Fix EXPLAIN ANALYZE exec when query returns no cols (#4672)

We do not include dummy column if original task didn't return any
columns.
Otherwise, number of columns that original task returned wouldn't
match number of columns returned by worker_save_query_explain_analyze.

src/backend/distributed/planner/multi_explain.c

@@ -1512,8 +1512,16 @@ WrapQueryForExplainAnalyze(const char *queryString, TupleDesc tupleDesc)
 					 ExplainFormatStr(CurrentDistributedQueryExplainOptions.format));
 
 	StringInfo wrappedQuery = makeStringInfo();
+
+	/*
+	 * We do not include dummy column if original query didn't return any columns.
+	 * Otherwise, number of columns that original query returned wouldn't match
+	 * number of columns returned by worker_save_query_explain_analyze.
+	 */
+	char *workerSaveQueryFetchCols = (tupleDesc->natts == 0) ? "" : "*";
 	appendStringInfo(wrappedQuery,
-					 "SELECT * FROM worker_save_query_explain_analyze(%s, %s) AS (%s)",
+					 "SELECT %s FROM worker_save_query_explain_analyze(%s, %s) AS (%s)",
+					 workerSaveQueryFetchCols,
 					 quote_literal_cstr(queryString),
 					 quote_literal_cstr(explainOptions->data),
 					 columnDef->data);

src/test/regress/expected/multi_explain.out

@@ -2956,3 +2956,78 @@ Custom Scan (Citus Adaptive) (actual rows=1 loops=1)
 -- reset back
 reset citus.explain_analyze_sort_method;
 DROP TABLE explain_analyze_execution_time;
+CREATE SCHEMA multi_explain;
+SET search_path TO multi_explain;
+-- test EXPLAIN ANALYZE when original query returns no columns
+CREATE TABLE reference_table(a int);
+SELECT create_reference_table('reference_table');
+
+INSERT INTO reference_table VALUES (1);
+EXPLAIN :default_analyze_flags SELECT FROM reference_table;
+Custom Scan (Citus Adaptive) (actual rows=1 loops=1)
+  Task Count: 1
+  Tasks Shown: All
+  ->  Task
+        Node: host=localhost port=xxxxx dbname=regression
+        ->  Seq Scan on reference_table_570031 reference_table (actual rows=1 loops=1)
+CREATE TABLE distributed_table_1(a int, b int);
+SELECT create_distributed_table('distributed_table_1','a');
+
+INSERT INTO distributed_table_1 values (1,1);
+EXPLAIN :default_analyze_flags SELECT row_number() OVER() AS r FROM distributed_table_1;
+WindowAgg (actual rows=1 loops=1)
+  ->  Custom Scan (Citus Adaptive) (actual rows=1 loops=1)
+        Task Count: 2
+        Tasks Shown: One of 2
+        ->  Task
+              Node: host=localhost port=xxxxx dbname=regression
+              ->  Seq Scan on distributed_table_1_570032 distributed_table_xxx (actual rows=1 loops=1)
+CREATE TABLE distributed_table_2(a int, b int);
+SELECT create_distributed_table('distributed_table_2','a');
+
+INSERT INTO distributed_table_2 VALUES (1,1);
+EXPLAIN :default_analyze_flags
+WITH r AS (SELECT row_number() OVER () AS r FROM distributed_table_1)
+SELECT * FROM distributed_table_2
+JOIN r ON (r = distributed_table_2.b)
+LIMIT 3;
+Limit (actual rows=1 loops=1)
+  ->  Custom Scan (Citus Adaptive) (actual rows=1 loops=1)
+        ->  Distributed Subplan XXX_1
+              Intermediate Data Size: 14 bytes
+              Result destination: Send to 2 nodes
+              ->  WindowAgg (actual rows=1 loops=1)
+                    ->  Custom Scan (Citus Adaptive) (actual rows=1 loops=1)
+                          Task Count: 2
+                          Tasks Shown: One of 2
+                          ->  Task
+                                Node: host=localhost port=xxxxx dbname=regression
+                                ->  Seq Scan on distributed_table_1_570032 distributed_table_xxx (actual rows=1 loops=1)
+        Task Count: 2
+        Tuple data received from nodes: 3 bytes
+        Tasks Shown: One of 2
+        ->  Task
+              Tuple data received from node: 3 bytes
+              Node: host=localhost port=xxxxx dbname=regression
+              ->  Limit (actual rows=1 loops=1)
+                    ->  Nested Loop (actual rows=1 loops=1)
+                          Join Filter: (distributed_table_2.b = intermediate_result.r)
+                          ->  Function Scan on read_intermediate_result intermediate_result (actual rows=1 loops=1)
+                          ->  Seq Scan on distributed_table_2_570034 distributed_table_xxx (actual rows=1 loops=1)
+EXPLAIN :default_analyze_flags SELECT FROM (SELECT * FROM reference_table) subquery;
+Custom Scan (Citus Adaptive) (actual rows=1 loops=1)
+  Task Count: 1
+  Tasks Shown: All
+  ->  Task
+        Node: host=localhost port=xxxxx dbname=regression
+        ->  Seq Scan on reference_table_570031 reference_table (actual rows=1 loops=1)
+PREPARE dummy_prep_stmt(int) AS SELECT FROM distributed_table_1;
+EXPLAIN :default_analyze_flags EXECUTE dummy_prep_stmt(50);
+Custom Scan (Citus Adaptive) (actual rows=1 loops=1)
+  Task Count: 2
+  Tasks Shown: One of 2
+  ->  Task
+        Node: host=localhost port=xxxxx dbname=regression
+        ->  Seq Scan on distributed_table_1_570032 distributed_table_xxx (actual rows=1 loops=1)
+SET client_min_messages TO ERROR;
+DROP SCHEMA multi_explain CASCADE;

src/test/regress/sql/multi_explain.sql

@@ -1052,3 +1052,37 @@ EXPLAIN (COSTS FALSE, ANALYZE TRUE, TIMING FALSE, SUMMARY FALSE) select a, CASE
 -- reset back
 reset citus.explain_analyze_sort_method;
 DROP TABLE explain_analyze_execution_time;
+
+CREATE SCHEMA multi_explain;
+SET search_path TO multi_explain;
+
+-- test EXPLAIN ANALYZE when original query returns no columns
+CREATE TABLE reference_table(a int);
+SELECT create_reference_table('reference_table');
+INSERT INTO reference_table VALUES (1);
+
+EXPLAIN :default_analyze_flags SELECT FROM reference_table;
+
+CREATE TABLE distributed_table_1(a int, b int);
+SELECT create_distributed_table('distributed_table_1','a');
+INSERT INTO distributed_table_1 values (1,1);
+
+EXPLAIN :default_analyze_flags SELECT row_number() OVER() AS r FROM distributed_table_1;
+
+CREATE TABLE distributed_table_2(a int, b int);
+SELECT create_distributed_table('distributed_table_2','a');
+INSERT INTO distributed_table_2 VALUES (1,1);
+
+EXPLAIN :default_analyze_flags
+WITH r AS (SELECT row_number() OVER () AS r FROM distributed_table_1)
+SELECT * FROM distributed_table_2
+JOIN r ON (r = distributed_table_2.b)
+LIMIT 3;
+
+EXPLAIN :default_analyze_flags SELECT FROM (SELECT * FROM reference_table) subquery;
+
+PREPARE dummy_prep_stmt(int) AS SELECT FROM distributed_table_1;
+EXPLAIN :default_analyze_flags EXECUTE dummy_prep_stmt(50);
+
+SET client_min_messages TO ERROR;
+DROP SCHEMA multi_explain CASCADE;