Add basic regression tests for semi/antijoin recognition.

Add some simple tests that the planner recognizes all the
standard idioms for SEMI and ANTI joins.  Failure to optimize
in this way won't necessarily cause any visible change in
query results, so check the plans.  We had no similar coverage
before, at least for some variants of antijoin, as noted by
Richard Guo.

Discussion: https://postgr.es/m/CAMbWs4-mvPPCJ1W6iK6dD5HiNwoJdi6mZp=-7mE8N9Sh+cd0tQ@mail.gmail.com

diff --git a/src/test/regress/expected/join.out b/src/test/regress/expected/join.out
index 9b69a8c122128d41cacdf74b56b23881812787f2..b901d7299fad00e6310edc3bcae0b5329510b6e7 100644 (file)
--- a/src/test/regress/expected/join.out
+++ b/src/test/regress/expected/join.out
@@ -2650,6 +2650,69 @@ WHERE d.f1 IS NULL;
  9999
 (3 rows)
 
+--
+-- basic semijoin and antijoin recognition tests
+--
+explain (costs off)
+select a.* from tenk1 a
+where unique1 in (select unique2 from tenk1 b);
+                         QUERY PLAN                         
+------------------------------------------------------------
+ Hash Semi Join
+   Hash Cond: (a.unique1 = b.unique2)
+   ->  Seq Scan on tenk1 a
+   ->  Hash
+         ->  Index Only Scan using tenk1_unique2 on tenk1 b
+(5 rows)
+
+-- sadly, this is not an antijoin
+explain (costs off)
+select a.* from tenk1 a
+where unique1 not in (select unique2 from tenk1 b);
+                       QUERY PLAN                       
+--------------------------------------------------------
+ Seq Scan on tenk1 a
+   Filter: (NOT (hashed SubPlan 1))
+   SubPlan 1
+     ->  Index Only Scan using tenk1_unique2 on tenk1 b
+(4 rows)
+
+explain (costs off)
+select a.* from tenk1 a
+where exists (select 1 from tenk1 b where a.unique1 = b.unique2);
+                         QUERY PLAN                         
+------------------------------------------------------------
+ Hash Semi Join
+   Hash Cond: (a.unique1 = b.unique2)
+   ->  Seq Scan on tenk1 a
+   ->  Hash
+         ->  Index Only Scan using tenk1_unique2 on tenk1 b
+(5 rows)
+
+explain (costs off)
+select a.* from tenk1 a
+where not exists (select 1 from tenk1 b where a.unique1 = b.unique2);
+                         QUERY PLAN                         
+------------------------------------------------------------
+ Hash Anti Join
+   Hash Cond: (a.unique1 = b.unique2)
+   ->  Seq Scan on tenk1 a
+   ->  Hash
+         ->  Index Only Scan using tenk1_unique2 on tenk1 b
+(5 rows)
+
+explain (costs off)
+select a.* from tenk1 a left join tenk1 b on a.unique1 = b.unique2
+where b.unique2 is null;
+                         QUERY PLAN                         
+------------------------------------------------------------
+ Hash Anti Join
+   Hash Cond: (a.unique1 = b.unique2)
+   ->  Seq Scan on tenk1 a
+   ->  Hash
+         ->  Index Only Scan using tenk1_unique2 on tenk1 b
+(5 rows)
+
 --
 -- regression test for proper handling of outer joins within antijoins
 --

diff --git a/src/test/regress/sql/join.sql b/src/test/regress/sql/join.sql
index 27e7e741a15fc8cbb3973537bdda9cef805f8b54..ccbbe5454c5bd09d82dc735191b92bdede1d5b46 100644 (file)
--- a/src/test/regress/sql/join.sql
+++ b/src/test/regress/sql/join.sql
@@ -624,6 +624,31 @@ LEFT JOIN (
 ) AS d ON (a.f1 = d.f1)
 WHERE d.f1 IS NULL;
 
+--
+-- basic semijoin and antijoin recognition tests
+--
+
+explain (costs off)
+select a.* from tenk1 a
+where unique1 in (select unique2 from tenk1 b);
+
+-- sadly, this is not an antijoin
+explain (costs off)
+select a.* from tenk1 a
+where unique1 not in (select unique2 from tenk1 b);
+
+explain (costs off)
+select a.* from tenk1 a
+where exists (select 1 from tenk1 b where a.unique1 = b.unique2);
+
+explain (costs off)
+select a.* from tenk1 a
+where not exists (select 1 from tenk1 b where a.unique1 = b.unique2);
+
+explain (costs off)
+select a.* from tenk1 a left join tenk1 b on a.unique1 = b.unique2
+where b.unique2 is null;
+
 --
 -- regression test for proper handling of outer joins within antijoins
 --