Fix overeager pushdown of HAVING clauses when grouping sets are used.

In 61444bfb we started to allow HAVING clauses to be fully pushed down
into WHERE, even when grouping sets are in use. That turns out not to
work correctly, because grouping sets can "produce" NULLs, meaning that
filtering in WHERE and HAVING can have different results, even when no
aggregates or volatile functions are involved.

Instead only allow pushdown of empty grouping sets.

It'd be nice to do better, but the exact mechanics of deciding which
cases are safe are still being debated. It's important to give correct
results till we find a good solution, and such a solution might not be
appropriate for backpatching anyway.

Bug: #13863
Reported-By: 'wrb'
Diagnosed-By: Dean Rasheed
Author: Andrew Gierth
Reviewed-By: Dean Rasheed and Andres Freund
Discussion: 20160113183558.12989.56904@wrigleys.postgresql.org
Backpatch: 9.5, where grouping sets were introduced

src/backend/optimizer/plan/planner.c

@@ -653,13 +653,19 @@ subquery_planner(PlannerGlobal *glob, Query *parse,
 	 * In some cases we may want to transfer a HAVING clause into WHERE. We
 	 * cannot do so if the HAVING clause contains aggregates (obviously) or
 	 * volatile functions (since a HAVING clause is supposed to be executed
-	 * only once per group).  Also, it may be that the clause is so expensive
-	 * to execute that we're better off doing it only once per group, despite
-	 * the loss of selectivity.  This is hard to estimate short of doing the
-	 * entire planning process twice, so we use a heuristic: clauses
-	 * containing subplans are left in HAVING.  Otherwise, we move or copy the
-	 * HAVING clause into WHERE, in hopes of eliminating tuples before
-	 * aggregation instead of after.
+	 * only once per group).  We also can't do this if there are any nonempty
+	 * grouping sets; moving such a clause into WHERE would potentially change
+	 * the results, if any referenced column isn't present in all the grouping
+	 * sets.  (If there are only empty grouping sets, then the HAVING clause
+	 * must be degenerate as discussed below.)
+	 *
+	 * Also, it may be that the clause is so expensive to execute that we're
+	 * better off doing it only once per group, despite the loss of
+	 * selectivity.  This is hard to estimate short of doing the entire
+	 * planning process twice, so we use a heuristic: clauses containing
+	 * subplans are left in HAVING.  Otherwise, we move or copy the HAVING
+	 * clause into WHERE, in hopes of eliminating tuples before aggregation
+	 * instead of after.
 	 *
 	 * If the query has explicit grouping then we can simply move such a
 	 * clause into WHERE; any group that fails the clause will not be in the
@@ -679,7 +685,8 @@ subquery_planner(PlannerGlobal *glob, Query *parse,
 	{
 		Node	   *havingclause = (Node *) lfirst(l);
 
-		if (contain_agg_clause(havingclause) ||
+		if ((parse->groupClause && parse->groupingSets) ||
+			contain_agg_clause(havingclause) ||
 			contain_volatile_functions(havingclause) ||
 			contain_subplans(havingclause))
 		{

src/test/regress/expected/groupingsets.out

@@ -607,6 +607,60 @@ having exists (select 1 from onek b where sum(distinct a.four) = b.four);
    9 |   3
 (25 rows)
 
+-- Tests around pushdown of HAVING clauses, partially testing against previous bugs
+select a,count(*) from gstest2 group by rollup(a) order by a;
+ a | count 
+---+-------
+ 1 |     8
+ 2 |     1
+   |     9
+(3 rows)
+
+select a,count(*) from gstest2 group by rollup(a) having a is distinct from 1 order by a;
+ a | count 
+---+-------
+ 2 |     1
+   |     9
+(2 rows)
+
+explain (costs off)
+  select a,count(*) from gstest2 group by rollup(a) having a is distinct from 1 order by a;
+            QUERY PLAN            
+----------------------------------
+ GroupAggregate
+   Group Key: a
+   Group Key: ()
+   Filter: (a IS DISTINCT FROM 1)
+   ->  Sort
+         Sort Key: a
+         ->  Seq Scan on gstest2
+(7 rows)
+
+select v.c, (select count(*) from gstest2 group by () having v.c)
+  from (values (false),(true)) v(c) order by v.c;
+ c | count 
+---+-------
+ f |      
+ t |     9
+(2 rows)
+
+explain (costs off)
+  select v.c, (select count(*) from gstest2 group by () having v.c)
+    from (values (false),(true)) v(c) order by v.c;
+                        QUERY PLAN                         
+-----------------------------------------------------------
+ Sort
+   Sort Key: "*VALUES*".column1
+   ->  Values Scan on "*VALUES*"
+         SubPlan 1
+           ->  Aggregate
+                 Group Key: ()
+                 Filter: "*VALUES*".column1
+                 ->  Result
+                       One-Time Filter: "*VALUES*".column1
+                       ->  Seq Scan on gstest2
+(10 rows)
+
 -- HAVING with GROUPING queries
 select ten, grouping(ten) from onek
 group by grouping sets(ten) having grouping(ten) >= 0

src/test/regress/sql/groupingsets.sql

@@ -183,6 +183,18 @@ select ten, sum(distinct four) from onek a
 group by grouping sets((ten,four),(ten))
 having exists (select 1 from onek b where sum(distinct a.four) = b.four);
 
+-- Tests around pushdown of HAVING clauses, partially testing against previous bugs
+select a,count(*) from gstest2 group by rollup(a) order by a;
+select a,count(*) from gstest2 group by rollup(a) having a is distinct from 1 order by a;
+explain (costs off)
+  select a,count(*) from gstest2 group by rollup(a) having a is distinct from 1 order by a;
+
+select v.c, (select count(*) from gstest2 group by () having v.c)
+  from (values (false),(true)) v(c) order by v.c;
+explain (costs off)
+  select v.c, (select count(*) from gstest2 group by () having v.c)
+    from (values (false),(true)) v(c) order by v.c;
+
 -- HAVING with GROUPING queries
 select ten, grouping(ten) from onek
 group by grouping sets(ten) having grouping(ten) >= 0