Fix bug with tree splicing.

marmbrus · Aug 8, 2014 · 4c4dc19 · 4c4dc19
1 parent ebb267e
commit 4c4dc19
Show file tree

Hide file tree

Showing 2 changed files with 28 additions and 2 deletions.
diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveStrategies.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveStrategies.scala
@@ -45,6 +45,9 @@ private[hive] trait HiveStrategies {
    *
    * TODO: Much of this logic is duplicated in HiveTableScan.  Ideally we would do some refactoring
    * but since this is after the code freeze for 1.1 all logic is here to minimize disruption.
+   *
+   * Other issues:
+   *  - Much of this logic assumes case insensitive resolution.
    */
   @Experimental
   object ParquetConversion extends Strategy {
@@ -60,8 +63,14 @@ private[hive] trait HiveStrategies {
           })
     }
 
-    implicit class PhysicalPlanHacks(s: SparkPlan) {
-      def fakeOutput(newOutput: Seq[Attribute]) = OutputFaker(newOutput, s)
+    implicit class PhysicalPlanHacks(originalPlan: SparkPlan) {
+      def fakeOutput(newOutput: Seq[Attribute]) =
+        OutputFaker(
+          originalPlan.output.map(a =>
+            newOutput.find(a.name.toLowerCase == _.name.toLowerCase)
+              .getOrElse(
+                sys.error(s"Can't find attribute $a to fake in set ${newOutput.mkString(",")}"))),
+          originalPlan)
     }
 
     def apply(plan: LogicalPlan): Seq[SparkPlan] = plan match {

diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/parquet/ParquetMetastoreSuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/parquet/ParquetMetastoreSuite.scala
@@ -103,6 +103,23 @@ class ParquetMetastoreSuite extends QueryTest with BeforeAndAfterAll {
     )
   }
 
+  test("project partitioning and non-partitioning columns") {
+    checkAnswer(
+      sql("SELECT stringField, p, count(intField) " +
+        "FROM partitioned_parquet GROUP BY p, stringField"),
+      ("part-1", 1, 10) ::
+      ("part-2", 2, 10) ::
+      ("part-3", 3, 10) ::
+      ("part-4", 4, 10) ::
+      ("part-5", 5, 10) ::
+      ("part-6", 6, 10) ::
+      ("part-7", 7, 10) ::
+      ("part-8", 8, 10) ::
+      ("part-9", 9, 10) ::
+      ("part-10", 10, 10) :: Nil
+    )
+  }
+
   test("simple count") {
     checkAnswer(
       sql("SELECT COUNT(*) FROM partitioned_parquet"),