Github user chenghao-intel commented on a diff in the pull request:
https://github.com/apache/spark/pull/6938#discussion_r33032295
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/math.scala
---
@@ -312,3 +315,90 @@ case class Logarithm(left: Expression, right:
Expression)
"""
}
}
+
+case class Round(child: Expression, scale: Expression) extends Expression {
+
+ def this(child: Expression) = {
+ this(child, Literal(0))
+ }
+
+ def children: Seq[Expression] = Seq(child, scale)
+
+ def nullable: Boolean = true
+
+ private lazy val scaleV = scale.asInstanceOf[Literal].value
+ private lazy val _scale = if (scaleV != null) scaleV.asInstanceOf[Int]
else 0
+
+ override lazy val dataType: DataType = {
+ child.dataType match {
+ case StringType | BinaryType => DoubleType
+ case DecimalType.Fixed(p, s) => DecimalType(p, _scale)
+ case t => t
+ }
+ }
+
+ override def checkInputDataTypes(): TypeCheckResult = {
+ child.dataType match {
+ case _: NumericType | NullType | BinaryType | StringType => //
satisfy requirement
+ case dt =>
+ return TypeCheckFailure(s"Only numeric, string or binary data
types" +
+ s" are allowed for ROUND function, got $dt")
+ }
+ scale match {
+ case Literal(value, LongType) =>
+ if (value.asInstanceOf[Long] < Int.MinValue ||
value.asInstanceOf[Long] > Int.MaxValue) {
+ return TypeCheckFailure("ROUND scale argument out of allowed
range")
+ }
+ case Literal(_, _: IntegralType) | Literal(_, NullType) => //
satisfy requirement
+ case child =>
+ if (child.find { case _: AttributeReference => true; case _ =>
false } != None) {
+ return TypeCheckFailure("Only Integral Literal or Null Literal "
+
+ s"are allowed for ROUND scale arguments, got
${child.dataType}")
+ }
+ }
+ TypeCheckSuccess
+ }
+
+ def eval(input: InternalRow): Any = {
+ val evalE = child.eval(input)
+
+ if (evalE == null || scaleV == null) return null
+
+ children(0).dataType match {
+ case decimalType: DecimalType =>
+ val decimal = evalE.asInstanceOf[Decimal]
+ if (decimal.changePrecision(decimal.precision, _scale)) decimal
else null
+ case ByteType =>
+ round(evalE.asInstanceOf[Byte], _scale)
+ case ShortType =>
+ round(evalE.asInstanceOf[Short], _scale)
+ case IntegerType =>
+ round(evalE.asInstanceOf[Int], _scale)
+ case LongType =>
+ round(evalE.asInstanceOf[Long], _scale)
+ case FloatType =>
+ round(evalE.asInstanceOf[Float], _scale)
+ case DoubleType =>
+ round(evalE.asInstanceOf[Double], _scale)
+ case StringType =>
+ round(evalE.asInstanceOf[UTF8String].toString, _scale)
+ case BinaryType =>
+
round(UTF8String.fromBytes(evalE.asInstanceOf[Array[Byte]]).toString, _scale)
+ }
+ }
+
+ private def round[T](input: T, scale: Int)(implicit bdc:
BigDecimalConverter[T]): T = {
+ input match {
--- End diff --
As `T` is a fix value during the runtime, make the `round` return a
function instead? Probably will be more efficient in performance.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
