chore(planner): improve physical join (#16962)

* chore(planner): improve physical join

* chore(test): add sqllogictest

* chore(code): refine comments.

* chore(test): update sqllogictest

* chore(test):  test iceberg

* chore(test): test iceberg

* chore(planner): fix join type

* chore(test): update sqllogictest

* chore(test): update sqllogictest

---------

Co-authored-by: Bohu <overred.shuttler@gmail.com>

Author: Dousir9

src/query/sql/src/executor/physical_plans/physical_join.rs

@@ -46,8 +46,21 @@ pub fn physical_join(join: &Join, s_expr: &SExpr) -> Result<PhysicalJoinType> {
         return Ok(PhysicalJoinType::Hash);
     }
 
-    let left_prop = RelExpr::with_s_expr(s_expr.child(1)?).derive_relational_prop()?;
-    let right_prop = RelExpr::with_s_expr(s_expr.child(0)?).derive_relational_prop()?;
+    let left_rel_expr = RelExpr::with_s_expr(s_expr.child(0)?);
+    let right_rel_expr = RelExpr::with_s_expr(s_expr.child(1)?);
+    if matches!(
+        right_rel_expr
+            .derive_cardinality()?
+            .statistics
+            .precise_cardinality,
+        Some(1)
+    ) {
+        // If the output rows of build side is equal to 1, we use CROSS JOIN + FILTER instead of MERGE JOIN.
+        return Ok(PhysicalJoinType::Hash);
+    }
+
+    let left_prop = left_rel_expr.derive_relational_prop()?;
+    let right_prop = right_rel_expr.derive_relational_prop()?;
     let mut range_conditions = vec![];
     let mut other_conditions = vec![];
     for condition in join.non_equi_conditions.iter() {

tests/sqllogictests/suites/mode/standalone/explain/join.test

@@ -940,3 +940,47 @@ EvalScalar
 
 statement ok
 drop table t1;
+
+statement ok
+CREATE OR REPLACE TABLE t1(a INT);
+
+statement ok
+CREATE OR REPLACE TABLE t2(a INT);
+
+statement ok
+INSERT INTO t1 VALUES(1), (2), (3), (4);
+
+statement ok
+INSERT INTO t2 VALUES(1);
+
+query T
+EXPLAIN SELECT * FROM t1 WHERE a >= (SELECT MAX(a) FROM t2);
+----
+HashJoin
+├── output columns: [t1.a (#0)]
+├── join type: INNER
+├── build keys: []
+├── probe keys: []
+├── filters: [t1.a (#0) >= CAST(scalar_subquery_2 (#2) AS Int32 NULL)]
+├── estimated rows: 4.00
+├── EvalScalar(Build)
+│   ├── output columns: [MAX(a) (#2)]
+│   ├── expressions: [1]
+│   ├── estimated rows: 1.00
+│   └── DummyTableScan
+└── TableScan(Probe)
+    ├── table: default.default.t1
+    ├── output columns: [a (#0)]
+    ├── read rows: 4
+    ├── read size: < 1 KiB
+    ├── partitions total: 1
+    ├── partitions scanned: 1
+    ├── pruning stats: [segments: <range pruning: 1 to 1>, blocks: <range pruning: 1 to 1>]
+    ├── push downs: [filters: [], limit: NONE]
+    └── estimated rows: 4.00
+
+statement ok
+DROP TABLE t1;
+
+statement ok
+DROP TABLE t2;

tests/sqllogictests/suites/mode/standalone/explain_native/join.test

@@ -634,3 +634,47 @@ set enable_cbo = 1
 
 statement ok
 drop table t1
+
+statement ok
+CREATE OR REPLACE TABLE t1(a INT);
+
+statement ok
+CREATE OR REPLACE TABLE t2(a INT);
+
+statement ok
+INSERT INTO t1 VALUES(1), (2), (3), (4);
+
+statement ok
+INSERT INTO t2 VALUES(1);
+
+query T
+EXPLAIN SELECT * FROM t1 WHERE a >= (SELECT MAX(a) FROM t2);
+----
+HashJoin
+├── output columns: [t1.a (#0)]
+├── join type: INNER
+├── build keys: []
+├── probe keys: []
+├── filters: [t1.a (#0) >= CAST(scalar_subquery_2 (#2) AS Int32 NULL)]
+├── estimated rows: 4.00
+├── EvalScalar(Build)
+│   ├── output columns: [MAX(a) (#2)]
+│   ├── expressions: [1]
+│   ├── estimated rows: 1.00
+│   └── DummyTableScan
+└── TableScan(Probe)
+    ├── table: default.default.t1
+    ├── output columns: [a (#0)]
+    ├── read rows: 4
+    ├── read size: < 1 KiB
+    ├── partitions total: 1
+    ├── partitions scanned: 1
+    ├── pruning stats: [segments: <range pruning: 1 to 1>, blocks: <range pruning: 1 to 1>]
+    ├── push downs: [filters: [], limit: NONE]
+    └── estimated rows: 4.00
+
+statement ok
+DROP TABLE t1;
+
+statement ok
+DROP TABLE t2;

tests/sqllogictests/suites/tpch/join_order.test

@@ -532,8 +532,8 @@ group by
 order by
     value desc limit 100;
 ----
-RangeJoin: INNER
-├── Left
+HashJoin: INNER
+├── Build
 │   └── HashJoin: INNER
 │       ├── Build
 │       │   └── HashJoin: INNER
@@ -543,7 +543,7 @@ RangeJoin: INNER
 │       │           └── Scan: default.tpch_test.supplier (#4) (read rows: 1000)
 │       └── Probe
 │           └── Scan: default.tpch_test.partsupp (#3) (read rows: 80000)
-└── Right
+└── Probe
     └── HashJoin: INNER
         ├── Build
         │   └── HashJoin: INNER
@@ -1104,10 +1104,10 @@ order by
 ----
 HashJoin: RIGHT ANTI
 ├── Build
-│   └── RangeJoin: INNER
-│       ├── Left
+│   └── HashJoin: INNER
+│       ├── Build
 │       │   └── Scan: default.tpch_test.customer (#1) (read rows: 15000)
-│       └── Right
+│       └── Probe
 │           └── Scan: default.tpch_test.customer (#0) (read rows: 15000)
 └── Probe
     └── Scan: default.tpch_test.orders (#2) (read rows: 150000)