Github user kiszk commented on a diff in the pull request:
https://github.com/apache/spark/pull/21045#discussion_r191260626
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/collectionOperations.scala
---
@@ -127,6 +127,165 @@ case class MapKeys(child: Expression)
override def prettyName: String = "map_keys"
}
+@ExpressionDescription(
+ usage = """_FUNC_(a1, a2, ...) - Returns a merged array containing in
the N-th position the
+ N-th value of each array given.""",
+ examples = """
+ Examples:
+ > SELECT _FUNC_(array(1, 2, 3), array(2, 3, 4));
+ [[1, 2], [2, 3], [3, 4]]
+ > SELECT _FUNC_(array(1, 2), array(2, 3), array(3, 4));
+ [[1, 2, 3], [2, 3, 4]]
+ """,
+ since = "2.4.0")
+case class Zip(children: Seq[Expression]) extends Expression with
ExpectsInputTypes {
+
+ override def inputTypes: Seq[AbstractDataType] =
Seq.fill(children.length)(ArrayType)
+
+ override def dataType: DataType = ArrayType(mountSchema)
+
+ override def nullable: Boolean = children.forall(_.nullable)
+
+ private lazy val arrayTypes =
children.map(_.dataType.asInstanceOf[ArrayType])
+
+ private lazy val arrayElementTypes = arrayTypes.map(_.elementType)
+
+ def mountSchema: StructType = {
+ val fields = children.zip(arrayElementTypes).zipWithIndex.map {
+ case ((expr: NamedExpression, elementType), _) =>
+ StructField(expr.name, elementType, nullable = true)
+ case ((_, elementType), idx) =>
+ StructField(s"$idx", elementType, nullable = true)
+ }
+ StructType(fields)
+ }
+
+ override def doGenCode(ctx: CodegenContext, ev: ExprCode): ExprCode = {
+ val numberOfArrays: Int = children.length
+ val genericArrayData = classOf[GenericArrayData].getName
+ val genericInternalRow = classOf[GenericInternalRow].getName
+ val arrVals = ctx.freshName("arrVals")
+ val arrCardinality = ctx.freshName("arrCardinality")
+ val biggestCardinality = ctx.freshName("biggestCardinality")
+ val storedArrTypes = ctx.freshName("storedArrTypes")
+ val returnNull = ctx.freshName("returnNull")
+ val evals = children.map(_.genCode(ctx))
+
+ val inputs = evals.zipWithIndex.map { case (eval, index) =>
+ s"""
+ |${eval.code}
+ |if (!${eval.isNull}) {
+ | $arrVals[$index] = ${eval.value};
+ | $arrCardinality[$index] = ${eval.value}.numElements();
+ |} else {
+ | $arrVals[$index] = null;
+ | $arrCardinality[$index] = 0;
+ | $returnNull[0] = true;
+ |}
+ |$storedArrTypes[$index] = "${arrayElementTypes(index)}";
--- End diff --
In simple cases, since we know only a data type of all children before
execution, we may not need to use `$storedArrTypes`. However, I may miss
something.
Would it be possible to show an example test case that requires to pick the
correct `getValue` by using `$storedArrTypes`?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
