[SPARK-5049][SQL] Fix ordering of partition columns in ParquetTableScan

Followup to #3870.  Props to rahulaggarwalguavus for identifying the issue.

Author: Michael Armbrust <michael@databricks.com>

Closes #3990 from marmbrus/SPARK-5049 and squashes the following commits:

dd03e4e [Michael Armbrust] Fill in the partition values of parquet scans instead of using JoinedRow

2 files changed, 29 insertions, 18 deletions

diff --git a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetRelation.scala b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetRelation.scala
index b237a07c72..2835dc3408 100644
--- a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetRelation.scala
+++ b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetRelation.scala
@@ -28,7 +28,7 @@ import parquet.schema.MessageType
 
 import org.apache.spark.sql.SQLContext
 import org.apache.spark.sql.catalyst.analysis.{MultiInstanceRelation, UnresolvedException}
-import org.apache.spark.sql.catalyst.expressions.Attribute
+import org.apache.spark.sql.catalyst.expressions.{AttributeMap, Attribute}
 import org.apache.spark.sql.catalyst.plans.logical.{LeafNode, LogicalPlan, Statistics}
 
 /**
@@ -67,6 +67,8 @@ private[sql] case class ParquetRelation(
       conf,
       sqlContext.isParquetBinaryAsString)
 
+  lazy val attributeMap = AttributeMap(output.map(o => o -> o))
+
   override def newInstance() = ParquetRelation(path, conf, sqlContext).asInstanceOf[this.type]
 
   // Equals must also take into account the output attributes so that we can distinguish between
diff --git a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTableOperations.scala b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTableOperations.scala
index 96bace1769..f5487740d3 100644
--- a/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTableOperations.scala
+++ b/sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTableOperations.scala
@@ -64,18 +64,17 @@ case class ParquetTableScan(
   // The resolution of Parquet attributes is case sensitive, so we resolve the original attributes
   // by exprId. note: output cannot be transient, see
   // https://issues.apache.org/jira/browse/SPARK-1367
-  val normalOutput =
-    attributes
-      .filterNot(a => relation.partitioningAttributes.map(_.exprId).contains(a.exprId))
-      .flatMap(a => relation.output.find(o => o.exprId == a.exprId))
+  val output = attributes.map(relation.attributeMap)
 
-  val partOutput =
-    attributes.flatMap(a => relation.partitioningAttributes.find(o => o.exprId == a.exprId))
+  // A mapping of ordinals partitionRow -> finalOutput.
+  val requestedPartitionOrdinals = {
+    val partitionAttributeOrdinals = AttributeMap(relation.partitioningAttributes.zipWithIndex)
 
-  def output = partOutput ++ normalOutput
-
-  assert(normalOutput.size + partOutput.size == attributes.size,
-    s"$normalOutput + $partOutput != $attributes, ${relation.output}")
+    attributes.zipWithIndex.flatMap {
+      case (attribute, finalOrdinal) =>
+        partitionAttributeOrdinals.get(attribute).map(_ -> finalOrdinal)
+    }
+  }.toArray
 
   override def execute(): RDD[Row] = {
     import parquet.filter2.compat.FilterCompat.FilterPredicateCompat
@@ -97,7 +96,7 @@ case class ParquetTableScan(
     // Store both requested and original schema in `Configuration`
     conf.set(
       RowReadSupport.SPARK_ROW_REQUESTED_SCHEMA,
-      ParquetTypesConverter.convertToString(normalOutput))
+      ParquetTypesConverter.convertToString(output))
     conf.set(
       RowWriteSupport.SPARK_ROW_SCHEMA,
       ParquetTypesConverter.convertToString(relation.output))
@@ -125,7 +124,7 @@ case class ParquetTableScan(
         classOf[Row],
         conf)
 
-    if (partOutput.nonEmpty) {
+    if (requestedPartitionOrdinals.nonEmpty) {
       baseRDD.mapPartitionsWithInputSplit { case (split, iter) =>
         val partValue = "([^=]+)=([^=]+)".r
         val partValues =
@@ -138,15 +137,25 @@ case class ParquetTableScan(
               case _ => None
             }.toMap
 
+        // Convert the partitioning attributes into the correct types
         val partitionRowValues =
-          partOutput.map(a => Cast(Literal(partValues(a.name)), a.dataType).eval(EmptyRow))
+          relation.partitioningAttributes
+            .map(a => Cast(Literal(partValues(a.name)), a.dataType).eval(EmptyRow))
 
         new Iterator[Row] {
-          private[this] val joinedRow = new JoinedRow5(Row(partitionRowValues:_*), null)
-
           def hasNext = iter.hasNext
-
-          def next() = joinedRow.withRight(iter.next()._2)
+          def next() = {
+            val row = iter.next()._2.asInstanceOf[SpecificMutableRow]
+
+            // Parquet will leave partitioning columns empty, so we fill them in here.
+            var i = 0
+            while (i < requestedPartitionOrdinals.size) {
+              row(requestedPartitionOrdinals(i)._2) =
+                partitionRowValues(requestedPartitionOrdinals(i)._1)
+              i += 1
+            }
+            row
+          }
         }
       }
     } else {