Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commita536ed5

Browse files
committed
Make use of statistics on index expressions. There are still some
corner cases that could stand improvement, but it does all the basicstuff. A byproduct is that the selectivity routines are no longerconstrained to working on simple Vars; we might in future be able toimprove the behavior for subexpressions that don't match indexes.
1 parentd372bba commita536ed5

File tree

5 files changed

+1062
-979
lines changed

5 files changed

+1062
-979
lines changed

‎src/backend/optimizer/path/costsize.c

Lines changed: 3 additions & 178 deletions
Original file line numberDiff line numberDiff line change
@@ -49,7 +49,7 @@
4949
* Portions Copyright (c) 1994, Regents of the University of California
5050
*
5151
* IDENTIFICATION
52-
* $PostgreSQL: pgsql/src/backend/optimizer/path/costsize.c,v 1.124 2004/02/03 17:34:03 tgl Exp $
52+
* $PostgreSQL: pgsql/src/backend/optimizer/path/costsize.c,v 1.125 2004/02/17 00:52:53 tgl Exp $
5353
*
5454
*-------------------------------------------------------------------------
5555
*/
@@ -102,8 +102,6 @@ boolenable_mergejoin = true;
102102
boolenable_hashjoin= true;
103103

104104

105-
staticSelectivityestimate_hash_bucketsize(Query*root,Var*var,
106-
intnbuckets);
107105
staticboolcost_qual_eval_walker(Node*node,QualCost*total);
108106
staticSelectivityapprox_selectivity(Query*root,List*quals,
109107
JoinTypejointype);
@@ -1152,7 +1150,7 @@ cost_hashjoin(HashPath *path, Query *root)
11521150
/* not cached yet */
11531151
thisbucketsize=
11541152
estimate_hash_bucketsize(root,
1155-
(Var*)get_rightop(restrictinfo->clause),
1153+
get_rightop(restrictinfo->clause),
11561154
virtualbuckets);
11571155
restrictinfo->right_bucketsize=thisbucketsize;
11581156
}
@@ -1168,7 +1166,7 @@ cost_hashjoin(HashPath *path, Query *root)
11681166
/* not cached yet */
11691167
thisbucketsize=
11701168
estimate_hash_bucketsize(root,
1171-
(Var*)get_leftop(restrictinfo->clause),
1169+
get_leftop(restrictinfo->clause),
11721170
virtualbuckets);
11731171
restrictinfo->left_bucketsize=thisbucketsize;
11741172
}
@@ -1249,179 +1247,6 @@ cost_hashjoin(HashPath *path, Query *root)
12491247
path->jpath.path.total_cost=startup_cost+run_cost;
12501248
}
12511249

1252-
/*
1253-
* Estimate hash bucketsize fraction (ie, number of entries in a bucket
1254-
* divided by total tuples in relation) if the specified Var is used
1255-
* as a hash key.
1256-
*
1257-
* XXX This is really pretty bogus since we're effectively assuming that the
1258-
* distribution of hash keys will be the same after applying restriction
1259-
* clauses as it was in the underlying relation. However, we are not nearly
1260-
* smart enough to figure out how the restrict clauses might change the
1261-
* distribution, so this will have to do for now.
1262-
*
1263-
* We are passed the number of buckets the executor will use for the given
1264-
* input relation.If the data were perfectly distributed, with the same
1265-
* number of tuples going into each available bucket, then the bucketsize
1266-
* fraction would be 1/nbuckets. But this happy state of affairs will occur
1267-
* only if (a) there are at least nbuckets distinct data values, and (b)
1268-
* we have a not-too-skewed data distribution.Otherwise the buckets will
1269-
* be nonuniformly occupied. If the other relation in the join has a key
1270-
* distribution similar to this one's, then the most-loaded buckets are
1271-
* exactly those that will be probed most often. Therefore, the "average"
1272-
* bucket size for costing purposes should really be taken as something close
1273-
* to the "worst case" bucket size. We try to estimate this by adjusting the
1274-
* fraction if there are too few distinct data values, and then scaling up
1275-
* by the ratio of the most common value's frequency to the average frequency.
1276-
*
1277-
* If no statistics are available, use a default estimate of 0.1. This will
1278-
* discourage use of a hash rather strongly if the inner relation is large,
1279-
* which is what we want. We do not want to hash unless we know that the
1280-
* inner rel is well-dispersed (or the alternatives seem much worse).
1281-
*/
1282-
staticSelectivity
1283-
estimate_hash_bucketsize(Query*root,Var*var,intnbuckets)
1284-
{
1285-
Oidrelid;
1286-
RelOptInfo*rel;
1287-
HeapTupletuple;
1288-
Form_pg_statisticstats;
1289-
doubleestfract,
1290-
ndistinct,
1291-
mcvfreq,
1292-
avgfreq;
1293-
float4*numbers;
1294-
intnnumbers;
1295-
1296-
/* Ignore any binary-compatible relabeling */
1297-
if (var&&IsA(var,RelabelType))
1298-
var= (Var*) ((RelabelType*)var)->arg;
1299-
1300-
/*
1301-
* Lookup info about var's relation and attribute; if none available,
1302-
* return default estimate.
1303-
*/
1304-
if (var==NULL|| !IsA(var,Var))
1305-
return0.1;
1306-
1307-
relid=getrelid(var->varno,root->rtable);
1308-
if (relid==InvalidOid)
1309-
return0.1;
1310-
1311-
rel=find_base_rel(root,var->varno);
1312-
1313-
if (rel->tuples <=0.0||rel->rows <=0.0)
1314-
return0.1;/* ensure we can divide below */
1315-
1316-
tuple=SearchSysCache(STATRELATT,
1317-
ObjectIdGetDatum(relid),
1318-
Int16GetDatum(var->varattno),
1319-
0,0);
1320-
if (!HeapTupleIsValid(tuple))
1321-
{
1322-
/*
1323-
* If the attribute is known unique because of an index,
1324-
* we can treat it as well-distributed.
1325-
*/
1326-
if (has_unique_index(rel,var->varattno))
1327-
return1.0 / (double)nbuckets;
1328-
1329-
/*
1330-
* Perhaps the Var is a system attribute; if so, it will have no
1331-
* entry in pg_statistic, but we may be able to guess something
1332-
* about its distribution anyway.
1333-
*/
1334-
switch (var->varattno)
1335-
{
1336-
caseObjectIdAttributeNumber:
1337-
caseSelfItemPointerAttributeNumber:
1338-
/* these are unique, so buckets should be well-distributed */
1339-
return1.0 / (double)nbuckets;
1340-
caseTableOidAttributeNumber:
1341-
/* hashing this is a terrible idea... */
1342-
return1.0;
1343-
}
1344-
return0.1;
1345-
}
1346-
stats= (Form_pg_statistic)GETSTRUCT(tuple);
1347-
1348-
/*
1349-
* Obtain number of distinct data values in raw relation.
1350-
*/
1351-
ndistinct=stats->stadistinct;
1352-
if (ndistinct<0.0)
1353-
ndistinct=-ndistinct*rel->tuples;
1354-
1355-
if (ndistinct <=0.0)/* ensure we can divide */
1356-
{
1357-
ReleaseSysCache(tuple);
1358-
return0.1;
1359-
}
1360-
1361-
/* Also compute avg freq of all distinct data values in raw relation */
1362-
avgfreq= (1.0-stats->stanullfrac) /ndistinct;
1363-
1364-
/*
1365-
* Adjust ndistinct to account for restriction clauses. Observe we
1366-
* are assuming that the data distribution is affected uniformly by
1367-
* the restriction clauses!
1368-
*
1369-
* XXX Possibly better way, but much more expensive: multiply by
1370-
* selectivity of rel's restriction clauses that mention the target
1371-
* Var.
1372-
*/
1373-
ndistinct *=rel->rows /rel->tuples;
1374-
1375-
/*
1376-
* Initial estimate of bucketsize fraction is 1/nbuckets as long as
1377-
* the number of buckets is less than the expected number of distinct
1378-
* values; otherwise it is 1/ndistinct.
1379-
*/
1380-
if (ndistinct> (double)nbuckets)
1381-
estfract=1.0 / (double)nbuckets;
1382-
else
1383-
estfract=1.0 /ndistinct;
1384-
1385-
/*
1386-
* Look up the frequency of the most common value, if available.
1387-
*/
1388-
mcvfreq=0.0;
1389-
1390-
if (get_attstatsslot(tuple,var->vartype,var->vartypmod,
1391-
STATISTIC_KIND_MCV,InvalidOid,
1392-
NULL,NULL,&numbers,&nnumbers))
1393-
{
1394-
/*
1395-
* The first MCV stat is for the most common value.
1396-
*/
1397-
if (nnumbers>0)
1398-
mcvfreq=numbers[0];
1399-
free_attstatsslot(var->vartype,NULL,0,
1400-
numbers,nnumbers);
1401-
}
1402-
1403-
/*
1404-
* Adjust estimated bucketsize upward to account for skewed
1405-
* distribution.
1406-
*/
1407-
if (avgfreq>0.0&&mcvfreq>avgfreq)
1408-
estfract *=mcvfreq /avgfreq;
1409-
1410-
/*
1411-
* Clamp bucketsize to sane range (the above adjustment could easily
1412-
* produce an out-of-range result). We set the lower bound a little
1413-
* above zero, since zero isn't a very sane result.
1414-
*/
1415-
if (estfract<1.0e-6)
1416-
estfract=1.0e-6;
1417-
elseif (estfract>1.0)
1418-
estfract=1.0;
1419-
1420-
ReleaseSysCache(tuple);
1421-
1422-
return (Selectivity)estfract;
1423-
}
1424-
14251250

14261251
/*
14271252
* cost_qual_eval

‎src/backend/optimizer/util/relnode.c

Lines changed: 2 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -8,7 +8,7 @@
88
*
99
*
1010
* IDENTIFICATION
11-
* $PostgreSQL: pgsql/src/backend/optimizer/util/relnode.c,v 1.54 2003/12/08 18:19:58 tgl Exp $
11+
* $PostgreSQL: pgsql/src/backend/optimizer/util/relnode.c,v 1.55 2004/02/17 00:52:53 tgl Exp $
1212
*
1313
*-------------------------------------------------------------------------
1414
*/
@@ -214,12 +214,8 @@ find_base_rel(Query *root, int relid)
214214
* find_join_rel
215215
* Returns relation entry corresponding to 'relids' (a set of RT indexes),
216216
* or NULL if none exists. This is for join relations.
217-
*
218-
* Note: there is probably no good reason for this to be called from
219-
* anywhere except build_join_rel, but keep it as a separate routine
220-
* just in case.
221217
*/
222-
staticRelOptInfo*
218+
RelOptInfo*
223219
find_join_rel(Query*root,Relidsrelids)
224220
{
225221
List*joinrels;

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp