[SPARK-26576][SQL] Broadcast hint not applied to partitioned table
authorJohn Zhuge <jzhuge@apache.org>
Sun, 13 Jan 2019 23:36:40 +0000 (15:36 -0800)
committergatorsmile <gatorsmile@gmail.com>
Sun, 13 Jan 2019 23:36:40 +0000 (15:36 -0800)
## What changes were proposed in this pull request?

Make sure broadcast hint is applied to partitioned tables.

## How was this patch tested?

- A new unit test in PruneFileSourcePartitionsSuite
- Unit test suites touched by SPARK-14581: JoinOptimizationSuite, FilterPushdownSuite, ColumnPruningSuite, and PruneFiltersSuite

Closes #23507 from jzhuge/SPARK-26576.

Closes #23530 from jzhuge/SPARK-26576-master.

Authored-by: John Zhuge <jzhuge@apache.org>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/PruneFileSourcePartitionsSuite.scala

index 9438418..6b2d0c6 100644 (file)
@@ -17,6 +17,8 @@
 
 package org.apache.spark.sql.hive.execution
 
+import org.scalatest.Matchers._
+
 import org.apache.spark.sql.QueryTest
 import org.apache.spark.sql.catalyst.TableIdentifier
 import org.apache.spark.sql.catalyst.dsl.expressions._
@@ -25,7 +27,10 @@ import org.apache.spark.sql.catalyst.plans.logical.{Filter, LogicalPlan, Project
 import org.apache.spark.sql.catalyst.rules.RuleExecutor
 import org.apache.spark.sql.execution.datasources.{CatalogFileIndex, HadoopFsRelation, LogicalRelation, PruneFileSourcePartitions}
 import org.apache.spark.sql.execution.datasources.parquet.ParquetFileFormat
+import org.apache.spark.sql.execution.joins.BroadcastHashJoinExec
+import org.apache.spark.sql.functions.broadcast
 import org.apache.spark.sql.hive.test.TestHiveSingleton
+import org.apache.spark.sql.internal.SQLConf
 import org.apache.spark.sql.test.SQLTestUtils
 import org.apache.spark.sql.types.StructType
 
@@ -91,4 +96,15 @@ class PruneFileSourcePartitionsSuite extends QueryTest with SQLTestUtils with Te
       assert(size2 < tableStats.get.sizeInBytes)
     }
   }
+
+  test("SPARK-26576 Broadcast hint not applied to partitioned table") {
+    withTable("tbl") {
+      withSQLConf(SQLConf.AUTO_BROADCASTJOIN_THRESHOLD.key -> "-1") {
+        spark.range(10).selectExpr("id", "id % 3 as p").write.partitionBy("p").saveAsTable("tbl")
+        val df = spark.table("tbl")
+        val qe = df.join(broadcast(df), "p").queryExecution
+        qe.sparkPlan.collect { case j: BroadcastHashJoinExec => j } should have size 1
+      }
+    }
+  }
 }