Github user DylanGuedes commented on a diff in the pull request: https://github.com/apache/spark/pull/21045#discussion_r191412802 --- 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 -- Suppose that you have three inputs, where the first is an Array[Int], the second an Array[String], and third an Array[Long]. How do you, in runtime, uses `getInt(s"$arrVals[$j]", IntegerType, i)` only when filling **the first position of each row**, `getString(s"$arrVals[$j]", StringType, i)` only when filling the second position of each row and `getInt(s"$arrVals[$j]", LongType, i)` only when filling the third? My solution is: I hardcodded a condition: if the stored arrayType for the current array is "StringType", when I use `getValue(..., StringType, i)`, if it is "DoubleType", then I use `getValue(..., DoubleType, i)`, etc.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org