feat: Add HashJoin support for BuildRight

core/Cargo.toml

@@ -66,10 +66,10 @@ itertools = "0.11.0"
 chrono = { version = "0.4", default-features = false, features = ["clock"] }
 chrono-tz = { version = "0.8" }
 paste = "1.0.14"
-datafusion-common = { git = "https://github.com/viirya/arrow-datafusion.git", rev = "57b3be4" }
-datafusion = { default-features = false, git = "https://github.com/viirya/arrow-datafusion.git", rev = "57b3be4", features = ["unicode_expressions", "crypto_expressions"] }
-datafusion-functions = { git = "https://github.com/viirya/arrow-datafusion.git", rev = "57b3be4", features = ["crypto_expressions"]}
-datafusion-physical-expr = { git = "https://github.com/viirya/arrow-datafusion.git", rev = "57b3be4", default-features = false, features = ["unicode_expressions"] }
+datafusion-common = { git = "https://github.com/viirya/arrow-datafusion.git", rev = "debb2f2" }
+datafusion = { default-features = false, git = "https://github.com/viirya/arrow-datafusion.git", rev = "debb2f2", features = ["unicode_expressions", "crypto_expressions"] }
+datafusion-functions = { git = "https://github.com/viirya/arrow-datafusion.git", rev = "debb2f2", features = ["crypto_expressions"] }
+datafusion-physical-expr = { git = "https://github.com/viirya/arrow-datafusion.git", rev = "debb2f2", default-features = false, features = ["unicode_expressions"] }
 unicode-segmentation = "^1.10.1"
 once_cell = "1.18.0"
 regex = "1.9.6"

core/src/execution/datafusion/planner.rs

@@ -37,6 +37,7 @@ use datafusion::{
         },
         AggregateExpr, PhysicalExpr, PhysicalSortExpr, ScalarFunctionExpr,
     },
+    physical_optimizer::join_selection::swap_hash_join,
     physical_plan::{
         aggregates::{AggregateMode as DFAggregateMode, PhysicalGroupBy},
         filter::FilterExec,
@@ -90,7 +91,7 @@ use crate::{
             agg_expr::ExprStruct as AggExprStruct, expr::ExprStruct, literal::Value, AggExpr, Expr,
             ScalarFunc,
         },
-        spark_operator::{operator::OpStruct, JoinType, Operator},
+        spark_operator::{operator::OpStruct, BuildSide, JoinType, Operator},
         spark_partitioning::{partitioning::PartitioningStruct, Partitioning as SparkPartitioning},
     },
 };
@@ -966,7 +967,7 @@ impl PhysicalPlanner {
                     join.join_type,
                     &join.condition,
                 )?;
-                let join = Arc::new(HashJoinExec::try_new(
+                let hash_join = Arc::new(HashJoinExec::try_new(
                     join_params.left,
                     join_params.right,
                     join_params.join_on,
@@ -978,7 +979,15 @@ impl PhysicalPlanner {
                     // `EqualNullSafe`, Spark will rewrite it during planning.
                     false,
                 )?);
-                Ok((scans, join))
+
+                // If the hash join is build right, we need to swap the left and right
+                let hash_join = if join.build_side == BuildSide::BuildLeft as i32 {
+                    hash_join
+                } else {
+                    swap_hash_join(hash_join.as_ref(), PartitionMode::Partitioned)?
+                };
+
+                Ok((scans, hash_join))
             }
         }
     }

core/src/execution/proto/operator.proto

@@ -95,6 +95,7 @@ message HashJoin {
   repeated spark.spark_expression.Expr right_join_keys = 2;
   JoinType join_type = 3;
   optional spark.spark_expression.Expr condition = 4;
+  BuildSide build_side = 5;
 }
 
 message SortMergeJoin {
@@ -114,3 +115,8 @@ enum JoinType {
   LeftAnti = 6;
   RightAnti = 7;
 }
+
+enum BuildSide {
+  BuildLeft = 0;
+  BuildRight = 1;
+}

dev/diffs/3.4.2.diff

@@ -352,7 +352,7 @@ index 7dec558f8df..840dda15033 100644
      assert(exchanges.size == 2)
    }
 diff --git a/sql/core/src/test/scala/org/apache/spark/sql/DynamicPartitionPruningSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/DynamicPartitionPruningSuite.scala
-index f33432ddb6f..060f874ea72 100644
+index f33432ddb6f..9cf7a9dd4e3 100644
 --- a/sql/core/src/test/scala/org/apache/spark/sql/DynamicPartitionPruningSuite.scala
 +++ b/sql/core/src/test/scala/org/apache/spark/sql/DynamicPartitionPruningSuite.scala
 @@ -22,6 +22,7 @@ import org.scalatest.GivenWhenThen
@@ -373,7 +373,37 @@ index f33432ddb6f..060f874ea72 100644
        case _ => Nil
      }
    }
-@@ -1187,7 +1191,8 @@ abstract class DynamicPartitionPruningSuiteBase
+@@ -665,7 +669,8 @@ abstract class DynamicPartitionPruningSuiteBase
+     }
+   }
+
+-  test("partition pruning in broadcast hash joins with aliases") {
++  test("partition pruning in broadcast hash joins with aliases",
++    IgnoreComet("TODO: Support SubqueryBroadcastExec in Comet: #242")) {
+     Given("alias with simple join condition, using attribute names only")
+     withSQLConf(SQLConf.DYNAMIC_PARTITION_PRUNING_REUSE_BROADCAST_ONLY.key -> "true") {
+       val df = sql(
+@@ -755,7 +760,8 @@ abstract class DynamicPartitionPruningSuiteBase
+     }
+   }
+
+-  test("partition pruning in broadcast hash joins") {
++  test("partition pruning in broadcast hash joins",
++    IgnoreComet("TODO: Support SubqueryBroadcastExec in Comet: #242")) {
+     Given("disable broadcast pruning and disable subquery duplication")
+     withSQLConf(
+       SQLConf.DYNAMIC_PARTITION_PRUNING_REUSE_BROADCAST_ONLY.key -> "true",
+@@ -990,7 +996,8 @@ abstract class DynamicPartitionPruningSuiteBase
+     }
+   }
+
+-  test("different broadcast subqueries with identical children") {
++  test("different broadcast subqueries with identical children",
++    IgnoreComet("TODO: Support SubqueryBroadcastExec in Comet: #242")) {
+     withSQLConf(SQLConf.DYNAMIC_PARTITION_PRUNING_REUSE_BROADCAST_ONLY.key -> "true") {
+       withTable("fact", "dim") {
+         spark.range(100).select(
+@@ -1187,7 +1194,8 @@ abstract class DynamicPartitionPruningSuiteBase
      }
    }
 
@@ -383,7 +413,7 @@ index f33432ddb6f..060f874ea72 100644
      withSQLConf(SQLConf.DYNAMIC_PARTITION_PRUNING_REUSE_BROADCAST_ONLY.key -> "true") {
        val df = sql(
          """
-@@ -1238,7 +1243,8 @@ abstract class DynamicPartitionPruningSuiteBase
+@@ -1238,7 +1246,8 @@ abstract class DynamicPartitionPruningSuiteBase
      }
    }
 
@@ -393,7 +423,27 @@ index f33432ddb6f..060f874ea72 100644
      Given("dynamic pruning filter on the build side")
      withSQLConf(SQLConf.DYNAMIC_PARTITION_PRUNING_REUSE_BROADCAST_ONLY.key -> "true") {
        val df = sql(
-@@ -1485,7 +1491,7 @@ abstract class DynamicPartitionPruningSuiteBase
+@@ -1311,7 +1320,8 @@ abstract class DynamicPartitionPruningSuiteBase
+     }
+   }
+
+-  test("SPARK-32817: DPP throws error when the broadcast side is empty") {
++  test("SPARK-32817: DPP throws error when the broadcast side is empty",
++    IgnoreComet("TODO: Support SubqueryBroadcastExec in Comet: #242")) {
+     withSQLConf(
+       SQLConf.DYNAMIC_PARTITION_PRUNING_ENABLED.key -> "true",
+       SQLConf.DYNAMIC_PARTITION_PRUNING_REUSE_BROADCAST_ONLY.key -> "true",
+@@ -1470,7 +1480,8 @@ abstract class DynamicPartitionPruningSuiteBase
+     checkAnswer(df, Row(3, 2) :: Row(3, 2) :: Row(3, 2) :: Row(3, 2) :: Nil)
+   }
+
+-  test("SPARK-36444: Remove OptimizeSubqueries from batch of PartitionPruning") {
++  test("SPARK-36444: Remove OptimizeSubqueries from batch of PartitionPruning",
++    IgnoreComet("TODO: Support SubqueryBroadcastExec in Comet: #242")) {
+     withSQLConf(SQLConf.DYNAMIC_PARTITION_PRUNING_ENABLED.key -> "true") {
+       val df = sql(
+         """
+@@ -1485,7 +1496,7 @@ abstract class DynamicPartitionPruningSuiteBase
    }
 
    test("SPARK-38148: Do not add dynamic partition pruning if there exists static partition " +
@@ -402,7 +452,37 @@ index f33432ddb6f..060f874ea72 100644
      withSQLConf(SQLConf.DYNAMIC_PARTITION_PRUNING_ENABLED.key -> "true") {
        Seq(
          "f.store_id = 1" -> false,
-@@ -1729,6 +1735,8 @@ abstract class DynamicPartitionPruningV1Suite extends DynamicPartitionPruningDat
+@@ -1557,7 +1568,8 @@ abstract class DynamicPartitionPruningSuiteBase
+     }
+   }
+
+-  test("SPARK-38674: Remove useless deduplicate in SubqueryBroadcastExec") {
++  test("SPARK-38674: Remove useless deduplicate in SubqueryBroadcastExec",
++    IgnoreComet("TODO: Support SubqueryBroadcastExec in Comet: #242")) {
+     withTable("duplicate_keys") {
+       withSQLConf(SQLConf.DYNAMIC_PARTITION_PRUNING_ENABLED.key -> "true") {
+         Seq[(Int, String)]((1, "NL"), (1, "NL"), (3, "US"), (3, "US"), (3, "US"))
+@@ -1588,7 +1600,8 @@ abstract class DynamicPartitionPruningSuiteBase
+     }
+   }
+
+-  test("SPARK-39338: Remove dynamic pruning subquery if pruningKey's references is empty") {
++  test("SPARK-39338: Remove dynamic pruning subquery if pruningKey's references is empty",
++    IgnoreComet("TODO: Support SubqueryBroadcastExec in Comet: #242")) {
+     withSQLConf(SQLConf.DYNAMIC_PARTITION_PRUNING_ENABLED.key -> "true") {
+       val df = sql(
+         """
+@@ -1617,7 +1630,8 @@ abstract class DynamicPartitionPruningSuiteBase
+     }
+   }
+
+-  test("SPARK-39217: Makes DPP support the pruning side has Union") {
++  test("SPARK-39217: Makes DPP support the pruning side has Union",
++    IgnoreComet("TODO: Support SubqueryBroadcastExec in Comet: #242")) {
+     withSQLConf(SQLConf.DYNAMIC_PARTITION_PRUNING_ENABLED.key -> "true") {
+       val df = sql(
+         """
+@@ -1729,6 +1743,8 @@ abstract class DynamicPartitionPruningV1Suite extends DynamicPartitionPruningDat
                case s: BatchScanExec =>
                  // we use f1 col for v2 tables due to schema pruning
                  s.output.exists(_.exists(_.argString(maxFields = 100).contains("f1")))
@@ -966,13 +1046,15 @@ index 4b3d3a4b805..56e1e0e6f16 100644
 
    setupTestData()
 diff --git a/sql/core/src/test/scala/org/apache/spark/sql/execution/RemoveRedundantProjectsSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/execution/RemoveRedundantProjectsSuite.scala
-index 9e9d717db3b..91a4f9a38d5 100644
+index 9e9d717db3b..c1a7caf56e0 100644
 --- a/sql/core/src/test/scala/org/apache/spark/sql/execution/RemoveRedundantProjectsSuite.scala
 +++ b/sql/core/src/test/scala/org/apache/spark/sql/execution/RemoveRedundantProjectsSuite.scala
-@@ -18,6 +18,7 @@
+@@ -17,7 +17,8 @@
+
  package org.apache.spark.sql.execution
 
- import org.apache.spark.sql.{DataFrame, QueryTest, Row}
+-import org.apache.spark.sql.{DataFrame, QueryTest, Row}
++import org.apache.spark.sql.{DataFrame, IgnoreComet, QueryTest, Row}
 +import org.apache.spark.sql.comet.CometProjectExec
  import org.apache.spark.sql.connector.SimpleWritableDataSource
  import org.apache.spark.sql.execution.adaptive.{AdaptiveSparkPlanHelper, DisableAdaptiveExecutionSuite, EnableAdaptiveExecutionSuite}
@@ -989,6 +1071,16 @@ index 9e9d717db3b..91a4f9a38d5 100644
        assert(actual == expected)
      }
    }
+@@ -112,7 +116,8 @@ abstract class RemoveRedundantProjectsSuiteBase
+     assertProjectExec(query, 1, 3)
+   }
+
+-  test("join with ordering requirement") {
++  test("join with ordering requirement",
++    IgnoreComet("TODO: Support SubqueryBroadcastExec in Comet: #242")) {
+     val query = "select * from (select key, a, c, b from testView) as t1 join " +
+       "(select key, a, b, c from testView) as t2 on t1.key = t2.key where t2.a > 50"
+     assertProjectExec(query, 2, 2)
 diff --git a/sql/core/src/test/scala/org/apache/spark/sql/execution/RemoveRedundantSortsSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/execution/RemoveRedundantSortsSuite.scala
 index 30ce940b032..0d3f6c6c934 100644
 --- a/sql/core/src/test/scala/org/apache/spark/sql/execution/RemoveRedundantSortsSuite.scala