Github user srinathshankar commented on a diff in the pull request:
https://github.com/apache/spark/pull/15637#discussion_r85993076
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/MapAggregate.scala
---
@@ -0,0 +1,332 @@
+/*
+ * 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 java.nio.ByteBuffer
+
+import scala.collection.immutable.TreeMap
+import scala.collection.mutable
+
+import com.google.common.primitives.{Doubles, Ints, Longs}
+
+import org.apache.spark.sql.catalyst.InternalRow
+import org.apache.spark.sql.catalyst.analysis.TypeCheckResult
+import
org.apache.spark.sql.catalyst.analysis.TypeCheckResult.{TypeCheckFailure,
TypeCheckSuccess}
+import org.apache.spark.sql.catalyst.expressions.{Expression,
ExpressionDescription}
+import org.apache.spark.sql.catalyst.util.ArrayBasedMapData
+import org.apache.spark.sql.types.{DataType, _}
+import org.apache.spark.unsafe.types.UTF8String
+
+/**
+ * The MapAggregate function for a column returns:
+ * 1. null if no non-null value exists.
+ * 2. (distinct non-null value, frequency) pairs of equi-width histogram
when the number of
+ * distinct non-null values is less than or equal to the specified maximum
number of bins.
+ * 3. an empty map otherwise.
+ *
+ * @param child child expression that can produce column value with
`child.eval(inputRow)`
+ * @param numBinsExpression The maximum number of bins.
+ */
+@ExpressionDescription(
+ usage =
+ """
+ _FUNC_(col, numBins) - Returns 1. null if no non-null value exists.
+ 2. (distinct non-null value, frequency) pairs of equi-width
histogram when the number of
+ distinct non-null values is less than or equal to the specified
maximum number of bins.
+ 3. an empty map otherwise.
+ """)
+case class MapAggregate(
+ child: Expression,
+ numBinsExpression: Expression,
+ override val mutableAggBufferOffset: Int,
+ override val inputAggBufferOffset: Int) extends
TypedImperativeAggregate[MapDigest] {
+
+ def this(child: Expression, numBinsExpression: Expression) = {
+ this(child, numBinsExpression, 0, 0)
+ }
+
+ // Mark as lazy so that numBinsExpression is not evaluated during tree
transformation.
+ private lazy val numBins: Int =
numBinsExpression.eval().asInstanceOf[Int]
+
+ override def inputTypes: Seq[AbstractDataType] = {
+ Seq(TypeCollection(NumericType, TimestampType, DateType, StringType),
IntegerType)
+ }
+
+ override def checkInputDataTypes(): TypeCheckResult = {
+ val defaultCheck = super.checkInputDataTypes()
+ if (defaultCheck.isFailure) {
+ defaultCheck
+ } else if (!numBinsExpression.foldable) {
+ TypeCheckFailure("The maximum number of bins provided must be a
constant literal")
+ } else if (numBins < 2) {
+ TypeCheckFailure(
+ "The maximum number of bins provided must be a positive integer
literal >= 2 " +
+ s"(current value = $numBins)")
+ } else {
+ TypeCheckSuccess
+ }
+ }
+
+ override def update(buffer: MapDigest, input: InternalRow): Unit = {
+ if (buffer.isInvalid) {
+ return
+ }
+ val evaluated = child.eval(input)
+ if (evaluated != null) {
+ buffer.update(child.dataType, evaluated, numBins)
+ }
+ }
+
+ override def merge(buffer: MapDigest, other: MapDigest): Unit = {
+ if (buffer.isInvalid) return
+ if (other.isInvalid) {
+ buffer.isInvalid = true
+ buffer.clear()
+ return
+ }
+ buffer.merge(other, numBins)
+ }
+
+ override def eval(buffer: MapDigest): Any = {
+ if (buffer.isInvalid) {
+ // return empty map
+ ArrayBasedMapData(Map.empty)
+ } else {
+ // sort the result to make it more readable
+ val sorted = buffer match {
+ case stringDigest: StringMapDigest => TreeMap[UTF8String,
Long](stringDigest.bins.toSeq: _*)
+ case numericDigest: NumericMapDigest => TreeMap[Double,
Long](numericDigest.bins.toSeq: _*)
+ }
+ if (sorted.isEmpty) {
+ // don't have non-null values
+ null
+ } else {
+ ArrayBasedMapData(sorted.keys.toArray, sorted.values.toArray)
+ }
+ }
+ }
+
+ override def serialize(buffer: MapDigest): Array[Byte] = {
+ buffer match {
+ case stringDigest: StringMapDigest =>
StringMapDigest.serialize(stringDigest)
+ case numericDigest: NumericMapDigest =>
NumericMapDigest.serialize(numericDigest)
+ }
+ }
+
+ override def deserialize(bytes: Array[Byte]): MapDigest = {
+ child.dataType match {
+ case StringType => StringMapDigest.deserialize(bytes)
+ case _ => NumericMapDigest.deserialize(bytes)
+ }
+ }
+
+ override def createAggregationBuffer(): MapDigest = {
+ child.dataType match {
+ case StringType => StringMapDigest()
+ case _ => NumericMapDigest()
+ }
+ }
+
+ override def withNewMutableAggBufferOffset(newMutableAggBufferOffset:
Int): MapAggregate = {
+ copy(mutableAggBufferOffset = newMutableAggBufferOffset)
+ }
+
+ override def withNewInputAggBufferOffset(newInputAggBufferOffset: Int):
MapAggregate = {
+ copy(inputAggBufferOffset = newInputAggBufferOffset)
+ }
+
+ override def nullable: Boolean = true
+
+ override def dataType: DataType = {
+ child.dataType match {
+ case StringType => MapType(StringType, LongType)
+ case _ => MapType(DoubleType, LongType)
+ }
+ }
+
+ override def children: Seq[Expression] = Seq(child, numBinsExpression)
+
+ override def prettyName: String = "map_aggregate"
+}
+
+trait MapDigest {
+ // Mark this MapDigest invalid when the size of the hashmap (ndv of the
column) exceeds numBins
+ var isInvalid: Boolean = false
+
+ def update(dataType: DataType, value: Any, numBins: Int): Unit
+ def merge(otherDigest: MapDigest, numBins: Int): Unit
+ def clear(): Unit
+
+ // Update baseMap and clear it when its size exceeds numBins.
+ def updateMap[T](baseMap: mutable.HashMap[T, Long], value: T, numBins:
Int): Unit = {
+ mergeMap(baseMap, mutable.HashMap(value -> 1L), numBins)
+ }
+
+ // Merge two maps and clear baseMap when its size exceeds numBins.
+ def mergeMap[T](
+ baseMap: mutable.HashMap[T, Long],
+ otherMap: mutable.HashMap[T, Long],
+ numBins: Int): Unit = {
+ otherMap.foreach { case (key, value) =>
+ if (baseMap.contains(key)) {
+ baseMap.update(key, baseMap(key) + value)
+ } else {
+ if (baseMap.size >= numBins) {
+ isInvalid = true
+ baseMap.clear()
+ return
+ } else {
+ baseMap.put(key, value)
+ }
+ }
+ }
+ }
+}
+
+// Digest class for column of string type.
+case class StringMapDigest(
+ bins: mutable.HashMap[UTF8String, Long] = mutable.HashMap.empty)
extends MapDigest {
+ def this(bins: mutable.HashMap[UTF8String, Long], isInvalid: Boolean) = {
+ this(bins)
+ this.isInvalid = isInvalid
+ }
+
+ override def update(dataType: DataType, value: Any, numBins: Int): Unit
= {
+ updateMap(bins, value.asInstanceOf[UTF8String], numBins)
+ }
+
+ override def merge(otherDigest: MapDigest, numBins: Int): Unit = {
+ mergeMap(baseMap = bins, otherMap =
otherDigest.asInstanceOf[StringMapDigest].bins, numBins)
+ }
+
+ override def clear(): Unit = bins.clear()
+}
+
+object StringMapDigest {
+
+ private final def length(obj: StringMapDigest): Int = {
+ // isInvalid, size of bins
+ var len: Int = Ints.BYTES + Ints.BYTES
+ obj.bins.foreach { case (key, value) =>
+ // length of key, key, value
+ len += Ints.BYTES + key.getBytes.length + Longs.BYTES
+ }
+ len
+ }
+
+ final def serialize(obj: StringMapDigest): Array[Byte] = {
+ val buffer = ByteBuffer.wrap(new Array(length(obj)))
+ buffer.putInt(if (obj.isInvalid) 0 else 1)
+ buffer.putInt(obj.bins.size)
+ obj.bins.foreach { case (key, value) =>
+ val bytes = key.getBytes
+ buffer.putInt(bytes.length)
+ buffer.put(bytes)
+ buffer.putLong(value)
+ }
+ buffer.array()
+ }
+
+ final def deserialize(bytes: Array[Byte]): StringMapDigest = {
+ val buffer = ByteBuffer.wrap(bytes)
+ val isInvalid = if (buffer.getInt == 0) true else false
+ val size = buffer.getInt
+ val bins = new mutable.HashMap[UTF8String, Long]
+ var i = 0
+ while (i < size) {
+ val keyLength = buffer.getInt
+ var j = 0
+ val keyBytes = new Array[Byte](keyLength)
+ while (j < keyLength) {
+ keyBytes(j) = buffer.get()
+ j += 1
+ }
+ val value = buffer.getLong
+ bins.put(UTF8String.fromBytes(keyBytes), value)
+ i += 1
+ }
+ new StringMapDigest(bins, isInvalid)
+ }
+}
+
+// Digest class for column of numeric type.
+case class NumericMapDigest(
+ bins: mutable.HashMap[Double, Long] = mutable.HashMap.empty) extends
MapDigest {
+ def this(bins: mutable.HashMap[Double, Long], isInvalid: Boolean) = {
+ this(bins)
+ this.isInvalid = isInvalid
+ }
+
+ override def update(dataType: DataType, value: Any, numBins: Int): Unit
= {
+ // Use Double to represent endpoints (in histograms) for simplicity
+ val doubleValue = dataType match {
--- End diff --
Couldn't you leave this type promotion to the engine ?
That is, for numeric types instead of calling
`mapaggregate(numericcol, bins)`
call
`mapaggregate(cast(numericcol as double), bins)`
That way you only have to deal with doubles in your aggregate.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
