Github user ueshin commented on a diff in the pull request:
https://github.com/apache/spark/pull/21061#discussion_r200541138
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/collectionOperations.scala
---
@@ -3261,3 +3261,339 @@ case class ArrayDistinct(child: Expression)
override def prettyName: String = "array_distinct"
}
+
+object ArraySetLike {
+ def throwUnionLengthOverflowException(length: Int): Unit = {
+ throw new RuntimeException(s"Unsuccessful try to union arrays with
$length " +
+ s"elements due to exceeding the array size limit " +
+ s"${ByteArrayMethods.MAX_ROUNDED_ARRAY_LENGTH}.")
+ }
+}
+
+
+abstract class ArraySetLike extends BinaryArrayExpressionWithImplicitCast {
+ override def dataType: DataType = left.dataType
+
+ override def checkInputDataTypes(): TypeCheckResult = {
+ val typeCheckResult = super.checkInputDataTypes()
+ if (typeCheckResult.isSuccess) {
+
TypeUtils.checkForOrderingExpr(dataType.asInstanceOf[ArrayType].elementType,
+ s"function $prettyName")
+ } else {
+ typeCheckResult
+ }
+ }
+
+ @transient protected lazy val ordering: Ordering[Any] =
+ TypeUtils.getInterpretedOrdering(elementType)
+
+ @transient protected lazy val elementTypeSupportEquals = elementType
match {
+ case BinaryType => false
+ case _: AtomicType => true
+ case _ => false
+ }
+}
+
+/**
+ * Returns an array of the elements in the union of x and y, without
duplicates
+ */
+@ExpressionDescription(
+ usage = """
+ _FUNC_(array1, array2) - Returns an array of the elements in the union
of array1 and array2,
+ without duplicates.
+ """,
+ examples = """
+ Examples:
+ > SELECT _FUNC_(array(1, 2, 3), array(1, 3, 5));
+ array(1, 2, 3, 5)
+ """,
+ since = "2.4.0")
+case class ArrayUnion(left: Expression, right: Expression) extends
ArraySetLike {
+ var hsInt: OpenHashSet[Int] = _
+ var hsLong: OpenHashSet[Long] = _
+
+ def assignInt(array: ArrayData, idx: Int, resultArray: ArrayData, pos:
Int): Boolean = {
+ val elem = array.getInt(idx)
+ if (!hsInt.contains(elem)) {
+ resultArray.setInt(pos, elem)
+ hsInt.add(elem)
+ true
+ } else {
+ false
+ }
+ }
+
+ def assignLong(array: ArrayData, idx: Int, resultArray: ArrayData, pos:
Int): Boolean = {
+ val elem = array.getLong(idx)
+ if (!hsLong.contains(elem)) {
+ resultArray.setLong(pos, elem)
+ hsLong.add(elem)
+ true
+ } else {
+ false
+ }
+ }
+
+ def evalIntLongPrimitiveType(
+ array1: ArrayData,
+ array2: ArrayData,
+ size: Int,
+ resultArray: ArrayData,
+ isLongType: Boolean): ArrayData = {
+ // store elements into resultArray
+ var foundNullElement = false
+ var pos = 0
+ Seq(array1, array2).foreach(array => {
+ var i = 0
+ while (i < array.numElements()) {
+ if (array.isNullAt(i)) {
+ if (!foundNullElement) {
+ resultArray.setNullAt(pos)
+ pos += 1
+ foundNullElement = true
+ }
+ } else {
+ val assigned = if (!isLongType) {
+ assignInt(array, i, resultArray, pos)
+ } else {
+ assignLong(array, i, resultArray, pos)
+ }
+ if (assigned) {
+ pos += 1
+ }
+ }
+ i += 1
+ }
+ })
+ resultArray
+ }
+
+ override def nullSafeEval(input1: Any, input2: Any): Any = {
+ val array1 = input1.asInstanceOf[ArrayData]
+ val array2 = input2.asInstanceOf[ArrayData]
+
+ if (elementTypeSupportEquals) {
+ elementType match {
+ case IntegerType =>
+ // avoid boxing of primitive int array elements
+ // calculate result array size
+ val hsSize = new OpenHashSet[Int]
+ var nullElementSize = 0
+ Seq(array1, array2).foreach(array => {
+ var i = 0
+ while (i < array.numElements()) {
+ if (hsSize.size + nullElementSize >
ByteArrayMethods.MAX_ROUNDED_ARRAY_LENGTH) {
+ ArraySetLike.throwUnionLengthOverflowException(hsSize.size)
+ }
+ if (array.isNullAt(i)) {
+ if (nullElementSize == 0) {
+ nullElementSize = 1
+ }
+ } else {
+ hsSize.add(array.getInt(i))
+ }
+ i += 1
+ }
+ })
+ val elements = hsSize.size + nullElementSize
+ hsInt = new OpenHashSet[Int]
+ val resultArray = if (UnsafeArrayData.useGenericArrayData(
+ IntegerType.defaultSize, elements)) {
+ new GenericArrayData(new Array[Any](elements))
+ } else {
+ UnsafeArrayData.forPrimitiveArray(
+ Platform.INT_ARRAY_OFFSET, elements,
IntegerType.defaultSize);
+ }
+ evalIntLongPrimitiveType(array1, array2, elements, resultArray,
false)
+ case LongType =>
+ // avoid boxing of primitive long array elements
+ // calculate result array size
+ val hsSize = new OpenHashSet[Long]
+ var nullElementSize = 0
+ Seq(array1, array2).foreach(array => {
+ var i = 0
+ while (i < array.numElements()) {
+ if (hsSize.size + nullElementSize>
ByteArrayMethods.MAX_ROUNDED_ARRAY_LENGTH) {
+ ArraySetLike.throwUnionLengthOverflowException(hsSize.size)
+ }
+ if (array.isNullAt(i)) {
+ if (nullElementSize == 0) {
+ nullElementSize = 1
+ }
+ } else {
+ hsSize.add(array.getLong(i))
+ }
+ i += 1
+ }
+ })
+ val elements = hsSize.size + nullElementSize
+ hsLong = new OpenHashSet[Long]
+ val resultArray = if (UnsafeArrayData.useGenericArrayData(
+ LongType.defaultSize, elements)) {
+ new GenericArrayData(new Array[Any](elements))
+ } else {
+ UnsafeArrayData.forPrimitiveArray(
+ Platform.LONG_ARRAY_OFFSET, elements, LongType.defaultSize);
+ }
+ evalIntLongPrimitiveType(array1, array2, elements, resultArray,
true)
+ case _ =>
+ val arrayBuffer = new scala.collection.mutable.ArrayBuffer[Any]
+ val hs = new OpenHashSet[Any]
+ var foundNullElement = false
+ Seq(array1, array2).foreach(array => {
--- End diff --
nit: `foreach { array =>`?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
