database/postgres
1
0
Fork 0
mirror of https://github.com/postgres/postgres.git synced 2025-06-23 14:01:44 +03:00
Code Activity

Redesign get_attstatsslot()/free_attstatsslot() for more safety and speed.

Browse Source 
The mess cleaned up in commit da0759600 is clear evidence that it's a
bug hazard to expect the caller of get_attstatsslot()/free_attstatsslot()
to provide the correct type OID for the array elements in the slot.
Moreover, we weren't even getting any performance benefit from that,
since get_attstatsslot() was extracting the real type OID from the array
anyway.  So we ought to get rid of that requirement; indeed, it would
make more sense for get_attstatsslot() to pass back the type OID it found,
in case the caller isn't sure what to expect, which is likely in binary-
compatible-operator cases.

Another problem with the current implementation is that if the stats array
element type is pass-by-reference, we incur a palloc/memcpy/pfree cycle
for each element.  That seemed acceptable when the code was written because
we were targeting O(10) array sizes --- but these days, stats arrays are
almost always bigger than that, sometimes much bigger.  We can save a
significant number of cycles by doing one palloc/memcpy/pfree of the whole
array.  Indeed, in the now-probably-common case where the array is toasted,
that happens anyway so this method is basically free.  (Note: although the
catcache code will inline any out-of-line toasted values, it doesn't
decompress them.  At the other end of the size range, it doesn't expand
short-header datums either.  In either case, DatumGetArrayTypeP would have
to make a copy.  We do end up using an extra array copy step if the element
type is pass-by-value and the array length is neither small enough for a
short header nor large enough to have suffered compression.  But that
seems like a very acceptable price for winning in pass-by-ref cases.)

Hence, redesign to take these insights into account.  While at it,
convert to an API in which we fill a struct rather than passing a bunch
of pointers to individual output arguments.  That will make it less
painful if we ever want further expansion of what get_attstatsslot can
pass back.

It's certainly arguable that this is new development and not something to
push post-feature-freeze.  However, I view it as primarily bug-proofing
and therefore something that's better to have sooner not later.  Since
we aren't quite at beta phase yet, let's put it in.

Discussion: https://postgr.es/m/16364.1494520862@sss.pgh.pa.us
This commit is contained in:
Tom Lane
2017-05-13 15:14:39 -04:00
parent 1848b73d45
commit 9aab83fc50
10 changed files with 434 additions and 599 deletions
Download Patch File Download Diff File Expand all files Collapse all files

417
src/backend/utils/adt/selfuncs.c
View File

@ -299,10 +299,7 @@ var_eq_const(VariableStatData *vardata, Oid operator,
(opfuncoid = get_opcode(operator))))
{
Form_pg_statistic stats;
Datum *values;
int nvalues;
float4 *numbers;
int nnumbers;
AttStatsSlot sslot;
bool match = false;
int i;
@ -315,30 +312,27 @@ var_eq_const(VariableStatData *vardata, Oid operator,
* don't like this, maybe you shouldn't be using eqsel for your
* operator...)
*/
if (get_attstatsslot(vardata->statsTuple,
vardata->atttype, vardata->atttypmod,
if (get_attstatsslot(&sslot, vardata->statsTuple,
STATISTIC_KIND_MCV, InvalidOid,
NULL,
&values, &nvalues,
&numbers, &nnumbers))
ATTSTATSSLOT_VALUES | ATTSTATSSLOT_NUMBERS))
{
FmgrInfo eqproc;
fmgr_info(opfuncoid, &eqproc);
for (i = 0; i < nvalues; i++)
for (i = 0; i < sslot.nvalues; i++)
{
/* be careful to apply operator right way 'round */
if (varonleft)
match = DatumGetBool(FunctionCall2Coll(&eqproc,
DEFAULT_COLLATION_OID,
values[i],
sslot.values[i],
constval));
else
match = DatumGetBool(FunctionCall2Coll(&eqproc,
DEFAULT_COLLATION_OID,
constval,
values[i]));
sslot.values[i]));
if (match)
break;
}
@ -346,9 +340,7 @@ var_eq_const(VariableStatData *vardata, Oid operator,
else
{
/* no most-common-value info available */
values = NULL;
numbers = NULL;
i = nvalues = nnumbers = 0;
i = 0; /* keep compiler quiet */
}
if (match)
@ -357,7 +349,7 @@ var_eq_const(VariableStatData *vardata, Oid operator,
* Constant is "=" to this common value. We know selectivity
* exactly (or as exactly as ANALYZE could calculate it, anyway).
*/
selec = numbers[i];
selec = sslot.numbers[i];
}
else
{
@ -369,8 +361,8 @@ var_eq_const(VariableStatData *vardata, Oid operator,
double sumcommon = 0.0;
double otherdistinct;
for (i = 0; i < nnumbers; i++)
sumcommon += numbers[i];
for (i = 0; i < sslot.nnumbers; i++)
sumcommon += sslot.numbers[i];
selec = 1.0 - sumcommon - stats->stanullfrac;
CLAMP_PROBABILITY(selec);
@ -379,7 +371,8 @@ var_eq_const(VariableStatData *vardata, Oid operator,
* all the not-common values share this remaining fraction
* equally, so we divide by the number of other distinct values.
*/
otherdistinct = get_variable_numdistinct(vardata, &isdefault) - nnumbers;
otherdistinct = get_variable_numdistinct(vardata, &isdefault) -
sslot.nnumbers;
if (otherdistinct > 1)
selec /= otherdistinct;
@ -387,12 +380,11 @@ var_eq_const(VariableStatData *vardata, Oid operator,
* Another cross-check: selectivity shouldn't be estimated as more
* than the least common "most common value".
*/
if (nnumbers > 0 && selec > numbers[nnumbers - 1])
selec = numbers[nnumbers - 1];
if (sslot.nnumbers > 0 && selec > sslot.numbers[sslot.nnumbers - 1])
selec = sslot.numbers[sslot.nnumbers - 1];
}
free_attstatsslot(vardata->atttype, values, nvalues,
numbers, nnumbers);
free_attstatsslot(&sslot);
}
else
{
@ -435,8 +427,7 @@ var_eq_non_const(VariableStatData *vardata, Oid operator,
{
Form_pg_statistic stats;
double ndistinct;
float4 *numbers;
int nnumbers;
AttStatsSlot sslot;
stats = (Form_pg_statistic) GETSTRUCT(vardata->statsTuple);
@ -459,16 +450,13 @@ var_eq_non_const(VariableStatData *vardata, Oid operator,
* Cross-check: selectivity should never be estimated as more than the
* most common value's.
*/
if (get_attstatsslot(vardata->statsTuple,
vardata->atttype, vardata->atttypmod,
if (get_attstatsslot(&sslot, vardata->statsTuple,
STATISTIC_KIND_MCV, InvalidOid,
NULL,
NULL, NULL,
&numbers, &nnumbers))
ATTSTATSSLOT_NUMBERS))
{
if (nnumbers > 0 && selec > numbers[0])
selec = numbers[0];
free_attstatsslot(vardata->atttype, NULL, 0, numbers, nnumbers);
if (sslot.nnumbers > 0 && selec > sslot.numbers[0])
selec = sslot.numbers[0];
free_attstatsslot(&sslot);
}
}
else
@ -620,10 +608,7 @@ mcv_selectivity(VariableStatData *vardata, FmgrInfo *opproc,
{
double mcv_selec,
sumcommon;
Datum *values;
int nvalues;
float4 *numbers;
int nnumbers;
AttStatsSlot sslot;
int i;
mcv_selec = 0.0;
@ -631,29 +616,25 @@ mcv_selectivity(VariableStatData *vardata, FmgrInfo *opproc,
if (HeapTupleIsValid(vardata->statsTuple) &&
statistic_proc_security_check(vardata, opproc->fn_oid) &&
get_attstatsslot(vardata->statsTuple,
vardata->atttype, vardata->atttypmod,
get_attstatsslot(&sslot, vardata->statsTuple,
STATISTIC_KIND_MCV, InvalidOid,
NULL,
&values, &nvalues,
&numbers, &nnumbers))
ATTSTATSSLOT_VALUES | ATTSTATSSLOT_NUMBERS))
{
for (i = 0; i < nvalues; i++)
for (i = 0; i < sslot.nvalues; i++)
{
if (varonleft ?
DatumGetBool(FunctionCall2Coll(opproc,
DEFAULT_COLLATION_OID,
values[i],
sslot.values[i],
constval)) :
DatumGetBool(FunctionCall2Coll(opproc,
DEFAULT_COLLATION_OID,
constval,
values[i])))
mcv_selec += numbers[i];
sumcommon += numbers[i];
sslot.values[i])))
mcv_selec += sslot.numbers[i];
sumcommon += sslot.numbers[i];
}
free_attstatsslot(vardata->atttype, values, nvalues,
numbers, nnumbers);
free_attstatsslot(&sslot);
}
*sumcommonp = sumcommon;
@ -699,8 +680,7 @@ histogram_selectivity(VariableStatData *vardata, FmgrInfo *opproc,
int *hist_size)
{
double result;
Datum *values;
int nvalues;
AttStatsSlot sslot;
/* check sanity of parameters */
Assert(n_skip >= 0);
@ -708,37 +688,34 @@ histogram_selectivity(VariableStatData *vardata, FmgrInfo *opproc,
if (HeapTupleIsValid(vardata->statsTuple) &&
statistic_proc_security_check(vardata, opproc->fn_oid) &&
get_attstatsslot(vardata->statsTuple,
vardata->atttype, vardata->atttypmod,
get_attstatsslot(&sslot, vardata->statsTuple,
STATISTIC_KIND_HISTOGRAM, InvalidOid,
NULL,
&values, &nvalues,
NULL, NULL))
ATTSTATSSLOT_VALUES))
{
*hist_size = nvalues;
if (nvalues >= min_hist_size)
*hist_size = sslot.nvalues;
if (sslot.nvalues >= min_hist_size)
{
int nmatch = 0;
int i;
for (i = n_skip; i < nvalues - n_skip; i++)
for (i = n_skip; i < sslot.nvalues - n_skip; i++)
{
if (varonleft ?
DatumGetBool(FunctionCall2Coll(opproc,
DEFAULT_COLLATION_OID,
values[i],
sslot.values[i],
constval)) :
DatumGetBool(FunctionCall2Coll(opproc,
DEFAULT_COLLATION_OID,
constval,
values[i])))
sslot.values[i])))
nmatch++;
}
result = ((double) nmatch) / ((double) (nvalues - 2 * n_skip));
result = ((double) nmatch) / ((double) (sslot.nvalues - 2 * n_skip));
}
else
result = -1;
free_attstatsslot(vardata->atttype, values, nvalues, NULL, 0);
free_attstatsslot(&sslot);
}
else
{
@ -768,9 +745,7 @@ ineq_histogram_selectivity(PlannerInfo *root,
Datum constval, Oid consttype)
{
double hist_selec;
Oid hist_op;
Datum *values;
int nvalues;
AttStatsSlot sslot;
hist_selec = -1.0;
@ -786,14 +761,11 @@ ineq_histogram_selectivity(PlannerInfo *root,
*/
if (HeapTupleIsValid(vardata->statsTuple) &&
statistic_proc_security_check(vardata, opproc->fn_oid) &&
get_attstatsslot(vardata->statsTuple,
vardata->atttype, vardata->atttypmod,
get_attstatsslot(&sslot, vardata->statsTuple,
STATISTIC_KIND_HISTOGRAM, InvalidOid,
&hist_op,
&values, &nvalues,
NULL, NULL))
ATTSTATSSLOT_VALUES))
{
if (nvalues > 1)
if (sslot.nvalues > 1)
{
/*
* Use binary search to find proper location, ie, the first slot
@ -812,7 +784,7 @@ ineq_histogram_selectivity(PlannerInfo *root,
*/
double histfrac;
int lobound = 0; /* first possible slot to search */
int hibound = nvalues; /* last+1 slot to search */
int hibound = sslot.nvalues; /* last+1 slot to search */
bool have_end = false;
/*
@ -821,12 +793,12 @@ ineq_histogram_selectivity(PlannerInfo *root,
* one of them to be updated, so we deal with that within the
* loop.)
*/
if (nvalues == 2)
if (sslot.nvalues == 2)
have_end = get_actual_variable_range(root,
vardata,
hist_op,
&values[0],
&values[1]);
sslot.staop,
&sslot.values[0],
&sslot.values[1]);
while (lobound < hibound)
{
@ -838,22 +810,22 @@ ineq_histogram_selectivity(PlannerInfo *root,
* histogram entry, first try to replace it with the actual
* current min or max (unless we already did so above).
*/
if (probe == 0 && nvalues > 2)
if (probe == 0 && sslot.nvalues > 2)
have_end = get_actual_variable_range(root,
vardata,
hist_op,
&values[0],
sslot.staop,
&sslot.values[0],
NULL);
else if (probe == nvalues - 1 && nvalues > 2)
else if (probe == sslot.nvalues - 1 && sslot.nvalues > 2)
have_end = get_actual_variable_range(root,
vardata,
hist_op,
sslot.staop,
NULL,
&values[probe]);
&sslot.values[probe]);
ltcmp = DatumGetBool(FunctionCall2Coll(opproc,
DEFAULT_COLLATION_OID,
values[probe],
sslot.values[probe],
constval));
if (isgt)
ltcmp = !ltcmp;
@ -868,7 +840,7 @@ ineq_histogram_selectivity(PlannerInfo *root,
/* Constant is below lower histogram boundary. */
histfrac = 0.0;
}
else if (lobound >= nvalues)
else if (lobound >= sslot.nvalues)
{
/* Constant is above upper histogram boundary. */
histfrac = 1.0;
@ -889,7 +861,7 @@ ineq_histogram_selectivity(PlannerInfo *root,
* interpolation within this bin.
*/
if (convert_to_scalar(constval, consttype, &val,
values[i - 1], values[i],
sslot.values[i - 1], sslot.values[i],
vardata->vartype,
&low, &high))
{
@ -936,7 +908,7 @@ ineq_histogram_selectivity(PlannerInfo *root,
* binfrac partial bin below the constant.
*/
histfrac = (double) (i - 1) + binfrac;
histfrac /= (double) (nvalues - 1);
histfrac /= (double) (sslot.nvalues - 1);
}
/*
@ -964,7 +936,7 @@ ineq_histogram_selectivity(PlannerInfo *root,
}
}
free_attstatsslot(vardata->atttype, values, nvalues, NULL, 0);
free_attstatsslot(&sslot);
}
return hist_selec;
@ -1517,21 +1489,15 @@ booltestsel(PlannerInfo *root, BoolTestType booltesttype, Node *arg,
{
Form_pg_statistic stats;
double freq_null;
Datum *values;
int nvalues;
float4 *numbers;
int nnumbers;
AttStatsSlot sslot;
stats = (Form_pg_statistic) GETSTRUCT(vardata.statsTuple);
freq_null = stats->stanullfrac;
if (get_attstatsslot(vardata.statsTuple,
vardata.atttype, vardata.atttypmod,
if (get_attstatsslot(&sslot, vardata.statsTuple,
STATISTIC_KIND_MCV, InvalidOid,
NULL,
&values, &nvalues,
&numbers, &nnumbers)
&& nnumbers > 0)
ATTSTATSSLOT_VALUES | ATTSTATSSLOT_NUMBERS)
&& sslot.nnumbers > 0)
{
double freq_true;
double freq_false;
@ -1539,10 +1505,10 @@ booltestsel(PlannerInfo *root, BoolTestType booltesttype, Node *arg,
/*
* Get first MCV frequency and derive frequency for true.
*/
if (DatumGetBool(values[0]))
freq_true = numbers[0];
if (DatumGetBool(sslot.values[0]))
freq_true = sslot.numbers[0];
else
freq_true = 1.0 - numbers[0] - freq_null;
freq_true = 1.0 - sslot.numbers[0] - freq_null;
/*
* Next derive frequency for false. Then use these as appropriate
@ -1583,8 +1549,7 @@ booltestsel(PlannerInfo *root, BoolTestType booltesttype, Node *arg,
break;
}
free_attstatsslot(vardata.atttype, values, nvalues,
numbers, nnumbers);
free_attstatsslot(&sslot);
}
else
{
@ -2274,34 +2239,26 @@ eqjoinsel_inner(Oid operator,
Form_pg_statistic stats1 = NULL;
Form_pg_statistic stats2 = NULL;
bool have_mcvs1 = false;
Datum *values1 = NULL;
int nvalues1 = 0;
float4 *numbers1 = NULL;
int nnumbers1 = 0;
bool have_mcvs2 = false;
Datum *values2 = NULL;
int nvalues2 = 0;
float4 *numbers2 = NULL;
int nnumbers2 = 0;
AttStatsSlot sslot1;
AttStatsSlot sslot2;
nd1 = get_variable_numdistinct(vardata1, &isdefault1);
nd2 = get_variable_numdistinct(vardata2, &isdefault2);
opfuncoid = get_opcode(operator);
memset(&sslot1, 0, sizeof(sslot1));
memset(&sslot2, 0, sizeof(sslot2));
if (HeapTupleIsValid(vardata1->statsTuple))
{
/* note we allow use of nullfrac regardless of security check */
stats1 = (Form_pg_statistic) GETSTRUCT(vardata1->statsTuple);
if (statistic_proc_security_check(vardata1, opfuncoid))
have_mcvs1 = get_attstatsslot(vardata1->statsTuple,
vardata1->atttype,
vardata1->atttypmod,
STATISTIC_KIND_MCV,
InvalidOid,
NULL,
&values1, &nvalues1,
&numbers1, &nnumbers1);
have_mcvs1 = get_attstatsslot(&sslot1, vardata1->statsTuple,
STATISTIC_KIND_MCV, InvalidOid,
ATTSTATSSLOT_VALUES | ATTSTATSSLOT_NUMBERS);
}
if (HeapTupleIsValid(vardata2->statsTuple))
@ -2309,14 +2266,9 @@ eqjoinsel_inner(Oid operator,
/* note we allow use of nullfrac regardless of security check */
stats2 = (Form_pg_statistic) GETSTRUCT(vardata2->statsTuple);
if (statistic_proc_security_check(vardata2, opfuncoid))
have_mcvs2 = get_attstatsslot(vardata2->statsTuple,
vardata2->atttype,
vardata2->atttypmod,
STATISTIC_KIND_MCV,
InvalidOid,
NULL,
&values2, &nvalues2,
&numbers2, &nnumbers2);
have_mcvs2 = get_attstatsslot(&sslot2, vardata2->statsTuple,
STATISTIC_KIND_MCV, InvalidOid,
ATTSTATSSLOT_VALUES | ATTSTATSSLOT_NUMBERS);
}
if (have_mcvs1 && have_mcvs2)
@ -2351,8 +2303,8 @@ eqjoinsel_inner(Oid operator,
nmatches;
fmgr_info(opfuncoid, &eqproc);
hasmatch1 = (bool *) palloc0(nvalues1 * sizeof(bool));
hasmatch2 = (bool *) palloc0(nvalues2 * sizeof(bool));
hasmatch1 = (bool *) palloc0(sslot1.nvalues * sizeof(bool));
hasmatch2 = (bool *) palloc0(sslot2.nvalues * sizeof(bool));
/*
* Note we assume that each MCV will match at most one member of the
@ -2362,21 +2314,21 @@ eqjoinsel_inner(Oid operator,
*/
matchprodfreq = 0.0;
nmatches = 0;
for (i = 0; i < nvalues1; i++)
for (i = 0; i < sslot1.nvalues; i++)
{
int j;
for (j = 0; j < nvalues2; j++)
for (j = 0; j < sslot2.nvalues; j++)
{
if (hasmatch2[j])
continue;
if (DatumGetBool(FunctionCall2Coll(&eqproc,
DEFAULT_COLLATION_OID,
values1[i],
values2[j])))
sslot1.values[i],
sslot2.values[j])))
{
hasmatch1[i] = hasmatch2[j] = true;
matchprodfreq += numbers1[i] * numbers2[j];
matchprodfreq += sslot1.numbers[i] * sslot2.numbers[j];
nmatches++;
break;
}
@ -2385,22 +2337,22 @@ eqjoinsel_inner(Oid operator,
CLAMP_PROBABILITY(matchprodfreq);
/* Sum up frequencies of matched and unmatched MCVs */
matchfreq1 = unmatchfreq1 = 0.0;
for (i = 0; i < nvalues1; i++)
for (i = 0; i < sslot1.nvalues; i++)
{
if (hasmatch1[i])
matchfreq1 += numbers1[i];
matchfreq1 += sslot1.numbers[i];
else
unmatchfreq1 += numbers1[i];
unmatchfreq1 += sslot1.numbers[i];
}
CLAMP_PROBABILITY(matchfreq1);
CLAMP_PROBABILITY(unmatchfreq1);
matchfreq2 = unmatchfreq2 = 0.0;
for (i = 0; i < nvalues2; i++)
for (i = 0; i < sslot2.nvalues; i++)
{
if (hasmatch2[i])
matchfreq2 += numbers2[i];
matchfreq2 += sslot2.numbers[i];
else
unmatchfreq2 += numbers2[i];
unmatchfreq2 += sslot2.numbers[i];
}
CLAMP_PROBABILITY(matchfreq2);
CLAMP_PROBABILITY(unmatchfreq2);
@ -2425,15 +2377,15 @@ eqjoinsel_inner(Oid operator,
* MCVs plus non-MCV values.
*/
totalsel1 = matchprodfreq;
if (nd2 > nvalues2)
totalsel1 += unmatchfreq1 * otherfreq2 / (nd2 - nvalues2);
if (nd2 > sslot2.nvalues)
totalsel1 += unmatchfreq1 * otherfreq2 / (nd2 - sslot2.nvalues);
if (nd2 > nmatches)
totalsel1 += otherfreq1 * (otherfreq2 + unmatchfreq2) /
(nd2 - nmatches);
/* Same estimate from the point of view of relation 2. */
totalsel2 = matchprodfreq;
if (nd1 > nvalues1)
totalsel2 += unmatchfreq2 * otherfreq1 / (nd1 - nvalues1);
if (nd1 > sslot1.nvalues)
totalsel2 += unmatchfreq2 * otherfreq1 / (nd1 - sslot1.nvalues);
if (nd1 > nmatches)
totalsel2 += otherfreq2 * (otherfreq1 + unmatchfreq1) /
(nd1 - nmatches);
@ -2478,12 +2430,8 @@ eqjoinsel_inner(Oid operator,
selec /= nd2;
}
if (have_mcvs1)
free_attstatsslot(vardata1->atttype, values1, nvalues1,
numbers1, nnumbers1);
if (have_mcvs2)
free_attstatsslot(vardata2->atttype, values2, nvalues2,
numbers2, nnumbers2);
free_attstatsslot(&sslot1);
free_attstatsslot(&sslot2);
return selec;
}
@ -2508,21 +2456,18 @@ eqjoinsel_semi(Oid operator,
Oid opfuncoid;
Form_pg_statistic stats1 = NULL;
bool have_mcvs1 = false;
Datum *values1 = NULL;
int nvalues1 = 0;
float4 *numbers1 = NULL;
int nnumbers1 = 0;
bool have_mcvs2 = false;
Datum *values2 = NULL;
int nvalues2 = 0;
float4 *numbers2 = NULL;
int nnumbers2 = 0;
AttStatsSlot sslot1;
AttStatsSlot sslot2;
nd1 = get_variable_numdistinct(vardata1, &isdefault1);
nd2 = get_variable_numdistinct(vardata2, &isdefault2);
opfuncoid = OidIsValid(operator) ? get_opcode(operator) : InvalidOid;
memset(&sslot1, 0, sizeof(sslot1));
memset(&sslot2, 0, sizeof(sslot2));
/*
* We clamp nd2 to be not more than what we estimate the inner relation's
* size to be. This is intuitively somewhat reasonable since obviously
@ -2561,27 +2506,18 @@ eqjoinsel_semi(Oid operator,
/* note we allow use of nullfrac regardless of security check */
stats1 = (Form_pg_statistic) GETSTRUCT(vardata1->statsTuple);
if (statistic_proc_security_check(vardata1, opfuncoid))
have_mcvs1 = get_attstatsslot(vardata1->statsTuple,
vardata1->atttype,
vardata1->atttypmod,
STATISTIC_KIND_MCV,
InvalidOid,
NULL,
&values1, &nvalues1,
&numbers1, &nnumbers1);
have_mcvs1 = get_attstatsslot(&sslot1, vardata1->statsTuple,
STATISTIC_KIND_MCV, InvalidOid,
ATTSTATSSLOT_VALUES | ATTSTATSSLOT_NUMBERS);
}
if (HeapTupleIsValid(vardata2->statsTuple) &&
statistic_proc_security_check(vardata2, opfuncoid))
{
have_mcvs2 = get_attstatsslot(vardata2->statsTuple,
vardata2->atttype,
vardata2->atttypmod,
STATISTIC_KIND_MCV,
InvalidOid,
NULL,
&values2, &nvalues2,
&numbers2, &nnumbers2);
have_mcvs2 = get_attstatsslot(&sslot2, vardata2->statsTuple,
STATISTIC_KIND_MCV, InvalidOid,
ATTSTATSSLOT_VALUES);
/* note: currently don't need stanumbers from RHS */
}
if (have_mcvs1 && have_mcvs2 && OidIsValid(operator))
@ -2607,15 +2543,15 @@ eqjoinsel_semi(Oid operator,
/*
* The clamping above could have resulted in nd2 being less than
* nvalues2; in which case, we assume that precisely the nd2 most
* common values in the relation will appear in the join input, and so
* compare to only the first nd2 members of the MCV list. Of course
* this is frequently wrong, but it's the best bet we can make.
* sslot2.nvalues; in which case, we assume that precisely the nd2
* most common values in the relation will appear in the join input,
* and so compare to only the first nd2 members of the MCV list. Of
* course this is frequently wrong, but it's the best bet we can make.
*/
clamped_nvalues2 = Min(nvalues2, nd2);
clamped_nvalues2 = Min(sslot2.nvalues, nd2);
fmgr_info(opfuncoid, &eqproc);
hasmatch1 = (bool *) palloc0(nvalues1 * sizeof(bool));
hasmatch1 = (bool *) palloc0(sslot1.nvalues * sizeof(bool));
hasmatch2 = (bool *) palloc0(clamped_nvalues2 * sizeof(bool));
/*
@ -2625,7 +2561,7 @@ eqjoinsel_semi(Oid operator,
* and because the math wouldn't add up...
*/
nmatches = 0;
for (i = 0; i < nvalues1; i++)
for (i = 0; i < sslot1.nvalues; i++)
{
int j;
@ -2635,8 +2571,8 @@ eqjoinsel_semi(Oid operator,
continue;
if (DatumGetBool(FunctionCall2Coll(&eqproc,
DEFAULT_COLLATION_OID,
values1[i],
values2[j])))
sslot1.values[i],
sslot2.values[j])))
{
hasmatch1[i] = hasmatch2[j] = true;
nmatches++;
@ -2646,10 +2582,10 @@ eqjoinsel_semi(Oid operator,
}
/* Sum up frequencies of matched MCVs */
matchfreq1 = 0.0;
for (i = 0; i < nvalues1; i++)
for (i = 0; i < sslot1.nvalues; i++)
{
if (hasmatch1[i])
matchfreq1 += numbers1[i];
matchfreq1 += sslot1.numbers[i];
}
CLAMP_PROBABILITY(matchfreq1);
pfree(hasmatch1);
@ -2704,12 +2640,8 @@ eqjoinsel_semi(Oid operator,
selec = 0.5 * (1.0 - nullfrac1);
}
if (have_mcvs1)
free_attstatsslot(vardata1->atttype, values1, nvalues1,
numbers1, nnumbers1);
if (have_mcvs2)
free_attstatsslot(vardata2->atttype, values2, nvalues2,
numbers2, nnumbers2);
free_attstatsslot(&sslot1);
free_attstatsslot(&sslot2);
return selec;
}
@ -3629,8 +3561,7 @@ estimate_hash_bucketsize(PlannerInfo *root, Node *hashkey, double nbuckets)
mcvfreq,
avgfreq;
bool isdefault;
float4 *numbers;
int nnumbers;
AttStatsSlot sslot;
examine_variable(root, hashkey, 0, &vardata);
@ -3689,20 +3620,16 @@ estimate_hash_bucketsize(PlannerInfo *root, Node *hashkey, double nbuckets)
if (HeapTupleIsValid(vardata.statsTuple))
{
if (get_attstatsslot(vardata.statsTuple,
vardata.atttype, vardata.atttypmod,
if (get_attstatsslot(&sslot, vardata.statsTuple,
STATISTIC_KIND_MCV, InvalidOid,
NULL,
NULL, NULL,
&numbers, &nnumbers))
ATTSTATSSLOT_NUMBERS))
{
/*
* The first MCV stat is for the most common value.
*/
if (nnumbers > 0)
mcvfreq = numbers[0];
free_attstatsslot(vardata.atttype, NULL, 0,
numbers, nnumbers);
if (sslot.nnumbers > 0)
mcvfreq = sslot.numbers[0];
free_attstatsslot(&sslot);
}
}
@ -4572,7 +4499,7 @@ get_join_variables(PlannerInfo *root, List *args, SpecialJoinInfo *sjinfo,
* freefunc: pointer to a function to release statsTuple with.
* vartype: exposed type of the expression; this should always match
* the declared input type of the operator we are estimating for.
* atttype, atttypmod: type data to pass to get_attstatsslot(). This is
* atttype, atttypmod: actual type/typmod of the "var" expression. This is
* commonly the same as the exposed type of the variable argument,
* but can be different in binary-compatible-type cases.
* isunique: TRUE if we were able to match the var to a unique index or a
@ -5125,8 +5052,7 @@ get_variable_range(PlannerInfo *root, VariableStatData *vardata, Oid sortop,
int16 typLen;
bool typByVal;
Oid opfuncoid;
Datum *values;
int nvalues;
AttStatsSlot sslot;
int i;
/*
@ -5167,29 +5093,23 @@ get_variable_range(PlannerInfo *root, VariableStatData *vardata, Oid sortop,
* the one we want, fail --- this suggests that there is data we can't
* use.
*/
if (get_attstatsslot(vardata->statsTuple,
vardata->atttype, vardata->atttypmod,
if (get_attstatsslot(&sslot, vardata->statsTuple,
STATISTIC_KIND_HISTOGRAM, sortop,
NULL,
&values, &nvalues,
NULL, NULL))
ATTSTATSSLOT_VALUES))
{
if (nvalues > 0)
if (sslot.nvalues > 0)
{
tmin = datumCopy(values[0], typByVal, typLen);
tmax = datumCopy(values[nvalues - 1], typByVal, typLen);
tmin = datumCopy(sslot.values[0], typByVal, typLen);
tmax = datumCopy(sslot.values[sslot.nvalues - 1], typByVal, typLen);
have_data = true;
}
free_attstatsslot(vardata->atttype, values, nvalues, NULL, 0);
free_attstatsslot(&sslot);
}
else if (get_attstatsslot(vardata->statsTuple,
vardata->atttype, vardata->atttypmod,
else if (get_attstatsslot(&sslot, vardata->statsTuple,
STATISTIC_KIND_HISTOGRAM, InvalidOid,
NULL,
&values, &nvalues,
NULL, NULL))
0))
{
free_attstatsslot(vardata->atttype, values, nvalues, NULL, 0);
free_attstatsslot(&sslot);
return false;
}
@ -5199,12 +5119,9 @@ get_variable_range(PlannerInfo *root, VariableStatData *vardata, Oid sortop,
* the MCVs. However, usually the MCVs will not be the extreme values, so
* avoid unnecessary data copying.
*/
if (get_attstatsslot(vardata->statsTuple,
vardata->atttype, vardata->atttypmod,
if (get_attstatsslot(&sslot, vardata->statsTuple,
STATISTIC_KIND_MCV, InvalidOid,
NULL,
&values, &nvalues,
NULL, NULL))
ATTSTATSSLOT_VALUES))
{
bool tmin_is_mcv = false;
bool tmax_is_mcv = false;
@ -5212,26 +5129,26 @@ get_variable_range(PlannerInfo *root, VariableStatData *vardata, Oid sortop,
fmgr_info(opfuncoid, &opproc);
for (i = 0; i < nvalues; i++)
for (i = 0; i < sslot.nvalues; i++)
{
if (!have_data)
{
tmin = tmax = values[i];
tmin = tmax = sslot.values[i];
tmin_is_mcv = tmax_is_mcv = have_data = true;
continue;
}
if (DatumGetBool(FunctionCall2Coll(&opproc,
DEFAULT_COLLATION_OID,
values[i], tmin)))
sslot.values[i], tmin)))
{
tmin = values[i];
tmin = sslot.values[i];
tmin_is_mcv = true;
}
if (DatumGetBool(FunctionCall2Coll(&opproc,
DEFAULT_COLLATION_OID,
tmax, values[i])))
tmax, sslot.values[i])))
{
tmax = values[i];
tmax = sslot.values[i];
tmax_is_mcv = true;
}
}
@ -5239,7 +5156,7 @@ get_variable_range(PlannerInfo *root, VariableStatData *vardata, Oid sortop,
tmin = datumCopy(tmin, typByVal, typLen);
if (tmax_is_mcv)
tmax = datumCopy(tmax, typByVal, typLen);
free_attstatsslot(vardata->atttype, values, nvalues, NULL, 0);
free_attstatsslot(&sslot);
}
*min = tmin;
@ -6979,25 +6896,21 @@ btcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
if (HeapTupleIsValid(vardata.statsTuple))
{
Oid sortop;
float4 *numbers;
int nnumbers;
AttStatsSlot sslot;
sortop = get_opfamily_member(index->opfamily[0],
index->opcintype[0],
index->opcintype[0],
BTLessStrategyNumber);
if (OidIsValid(sortop) &&
get_attstatsslot(vardata.statsTuple, InvalidOid, 0,
STATISTIC_KIND_CORRELATION,
sortop,
NULL,
NULL, NULL,
&numbers, &nnumbers))
get_attstatsslot(&sslot, vardata.statsTuple,
STATISTIC_KIND_CORRELATION, sortop,
ATTSTATSSLOT_NUMBERS))
{
double varCorrelation;
Assert(nnumbers == 1);
varCorrelation = numbers[0];
Assert(sslot.nnumbers == 1);
varCorrelation = sslot.numbers[0];
if (index->reverse_sort[0])
varCorrelation = -varCorrelation;
@ -7007,7 +6920,7 @@ btcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
else
costs.indexCorrelation = varCorrelation;
free_attstatsslot(InvalidOid, NULL, 0, numbers, nnumbers);
free_attstatsslot(&sslot);
}
}
@ -7920,25 +7833,21 @@ brincostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
if (HeapTupleIsValid(vardata.statsTuple))
{
float4 *numbers;
int nnumbers;
AttStatsSlot sslot;
if (get_attstatsslot(vardata.statsTuple, InvalidOid, 0,
STATISTIC_KIND_CORRELATION,
InvalidOid,
NULL,
NULL, NULL,
&numbers, &nnumbers))
if (get_attstatsslot(&sslot, vardata.statsTuple,
STATISTIC_KIND_CORRELATION, InvalidOid,
ATTSTATSSLOT_NUMBERS))
{
double varCorrelation = 0.0;
if (nnumbers > 0)
varCorrelation = Abs(numbers[0]);
if (sslot.nnumbers > 0)
varCorrelation = Abs(sslot.numbers[0]);
if (varCorrelation > *indexCorrelation)
*indexCorrelation = varCorrelation;
free_attstatsslot(InvalidOid, NULL, 0, numbers, nnumbers);
free_attstatsslot(&sslot);
}
}