rdelaage
/
postgresql
mirror of https://git.postgresql.org/git/postgresql.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Fix estimate_num_groups() to assume that GROUP BY expressions yielding boolean 

results always contribute two groups, regardless of the expression contents.
This is very substantially more accurate than the regular heuristic for
certain boolean tests like "col IS NULL".  Per gripe from Sam Mason.

Back-patch to all supported releases, since the behavior of
estimate_num_groups() hasn't changed all that much since 7.4.
This commit is contained in:
Tom Lane 2008-07-07 20:24:55 +00:00
parent c50838533b
commit 170063cd1e
1 changed files with 31 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

42
src/backend/utils/adt/selfuncs.c
View File

@ -15,7 +15,7 @@
*
*
* IDENTIFICATION
* $PostgreSQL: pgsql/src/backend/utils/adt/selfuncs.c,v 1.249 2008/05/12 00:00:51 alvherre Exp $
* $PostgreSQL: pgsql/src/backend/utils/adt/selfuncs.c,v 1.250 2008/07/07 20:24:55 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@ -2645,7 +2645,11 @@ add_unique_group_var(PlannerInfo *root, List *varinfos,
* case (all possible cross-product terms actually appear as groups) since
* very often the grouped-by Vars are highly correlated. Our current approach
* is as follows:
* 1. Reduce the given expressions to a list of unique Vars used. For
* 1. Expressions yielding boolean are assumed to contribute two groups,
* independently of their content, and are ignored in the subsequent
* steps. This is mainly because tests like "col IS NULL" break the
* heuristic used in step 2 especially badly.
* 2. Reduce the given expressions to a list of unique Vars used. For
* example, GROUP BY a, a + b is treated the same as GROUP BY a, b.
* It is clearly correct not to count the same Var more than once.
* It is also reasonable to treat f(x) the same as x: f() cannot
@ -2655,14 +2659,14 @@ add_unique_group_var(PlannerInfo *root, List *varinfos,
* As a special case, if a GROUP BY expression can be matched to an
* expressional index for which we have statistics, then we treat the
* whole expression as though it were just a Var.
* 2. If the list contains Vars of different relations that are known equal
* 3. If the list contains Vars of different relations that are known equal
* due to equivalence classes, then drop all but one of the Vars from each
* known-equal set, keeping the one with smallest estimated # of values
* (since the extra values of the others can't appear in joined rows).
* Note the reason we only consider Vars of different relations is that
* if we considered ones of the same rel, we'd be double-counting the
* restriction selectivity of the equality in the next step.
* 3. For Vars within a single source rel, we multiply together the numbers
* 4. For Vars within a single source rel, we multiply together the numbers
* of values, clamp to the number of rows in the rel (divided by 10 if
* more than one Var), and then multiply by the selectivity of the
* restriction clauses for that rel. When there's more than one Var,
@ -2673,7 +2677,7 @@ add_unique_group_var(PlannerInfo *root, List *varinfos,
* by the restriction selectivity is effectively assuming that the
* restriction clauses are independent of the grouping, which is a crummy
* assumption, but it's hard to do better.
* 4. If there are Vars from multiple rels, we repeat step 3 for each such
* 5. If there are Vars from multiple rels, we repeat step 4 for each such
* rel, and multiply the results together.
* Note that rels not containing grouped Vars are ignored completely, as are
* join clauses. Such rels cannot increase the number of groups, and we
@ -2691,11 +2695,14 @@ estimate_num_groups(PlannerInfo *root, List *groupExprs, double input_rows)
Assert(groupExprs != NIL);
/*
* Steps 1/2: find the unique Vars used, treating an expression as a Var
* Count groups derived from boolean grouping expressions. For other
* expressions, find the unique Vars used, treating an expression as a Var
* if we can find stats for it. For each one, record the statistical
* estimate of number of distinct values (total in its table, without
* regard for filtering).
*/
numdistinct = 1.0;
foreach(l, groupExprs)
{
Node *groupexpr = (Node *) lfirst(l);
@ -2703,6 +2710,13 @@ estimate_num_groups(PlannerInfo *root, List *groupExprs, double input_rows)
List *varshere;
ListCell *l2;
/* Short-circuit for expressions returning boolean */
if (exprType(groupexpr) == BOOLOID)
{
numdistinct *= 2.0;
continue;
}
/*
* If examine_variable is able to deduce anything about the GROUP BY
* expression, treat it as a single variable even if it's really more
@ -2749,20 +2763,26 @@ estimate_num_groups(PlannerInfo *root, List *groupExprs, double input_rows)
}
}
/* If now no Vars, we must have an all-constant GROUP BY list. */
/*
* If now no Vars, we must have an all-constant or all-boolean GROUP BY
* list.
*/
if (varinfos == NIL)
return 1.0;
{
/* Guard against out-of-range answers */
if (numdistinct > input_rows)
numdistinct = input_rows;
return numdistinct;
}
/*
* Steps 3/4: group Vars by relation and estimate total numdistinct.
* Group Vars by relation and estimate total numdistinct.
*
* For each iteration of the outer loop, we process the frontmost Var in
* varinfos, plus all other Vars in the same relation. We remove these
* Vars from the newvarinfos list for the next iteration. This is the
* easiest way to group Vars of same rel together.
*/
numdistinct = 1.0;
do
{
GroupVarInfo *varinfo1 = (GroupVarInfo *) linitial(varinfos);