Update comments, checks

src/backend/distributed/planner/deparse_shard_query.c

@@ -222,8 +222,8 @@ DefineQualsForShardInterval(RelationShard *relationShard, int attnum, int rtinde
 
 	/*
 	 * Add constraints for the relation identified by rtindex, specifically on its column at attnum.
-	 * Create a Var node representing this column, which will be used to compare against the partition
-	 * column for shard interval qualification.
+	 * Create a Var node representing this column, which will be used to compare against the bounds
+	 * from the partition column of shard interval.
 	 */
 
 	Var *outerTablePartitionColumnVar = makeVar(
@@ -371,6 +371,7 @@ UpdateWhereClauseForOuterJoinWalker(Node *node, List *relationShardList)
  *            outerRtIndex : RT index whose column we will constrain,
  *            outerRte / innerRte,
  *            attnum       : attribute number (partition column) on outer side.
+ *                           This is compared to partition column of innerRte.
  *   3. Find the RelationShard for the inner distributed table (innerRte->relid)
  *      in relationShardList; skip if absent (no fixed shard chosen).
  *   4. Build the shard qualification with DefineQualsForShardInterval():
@@ -385,7 +386,7 @@ UpdateWhereClauseForOuterJoinWalker(Node *node, List *relationShardList)
 void
 UpdateWhereClauseForOuterJoin(Query *query, List *relationShardList)
 {
-	if (query == NULL || query->jointree == NULL || query->jointree->fromlist == NIL)
+	if (query == NULL)
 	{
 		return;
 	}

src/backend/distributed/planner/multi_physical_planner.c

@@ -2357,8 +2357,6 @@ QueryPushdownSqlTaskList(Query *query, uint64 jobId,
  *
  *  The function also sets outerPartHasDistributedTable if the outer part
  *  of the corresponding join has a distributed table.
- *
- * The function returns true only if all three conditions above hold true.
  */
 static bool
 IsInnerTableOfOuterJoin(RelationRestriction *relationRestriction,
@@ -2598,17 +2596,11 @@ QueryPushdownTaskCreate(Query *originalQuery, int shardIndex,
 	 */
 	UpdateRelationToShardNames((Node *) taskQuery, relationShardList);
 
-	if (updateQualsForOuterJoin)
-	{
-		UpdateWhereClauseForOuterJoinWalker((Node *) taskQuery, relationShardList);
-	}
-
 	/*
 	 * Ands are made implicit during shard pruning, as predicate comparison and
 	 * refutation depend on it being so. We need to make them explicit again so
 	 * that the query string is generated as (...) AND (...) as opposed to
 	 * (...), (...).
-	 * TODO: do we need to run this before adding quals?
 	 */
 	if (taskQuery->jointree->quals != NULL && IsA(taskQuery->jointree->quals, List))
 	{
@@ -2616,6 +2608,12 @@ QueryPushdownTaskCreate(Query *originalQuery, int shardIndex,
 			(List *) taskQuery->jointree->quals);
 	}
 
+	if (updateQualsForOuterJoin)
+	{
+		UpdateWhereClauseForOuterJoinWalker((Node *) taskQuery, relationShardList);
+	}
+
+
 	Task *subqueryTask = CreateBasicTask(jobId, taskId, taskType, NULL);
 
 	if ((taskType == MODIFY_TASK && !modifyRequiresCoordinatorEvaluation) ||

src/backend/distributed/shared_library_init.c

@@ -2334,7 +2334,6 @@ RegisterCitusConfigVariables(void)
 		GUC_UNIT_BYTE | GUC_NO_SHOW_ALL | GUC_NOT_IN_SAMPLE,
 		NULL, NULL, NULL);
 
-
 	DefineCustomIntVariable(
 		"citus.recover_2pc_interval",
 		gettext_noop("Sets the time to wait between recovering 2PCs."),

src/test/regress/expected/multi_subquery_in_where_reference_clause.out

@@ -306,7 +306,7 @@ ORDER BY 1, 2;
        5 |       5
 (2 rows)
 
--- reference table LEFT JOIN distributed table in WHERE is ok
+-- reference table LEFT JOIN distributed table in WHERE is still not ok
 SELECT user_id, value_2 FROM users_table WHERE
   value_1 > 1 AND value_1 < 3
   AND value_2 >= 5

src/test/regress/sql/multi_subquery_in_where_reference_clause.sql

@@ -260,7 +260,7 @@ SELECT user_id, value_2 FROM users_table WHERE
 )
 ORDER BY 1, 2;
 
--- reference table LEFT JOIN distributed table in WHERE is ok
+-- reference table LEFT JOIN distributed table in WHERE is still not ok
 SELECT user_id, value_2 FROM users_table WHERE
   value_1 > 1 AND value_1 < 3
   AND value_2 >= 5