Prevent cross join without any target list entries (#4750)

/*
 * The physical planner assumes that all worker queries would have
 * target list entries based on the fact that at least the column
 * on the JOINs have to be on the target list. However, there is
 * an exception to that if there is a cartesian product join and
 * there is no additional target list entries belong to one side
 * of the JOIN. Once we support cartesian product join, we should
 * remove this error.
 */

src/backend/distributed/planner/multi_physical_planner.c

@@ -824,7 +824,21 @@ static List *
 QueryTargetList(MultiNode *multiNode)
 {
 	List *projectNodeList = FindNodesOfType(multiNode, T_MultiProject);
-	Assert(list_length(projectNodeList) > 0);
+	if (list_length(projectNodeList) == 0)
+	{
+		/*
+		 * The physical planner assumes that all worker queries would have
+		 * target list entries based on the fact that at least the column
+		 * on the JOINs have to be on the target list. However, there is
+		 * an exception to that if there is a cartesian product join and
+		 * there is no additional target list entries belong to one side
+		 * of the JOIN. Once we support cartesian product join, we should
+		 * remove this error.
+		 */
+		ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+						errmsg("cannot perform distributed planning on this query"),
+						errdetail("Cartesian products are currently unsupported")));
+	}
 
 	MultiProject *topProjectNode = (MultiProject *) linitial(projectNodeList);
 	List *columnList = topProjectNode->columnList;

src/test/regress/expected/cross_join.out

@@ -251,3 +251,48 @@ ORDER BY 1,2,3,4 LIMIT 5;
        1 | Wed Nov 22 22:51:43.132261 2017 |       4 |       0 |       3 |
 (5 rows)
 
+-- we don't support cross JOINs between distributed tables
+-- and without target list entries
+CREATE TABLE dist1(c0 int);
+CREATE TABLE dist2(c0 int);
+CREATE TABLE dist3(c0 int , c1 int);
+CREATE TABLE dist4(c0 int , c1 int);
+SELECT create_distributed_table('dist1', 'c0');
+ create_distributed_table
+---------------------------------------------------------------------
+
+(1 row)
+
+SELECT create_distributed_table('dist2', 'c0');
+ create_distributed_table
+---------------------------------------------------------------------
+
+(1 row)
+
+SELECT create_distributed_table('dist3', 'c1');
+ create_distributed_table
+---------------------------------------------------------------------
+
+(1 row)
+
+SELECT create_distributed_table('dist4', 'c1');
+ create_distributed_table
+---------------------------------------------------------------------
+
+(1 row)
+
+SELECT dist2.c0 FROM dist1, dist3, dist4, dist2 WHERE (dist3.c0) IN (dist4.c0);
+ERROR:  cannot perform distributed planning on this query
+DETAIL:  Cartesian products are currently unsupported
+SELECT 1 FROM dist3, dist4, dist2 WHERE (dist3.c0) IN (dist4.c0);
+ERROR:  cannot perform distributed planning on this query
+DETAIL:  Cartesian products are currently unsupported
+SELECT  FROM dist3, dist4, dist2 WHERE (dist3.c0) IN (dist4.c0);
+ERROR:  cannot perform distributed planning on this query
+DETAIL:  Cartesian products are currently unsupported
+SELECT dist2.c0 FROM dist3, dist4, dist2 WHERE (dist3.c0) IN (dist4.c0);
+ERROR:  cannot perform distributed planning on this query
+DETAIL:  Cartesian products are currently unsupported
+SELECT dist2.* FROM dist3, dist4, dist2 WHERE (dist3.c0) IN (dist4.c0);
+ERROR:  cannot perform distributed planning on this query
+DETAIL:  Cartesian products are currently unsupported

src/test/regress/sql/cross_join.sql

@@ -128,4 +128,20 @@ FROM
 	LEFT JOIN users_table USING (user_id)
 ORDER BY 1,2,3,4 LIMIT 5;
 
+-- we don't support cross JOINs between distributed tables
+-- and without target list entries
+CREATE TABLE dist1(c0 int);
+CREATE TABLE dist2(c0 int);
+CREATE TABLE dist3(c0 int , c1 int);
+CREATE TABLE dist4(c0 int , c1 int);
 
+SELECT create_distributed_table('dist1', 'c0');
+SELECT create_distributed_table('dist2', 'c0');
+SELECT create_distributed_table('dist3', 'c1');
+SELECT create_distributed_table('dist4', 'c1');
+
+SELECT dist2.c0 FROM dist1, dist3, dist4, dist2 WHERE (dist3.c0) IN (dist4.c0);
+SELECT 1 FROM dist3, dist4, dist2 WHERE (dist3.c0) IN (dist4.c0);
+SELECT  FROM dist3, dist4, dist2 WHERE (dist3.c0) IN (dist4.c0);
+SELECT dist2.c0 FROM dist3, dist4, dist2 WHERE (dist3.c0) IN (dist4.c0);
+SELECT dist2.* FROM dist3, dist4, dist2 WHERE (dist3.c0) IN (dist4.c0);