Duplicate grouping on worker whenever possible

This is possible whenever we aren't pulling up intermediate rows

We want to do this because this was done in 9.2,
some queries rely on the performance of grouping causing distinct values

This change was introduced when implementing window functions on coordinator

src/backend/distributed/planner/multi_logical_optimizer.c

@@ -328,9 +328,6 @@ static bool HasOrderByAggregate(List *sortClauseList, List *targetList);
 static bool HasOrderByNonCommutativeAggregate(List *sortClauseList, List *targetList);
 static bool HasOrderByComplexExpression(List *sortClauseList, List *targetList);
 static bool HasOrderByHllType(List *sortClauseList, List *targetList);
-static bool ShouldPushDownGroupingToWorker(MultiExtendedOp *opNode,
-										   ExtendedOpNodeProperties *
-										   extendedOpNodeProperties);
 static bool ShouldProcessDistinctOrderAndLimitForWorker(
 	ExtendedOpNodeProperties *extendedOpNodeProperties,
 	bool pushingDownOriginalGrouping,
@@ -2223,7 +2220,7 @@ WorkerExtendedOpNode(MultiExtendedOp *originalOpNode,
 	/* targetProjectionNumber starts from 1 */
 	queryTargetList.targetProjectionNumber = 1;
 
-	if (ShouldPushDownGroupingToWorker(originalOpNode, extendedOpNodeProperties))
+	if (!extendedOpNodeProperties->pullUpIntermediateRows)
 	{
 		queryGroupClause.groupClauseList = copyObject(originalGroupClauseList);
 	}
@@ -4717,46 +4714,6 @@ HasOrderByHllType(List *sortClauseList, List *targetList)
 }
 
 
-/*
- * ShouldPushDownGroupingToWorker returns whether we push down GROUP BY.
- * This may return true even when GROUP BY is necessary on master.
- */
-static bool
-ShouldPushDownGroupingToWorker(MultiExtendedOp *opNode,
-							   ExtendedOpNodeProperties *extendedOpNodeProperties)
-{
-	if (extendedOpNodeProperties->pushDownGroupingAndHaving)
-	{
-		return true;
-	}
-
-	if (extendedOpNodeProperties->pullUpIntermediateRows)
-	{
-		return false;
-	}
-
-	/*
-	 * Duplicate grouping if we have LIMIT without HAVING, as this can
-	 * often result in LIMIT being pushed down.
-	 */
-	if (opNode->havingQual == NULL && opNode->limitCount != NULL)
-	{
-		return true;
-	}
-
-	/*
-	 * If aggregates are being split across worker & master, so must grouping.
-	 */
-	if (contain_aggs_of_level(opNode->havingQual, 0) ||
-		contain_aggs_of_level((Node *) opNode->targetList, 0))
-	{
-		return true;
-	}
-
-	return false;
-}
-
-
 /*
  * ShouldProcessDistinctOrderAndLimitForWorker returns whether
  * ProcessDistinctClauseForWorkerQuery should be called. If not,

src/test/regress/expected/multi_explain.out

@@ -453,8 +453,11 @@ HashAggregate
         Tasks Shown: One of 2
         ->  Task
               Node: host=localhost port=xxxxx dbname=regression
-              ->  Seq Scan on public.lineitem_290000 lineitem
+              ->  HashAggregate
                     Output: l_quantity, l_quantity
+                    Group Key: lineitem.l_quantity
+                    ->  Seq Scan on public.lineitem_290000 lineitem
+                          Output: l_orderkey, l_partkey, l_suppkey, l_linenumber, l_quantity, l_extendedprice, l_discount, l_tax, l_returnflag, l_linestatus, l_shipdate, l_commitdate, l_receiptdate, l_shipinstruct, l_shipmode, l_comment
 -- Subquery pushdown tests with explain
 EXPLAIN (COSTS OFF)
 SELECT

src/test/regress/expected/multi_view.out

@@ -758,6 +758,8 @@ EXPLAIN (COSTS FALSE) SELECT user_id FROM recent_selected_users GROUP BY 1 ORDER
                Tasks Shown: One of 4
                ->  Task
                      Node: host=localhost port=xxxxx dbname=regression
+                     ->  HashAggregate
+                           Group Key: users_table.user_id
                            ->  Nested Loop
                                  Join Filter: (users_table.user_id = users_table_1.user_id)
                                  ->  Sort
@@ -768,7 +770,7 @@ EXPLAIN (COSTS FALSE) SELECT user_id FROM recent_selected_users GROUP BY 1 ORDER
                                              ->  Seq Scan on users_table_1400256 users_table_1
                                  ->  Seq Scan on users_table_1400256 users_table
                                        Filter: ((value_1 >= 1) AND (value_1 < 3))
-(19 rows)
+(21 rows)
 
 EXPLAIN (COSTS FALSE) SELECT *
 	FROM (