[SPARK-7289] handle project -> limit -> sort efficiently

make the `TakeOrdered` strategy and operator more general, such that it can optionally handle a projection when necessary

Author: Wenchen Fan <[email protected]>

Closes apache#6780 from cloud-fan/limit and squashes the following commits:

34aa07b [Wenchen Fan] revert
07d5456 [Wenchen Fan] clean closure
20821ec [Wenchen Fan] fix
3676a82 [Wenchen Fan] address comments
b558549 [Wenchen Fan] address comments
214842b [Wenchen Fan] fix style
2d8be83 [Wenchen Fan] add LimitPushDown
948f740 [Wenchen Fan] fix existing

Author: cloud-fan

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala

@@ -39,19 +39,22 @@ object DefaultOptimizer extends Optimizer {
     Batch("Distinct", FixedPoint(100),
       ReplaceDistinctWithAggregate) ::
     Batch("Operator Optimizations", FixedPoint(100),
-      UnionPushdown,
-      CombineFilters,
+      // Operator push down
+      UnionPushDown,
+      PushPredicateThroughJoin,
       PushPredicateThroughProject,
       PushPredicateThroughGenerate,
       ColumnPruning,
+      // Operator combine
       ProjectCollapsing,
+      CombineFilters,
       CombineLimits,
+      // Constant folding
       NullPropagation,
       OptimizeIn,
       ConstantFolding,
       LikeSimplification,
       BooleanSimplification,
-      PushPredicateThroughJoin,
       RemovePositive,
       SimplifyFilters,
       SimplifyCasts,
@@ -63,25 +66,25 @@ object DefaultOptimizer extends Optimizer {
 }
 
 /**
-  *  Pushes operations to either side of a Union.
-  */
-object UnionPushdown extends Rule[LogicalPlan] {
+ * Pushes operations to either side of a Union.
+ */
+object UnionPushDown extends Rule[LogicalPlan] {
 
   /**
-    *  Maps Attributes from the left side to the corresponding Attribute on the right side.
-    */
-  def buildRewrites(union: Union): AttributeMap[Attribute] = {
+   * Maps Attributes from the left side to the corresponding Attribute on the right side.
+   */
+  private def buildRewrites(union: Union): AttributeMap[Attribute] = {
     assert(union.left.output.size == union.right.output.size)
 
     AttributeMap(union.left.output.zip(union.right.output))
   }
 
   /**
-    *  Rewrites an expression so that it can be pushed to the right side of a Union operator.
-    *  This method relies on the fact that the output attributes of a union are always equal
-    *  to the left child's output.
-    */
-  def pushToRight[A <: Expression](e: A, rewrites: AttributeMap[Attribute]): A = {
+   * Rewrites an expression so that it can be pushed to the right side of a Union operator.
+   * This method relies on the fact that the output attributes of a union are always equal
+   * to the left child's output.
+   */
+  private def pushToRight[A <: Expression](e: A, rewrites: AttributeMap[Attribute]) = {
     val result = e transform {
       case a: Attribute => rewrites(a)
     }
@@ -108,7 +111,6 @@ object UnionPushdown extends Rule[LogicalPlan] {
   }
 }
 
-
 /**
  * Attempts to eliminate the reading of unneeded columns from the query plan using the following
  * transformations:
@@ -117,7 +119,6 @@ object UnionPushdown extends Rule[LogicalPlan] {
  *   - Aggregate
  *   - Project <- Join
  *   - LeftSemiJoin
- *  - Performing alias substitution.
  */
 object ColumnPruning extends Rule[LogicalPlan] {
   def apply(plan: LogicalPlan): LogicalPlan = plan transform {
@@ -159,10 +160,11 @@ object ColumnPruning extends Rule[LogicalPlan] {
 
       Join(left, prunedChild(right, allReferences), LeftSemi, condition)
 
+    // Push down project through limit, so that we may have chance to push it further.
     case Project(projectList, Limit(exp, child)) =>
       Limit(exp, Project(projectList, child))
 
-    // push down project if possible when the child is sort
+    // Push down project if possible when the child is sort
     case p @ Project(projectList, s @ Sort(_, _, grandChild))
       if s.references.subsetOf(p.outputSet) =>
       s.copy(child = Project(projectList, grandChild))
@@ -181,8 +183,8 @@ object ColumnPruning extends Rule[LogicalPlan] {
 }
 
 /**
- * Combines two adjacent [[Project]] operators into one, merging the
- * expressions into one single expression.
+ * Combines two adjacent [[Project]] operators into one and perform alias substitution,
+ * merging the expressions into one single expression.
  */
 object ProjectCollapsing extends Rule[LogicalPlan] {
 
@@ -222,10 +224,10 @@ object ProjectCollapsing extends Rule[LogicalPlan] {
 object LikeSimplification extends Rule[LogicalPlan] {
   // if guards below protect from escapes on trailing %.
   // Cases like "something\%" are not optimized, but this does not affect correctness.
-  val startsWith = "([^_%]+)%".r
-  val endsWith = "%([^_%]+)".r
-  val contains = "%([^_%]+)%".r
-  val equalTo = "([^_%]*)".r
+  private val startsWith = "([^_%]+)%".r
+  private val endsWith = "%([^_%]+)".r
+  private val contains = "%([^_%]+)%".r
+  private val equalTo = "([^_%]*)".r
 
   def apply(plan: LogicalPlan): LogicalPlan = plan transformAllExpressions {
     case Like(l, Literal(utf, StringType)) =>
@@ -497,7 +499,7 @@ object PushPredicateThroughProject extends Rule[LogicalPlan] {
         grandChild))
   }
 
-  def replaceAlias(condition: Expression, sourceAliases: Map[Attribute, Expression]): Expression = {
+  private def replaceAlias(condition: Expression, sourceAliases: Map[Attribute, Expression]) = {
     condition transform {
       case a: AttributeReference => sourceAliases.getOrElse(a, a)
     }
@@ -682,7 +684,7 @@ object DecimalAggregates extends Rule[LogicalPlan] {
   import Decimal.MAX_LONG_DIGITS
 
   /** Maximum number of decimal digits representable precisely in a Double */
-  val MAX_DOUBLE_DIGITS = 15
+  private val MAX_DOUBLE_DIGITS = 15
 
   def apply(plan: LogicalPlan): LogicalPlan = plan transformAllExpressions {
     case Sum(e @ DecimalType.Expression(prec, scale)) if prec + 10 <= MAX_LONG_DIGITS =>

sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/UnionPushdownSuite.scala

@@ -24,13 +24,13 @@ import org.apache.spark.sql.catalyst.rules._
 import org.apache.spark.sql.catalyst.dsl.plans._
 import org.apache.spark.sql.catalyst.dsl.expressions._
 
-class UnionPushdownSuite extends PlanTest {
+class UnionPushDownSuite extends PlanTest {
   object Optimize extends RuleExecutor[LogicalPlan] {
     val batches =
       Batch("Subqueries", Once,
         EliminateSubQueries) ::
       Batch("Union Pushdown", Once,
-        UnionPushdown) :: Nil
+        UnionPushDown) :: Nil
   }
 
   val testRelation = LocalRelation('a.int, 'b.int, 'c.int)

sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala

@@ -858,7 +858,7 @@ class SQLContext(@transient val sparkContext: SparkContext)
       experimental.extraStrategies ++ (
       DataSourceStrategy ::
       DDLStrategy ::
-      TakeOrdered ::
+      TakeOrderedAndProject ::
       HashAggregation ::
       LeftSemiJoin ::
       HashJoin ::

sql/core/src/main/scala/org/apache/spark/sql/execution/SparkPlan.scala

@@ -169,7 +169,6 @@ abstract class SparkPlan extends QueryPlan[SparkPlan] with Logging with Serializ
     log.debug(
       s"Creating MutableProj: $expressions, inputSchema: $inputSchema, codegen:$codegenEnabled")
     if(codegenEnabled && expressions.forall(_.isThreadSafe)) {
-
       GenerateMutableProjection.generate(expressions, inputSchema)
     } else {
       () => new InterpretedMutableProjection(expressions, inputSchema)

sql/core/src/main/scala/org/apache/spark/sql/execution/SparkStrategies.scala

@@ -213,10 +213,14 @@ private[sql] abstract class SparkStrategies extends QueryPlanner[SparkPlan] {
   protected lazy val singleRowRdd =
     sparkContext.parallelize(Seq(new GenericRow(Array[Any]()): InternalRow), 1)
 
-  object TakeOrdered extends Strategy {
+  object TakeOrderedAndProject extends Strategy {
     def apply(plan: LogicalPlan): Seq[SparkPlan] = plan match {
       case logical.Limit(IntegerLiteral(limit), logical.Sort(order, true, child)) =>
-        execution.TakeOrdered(limit, order, planLater(child)) :: Nil
+        execution.TakeOrderedAndProject(limit, order, None, planLater(child)) :: Nil
+      case logical.Limit(
+             IntegerLiteral(limit),
+             logical.Project(projectList, logical.Sort(order, true, child))) =>
+        execution.TakeOrderedAndProject(limit, order, Some(projectList), planLater(child)) :: Nil
       case _ => Nil
     }
   }

sql/core/src/main/scala/org/apache/spark/sql/execution/basicOperators.scala

@@ -39,8 +39,8 @@ case class Project(projectList: Seq[NamedExpression], child: SparkPlan) extends
   @transient lazy val buildProjection = newMutableProjection(projectList, child.output)
 
   protected override def doExecute(): RDD[InternalRow] = child.execute().mapPartitions { iter =>
-    val resuableProjection = buildProjection()
-    iter.map(resuableProjection)
+    val reusableProjection = buildProjection()
+    iter.map(reusableProjection)
   }
 
   override def outputOrdering: Seq[SortOrder] = child.outputOrdering
@@ -147,21 +147,32 @@ case class Limit(limit: Int, child: SparkPlan)
 
 /**
  * :: DeveloperApi ::
- * Take the first limit elements as defined by the sortOrder. This is logically equivalent to
- * having a [[Limit]] operator after a [[Sort]] operator. This could have been named TopK, but
- * Spark's top operator does the opposite in ordering so we name it TakeOrdered to avoid confusion.
+ * Take the first limit elements as defined by the sortOrder, and do projection if needed.
+ * This is logically equivalent to having a [[Limit]] operator after a [[Sort]] operator,
+ * or having a [[Project]] operator between them.
+ * This could have been named TopK, but Spark's top operator does the opposite in ordering
+ * so we name it TakeOrdered to avoid confusion.
  */
 @DeveloperApi
-case class TakeOrdered(limit: Int, sortOrder: Seq[SortOrder], child: SparkPlan) extends UnaryNode {
+case class TakeOrderedAndProject(
+    limit: Int,
+    sortOrder: Seq[SortOrder],
+    projectList: Option[Seq[NamedExpression]],
+    child: SparkPlan) extends UnaryNode {
 
   override def output: Seq[Attribute] = child.output
 
   override def outputPartitioning: Partitioning = SinglePartition
 
   private val ord: RowOrdering = new RowOrdering(sortOrder, child.output)
 
-  private def collectData(): Array[InternalRow] =
-    child.execute().map(_.copy()).takeOrdered(limit)(ord)
+  // TODO: remove @transient after figure out how to clean closure at InsertIntoHiveTable.
+  @transient private val projection = projectList.map(new InterpretedProjection(_, child.output))
+
+  private def collectData(): Array[InternalRow] = {
+    val data = child.execute().map(_.copy()).takeOrdered(limit)(ord)
+    projection.map(data.map(_)).getOrElse(data)
+  }
 
   override def executeCollect(): Array[Row] = {
     val converter = CatalystTypeConverters.createToScalaConverter(schema)

sql/core/src/test/scala/org/apache/spark/sql/execution/PlannerSuite.scala

@@ -141,4 +141,10 @@ class PlannerSuite extends SparkFunSuite {
 
     setConf(SQLConf.AUTO_BROADCASTJOIN_THRESHOLD, origThreshold)
   }
+
+  test("efficient limit -> project -> sort") {
+    val query = testData.sort('key).select('value).limit(2).logicalPlan
+    val planned = planner.TakeOrderedAndProject(query)
+    assert(planned.head.isInstanceOf[execution.TakeOrderedAndProject])
+  }
 }

sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveContext.scala

@@ -442,7 +442,7 @@ class HiveContext(sc: SparkContext) extends SQLContext(sc) {
       HiveCommandStrategy(self),
       HiveDDLStrategy,
       DDLStrategy,
-      TakeOrdered,
+      TakeOrderedAndProject,
       ParquetOperations,
       InMemoryScans,
       ParquetConversion, // Must be before HiveTableScans