Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
@@ -0,0 +1,174 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/

package org.apache.spark.sql.catalyst.expressions.aggregate

import org.apache.spark.sql.catalyst.InternalRow
import org.apache.spark.sql.catalyst.expressions._
import org.apache.spark.sql.catalyst.expressions.codegen.{GeneratedExpressionCode, CodeGenContext, CodegenFallback}
import org.apache.spark.sql.types.{AbstractDataType, DataType}
import org.apache.spark.util.collection.OpenHashSet

/**
* Fallback operator for distinct operators. This will be used when a user issues multiple
* different distinct expressions in a query.
*
* The operator uses the OpenHashSetUDT for de-duplicating values. It is, as a result, not possible
* to use UnsafeRow based aggregation.
*/
case class DistinctAggregateFallback(function: AggregateFunction2) extends DeclarativeAggregate {
override def inputTypes: Seq[AbstractDataType] = function.inputTypes
override def nullable: Boolean = function.nullable
override def dataType: DataType = function.dataType
override def children: Seq[Expression] = Seq(function)

private[this] val input = function.children match {
case child :: Nil => child
case children => CreateStruct(children) // TODO can we test this?
}
private[this] val items = AttributeReference("itemSet", new OpenHashSetUDT(input.dataType))()

override def aggBufferAttributes: Seq[AttributeReference] = Seq(items)
override val initialValues: Seq[Expression] = Seq(NewSet(input.dataType))
override val updateExpressions: Seq[Expression] = Seq(AddItemToSet(input, items))
override val mergeExpressions: Seq[Expression] = Seq(CombineSets(items.left, items.right))
override val evaluateExpression: Expression = function match {
case f: Count => CountSet(items)
case f: DeclarativeAggregate => ReduceSetUsingDeclarativeAggregate(items, f)
case f: ImperativeAggregate => ReduceSetUsingImperativeAggregate(items, f)
}
}

case class ReduceSetUsingImperativeAggregate(left: Expression, right: ImperativeAggregate)
extends BinaryExpression with CodegenFallback {

override def dataType: DataType = right.dataType

private[this] val single = right.children.size == 1

// TODO can we assume that the offsets are 0 when we haven't touched them yet?
private[this] val function = right
.withNewInputAggBufferOffset(0)
.withNewMutableAggBufferOffset(0)

@transient private[this] lazy val buffer =
new SpecificMutableRow(right.aggBufferAttributes.map(_.dataType))

@transient private[this] lazy val singleValueInput = new GenericMutableRow(1)

override def eval(input: InternalRow): Any = {
val result = left.eval(input).asInstanceOf[OpenHashSet[Any]]
if (result != null) {
right.initialize(buffer)
val iterator = result.iterator
if (single) {
while (iterator.hasNext) {
singleValueInput.update(0, iterator.next())
function.update(buffer, singleValueInput)
}
} else {
while (iterator.hasNext) {
function.update(buffer, iterator.next().asInstanceOf[InternalRow])
}
}
function.eval(buffer)
} else null
}
}

case class ReduceSetUsingDeclarativeAggregate(left: Expression, right: DeclarativeAggregate)
extends Expression with CodegenFallback {
override def children: Seq[Expression] = Seq(left)
override def nullable: Boolean = right.nullable
override def dataType: DataType = right.dataType

private[this] val single = right.children.size == 1

private[this] val inputOrdinal = right.children.size

@transient private[this] lazy val initial =
InterpretedMutableProjection(right.initialValues).target(buffer)

@transient private[this] lazy val update = {
val boundRefs = (right.aggBufferAttributes ++ right.children).zipWithIndex.map {
case (e, i) => (e, new BoundReference(i, e.dataType, e.nullable))
}.toMap
val boundExpressions = right.updateExpressions.map(_.transform(boundRefs))
new InterpretedMutableProjection(boundExpressions).target(buffer)
}

@transient private[this] lazy val evaluate =
BindReferences.bindReference(right.evaluateExpression, right.aggBufferAttributes)

@transient private[this] lazy val buffer = {
val singleType = if (single) {
Seq(right.children.head.dataType)
} else {
Seq.empty
}
new SpecificMutableRow(right.inputAggBufferAttributes.map(_.dataType) ++ singleType)
}

@transient private[this] lazy val joinRow = new JoinedRow

override def eval(input: InternalRow): Any = {
val result = left.eval(input).asInstanceOf[OpenHashSet[Any]]
if (result != null) {
initial(EmptyRow)
val iterator = result.iterator
if (single) {
while (iterator.hasNext) {
buffer.update(inputOrdinal, iterator.next())
update(buffer)
}
} else {
while (iterator.hasNext) {
joinRow(buffer, iterator.next().asInstanceOf[InternalRow])
update(joinRow)
}
}
evaluate.eval(buffer)
} else null
}
}

/** Operator that drops a row when it contains any nulls. */
case class DropAnyNull(child: Expression) extends UnaryExpression {
override def nullable: Boolean = true
override def dataType: DataType = child.dataType
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'll remove this in the next iteration...

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Removed....


protected override def nullSafeEval(input: Any): InternalRow = {
val row = input.asInstanceOf[InternalRow]
if (row.anyNull) {
null
} else {
row
}
}

override def genCode(ctx: CodeGenContext, ev: GeneratedExpressionCode): String = {
nullSafeCodeGen(ctx, ev, eval => {
s"""
if ($eval.anyNull) {
${ev.isNull} = true;
} else {
${ev.value} = $eval
}
"""
})
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -41,6 +41,16 @@ object Utils {

private def doConvert(plan: LogicalPlan): Option[Aggregate] = plan match {
case p: Aggregate if supportsGroupingKeySchema(p) =>

var distinctExpressionSet: Seq[Expression] = null
def addDistinct(children: Seq[Expression], function: AggregateFunction2) = {
if (distinctExpressionSet == null || distinctExpressionSet == children) {
distinctExpressionSet = children
(true, function)
} else {
(false, DistinctAggregateFallback(function))
}
}
val converted = p.transformExpressionsDown {
case expressions.Average(child) =>
aggregate.AggregateExpression2(
Expand All @@ -56,10 +66,20 @@ object Utils {

// We do not support multiple COUNT DISTINCT columns for now.
case expressions.CountDistinct(children) if children.length == 1 =>
val (isDistinct, function) = addDistinct(children, aggregate.Count(children.head))
aggregate.AggregateExpression2(
aggregateFunction = function,
mode = aggregate.Complete,
isDistinct = isDistinct)

// Always use the fallback distinct operator when we have to deal with multiple children.
case expressions.CountDistinct(children) =>
val function = DistinctAggregateFallback(
aggregate.Count(DropAnyNull(CreateStruct(children))))
aggregate.AggregateExpression2(
aggregateFunction = aggregate.Count(children.head),
aggregateFunction = function,
mode = aggregate.Complete,
isDistinct = true)
isDistinct = false)

case expressions.First(child, ignoreNulls) =>
aggregate.AggregateExpression2(
Expand Down Expand Up @@ -110,17 +130,19 @@ object Utils {
isDistinct = false)

case expressions.SumDistinct(child) =>
val (isDistinct, function) = addDistinct(Seq(child), aggregate.Sum(child))
aggregate.AggregateExpression2(
aggregateFunction = aggregate.Sum(child),
aggregateFunction = function,
mode = aggregate.Complete,
isDistinct = true)
isDistinct = isDistinct)

case expressions.ApproxCountDistinct(child, rsd) =>
aggregate.AggregateExpression2(
aggregateFunction = aggregate.HyperLogLogPlusPlus(child, rsd),
mode = aggregate.Complete,
isDistinct = false)
}

// Check if there is any expressions.AggregateExpression1 left.
// If so, we cannot convert this plan.
val hasAggregateExpression1 = converted.aggregateExpressions.exists { expr =>
Expand All @@ -131,21 +153,7 @@ object Utils {
}.isDefined
}

// Check if there are multiple distinct columns.
val aggregateExpressions = converted.aggregateExpressions.flatMap { expr =>
expr.collect {
case agg: AggregateExpression2 => agg
}
}.toSet.toSeq
val functionsWithDistinct = aggregateExpressions.filter(_.isDistinct)
val hasMultipleDistinctColumnSets =
if (functionsWithDistinct.map(_.aggregateFunction.children).distinct.length > 1) {
true
} else {
false
}

if (!hasAggregateExpression1 && !hasMultipleDistinctColumnSets) Some(converted) else None
if (!hasAggregateExpression1) Some(converted) else None

case other => None
}
Expand All @@ -170,7 +178,7 @@ object Utils {
}
}
val errorMessage =
s"${invalidFunctions} implemented based on the new Aggregate Function " +
s"$invalidFunctions implemented based on the new Aggregate Function " +
s"interface and it cannot be used with functions implemented based on " +
s"the old Aggregate Function interface."
throw new AnalysisException(errorMessage)
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,8 @@

package org.apache.spark.sql.catalyst.expressions.codegen

import org.apache.spark.util.collection.OpenHashSet

import scala.collection.mutable
import scala.collection.mutable.ArrayBuffer
import scala.language.existentials
Expand Down Expand Up @@ -127,6 +129,7 @@ class CodeGenContext {
case BinaryType => s"$input.getBinary($ordinal)"
case CalendarIntervalType => s"$input.getInterval($ordinal)"
case t: StructType => s"$input.getStruct($ordinal, ${t.size})"
case _: OpenHashSetUDT => s"($jt)$input.get($ordinal, null)"
case _: ArrayType => s"$input.getArray($ordinal)"
case _: MapType => s"$input.getMap($ordinal)"
case NullType => "null"
Expand Down Expand Up @@ -175,11 +178,12 @@ class CodeGenContext {
case BinaryType => "byte[]"
case StringType => "UTF8String"
case CalendarIntervalType => "CalendarInterval"
case dt: OpenHashSetUDT if dt.elementType == IntegerType => classOf[IntegerHashSet].getName
case dt: OpenHashSetUDT if dt.elementType == LongType => classOf[LongHashSet].getName
case dt: OpenHashSetUDT => s"${classOf[OpenHashSet[_]].getName}<${boxedType(dt.elementType)}>"
case _: StructType => "InternalRow"
case _: ArrayType => "ArrayData"
case _: MapType => "MapData"
case dt: OpenHashSetUDT if dt.elementType == IntegerType => classOf[IntegerHashSet].getName
case dt: OpenHashSetUDT if dt.elementType == LongType => classOf[LongHashSet].getName
case udt: UserDefinedType[_] => javaType(udt.sqlType)
case ObjectType(cls) if cls.isArray => s"${javaType(ObjectType(cls.getComponentType))}[]"
case ObjectType(cls) => cls.getName
Expand Down