level; in.level = level;
in.returnData = so->want_itup; in.returnData = so->want_itup;
in.leafDatum = leafDatum; in.leafDatum = leafDatum;
/* Apply each leaf consistency check, working in the temp context */ out.leafValue = (Datum) 0;
oldCtx = MemoryContextSwitchTo(so->tempCxt); out.recheck = false;
procinfo = index_getprocinfo(index, 1, SPGIST_LEAF_CONSISTENT_PROC); procinfo = index_getprocinfo(index, 1, SPGIST_LEAF_CONSISTENT_PROC);
result = DatumGetBool(FunctionCall2Coll(procinfo,
index->rd_indcollation[0],
PointerGetDatum(&in),
PointerGetDatum(&out)));
for (i = 0; i < so->numberOfKeys; i++) *leafValue = out.leafValue;
{ *recheck = out.recheck;
ScanKey skey = &so->keyData[i];
/* Assume SPGiST-indexable operators are strict */
if (skey->sk_flags & SK_ISNULL)
{
result = false;
break;
}
in.strategy = skey->sk_strategy;
in.query = skey->sk_argument;
out.leafValue = (Datum) 0;
out.recheck = false;
result = DatumGetBool(FunctionCall2Coll(procinfo,
skey->sk_collation,
PointerGetDatum(&in),
PointerGetDatum(&out)));
*leafValue = out.leafValue;
*recheck |= out.recheck;
if (!result)
break;
}
MemoryContextSwitchTo(oldCtx); MemoryContextSwitchTo(oldCtx);
return result; return result;
@ -349,8 +421,13 @@ redirect:
else /* page is inner */ else /* page is inner */
{ {
SpGistInnerTuple innerTuple; SpGistInnerTuple innerTuple;
spgInnerConsistentIn in;
spgInnerConsistentOut out;
FmgrInfo *procinfo;
SpGistNodeTuple *nodes;
SpGistNodeTuple node; SpGistNodeTuple node;
int i; int i;
MemoryContext oldCtx;
innerTuple = (SpGistInnerTuple) PageGetItem(page, innerTuple = (SpGistInnerTuple) PageGetItem(page,
PageGetItemId(page, offset)); PageGetItemId(page, offset));
@ -368,144 +445,68 @@ redirect:
innerTuple->tupstate); innerTuple->tupstate);
} }
if (so->numberOfKeys == 0) /* use temp context for calling inner_consistent */
oldCtx = MemoryContextSwitchTo(so->tempCxt);
in.scankeys = so->keyData;
in.nkeys = so->numberOfKeys;
in.reconstructedValue = stackEntry->reconstructedValue;
in.level = stackEntry->level;
in.returnData = so->want_itup;
in.allTheSame = innerTuple->allTheSame;
in.hasPrefix = (innerTuple->prefixSize > 0);
in.prefixDatum = SGITDATUM(innerTuple, &so->state);
in.nNodes = innerTuple->nNodes;
in.nodeLabels = spgExtractNodeLabels(&so->state, innerTuple);
/* collect node pointers */
nodes = (SpGistNodeTuple *) palloc(sizeof(SpGistNodeTuple) * in.nNodes);
SGITITERATE(innerTuple, i, node)
{ {
/* nodes[i] = node;
* This case cannot happen at the moment, because we don't
* set pg_am.amoptionalkey for SP-GiST. In order for full
* index scans to produce correct answers, we'd need to
* index nulls, which we don't.
*/
Assert(false);
#ifdef NOT_USED
/*
* A full index scan could be done approximately like this,
* but note that reconstruction of indexed values would be
* impossible unless the API for inner_consistent is changed.
*/
SGITITERATE(innerTuple, i, node)
{
if (ItemPointerIsValid(&node->t_tid))
{
ScanStackEntry *newEntry = palloc(sizeof(ScanStackEntry));
newEntry->ptr = node->t_tid;
newEntry->level = -1;
newEntry->reconstructedValue = (Datum) 0;
so->scanStack = lcons(newEntry, so->scanStack);
}
}
#endif
} }
else
memset(&out, 0, sizeof(out));
procinfo = index_getprocinfo(index, 1, SPGIST_INNER_CONSISTENT_PROC);
FunctionCall2Coll(procinfo,
index->rd_indcollation[0],
PointerGetDatum(&in),
PointerGetDatum(&out));
MemoryContextSwitchTo(oldCtx);
/* If allTheSame, they should all or none of 'em match */
if (innerTuple->allTheSame)
if (out.nNodes != 0 && out.nNodes != in.nNodes)
elog(ERROR, "inconsistent inner_consistent results for allTheSame inner tuple");
for (i = 0; i < out.nNodes; i++)
{ {
spgInnerConsistentIn in; int nodeN = out.nodeNumbers[i];
spgInnerConsistentOut out;
FmgrInfo *procinfo;
SpGistNodeTuple *nodes;
int *andMap;
int *levelAdds;
Datum *reconstructedValues;
int j,
nMatches = 0;
MemoryContext oldCtx;
/* use temp context for calling inner_consistent */ Assert(nodeN >= 0 && nodeN < in.nNodes);
oldCtx = MemoryContextSwitchTo(so->tempCxt); if (ItemPointerIsValid(&nodes[nodeN]->t_tid))
/* set up values that are the same for all scankeys */
in.reconstructedValue = stackEntry->reconstructedValue;
in.level = stackEntry->level;
in.returnData = so->want_itup;
in.allTheSame = innerTuple->allTheSame;
in.hasPrefix = (innerTuple->prefixSize > 0);
in.prefixDatum = SGITDATUM(innerTuple, &so->state);
in.nNodes = innerTuple->nNodes;
in.nodeLabels = spgExtractNodeLabels(&so->state, innerTuple);
/* collect node pointers */
nodes = (SpGistNodeTuple *) palloc(sizeof(SpGistNodeTuple) * in.nNodes);
SGITITERATE(innerTuple, i, node)
{ {
nodes[i] = node; ScanStackEntry *newEntry;
}
andMap = (int *) palloc0(sizeof(int) * in.nNodes); /* Create new work item for this node */
levelAdds = (int *) palloc0(sizeof(int) * in.nNodes); newEntry = palloc(sizeof(ScanStackEntry));
reconstructedValues = (Datum *) palloc0(sizeof(Datum) * in.nNodes); newEntry->ptr = nodes[nodeN]->t_tid;
if (out.levelAdds)
newEntry->level = stackEntry->level + out.levelAdds[i];
else
newEntry->level = stackEntry->level;
/* Must copy value out of temp context */
if (out.reconstructedValues)
newEntry->reconstructedValue =
datumCopy(out.reconstructedValues[i],
so->state.attType.attbyval,
so->state.attType.attlen);
else
newEntry->reconstructedValue = (Datum) 0;
procinfo = index_getprocinfo(index, 1, SPGIST_INNER_CONSISTENT_PROC); so->scanStack = lcons(newEntry, so->scanStack);
for (j = 0; j < so->numberOfKeys; j++)
{
ScanKey skey = &so->keyData[j];
/* Assume SPGiST-indexable operators are strict */
if (skey->sk_flags & SK_ISNULL)
{
nMatches = 0;
break;
}
in.strategy = skey->sk_strategy;
in.query = skey->sk_argument;
memset(&out, 0, sizeof(out));
FunctionCall2Coll(procinfo,
skey->sk_collation,
PointerGetDatum(&in),
PointerGetDatum(&out));
/* If allTheSame, they should all or none of 'em match */
if (innerTuple->allTheSame)
if (out.nNodes != 0 && out.nNodes != in.nNodes)
elog(ERROR, "inconsistent inner_consistent results for allTheSame inner tuple");
nMatches = 0;
for (i = 0; i < out.nNodes; i++)
{
int nodeN = out.nodeNumbers[i];
andMap[nodeN]++;
if (andMap[nodeN] == j + 1)
nMatches++;
if (out.levelAdds)
levelAdds[nodeN] = out.levelAdds[i];
if (out.reconstructedValues)
reconstructedValues[nodeN] = out.reconstructedValues[i];
}
/* quit as soon as all nodes have failed some qual */
if (nMatches == 0)
break;
}
MemoryContextSwitchTo(oldCtx);
if (nMatches > 0)
{
for (i = 0; i < in.nNodes; i++)
{
if (andMap[i] == so->numberOfKeys &&
ItemPointerIsValid(&nodes[i]->t_tid))
{
ScanStackEntry *newEntry;
/* Create new work item for this node */
newEntry = palloc(sizeof(ScanStackEntry));
newEntry->ptr = nodes[i]->t_tid;
newEntry->level = stackEntry->level + levelAdds[i];
/* Must copy value out of temp context */
newEntry->reconstructedValue =
datumCopy(reconstructedValues[i],
so->state.attType.attbyval,
so->state.attType.attlen);
so->scanStack = lcons(newEntry, so->scanStack);
}
}
} }
} }
} }
@ -536,10 +537,7 @@ spggetbitmap(PG_FUNCTION_ARGS)
TIDBitmap *tbm = (TIDBitmap *) PG_GETARG_POINTER(1); TIDBitmap *tbm = (TIDBitmap *) PG_GETARG_POINTER(1);
SpGistScanOpaque so = (SpGistScanOpaque) scan->opaque; SpGistScanOpaque so = (SpGistScanOpaque) scan->opaque;
/* Copy scankey to *so so we don't need to pass it around separately */ /* Copy want_itup to *so so we don't need to pass it around separately */
so->numberOfKeys = scan->numberOfKeys;
so->keyData = scan->keyData;
/* Ditto for the want_itup flag */
so->want_itup = false; so->want_itup = false;
so->tbm = tbm; so->tbm = tbm;
@ -583,10 +581,7 @@ spggettuple(PG_FUNCTION_ARGS)
if (dir != ForwardScanDirection) if (dir != ForwardScanDirection)
elog(ERROR, "SP-GiST only supports forward scan direction"); elog(ERROR, "SP-GiST only supports forward scan direction");
/* Copy scankey to *so so we don't need to pass it around separately */ /* Copy want_itup to *so so we don't need to pass it around separately */
so->numberOfKeys = scan->numberOfKeys;
so->keyData = scan->keyData;
/* Ditto for the want_itup flag */
so->want_itup = scan->xs_want_itup; so->want_itup = scan->xs_want_itup;
for (;;) for (;;)

View File

@ -362,25 +362,12 @@ spg_text_inner_consistent(PG_FUNCTION_ARGS)
{ {
spgInnerConsistentIn *in = (spgInnerConsistentIn *) PG_GETARG_POINTER(0); spgInnerConsistentIn *in = (spgInnerConsistentIn *) PG_GETARG_POINTER(0);
spgInnerConsistentOut *out = (spgInnerConsistentOut *) PG_GETARG_POINTER(1); spgInnerConsistentOut *out = (spgInnerConsistentOut *) PG_GETARG_POINTER(1);
StrategyNumber strategy = in->strategy; bool collate_is_c = lc_collate_is_c(PG_GET_COLLATION());
text *inText;
int inSize;
int i;
text *reconstrText = NULL; text *reconstrText = NULL;
int maxReconstrLen = 0; int maxReconstrLen = 0;
text *prefixText = NULL; text *prefixText = NULL;
int prefixSize = 0; int prefixSize = 0;
int i;
/*
* If it's a collation-aware operator, but the collation is C, we can
* treat it as non-collation-aware.
*/
if (strategy > 10 &&
lc_collate_is_c(PG_GET_COLLATION()))
strategy -= 10;
inText = DatumGetTextPP(in->query);
inSize = VARSIZE_ANY_EXHDR(inText);
/* /*
* Reconstruct values represented at this tuple, including parent data, * Reconstruct values represented at this tuple, including parent data,
@ -431,8 +418,8 @@ spg_text_inner_consistent(PG_FUNCTION_ARGS)
{ {
uint8 nodeChar = DatumGetUInt8(in->nodeLabels[i]); uint8 nodeChar = DatumGetUInt8(in->nodeLabels[i]);
int thisLen; int thisLen;
int r; bool res = true;
bool res = false; int j;
/* If nodeChar is zero, don't include it in data */ /* If nodeChar is zero, don't include it in data */
if (nodeChar == '\0') if (nodeChar == '\0')
@ -443,38 +430,57 @@ spg_text_inner_consistent(PG_FUNCTION_ARGS)
thisLen = maxReconstrLen; thisLen = maxReconstrLen;
} }
r = memcmp(VARDATA(reconstrText), VARDATA_ANY(inText), for (j = 0; j < in->nkeys; j++)
Min(inSize, thisLen));
switch (strategy)
{ {
case BTLessStrategyNumber: StrategyNumber strategy = in->scankeys[j].sk_strategy;
case BTLessEqualStrategyNumber: text *inText;
if (r <= 0) int inSize;
res = true; int r;
break;
case BTEqualStrategyNumber: /*
if (r == 0 && inSize >= thisLen) * If it's a collation-aware operator, but the collation is C, we
res = true; * can treat it as non-collation-aware. With non-C collation we
break; * need to traverse whole tree :-( so there's no point in making
case BTGreaterEqualStrategyNumber: * any check here.
case BTGreaterStrategyNumber: */
if (r >= 0) if (strategy > 10)
res = true; {
break; if (collate_is_c)
case BTLessStrategyNumber + 10: strategy -= 10;
case BTLessEqualStrategyNumber + 10: else
case BTGreaterEqualStrategyNumber + 10: continue;
case BTGreaterStrategyNumber + 10: }
/*
* with non-C collation we need to traverse whole tree :-( inText = DatumGetTextPP(in->scankeys[j].sk_argument);
*/ inSize = VARSIZE_ANY_EXHDR(inText);
res = true;
break; r = memcmp(VARDATA(reconstrText), VARDATA_ANY(inText),
default: Min(inSize, thisLen));
elog(ERROR, "unrecognized strategy number: %d",
in->strategy); switch (strategy)
break; {
case BTLessStrategyNumber:
case BTLessEqualStrategyNumber:
if (r > 0)
res = false;
break;
case BTEqualStrategyNumber:
if (r != 0 || inSize < thisLen)
res = false;
break;
case BTGreaterEqualStrategyNumber:
case BTGreaterStrategyNumber:
if (r < 0)
res = false;
break;
default:
elog(ERROR, "unrecognized strategy number: %d",
in->scankeys[j].sk_strategy);
break;
}
if (!res)
break; /* no need to consider remaining conditions */
} }
if (res) if (res)
@ -496,16 +502,13 @@ spg_text_leaf_consistent(PG_FUNCTION_ARGS)
{ {
spgLeafConsistentIn *in = (spgLeafConsistentIn *) PG_GETARG_POINTER(0); spgLeafConsistentIn *in = (spgLeafConsistentIn *) PG_GETARG_POINTER(0);
spgLeafConsistentOut *out = (spgLeafConsistentOut *) PG_GETARG_POINTER(1); spgLeafConsistentOut *out = (spgLeafConsistentOut *) PG_GETARG_POINTER(1);
StrategyNumber strategy = in->strategy;
text *query = DatumGetTextPP(in->query);
int level = in->level; int level = in->level;
text *leafValue, text *leafValue,
*reconstrValue = NULL; *reconstrValue = NULL;
char *fullValue; char *fullValue;
int fullLen; int fullLen;
int queryLen;
int r;
bool res; bool res;
int j;
/* all tests are exact */ /* all tests are exact */
out->recheck = false; out->recheck = false;
@ -518,18 +521,8 @@ spg_text_leaf_consistent(PG_FUNCTION_ARGS)
Assert(level == 0 ? reconstrValue == NULL : Assert(level == 0 ? reconstrValue == NULL :
VARSIZE_ANY_EXHDR(reconstrValue) == level); VARSIZE_ANY_EXHDR(reconstrValue) == level);
/* Reconstruct the full string represented by this leaf tuple */
fullLen = level + VARSIZE_ANY_EXHDR(leafValue); fullLen = level + VARSIZE_ANY_EXHDR(leafValue);
queryLen = VARSIZE_ANY_EXHDR(query);
/*
* For an equality check, we needn't reconstruct fullValue if not same
* length; it can't match
*/
if (strategy == BTEqualStrategyNumber && queryLen != fullLen)
PG_RETURN_BOOL(false);
/* Else, reconstruct the full string represented by this leaf tuple */
if (VARSIZE_ANY_EXHDR(leafValue) == 0 && level > 0) if (VARSIZE_ANY_EXHDR(leafValue) == 0 && level > 0)
{ {
fullValue = VARDATA(reconstrValue); fullValue = VARDATA(reconstrValue);
@ -549,54 +542,67 @@ spg_text_leaf_consistent(PG_FUNCTION_ARGS)
out->leafValue = PointerGetDatum(fullText); out->leafValue = PointerGetDatum(fullText);
} }
/* Run the appropriate type of comparison */ /* Perform the required comparison(s) */
if (strategy > 10) res = true;
for (j = 0; j < in->nkeys; j++)
{ {
/* Collation-aware comparison */ StrategyNumber strategy = in->scankeys[j].sk_strategy;
strategy -= 10; text *query = DatumGetTextPP(in->scankeys[j].sk_argument);
int queryLen = VARSIZE_ANY_EXHDR(query);
int r;
/* If asserts are enabled, verify encoding of reconstructed string */ if (strategy > 10)
Assert(pg_verifymbstr(fullValue, fullLen, false)); {
/* Collation-aware comparison */
strategy -= 10;
r = varstr_cmp(fullValue, Min(queryLen, fullLen), /* If asserts enabled, verify encoding of reconstructed string */
VARDATA_ANY(query), Min(queryLen, fullLen), Assert(pg_verifymbstr(fullValue, fullLen, false));
PG_GET_COLLATION());
}
else
{
/* Non-collation-aware comparison */
r = memcmp(fullValue, VARDATA_ANY(query), Min(queryLen, fullLen));
}
if (r == 0) r = varstr_cmp(fullValue, Min(queryLen, fullLen),
{ VARDATA_ANY(query), Min(queryLen, fullLen),
if (queryLen > fullLen) PG_GET_COLLATION());
r = -1; }
else if (queryLen < fullLen) else
r = 1; {
} /* Non-collation-aware comparison */
r = memcmp(fullValue, VARDATA_ANY(query), Min(queryLen, fullLen));
}
switch (strategy) if (r == 0)
{ {
case BTLessStrategyNumber: if (queryLen > fullLen)
res = (r < 0); r = -1;
break; else if (queryLen < fullLen)
case BTLessEqualStrategyNumber: r = 1;
res = (r <= 0); }
break;
case BTEqualStrategyNumber: switch (strategy)
res = (r == 0); {
break; case BTLessStrategyNumber:
case BTGreaterEqualStrategyNumber: res = (r < 0);
res = (r >= 0); break;
break; case BTLessEqualStrategyNumber:
case BTGreaterStrategyNumber: res = (r <= 0);
res = (r > 0); break;
break; case BTEqualStrategyNumber:
default: res = (r == 0);
elog(ERROR, "unrecognized strategy number: %d", in->strategy); break;
res = false; case BTGreaterEqualStrategyNumber:
break; res = (r >= 0);
break;
case BTGreaterStrategyNumber:
res = (r > 0);
break;
default:
elog(ERROR, "unrecognized strategy number: %d",
in->scankeys[j].sk_strategy);
res = false;
break;
}
if (!res)
break; /* no need to consider remaining conditions */
} }
PG_RETURN_BOOL(res); PG_RETURN_BOOL(res);

View File

@ -128,8 +128,8 @@ typedef struct spgPickSplitOut
*/ */
typedef struct spgInnerConsistentIn typedef struct spgInnerConsistentIn
{ {
StrategyNumber strategy; /* operator strategy number */ ScanKey scankeys; /* array of operators and comparison values */
Datum query; /* operator's RHS value */ int nkeys; /* length of array */
Datum reconstructedValue; /* value reconstructed at parent */ Datum reconstructedValue; /* value reconstructed at parent */
int level; /* current level (counting from zero) */ int level; /* current level (counting from zero) */
@ -156,8 +156,8 @@ typedef struct spgInnerConsistentOut
*/ */
typedef struct spgLeafConsistentIn typedef struct spgLeafConsistentIn
{ {
StrategyNumber strategy; /* operator strategy number */ ScanKey scankeys; /* array of operators and comparison values */
Datum query; /* operator's RHS value */ int nkeys; /* length of array */
Datum reconstructedValue; /* value reconstructed at parent */ Datum reconstructedValue; /* value reconstructed at parent */
int level; /* current level (counting from zero) */ int level; /* current level (counting from zero) */

View File

@ -126,7 +126,11 @@ typedef struct SpGistScanOpaqueData
SpGistState state; /* see above */ SpGistState state; /* see above */
MemoryContext tempCxt; /* short-lived memory context */ MemoryContext tempCxt; /* short-lived memory context */
/* Index quals for scan (copied from IndexScanDesc for convenience) */ /* Control flags showing whether to search nulls and/or non-nulls */
bool searchNulls; /* scan matches (all) null entries */
bool searchNonNulls; /* scan matches (some) non-null entries */
/* Index quals to be passed to opclass (null-related quals removed) */
int numberOfKeys; /* number of index qualifier conditions */ int numberOfKeys; /* number of index qualifier conditions */
ScanKey keyData; /* array of index qualifier descriptors */ ScanKey keyData; /* array of index qualifier descriptors */