#include <float.h>
#include <math.h>
+#include "access/brin.h"
#include "access/gin.h"
#include "access/htup_details.h"
#include "access/sysattr.h"
#include "catalog/pg_operator.h"
#include "catalog/pg_opfamily.h"
#include "catalog/pg_statistic.h"
+#include "catalog/pg_statistic_ext.h"
#include "catalog/pg_type.h"
#include "executor/executor.h"
#include "mb/pg_wchar.h"
+#include "miscadmin.h"
#include "nodes/makefuncs.h"
#include "nodes/nodeFuncs.h"
#include "optimizer/clauses.h"
#include "parser/parse_clause.h"
#include "parser/parse_coerce.h"
#include "parser/parsetree.h"
+#include "statistics/statistics.h"
+#include "utils/acl.h"
#include "utils/builtins.h"
#include "utils/bytea.h"
#include "utils/date.h"
static double eqjoinsel_semi(Oid operator,
VariableStatData *vardata1, VariableStatData *vardata2,
RelOptInfo *inner_rel);
+static bool estimate_multivariate_ndistinct(PlannerInfo *root,
+ RelOptInfo *rel, List **varinfos, double *ndistinct);
static bool convert_to_scalar(Datum value, Oid valuetypid, double *scaledvalue,
Datum lobound, Datum hibound, Oid boundstypid,
double *scaledlobound, double *scaledhibound);
{
double selec;
bool isdefault;
+ Oid opfuncoid;
/*
* If the constant is NULL, assume operator is strict and return zero, ie,
if (vardata->isunique && vardata->rel && vardata->rel->tuples >= 1.0)
return 1.0 / vardata->rel->tuples;
- if (HeapTupleIsValid(vardata->statsTuple))
+ if (HeapTupleIsValid(vardata->statsTuple) &&
+ statistic_proc_security_check(vardata,
+ (opfuncoid = get_opcode(operator))))
{
Form_pg_statistic stats;
- Datum *values;
- int nvalues;
- float4 *numbers;
- int nnumbers;
+ AttStatsSlot sslot;
bool match = false;
int i;
* 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(get_opcode(operator), &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;
}
else
{
/* no most-common-value info available */
- values = NULL;
- numbers = NULL;
- i = nvalues = nnumbers = 0;
+ i = 0; /* keep compiler quiet */
}
if (match)
* 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
{
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);
* 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;
* 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
{
{
Form_pg_statistic stats;
double ndistinct;
- float4 *numbers;
- int nnumbers;
+ AttStatsSlot sslot;
stats = (Form_pg_statistic) GETSTRUCT(vardata->statsTuple);
* 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
{
double mcv_selec,
sumcommon;
- Datum *values;
- int nvalues;
- float4 *numbers;
- int nnumbers;
+ AttStatsSlot sslot;
int i;
mcv_selec = 0.0;
sumcommon = 0.0;
if (HeapTupleIsValid(vardata->statsTuple) &&
- get_attstatsslot(vardata->statsTuple,
- vardata->atttype, vardata->atttypmod,
+ statistic_proc_security_check(vardata, opproc->fn_oid) &&
+ 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;
int *hist_size)
{
double result;
- Datum *values;
- int nvalues;
+ AttStatsSlot sslot;
/* check sanity of parameters */
Assert(n_skip >= 0);
Assert(min_hist_size > 2 * n_skip);
if (HeapTupleIsValid(vardata->statsTuple) &&
- get_attstatsslot(vardata->statsTuple,
- vardata->atttype, vardata->atttypmod,
+ statistic_proc_security_check(vardata, opproc->fn_oid) &&
+ 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
{
Datum constval, Oid consttype)
{
double hist_selec;
- Oid hist_op;
- Datum *values;
- int nvalues;
+ AttStatsSlot sslot;
hist_selec = -1.0;
* the reverse way if isgt is TRUE.
*/
if (HeapTupleIsValid(vardata->statsTuple) &&
- get_attstatsslot(vardata->statsTuple,
- vardata->atttype, vardata->atttypmod,
+ statistic_proc_security_check(vardata, opproc->fn_oid) &&
+ 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
*/
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;
/*
* 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)
{
* 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;
/* 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;
* 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))
{
* binfrac partial bin below the constant.
*/
histfrac = (double) (i - 1) + binfrac;
- histfrac /= (double) (nvalues - 1);
+ histfrac /= (double) (sslot.nvalues - 1);
}
/*
}
}
- free_attstatsslot(vardata->atttype, values, nvalues, NULL, 0);
+ free_attstatsslot(&sslot);
}
return hist_selec;
{
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;
/*
* 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
break;
}
- free_attstatsslot(vardata.atttype, values, nvalues,
- numbers, nnumbers);
+ free_attstatsslot(&sslot);
}
else
{
double nd2;
bool isdefault1;
bool isdefault2;
+ Oid opfuncoid;
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);
- have_mcvs1 = get_attstatsslot(vardata1->statsTuple,
- vardata1->atttype,
- vardata1->atttypmod,
- STATISTIC_KIND_MCV,
- InvalidOid,
- NULL,
- &values1, &nvalues1,
- &numbers1, &nnumbers1);
+ if (statistic_proc_security_check(vardata1, opfuncoid))
+ have_mcvs1 = get_attstatsslot(&sslot1, vardata1->statsTuple,
+ STATISTIC_KIND_MCV, InvalidOid,
+ ATTSTATSSLOT_VALUES | ATTSTATSSLOT_NUMBERS);
}
if (HeapTupleIsValid(vardata2->statsTuple))
{
+ /* note we allow use of nullfrac regardless of security check */
stats2 = (Form_pg_statistic) GETSTRUCT(vardata2->statsTuple);
- have_mcvs2 = get_attstatsslot(vardata2->statsTuple,
- vardata2->atttype,
- vardata2->atttypmod,
- STATISTIC_KIND_MCV,
- InvalidOid,
- NULL,
- &values2, &nvalues2,
- &numbers2, &nnumbers2);
+ if (statistic_proc_security_check(vardata2, opfuncoid))
+ have_mcvs2 = get_attstatsslot(&sslot2, vardata2->statsTuple,
+ STATISTIC_KIND_MCV, InvalidOid,
+ ATTSTATSSLOT_VALUES | ATTSTATSSLOT_NUMBERS);
}
if (have_mcvs1 && have_mcvs2)
int i,
nmatches;
- fmgr_info(get_opcode(operator), &eqproc);
- hasmatch1 = (bool *) palloc0(nvalues1 * sizeof(bool));
- hasmatch2 = (bool *) palloc0(nvalues2 * sizeof(bool));
+ fmgr_info(opfuncoid, &eqproc);
+ 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
*/
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;
}
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);
* 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);
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;
}
*
* (Also used for anti join, which we are supposed to estimate the same way.)
* Caller has ensured that vardata1 is the LHS variable.
+ * Unlike eqjoinsel_inner, we have to cope with operator being InvalidOid.
*/
static double
eqjoinsel_semi(Oid operator,
double nd2;
bool isdefault1;
bool isdefault2;
+ 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
if (HeapTupleIsValid(vardata1->statsTuple))
{
+ /* note we allow use of nullfrac regardless of security check */
stats1 = (Form_pg_statistic) GETSTRUCT(vardata1->statsTuple);
- have_mcvs1 = get_attstatsslot(vardata1->statsTuple,
- vardata1->atttype,
- vardata1->atttypmod,
- STATISTIC_KIND_MCV,
- InvalidOid,
- NULL,
- &values1, &nvalues1,
- &numbers1, &nnumbers1);
+ if (statistic_proc_security_check(vardata1, opfuncoid))
+ have_mcvs1 = get_attstatsslot(&sslot1, vardata1->statsTuple,
+ STATISTIC_KIND_MCV, InvalidOid,
+ ATTSTATSSLOT_VALUES | ATTSTATSSLOT_NUMBERS);
}
- if (HeapTupleIsValid(vardata2->statsTuple))
+ 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))
/*
* 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(get_opcode(operator), &eqproc);
- hasmatch1 = (bool *) palloc0(nvalues1 * sizeof(bool));
+ fmgr_info(opfuncoid, &eqproc);
+ hasmatch1 = (bool *) palloc0(sslot1.nvalues * sizeof(bool));
hasmatch2 = (bool *) palloc0(clamped_nvalues2 * sizeof(bool));
/*
* 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;
continue;
if (DatumGetBool(FunctionCall2Coll(&eqproc,
DEFAULT_COLLATION_OID,
- values1[i],
- values2[j])))
+ sslot1.values[i],
+ sslot2.values[j])))
{
hasmatch1[i] = hasmatch2[j] = true;
nmatches++;
}
/* 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);
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;
}
{
GroupVarInfo *varinfo1 = (GroupVarInfo *) linitial(varinfos);
RelOptInfo *rel = varinfo1->rel;
- double reldistinct = varinfo1->ndistinct;
+ double reldistinct = 1;
double relmaxndistinct = reldistinct;
- int relvarcount = 1;
+ int relvarcount = 0;
List *newvarinfos = NIL;
+ List *relvarinfos = NIL;
/*
- * Get the product of numdistinct estimates of the Vars for this rel.
- * Also, construct new varinfos list of remaining Vars.
+ * Split the list of varinfos in two - one for the current rel, one
+ * for remaining Vars on other rels.
*/
+ relvarinfos = lcons(varinfo1, relvarinfos);
for_each_cell(l, lnext(list_head(varinfos)))
{
GroupVarInfo *varinfo2 = (GroupVarInfo *) lfirst(l);
if (varinfo2->rel == varinfo1->rel)
{
- reldistinct *= varinfo2->ndistinct;
- if (relmaxndistinct < varinfo2->ndistinct)
- relmaxndistinct = varinfo2->ndistinct;
- relvarcount++;
+ /* varinfos on current rel */
+ relvarinfos = lcons(varinfo2, relvarinfos);
}
else
{
}
}
+ /*
+ * Get the numdistinct estimate for the Vars of this rel. We
+ * iteratively search for multivariate n-distinct with maximum number
+ * of vars; assuming that each var group is independent of the others,
+ * we multiply them together. Any remaining relvarinfos after no more
+ * multivariate matches are found are assumed independent too, so
+ * their individual ndistinct estimates are multiplied also.
+ *
+ * While iterating, count how many separate numdistinct values we
+ * apply. We apply a fudge factor below, but only if we multiplied
+ * more than one such values.
+ */
+ while (relvarinfos)
+ {
+ double mvndistinct;
+
+ if (estimate_multivariate_ndistinct(root, rel, &relvarinfos,
+ &mvndistinct))
+ {
+ reldistinct *= mvndistinct;
+ if (relmaxndistinct < mvndistinct)
+ relmaxndistinct = mvndistinct;
+ relvarcount++;
+ }
+ else
+ {
+ foreach(l, relvarinfos)
+ {
+ GroupVarInfo *varinfo2 = (GroupVarInfo *) lfirst(l);
+
+ reldistinct *= varinfo2->ndistinct;
+ if (relmaxndistinct < varinfo2->ndistinct)
+ relmaxndistinct = varinfo2->ndistinct;
+ relvarcount++;
+ }
+
+ /* we're done with this relation */
+ relvarinfos = NIL;
+ }
+ }
+
/*
* Sanity check --- don't divide by zero if empty relation.
*/
- Assert(rel->reloptkind == RELOPT_BASEREL);
+ Assert(IS_SIMPLE_REL(rel));
if (rel->tuples > 0)
{
/*
mcvfreq,
avgfreq;
bool isdefault;
- float4 *numbers;
- int nnumbers;
+ AttStatsSlot sslot;
examine_variable(root, hashkey, 0, &vardata);
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);
}
}
*-------------------------------------------------------------------------
*/
+/*
+ * Find applicable ndistinct statistics for the given list of VarInfos (which
+ * must all belong to the given rel), and update *ndistinct to the estimate of
+ * the MVNDistinctItem that best matches. If a match it found, *varinfos is
+ * updated to remove the list of matched varinfos.
+ *
+ * Varinfos that aren't for simple Vars are ignored.
+ *
+ * Return TRUE if we're able to find a match, FALSE otherwise.
+ */
+static bool
+estimate_multivariate_ndistinct(PlannerInfo *root, RelOptInfo *rel,
+ List **varinfos, double *ndistinct)
+{
+ ListCell *lc;
+ Bitmapset *attnums = NULL;
+ int nmatches;
+ Oid statOid = InvalidOid;
+ MVNDistinct *stats;
+ Bitmapset *matched = NULL;
+
+ /* bail out immediately if the table has no extended statistics */
+ if (!rel->statlist)
+ return false;
+
+ /* Determine the attnums we're looking for */
+ foreach(lc, *varinfos)
+ {
+ GroupVarInfo *varinfo = (GroupVarInfo *) lfirst(lc);
+
+ Assert(varinfo->rel == rel);
+
+ if (IsA(varinfo->var, Var))
+ {
+ attnums = bms_add_member(attnums,
+ ((Var *) varinfo->var)->varattno);
+ }
+ }
+
+ /* look for the ndistinct statistics matching the most vars */
+ nmatches = 1; /* we require at least two matches */
+ foreach(lc, rel->statlist)
+ {
+ StatisticExtInfo *info = (StatisticExtInfo *) lfirst(lc);
+ Bitmapset *shared;
+ int nshared;
+
+ /* skip statistics of other kinds */
+ if (info->kind != STATS_EXT_NDISTINCT)
+ continue;
+
+ /* compute attnums shared by the vars and the statistics object */
+ shared = bms_intersect(info->keys, attnums);
+ nshared = bms_num_members(shared);
+
+ /*
+ * Does this statistics object match more columns than the currently
+ * best object? If so, use this one instead.
+ *
+ * XXX This should break ties using name of the object, or something
+ * like that, to make the outcome stable.
+ */
+ if (nshared > nmatches)
+ {
+ statOid = info->statOid;
+ nmatches = nshared;
+ matched = shared;
+ }
+ }
+
+ /* No match? */
+ if (statOid == InvalidOid)
+ return false;
+ Assert(nmatches > 1 && matched != NULL);
+
+ stats = statext_ndistinct_load(statOid);
+
+ /*
+ * If we have a match, search it for the specific item that matches (there
+ * must be one), and construct the output values.
+ */
+ if (stats)
+ {
+ int i;
+ List *newlist = NIL;
+ MVNDistinctItem *item = NULL;
+
+ /* Find the specific item that exactly matches the combination */
+ for (i = 0; i < stats->nitems; i++)
+ {
+ MVNDistinctItem *tmpitem = &stats->items[i];
+
+ if (bms_subset_compare(tmpitem->attrs, matched) == BMS_EQUAL)
+ {
+ item = tmpitem;
+ break;
+ }
+ }
+
+ /* make sure we found an item */
+ if (!item)
+ elog(ERROR, "corrupt MVNDistinct entry");
+
+ /* Form the output varinfo list, keeping only unmatched ones */
+ foreach(lc, *varinfos)
+ {
+ GroupVarInfo *varinfo = (GroupVarInfo *) lfirst(lc);
+ AttrNumber attnum;
+
+ if (!IsA(varinfo->var, Var))
+ {
+ newlist = lappend(newlist, varinfo);
+ continue;
+ }
+
+ attnum = ((Var *) varinfo->var)->varattno;
+ if (!bms_is_member(attnum, matched))
+ newlist = lappend(newlist, varinfo);
+ }
+
+ *varinfos = newlist;
+ *ndistinct = item->ndistinct;
+ return true;
+ }
+
+ return false;
+}
+
/*
* convert_to_scalar
* Convert non-NULL values of the indicated types to the comparison
* 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
* this query. (Caution: this should be trusted for statistical
* purposes only, since we do not check indimmediate nor verify that
* the exact same definition of equality applies.)
+ * acl_ok: TRUE if current user has permission to read the column(s)
+ * underlying the pg_statistic entry. This is consulted by
+ * statistic_proc_security_check().
*
* Caller is responsible for doing ReleaseVariableStats() before exiting.
*/
Int16GetDatum(pos + 1),
BoolGetDatum(false));
vardata->freefunc = ReleaseSysCache;
+
+ if (HeapTupleIsValid(vardata->statsTuple))
+ {
+ /* Get index's table for permission check */
+ RangeTblEntry *rte;
+
+ rte = planner_rt_fetch(index->rel->relid, root);
+ Assert(rte->rtekind == RTE_RELATION);
+
+ /*
+ * For simplicity, we insist on the whole
+ * table being selectable, rather than trying
+ * to identify which column(s) the index
+ * depends on.
+ */
+ vardata->acl_ok =
+ (pg_class_aclcheck(rte->relid, GetUserId(),
+ ACL_SELECT) == ACLCHECK_OK);
+ }
+ else
+ {
+ /* suppress leakproofness checks later */
+ vardata->acl_ok = true;
+ }
}
if (vardata->statsTuple)
break;
Int16GetDatum(var->varattno),
BoolGetDatum(rte->inh));
vardata->freefunc = ReleaseSysCache;
+
+ if (HeapTupleIsValid(vardata->statsTuple))
+ {
+ /* check if user has permission to read this column */
+ vardata->acl_ok =
+ (pg_class_aclcheck(rte->relid, GetUserId(),
+ ACL_SELECT) == ACLCHECK_OK) ||
+ (pg_attribute_aclcheck(rte->relid, var->varattno, GetUserId(),
+ ACL_SELECT) == ACLCHECK_OK);
+ }
+ else
+ {
+ /* suppress any possible leakproofness checks later */
+ vardata->acl_ok = true;
+ }
}
else if (rte->rtekind == RTE_SUBQUERY && !rte->inh)
{
}
}
+/*
+ * Check whether it is permitted to call func_oid passing some of the
+ * pg_statistic data in vardata. We allow this either if the user has SELECT
+ * privileges on the table or column underlying the pg_statistic data or if
+ * the function is marked leak-proof.
+ */
+bool
+statistic_proc_security_check(VariableStatData *vardata, Oid func_oid)
+{
+ if (vardata->acl_ok)
+ return true;
+
+ if (!OidIsValid(func_oid))
+ return false;
+
+ if (get_func_leakproof(func_oid))
+ return true;
+
+ ereport(DEBUG2,
+ (errmsg_internal("not using statistics because function \"%s\" is not leak-proof",
+ get_func_name(func_oid))));
+ return false;
+}
+
/*
* get_variable_numdistinct
* Estimate the number of distinct values of a variable.
bool have_data = false;
int16 typLen;
bool typByVal;
- Datum *values;
- int nvalues;
+ Oid opfuncoid;
+ AttStatsSlot sslot;
int i;
/*
return false;
}
+ /*
+ * If we can't apply the sortop to the stats data, just fail. In
+ * principle, if there's a histogram and no MCVs, we could return the
+ * histogram endpoints without ever applying the sortop ... but it's
+ * probably not worth trying, because whatever the caller wants to do with
+ * the endpoints would likely fail the security check too.
+ */
+ if (!statistic_proc_security_check(vardata,
+ (opfuncoid = get_opcode(sortop))))
+ return false;
+
get_typlenbyval(vardata->atttype, &typLen, &typByVal);
/*
* 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;
}
* 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;
FmgrInfo opproc;
- fmgr_info(get_opcode(sortop), &opproc);
+ 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;
}
}
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;
forboth(lcc, path->indexquals, lci, path->indexqualcols)
{
- RestrictInfo *rinfo = castNode(RestrictInfo, lfirst(lcc));
+ RestrictInfo *rinfo = lfirst_node(RestrictInfo, lcc);
int indexcol = lfirst_int(lci);
Expr *clause;
Node *leftop,
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;
else
costs.indexCorrelation = varCorrelation;
- free_attstatsslot(InvalidOid, NULL, 0, numbers, nnumbers);
+ free_attstatsslot(&sslot);
}
}
{
IndexOptInfo *index = path->indexinfo;
List *indexQuals = path->indexquals;
- List *indexOrderBys = path->indexorderbys;
double numPages = index->pages;
- double numTuples = index->tuples;
+ RelOptInfo *baserel = index->rel;
+ RangeTblEntry *rte = planner_rt_fetch(baserel->relid, root);
List *qinfos;
Cost spc_seq_page_cost;
Cost spc_random_page_cost;
- double qual_op_cost;
double qual_arg_cost;
+ double qualSelectivity;
+ BrinStatsData statsData;
+ double indexRanges;
+ double minimalRanges;
+ double estimatedRanges;
+ double selec;
+ Relation indexRel;
+ ListCell *l;
+ VariableStatData vardata;
- /* Do preliminary analysis of indexquals */
- qinfos = deconstruct_indexquals(path);
+ Assert(rte->rtekind == RTE_RELATION);
- /* fetch estimated page cost for tablespace containing index */
+ /* fetch estimated page cost for the tablespace containing the index */
get_tablespace_page_costs(index->reltablespace,
&spc_random_page_cost,
&spc_seq_page_cost);
/*
- * BRIN indexes are always read in full; use that as startup cost.
+ * Obtain some data from the index itself.
+ */
+ indexRel = index_open(index->indexoid, AccessShareLock);
+ brinGetStats(indexRel, &statsData);
+ index_close(indexRel, AccessShareLock);
+
+ /*
+ * Compute index correlation
*
- * XXX maybe only include revmap pages here?
+ * Because we can use all index quals equally when scanning, we can use
+ * the largest correlation (in absolute value) among columns used by the
+ * query. Start at zero, the worst possible case. If we cannot find any
+ * correlation statistics, we will keep it as 0.
*/
- *indexStartupCost = spc_seq_page_cost * numPages * loop_count;
+ *indexCorrelation = 0;
+
+ qinfos = deconstruct_indexquals(path);
+ foreach(l, qinfos)
+ {
+ IndexQualInfo *qinfo = (IndexQualInfo *) lfirst(l);
+ AttrNumber attnum = index->indexkeys[qinfo->indexcol];
+
+ /* attempt to lookup stats in relation for this index column */
+ if (attnum != 0)
+ {
+ /* Simple variable -- look to stats for the underlying table */
+ if (get_relation_stats_hook &&
+ (*get_relation_stats_hook) (root, rte, attnum, &vardata))
+ {
+ /*
+ * The hook took control of acquiring a stats tuple. If it
+ * did supply a tuple, it'd better have supplied a freefunc.
+ */
+ if (HeapTupleIsValid(vardata.statsTuple) && !vardata.freefunc)
+ elog(ERROR,
+ "no function provided to release variable stats with");
+ }
+ else
+ {
+ vardata.statsTuple =
+ SearchSysCache3(STATRELATTINH,
+ ObjectIdGetDatum(rte->relid),
+ Int16GetDatum(attnum),
+ BoolGetDatum(false));
+ vardata.freefunc = ReleaseSysCache;
+ }
+ }
+ else
+ {
+ /*
+ * Looks like we've found an expression column in the index. Let's
+ * see if there's any stats for it.
+ */
+
+ /* get the attnum from the 0-based index. */
+ attnum = qinfo->indexcol + 1;
+
+ if (get_index_stats_hook &&
+ (*get_index_stats_hook) (root, index->indexoid, attnum, &vardata))
+ {
+ /*
+ * The hook took control of acquiring a stats tuple. If it
+ * did supply a tuple, it'd better have supplied a freefunc.
+ */
+ if (HeapTupleIsValid(vardata.statsTuple) &&
+ !vardata.freefunc)
+ elog(ERROR, "no function provided to release variable stats with");
+ }
+ else
+ {
+ vardata.statsTuple = SearchSysCache3(STATRELATTINH,
+ ObjectIdGetDatum(index->indexoid),
+ Int16GetDatum(attnum),
+ BoolGetDatum(false));
+ vardata.freefunc = ReleaseSysCache;
+ }
+ }
+
+ if (HeapTupleIsValid(vardata.statsTuple))
+ {
+ AttStatsSlot sslot;
+
+ if (get_attstatsslot(&sslot, vardata.statsTuple,
+ STATISTIC_KIND_CORRELATION, InvalidOid,
+ ATTSTATSSLOT_NUMBERS))
+ {
+ double varCorrelation = 0.0;
+
+ if (sslot.nnumbers > 0)
+ varCorrelation = Abs(sslot.numbers[0]);
+
+ if (varCorrelation > *indexCorrelation)
+ *indexCorrelation = varCorrelation;
+
+ free_attstatsslot(&sslot);
+ }
+ }
+
+ ReleaseVariableStats(vardata);
+ }
+
+ qualSelectivity = clauselist_selectivity(root, indexQuals,
+ baserel->relid,
+ JOIN_INNER, NULL);
+
+ /* work out the actual number of ranges in the index */
+ indexRanges = Max(ceil((double) baserel->pages / statsData.pagesPerRange),
+ 1.0);
/*
- * To read a BRIN index there might be a bit of back and forth over
- * regular pages, as revmap might point to them out of sequential order;
- * calculate this as reading the whole index in random order.
+ * Now calculate the minimum possible ranges we could match with if all of
+ * the rows were in the perfect order in the table's heap.
*/
- *indexTotalCost = spc_random_page_cost * numPages * loop_count;
+ minimalRanges = ceil(indexRanges * qualSelectivity);
- *indexSelectivity =
- clauselist_selectivity(root, indexQuals,
- path->indexinfo->rel->relid,
- JOIN_INNER, NULL);
- *indexCorrelation = 1;
+ /*
+ * Now estimate the number of ranges that we'll touch by using the
+ * indexCorrelation from the stats. Careful not to divide by zero (note
+ * we're using the absolute value of the correlation).
+ */
+ if (*indexCorrelation < 1.0e-10)
+ estimatedRanges = indexRanges;
+ else
+ estimatedRanges = Min(minimalRanges / *indexCorrelation, indexRanges);
+
+ /* we expect to visit this portion of the table */
+ selec = estimatedRanges / indexRanges;
+
+ CLAMP_PROBABILITY(selec);
+
+ *indexSelectivity = selec;
/*
- * Add on index qual eval costs, much as in genericcostestimate.
+ * Compute the index qual costs, much as in genericcostestimate, to add to
+ * the index costs.
*/
qual_arg_cost = other_operands_eval_cost(root, qinfos) +
orderby_operands_eval_cost(root, path);
- qual_op_cost = cpu_operator_cost *
- (list_length(indexQuals) + list_length(indexOrderBys));
+ /*
+ * Compute the startup cost as the cost to read the whole revmap
+ * sequentially, including the cost to execute the index quals.
+ */
+ *indexStartupCost =
+ spc_seq_page_cost * statsData.revmapNumPages * loop_count;
*indexStartupCost += qual_arg_cost;
- *indexTotalCost += qual_arg_cost;
- *indexTotalCost += (numTuples * *indexSelectivity) * (cpu_index_tuple_cost + qual_op_cost);
- *indexPages = index->pages;
- /* XXX what about pages_per_range? */
+ /*
+ * To read a BRIN index there might be a bit of back and forth over
+ * regular pages, as revmap might point to them out of sequential order;
+ * calculate the total cost as reading the whole index in random order.
+ */
+ *indexTotalCost = *indexStartupCost +
+ spc_random_page_cost * (numPages - statsData.revmapNumPages) * loop_count;
+
+ /*
+ * Charge a small amount per range tuple which we expect to match to. This
+ * is meant to reflect the costs of manipulating the bitmap. The BRIN scan
+ * will set a bit for each page in the range when we find a matching
+ * range, so we must multiply the charge by the number of pages in the
+ * range.
+ */
+ *indexTotalCost += 0.1 * cpu_operator_cost * estimatedRanges *
+ statsData.pagesPerRange;
+
+ *indexPages = index->pages;
}