When estimating without benefit of MCV lists (suggesting that one or both

inputs is unique or nearly so), make eqjoinsel() clamp the ndistinct estimates
to be not more than the estimated number of rows coming from the input
relations.  This allows the estimate to change in response to the selectivity
of restriction conditions on the inputs.

This is a pretty narrow patch and maybe we should be more aggressive about
similarly clamping ndistinct in other cases; but I'm worried about
double-counting the effects of the restriction conditions.  However, it seems
to help for the case exhibited by Grzegorz Jaskiewicz (antijoin against a
small subset of a relation), so let's try this for awhile.

diff --git a/src/backend/utils/adt/selfuncs.c b/src/backend/utils/adt/selfuncs.c
index b659013bdec3fa61d4eb2a585431aa67ac0f8e67..d5a60256d12f994d5c4fb669c80d42202f110a2f 100644 (file)
--- a/src/backend/utils/adt/selfuncs.c
+++ b/src/backend/utils/adt/selfuncs.c
@@ -15,7 +15,7 @@
  *
  *
  * IDENTIFICATION
- *   $PostgreSQL: pgsql/src/backend/utils/adt/selfuncs.c,v 1.256 2008/10/21 20:42:53 tgl Exp $
+ *   $PostgreSQL: pgsql/src/backend/utils/adt/selfuncs.c,v 1.257 2008/10/23 00:24:50 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -2143,10 +2143,22 @@ eqjoinsel_inner(Oid operator,
         * XXX Can we be smarter if we have an MCV list for just one side? It
         * seems that if we assume equal distribution for the other side, we
         * end up with the same answer anyway.
+        *
+        * An additional hack we use here is to clamp the nd1 and nd2 values
+        * to not more than what we are estimating the input relation sizes
+        * to be, providing a crude correction for the selectivity of
+        * restriction clauses on those relations.  (We don't do that in the
+        * other path since there we are comparing the nd values to stats for
+        * the whole relations.)
         */
        double      nullfrac1 = stats1 ? stats1->stanullfrac : 0.0;
        double      nullfrac2 = stats2 ? stats2->stanullfrac : 0.0;
 
+       if (vardata1->rel)
+           nd1 = Min(nd1, vardata1->rel->rows);
+       if (vardata2->rel)
+           nd2 = Min(nd2, vardata2->rel->rows);
+
        selec = (1.0 - nullfrac1) * (1.0 - nullfrac2);
        if (nd1 > nd2)
            selec /= nd1;
@@ -2305,6 +2317,11 @@ eqjoinsel_semi(Oid operator,
         */
        double      nullfrac1 = stats1 ? stats1->stanullfrac : 0.0;
 
+       if (vardata1->rel)
+           nd1 = Min(nd1, vardata1->rel->rows);
+       if (vardata2->rel)
+           nd2 = Min(nd2, vardata2->rel->rows);
+
        if (nd1 <= nd2 || nd2 <= 0)
            selec = 1.0 - nullfrac1;
        else