[ORCA]: fallback to postgres optimizer when NL inner plan is index scan

ORCA may produce wrong plan for missing motion, see issue 567.

src/backend/gporca/libgpopt/src/operators/CPhysicalInnerIndexNLJoin.cpp

@@ -17,6 +17,7 @@
 #include "gpopt/base/CDistributionSpecHashed.h"
 #include "gpopt/base/CDistributionSpecNonSingleton.h"
 #include "gpopt/base/CDistributionSpecReplicated.h"
+#include "gpopt/exception.h"
 #include "gpopt/operators/CExpressionHandle.h"
 #include "gpopt/operators/CPredicateUtils.h"
 
@@ -118,6 +119,13 @@ CPhysicalInnerIndexNLJoin::Ped(CMemoryPool *mp, CExpressionHandle &exprhdl,
 	CEnfdDistribution::EDistributionMatching dmatch =
 		Edm(prppInput, child_index, pdrgpdpCtxt, ulDistrReq);
 
+	// FIXME: nestloop with inner index scan may produce wrong plan, see
+	// issue https://github.com/cloudberrydb/cloudberrydb/issues/567
+	// Fallback to postgres optimizer to avoid wrong plan. We should
+	// fix this issue and remove the following exception.
+	GPOS_RAISE(gpopt::ExmaGPOPT, gpopt::ExmiUnsupportedOp,
+			   GPOS_WSZ_LIT("Fallback: InnerIndexNestLoopJoin may have wrong plan"));
+
 	if (1 == child_index)
 	{
 		// inner (index-scan side) is requested for Any distribution,

src/test/regress/expected/aggregates_optimizer.out

@@ -1346,31 +1346,39 @@ explain (costs off) select a,c from t1 group by a,c,d;
 explain (costs off) select *
 from t1 inner join t2 on t1.a = t2.x and t1.b = t2.y
 group by t1.a,t1.b,t1.c,t1.d,t2.x,t2.y,t2.z;
-                      QUERY PLAN                       
--------------------------------------------------------
+INFO:  GPORCA failed to produce a plan, falling back to planner
+DETAIL:  Operator Fallback: InnerIndexNestLoopJoin may have wrong plan not supported
+                         QUERY PLAN                         
+------------------------------------------------------------
  Gather Motion 3:1  (slice1; segments: 3)
-   ->  Nested Loop
-         Join Filter: true
-         ->  Seq Scan on t1
-         ->  Index Scan using t2_pkey on t2
-               Index Cond: ((x = t1.a) AND (y = t1.b))
- Optimizer: Pivotal Optimizer (GPORCA)
-(7 rows)
+   ->  HashAggregate
+         Group Key: t1.a, t1.b, t2.x, t2.y
+         ->  Hash Join
+               Hash Cond: ((t2.x = t1.a) AND (t2.y = t1.b))
+               ->  Seq Scan on t2
+               ->  Hash
+                     ->  Seq Scan on t1
+ Optimizer: Postgres query optimizer
+(9 rows)
 
 -- Test case where t1 can be optimized but not t2
 explain (costs off) select t1.*,t2.x,t2.z
 from t1 inner join t2 on t1.a = t2.x and t1.b = t2.y
 group by t1.a,t1.b,t1.c,t1.d,t2.x,t2.z;
-                      QUERY PLAN                       
--------------------------------------------------------
+INFO:  GPORCA failed to produce a plan, falling back to planner
+DETAIL:  Operator Fallback: InnerIndexNestLoopJoin may have wrong plan not supported
+                         QUERY PLAN                         
+------------------------------------------------------------
  Gather Motion 3:1  (slice1; segments: 3)
-   ->  Nested Loop
-         Join Filter: true
-         ->  Seq Scan on t1
-         ->  Index Scan using t2_pkey on t2
-               Index Cond: ((x = t1.a) AND (y = t1.b))
- Optimizer: Pivotal Optimizer (GPORCA)
-(7 rows)
+   ->  HashAggregate
+         Group Key: t1.a, t1.b, t2.x, t2.z
+         ->  Hash Join
+               Hash Cond: ((t2.x = t1.a) AND (t2.y = t1.b))
+               ->  Seq Scan on t2
+               ->  Hash
+                     ->  Seq Scan on t1
+ Optimizer: Postgres query optimizer
+(9 rows)
 
 -- Cannot optimize when PK is deferrable
 explain (costs off) select * from t3 group by a,b,c;

src/test/regress/expected/co_nestloop_idxscan_optimizer.out

@@ -27,16 +27,14 @@ create index foo_id_idx on co_nestloop_idxscan.foo(id);
 explain select f.id from co_nestloop_idxscan.foo f, co_nestloop_idxscan.bar b where f.id = b.id;
                                      QUERY PLAN                                      
 -------------------------------------------------------------------------------------
- Gather Motion 3:1  (slice1; segments: 3)  (cost=0.00..499.13 rows=1 width=8)
-   ->  Nested Loop  (cost=0.00..499.13 rows=1 width=8)
-         Join Filter: true
-         ->  Seq Scan on bar  (cost=0.00..431.00 rows=1 width=8)
-         ->  Bitmap Heap Scan on foo  (cost=0.00..68.13 rows=1 width=8)
-               Recheck Cond: (id = bar.id)
-               ->  Bitmap Index Scan on foo_id_idx  (cost=0.00..0.00 rows=0 width=0)
-                     Index Cond: (id = bar.id)
- Optimizer: Pivotal Optimizer (GPORCA) version 3.72.0
-(9 rows)
+ Gather Motion 3:1  (slice1; segments: 3)  (cost=1.02..510.33 rows=6 width=8)
+   ->  Hash Join  (cost=1.02..510.25 rows=2 width=8)
+         Hash Cond: (f.id = b.id)
+         ->  Seq Scan on foo f  (cost=0.00..509.17 rows=17 width=8)
+         ->  Hash  (cost=1.01..1.01 rows=1 width=8)
+               ->  Seq Scan on bar b  (cost=0.00..1.01 rows=1 width=8)
+ Optimizer: Postgres query optimizer
+(7 rows)
 
 select f.id from co_nestloop_idxscan.foo f, co_nestloop_idxscan.bar b where f.id = b.id;
  id 
@@ -51,16 +49,15 @@ set enable_nestloop=on;
 explain select f.id from co_nestloop_idxscan.foo f, co_nestloop_idxscan.bar b where f.id = b.id;
                                      QUERY PLAN                                      
 -------------------------------------------------------------------------------------
- Gather Motion 3:1  (slice1; segments: 3)  (cost=0.00..499.13 rows=1 width=8)
-   ->  Nested Loop  (cost=0.00..499.13 rows=1 width=8)
-         Join Filter: true
-         ->  Seq Scan on bar  (cost=0.00..431.00 rows=1 width=8)
-         ->  Bitmap Heap Scan on foo  (cost=0.00..68.13 rows=1 width=8)
-               Recheck Cond: (id = bar.id)
-               ->  Bitmap Index Scan on foo_id_idx  (cost=0.00..0.00 rows=0 width=0)
-                     Index Cond: (id = bar.id)
- Optimizer: Pivotal Optimizer (GPORCA) version 3.72.0
-(9 rows)
+ Gather Motion 3:1  (slice1; segments: 3)  (cost=8.15..13.26 rows=6 width=8)
+   ->  Nested Loop  (cost=8.15..13.18 rows=2 width=8)
+         ->  Seq Scan on bar b  (cost=0.00..1.01 rows=1 width=8)
+         ->  Bitmap Heap Scan on foo f  (cost=8.15..12.16 rows=1 width=8)
+               Recheck Cond: (id = b.id)
+               ->  Bitmap Index Scan on foo_id_idx  (cost=0.00..8.15 rows=1 width=0)
+                     Index Cond: (id = b.id)
+ Optimizer: Postgres query optimizer
+(8 rows)
 
 select f.id from co_nestloop_idxscan.foo f, co_nestloop_idxscan.bar b where f.id = b.id;
  id 
@@ -76,16 +73,15 @@ set enable_seqscan = off;
 explain select f.id from co_nestloop_idxscan.bar b, co_nestloop_idxscan.foo f where f.id = b.id;
                                      QUERY PLAN                                      
 -------------------------------------------------------------------------------------
- Gather Motion 3:1  (slice1; segments: 3)  (cost=0.00..499.13 rows=1 width=8)
-   ->  Nested Loop  (cost=0.00..499.13 rows=1 width=8)
-         Join Filter: true
-         ->  Seq Scan on bar  (cost=0.00..431.00 rows=1 width=8)
-         ->  Bitmap Heap Scan on foo  (cost=0.00..68.13 rows=1 width=8)
-               Recheck Cond: (id = bar.id)
-               ->  Bitmap Index Scan on foo_id_idx  (cost=0.00..0.00 rows=0 width=0)
-                     Index Cond: (id = bar.id)
- Optimizer: Pivotal Optimizer (GPORCA) version 3.72.0
-(9 rows)
+ Gather Motion 3:1  (slice1; segments: 3)  (cost=10000000008.15..10000000013.26 rows=6 width=8)
+   ->  Nested Loop  (cost=10000000008.15..10000000013.18 rows=2 width=8)
+         ->  Seq Scan on bar b  (cost=10000000000.00..10000000001.01 rows=1 width=8)
+         ->  Bitmap Heap Scan on foo f  (cost=8.15..12.16 rows=1 width=8)
+               Recheck Cond: (id = b.id)
+               ->  Bitmap Index Scan on foo_id_idx  (cost=0.00..8.15 rows=1 width=0)
+                     Index Cond: (id = b.id)
+ Optimizer: Postgres query optimizer
+(8 rows)
 
 select f.id from co_nestloop_idxscan.foo f, co_nestloop_idxscan.bar b where f.id = b.id;
  id