improve comments/ error checks for CTEs and set operations

src/backend/distributed/planner/multi_router_planner.c

@@ -551,6 +551,19 @@ ErrorIfMultiTaskRouterSelectQueryUnsupported(Query *query)
 									  "INSERT ... SELECT queries")));
 		}
 
+		/*
+		 * We currently do not support CTEs. In order to handle CTEs, consider expanding
+		 * IsPartitionColumnRecursive() to handle CTEs.
+		 */
+		if (subquery->cteList != NULL)
+		{
+			ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+							errmsg("cannot perform distributed planning for the given "
+								   "modification"),
+							errdetail("Common table expressions are not allowed in "
+									  "INSERT ... SELECT queries")));
+		}
+
 		/* see comment on AddUninstantiatedPartitionColumnEqualityQual() */
 		if (subquery->setOperations != NULL)
 		{
@@ -655,32 +668,15 @@ AddUninstantiatedPartitionColumnEqualityQual(Query *originalQuery)
 
 	Assert(InsertSelectQuery(originalQuery));
 
-	/* we currently do not support CTEs */
-	if (originalQuery->cteList != NULL)
-	{
-		return;
-	}
-
 	subqueryEntry = ExtractSelectRangeTableEntry(originalQuery);
 	subquery = subqueryEntry->subquery;
 
 	/*
-	 * We currently not support the subquery with set operations for three reasons.
+	 * We currently not support the subquery with set operations. The main reason is that
-	 *
+	 * there is an "Assert(parse->jointree->quals == NULL);" on standard planner's execution
-	 *   (i) Adding only a single qual where there are more than one join trees
+	 * path (i.e., plan_set_operations).
-	 *   leads to an assertion failure on the standard planner (i.e.,
+	 * If we are to add uninstantiated equality qual to the query, we may end up hitting that
-	 *   Assert(parse->jointree->quals == NULL); on plan_set_operations()).
+	 * assertion, so it's better not to support for now.
-	 *   [THE ABOVE COMMENT IS TO EASE THE REVIEW, REMOVE LATER ON]
-	 *
-	 *   (ii)  There are potentially multiple jointree quals that we need to add
-	 *   the qual, and we haven't implemented that logic yet.
-	 *
-	 *   (iii) We cannot get the source tables OID via target entries resorigtbl field.
-	 *   This makes hard to check the colocation requirement of the source and target
-	 *   tables.
-	 *
-	 * Note that we do not allow set operations on the lower level's of the subquery
-	 * as well, which is handled on ErrorIfMultiTaskRouterSelectQueryUnsupported().
 	 */
 	if (subquery->setOperations != NULL)
 	{

src/test/regress/expected/multi_insert_select.out

@@ -640,10 +640,10 @@ DEBUG:  sent COMMIT over connection 13300010
 (1 row)
 
 -- TODO:: add hll and date_trunc 
-INSERT INTO agg_events (user_id, value_1_agg) 
+INSERT INTO agg_events (user_id, value_1_agg)
-SELECT 
+SELECT
-   user_id, sum(value_1 + value_2) 
+   user_id, sum(value_1 + value_2)
-FROM 
+FROM
    raw_events_first GROUP BY user_id;
 DEBUG:  StartTransactionCommand
 DEBUG:  StartTransaction
@@ -678,10 +678,10 @@ DEBUG:  sent COMMIT over connection 13300009
 DEBUG:  sent COMMIT over connection 13300010
 DEBUG:  sent COMMIT over connection 13300010
 --  FILTER CLAUSE
-INSERT INTO agg_events (user_id, value_1_agg) 
+INSERT INTO agg_events (user_id, value_1_agg)
-SELECT 
+SELECT
    user_id, sum(value_1 + value_2) FILTER (where value_3 = 15)
-FROM 
+FROM
    raw_events_first GROUP BY user_id;
 DEBUG:  StartTransactionCommand
 DEBUG:  StartTransaction
@@ -715,51 +715,15 @@ DEBUG:  sent COMMIT over connection 13300009
 DEBUG:  sent COMMIT over connection 13300009
 DEBUG:  sent COMMIT over connection 13300010
 DEBUG:  sent COMMIT over connection 13300010
--- a very simple UNION query
-INSERT INTO 
-  raw_events_first(user_id) 
-SELECT 
-  user_id 
-FROM 
-  ((SELECT user_id FROM raw_events_first) UNION 
-   (SELECT user_id FROM raw_events_second)) as foo;
-DEBUG:  StartTransactionCommand
-DEBUG:  StartTransaction
-DEBUG:  name: unnamed; blockState:       DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children: 
-ERROR:  cannot perform distributed planning for the given modification
-DETAIL:  Set operations are not allowed in INSERT ... SELECT queries
--- same query with slightly different syntax, but this time we cannot push it down
-INSERT INTO 
-  raw_events_first(user_id)
-  (SELECT user_id FROM raw_events_first) UNION 
-  (SELECT user_id FROM raw_events_first);
-DEBUG:  StartTransactionCommand
-DEBUG:  StartTransaction
-DEBUG:  name: unnamed; blockState:       DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children: 
-ERROR:  cannot perform distributed planning for the given modification
-DETAIL:  Set operations are not allowed in INSERT ... SELECT queries
--- similar query with a filter on two of the queries
-INSERT INTO 
-  raw_events_first(user_id) 
-SELECT 
-  user_id 
-FROM 
-  ((SELECT user_id FROM raw_events_first WHERE user_id = 15) UNION 
-   (SELECT user_id FROM raw_events_second where user_id = 17)) as foo;
-DEBUG:  StartTransactionCommand
-DEBUG:  StartTransaction
-DEBUG:  name: unnamed; blockState:       DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children: 
-ERROR:  cannot perform distributed planning for the given modification
-DETAIL:  Set operations are not allowed in INSERT ... SELECT queries
 -- TODO: UUIDs
 -- a test with reference table JOINs
-INSERT INTO 
+INSERT INTO
-  agg_events (user_id, value_1_agg) 
+  agg_events (user_id, value_1_agg)
-SELECT 
+SELECT
   raw_events_first.user_id, sum(value_1)
-FROM 
+FROM
-  reference_table, raw_events_first 
+  reference_table, raw_events_first
-WHERE 
+WHERE
   raw_events_first.user_id = reference_table.user_id
 GROUP BY
   raw_events_first.user_id;
@@ -795,6 +759,68 @@ DEBUG:  sent COMMIT over connection 13300009
 DEBUG:  sent COMMIT over connection 13300009
 DEBUG:  sent COMMIT over connection 13300010
 DEBUG:  sent COMMIT over connection 13300010
+-- We do not support CTEs
+WITH fist_table_agg AS
+  (SELECT sum(value_1) as v1_agg, user_id FROM raw_events_first GROUP BY user_id)
+INSERT INTO agg_events
+            (value_1_agg, user_id)
+            SELECT
+              v1_agg, user_id
+            FROM
+              fist_table_agg;
+DEBUG:  StartTransactionCommand
+DEBUG:  StartTransaction
+DEBUG:  name: unnamed; blockState:       DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children: 
+ERROR:  cannot perform distributed planning for the given modification
+DETAIL:  Common table expressions are not allowed in INSERT ... SELECT queries
+-- We do not support CTEs in the INSERT as well
+INSERT INTO agg_events
+  WITH sub_cte AS (SELECT 1)
+  SELECT
+    raw_events_first.user_id, (SELECT * FROM sub_cte)
+  FROM
+    raw_events_first;
+DEBUG:  StartTransactionCommand
+DEBUG:  StartTransaction
+DEBUG:  name: unnamed; blockState:       DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children: 
+ERROR:  cannot perform distributed planning for the given modification
+DETAIL:  Common table expressions are not allowed in INSERT ... SELECT queries
+-- We do not support any set operations
+INSERT INTO
+  raw_events_first(user_id)
+SELECT
+  user_id
+FROM
+  ((SELECT user_id FROM raw_events_first) UNION
+   (SELECT user_id FROM raw_events_second)) as foo;
+DEBUG:  StartTransactionCommand
+DEBUG:  StartTransaction
+DEBUG:  name: unnamed; blockState:       DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children: 
+ERROR:  cannot perform distributed planning for the given modification
+DETAIL:  Set operations are not allowed in INSERT ... SELECT queries
+-- We do not support any set operations
+INSERT INTO
+  raw_events_first(user_id)
+  (SELECT user_id FROM raw_events_first) INTERSECT
+  (SELECT user_id FROM raw_events_first);
+DEBUG:  StartTransactionCommand
+DEBUG:  StartTransaction
+DEBUG:  name: unnamed; blockState:       DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children: 
+ERROR:  cannot perform distributed planning for the given modification
+DETAIL:  Set operations are not allowed in INSERT ... SELECT queries
+-- We do not support any set operations
+INSERT INTO
+  raw_events_first(user_id)
+SELECT
+  user_id
+FROM
+  ((SELECT user_id FROM raw_events_first WHERE user_id = 15) EXCEPT
+   (SELECT user_id FROM raw_events_second where user_id = 17)) as foo;
+DEBUG:  StartTransactionCommand
+DEBUG:  StartTransaction
+DEBUG:  name: unnamed; blockState:       DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children: 
+ERROR:  cannot perform distributed planning for the given modification
+DETAIL:  Set operations are not allowed in INSERT ... SELECT queries
 -- unsupported JOIN
 INSERT INTO agg_events 
             (value_4_agg, 

src/test/regress/sql/multi_insert_select.sql

@@ -217,57 +217,74 @@ RETURNING user_id, value_1_agg;
 
 
 -- TODO:: add hll and date_trunc 
-INSERT INTO agg_events (user_id, value_1_agg) 
+INSERT INTO agg_events (user_id, value_1_agg)
-SELECT 
+SELECT
-   user_id, sum(value_1 + value_2) 
+   user_id, sum(value_1 + value_2)
-FROM 
+FROM
    raw_events_first GROUP BY user_id;
 
 --  FILTER CLAUSE
-INSERT INTO agg_events (user_id, value_1_agg) 
+INSERT INTO agg_events (user_id, value_1_agg)
-SELECT 
+SELECT
    user_id, sum(value_1 + value_2) FILTER (where value_3 = 15)
-FROM 
+FROM
    raw_events_first GROUP BY user_id;
 
--- a very simple UNION query
-INSERT INTO 
-  raw_events_first(user_id) 
-SELECT 
-  user_id 
-FROM 
-  ((SELECT user_id FROM raw_events_first) UNION 
-   (SELECT user_id FROM raw_events_second)) as foo;
-
--- same query with slightly different syntax, but this time we cannot push it down
-INSERT INTO 
-  raw_events_first(user_id)
-  (SELECT user_id FROM raw_events_first) UNION 
-  (SELECT user_id FROM raw_events_first);
-
--- similar query with a filter on two of the queries
-INSERT INTO 
-  raw_events_first(user_id) 
-SELECT 
-  user_id 
-FROM 
-  ((SELECT user_id FROM raw_events_first WHERE user_id = 15) UNION 
-   (SELECT user_id FROM raw_events_second where user_id = 17)) as foo;
-
 -- TODO: UUIDs
 
 -- a test with reference table JOINs
-INSERT INTO 
+INSERT INTO
-  agg_events (user_id, value_1_agg) 
+  agg_events (user_id, value_1_agg)
-SELECT 
+SELECT
   raw_events_first.user_id, sum(value_1)
-FROM 
+FROM
-  reference_table, raw_events_first 
+  reference_table, raw_events_first
-WHERE 
+WHERE
   raw_events_first.user_id = reference_table.user_id
 GROUP BY
   raw_events_first.user_id;
 
+-- We do not support CTEs
+WITH fist_table_agg AS
+  (SELECT sum(value_1) as v1_agg, user_id FROM raw_events_first GROUP BY user_id)
+INSERT INTO agg_events
+            (value_1_agg, user_id)
+            SELECT
+              v1_agg, user_id
+            FROM
+              fist_table_agg;
+
+-- We do not support CTEs in the INSERT as well
+INSERT INTO agg_events
+  WITH sub_cte AS (SELECT 1)
+  SELECT
+    raw_events_first.user_id, (SELECT * FROM sub_cte)
+  FROM
+    raw_events_first;
+
+-- We do not support any set operations
+INSERT INTO
+  raw_events_first(user_id)
+SELECT
+  user_id
+FROM
+  ((SELECT user_id FROM raw_events_first) UNION
+   (SELECT user_id FROM raw_events_second)) as foo;
+
+-- We do not support any set operations
+INSERT INTO
+  raw_events_first(user_id)
+  (SELECT user_id FROM raw_events_first) INTERSECT
+  (SELECT user_id FROM raw_events_first);
+
+-- We do not support any set operations
+INSERT INTO
+  raw_events_first(user_id)
+SELECT
+  user_id
+FROM
+  ((SELECT user_id FROM raw_events_first WHERE user_id = 15) EXCEPT
+   (SELECT user_id FROM raw_events_second where user_id = 17)) as foo;
 
 -- unsupported JOIN
 INSERT INTO agg_events