[AQUMV] Allow to use normal materialized views to answer query.

As we have enabled materialized view data status maintenance,
normal materialized views could also be used to answer query
if their data status is up to date.
There is no difference between IVM and normal materialized
views to answer query.

example, compute agg from a normal materialized view:

create table t1(c1 int, c2 int, c3 int) distributed by (c1);

create materialized view normal_mv_t1 as
  select c3 as mc3, c1 as mc1 from t1 where c1 > 90;

set enable_answer_query_using_materialized_views = on;
explain(costs off, verbose)
select count(c3) from t1 where c1 > 90;
                    QUERY PLAN
------------------------------------------------------
 Finalize Aggregate
   Output: count(mc3)
   ->  Gather Motion 3:1  (slice1; segments: 3)
         Output: (PARTIAL count(mc3))
         ->  Partial Aggregate
               Output: PARTIAL count(mc3)
               ->  Seq Scan on aqumv.normal_mv_t1
                     Output: mc3, mc1
 Optimizer: Postgres query optimizer

Authored-by: Zhang Mingli avamingli@gmail.com

Author: avamingli

src/backend/catalog/gp_matview_aux.c

@@ -421,6 +421,9 @@ bool
 MatviewUsableForAppendAgg(Oid mvoid)
 {
 	HeapTuple mvauxtup = SearchSysCacheCopy1(MVAUXOID, ObjectIdGetDatum(mvoid));
+	if (!HeapTupleIsValid(mvauxtup))
+		return false;
+
 	Form_gp_matview_aux auxform = (Form_gp_matview_aux) GETSTRUCT(mvauxtup);
 	return ((auxform->datastatus == MV_DATA_STATUS_UP_TO_DATE) || 
 			(auxform->datastatus == MV_DATA_STATUS_EXPIRED_INSERT_ONLY));
@@ -438,6 +441,11 @@ bool
 MatviewIsGeneralyUpToDate(Oid mvoid)
 {
 	HeapTuple mvauxtup = SearchSysCacheCopy1(MVAUXOID, ObjectIdGetDatum(mvoid));
+
+	/* Not a candidate we recorded. */
+	if (!HeapTupleIsValid(mvauxtup))
+		return false;
+
 	Form_gp_matview_aux auxform = (Form_gp_matview_aux) GETSTRUCT(mvauxtup);
 	return ((auxform->datastatus == MV_DATA_STATUS_UP_TO_DATE) || 
 			(auxform->datastatus == MV_DATA_STATUS_UP_REORGANIZED));

src/backend/optimizer/plan/aqumv.c

@@ -18,6 +18,7 @@
 #include "access/htup_details.h"
 #include "access/table.h"
 #include "catalog/catalog.h"
+#include "catalog/gp_matview_aux.h"
 #include "catalog/pg_class_d.h"
 #include "catalog/pg_inherits.h"
 #include "catalog/pg_rewrite.h"
@@ -167,12 +168,18 @@ answer_query_using_materialized_views(PlannerInfo *root,
 		matviewRel = table_open(rewrite_tup->ev_class, AccessShareLock);
 		need_close = true;
 
+		if (!RelationIsPopulated(matviewRel))
+			continue;
+
 		/*
 		 * AQUMV
 		 * Currently the data of IVM is always up-to-date if there were.
-		 * Take care of this when IVM defered-fefresh is supported.
+		 * Take care of this when IVM defered-fefresh is supported(in SERVERLESS mode).
+		 * 
+		 * Normal materialized views could also be used if its data is up to date.
 		 */
-		if (!(RelationIsIVM(matviewRel) && RelationIsPopulated(matviewRel)))
+		if (!RelationIsIVM(matviewRel) &&
+			!MatviewIsGeneralyUpToDate(RelationGetRelid(matviewRel)))
 			continue;
 
 		if (matviewRel->rd_rel->relhasrules == false ||

src/test/regress/expected/aqumv.out

@@ -988,6 +988,118 @@ select count(c2), count(*) from aqumv_t2 where c1 > 90;
 (1 row)
 
 abort;
+-- Test use normal materialized views
+create table t1(c1 int, c2 int, c3 int) distributed by (c1);
+insert into t1 select i, i+1, i+2 from generate_series(1, 100) i;
+insert into t1 values (91, NULL, 95);
+analyze t1;
+create materialized view normal_mv_t1 as
+  select c3 as mc3, c1 as mc1
+  from t1 where c1 > 90;
+NOTICE:  Table doesn't have 'DISTRIBUTED BY' clause -- Using column(s) named 'mc1' as the Cloudberry Database data distribution key for this table.
+HINT:  The 'DISTRIBUTED BY' clause determines the distribution of data. Make sure column(s) chosen are the optimal data distribution key to minimize skew.
+analyze normal_mv_t1;
+set enable_answer_query_using_materialized_views = off;
+explain(costs off, verbose)
+select count(c3) from t1 where c1 > 90;
+                                    QUERY PLAN                                     
+-----------------------------------------------------------------------------------
+ Finalize Aggregate
+   Output: count(c3)
+   ->  Gather Motion 3:1  (slice1; segments: 3)
+         Output: (PARTIAL count(c3))
+         ->  Partial Aggregate
+               Output: PARTIAL count(c3)
+               ->  Seq Scan on aqumv.t1
+                     Output: c1, c2, c3
+                     Filter: (t1.c1 > 90)
+ Settings: enable_answer_query_using_materialized_views = 'off', optimizer = 'off'
+ Optimizer: Postgres query optimizer
+(11 rows)
+
+select count(c3) from t1 where c1 > 90;
+ count 
+-------
+    11
+(1 row)
+
+set enable_answer_query_using_materialized_views = on;
+explain(costs off, verbose)
+select count(c3) from t1 where c1 > 90;
+                                    QUERY PLAN                                    
+----------------------------------------------------------------------------------
+ Finalize Aggregate
+   Output: count(mc3)
+   ->  Gather Motion 3:1  (slice1; segments: 3)
+         Output: (PARTIAL count(mc3))
+         ->  Partial Aggregate
+               Output: PARTIAL count(mc3)
+               ->  Seq Scan on aqumv.normal_mv_t1
+                     Output: mc3, mc1
+ Settings: enable_answer_query_using_materialized_views = 'on', optimizer = 'off'
+ Optimizer: Postgres query optimizer
+(10 rows)
+
+select count(c3) from t1 where c1 > 90;
+ count 
+-------
+    11
+(1 row)
+
+vacuum full t1;
+explain(costs off, verbose)
+select count(c3) from t1 where c1 > 90;
+                                    QUERY PLAN                                    
+----------------------------------------------------------------------------------
+ Finalize Aggregate
+   Output: count(mc3)
+   ->  Gather Motion 3:1  (slice1; segments: 3)
+         Output: (PARTIAL count(mc3))
+         ->  Partial Aggregate
+               Output: PARTIAL count(mc3)
+               ->  Seq Scan on aqumv.normal_mv_t1
+                     Output: mc3, mc1
+ Settings: enable_answer_query_using_materialized_views = 'on', optimizer = 'off'
+ Optimizer: Postgres query optimizer
+(10 rows)
+
+explain(costs off, verbose)
+select c3 from t1 where c1 > 90;
+                                    QUERY PLAN                                    
+----------------------------------------------------------------------------------
+ Gather Motion 3:1  (slice1; segments: 3)
+   Output: mc3
+   ->  Seq Scan on aqumv.normal_mv_t1
+         Output: mc3
+ Settings: enable_answer_query_using_materialized_views = 'on', optimizer = 'off'
+ Optimizer: Postgres query optimizer
+(6 rows)
+
+-- insert data after refresh
+insert into t1 values (91, NULL, 95);
+explain(costs off, verbose)
+select count(c3) from t1 where c1 > 90;
+                                    QUERY PLAN                                    
+----------------------------------------------------------------------------------
+ Finalize Aggregate
+   Output: count(c3)
+   ->  Gather Motion 3:1  (slice1; segments: 3)
+         Output: (PARTIAL count(c3))
+         ->  Partial Aggregate
+               Output: PARTIAL count(c3)
+               ->  Seq Scan on aqumv.t1
+                     Output: c1, c2, c3
+                     Filter: (t1.c1 > 90)
+ Settings: enable_answer_query_using_materialized_views = 'on', optimizer = 'off'
+ Optimizer: Postgres query optimizer
+(11 rows)
+
+select mvname, datastatus from gp_matview_aux where mvname = 'normal_mv_t1';
+    mvname    | datastatus 
+--------------+------------
+ normal_mv_t1 | e
+(1 row)
+
 -- Test Agg on IMMV who has less columns than origin table.
 begin;
 create table aqumv_t2(c1 int, c2 int, c3 int) distributed by (c1);
@@ -2471,5 +2583,7 @@ select c2, c3 from aqumv_t7 where c1 > 90 order by c2, c3 fetch first 3 rows wit
 abort;
 reset optimizer;
 reset enable_answer_query_using_materialized_views;
-drop table aqumv_t1 cascade;
+-- start_ignore
+drop schema aqumv cascade;
+-- end_ignore
 reset search_path;

src/test/regress/sql/aqumv.sql

@@ -291,6 +291,39 @@ select count(c2), count(*) from aqumv_t2 where c1 > 90;
 select count(c2), count(*) from aqumv_t2 where c1 > 90;
 abort;
 
+-- Test use normal materialized views
+create table t1(c1 int, c2 int, c3 int) distributed by (c1);
+insert into t1 select i, i+1, i+2 from generate_series(1, 100) i;
+insert into t1 values (91, NULL, 95);
+analyze t1;
+
+create materialized view normal_mv_t1 as
+  select c3 as mc3, c1 as mc1
+  from t1 where c1 > 90;
+analyze normal_mv_t1;
+
+set enable_answer_query_using_materialized_views = off;
+explain(costs off, verbose)
+select count(c3) from t1 where c1 > 90;
+select count(c3) from t1 where c1 > 90;
+set enable_answer_query_using_materialized_views = on;
+explain(costs off, verbose)
+select count(c3) from t1 where c1 > 90;
+select count(c3) from t1 where c1 > 90;
+
+vacuum full t1;
+explain(costs off, verbose)
+select count(c3) from t1 where c1 > 90;
+
+explain(costs off, verbose)
+select c3 from t1 where c1 > 90;
+
+-- insert data after refresh
+insert into t1 values (91, NULL, 95);
+explain(costs off, verbose)
+select count(c3) from t1 where c1 > 90;
+select mvname, datastatus from gp_matview_aux where mvname = 'normal_mv_t1';
+
 -- Test Agg on IMMV who has less columns than origin table.
 begin;
 create table aqumv_t2(c1 int, c2 int, c3 int) distributed by (c1);
@@ -595,5 +628,7 @@ abort;
 
 reset optimizer;
 reset enable_answer_query_using_materialized_views;
-drop table aqumv_t1 cascade;
+-- start_ignore
+drop schema aqumv cascade;
+-- end_ignore
 reset search_path;