Github user cloud-fan commented on a diff in the pull request: https://github.com/apache/spark/pull/10577#discussion_r49915065 --- Diff: sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/HiveTypeCoercion.scala --- @@ -200,41 +203,62 @@ object HiveTypeCoercion { */ object WidenSetOperationTypes extends Rule[LogicalPlan] { - private[this] def widenOutputTypes( - planName: String, - left: LogicalPlan, - right: LogicalPlan): (LogicalPlan, LogicalPlan) = { - require(left.output.length == right.output.length) + private def widenOutputTypes(children: Seq[LogicalPlan]): Seq[LogicalPlan] = { + require(children.forall(_.output.length == children.head.output.length)) - val castedTypes = left.output.zip(right.output).map { - case (lhs, rhs) if lhs.dataType != rhs.dataType => - findWiderTypeForTwo(lhs.dataType, rhs.dataType) - case other => None - } + // Get a sequence of data types, each of which is the widest type of this specific attribute + // in all the children + val castedTypes: Seq[DataType] = + getCastedTypes(children, attrIndex = 0, mutable.Queue[DataType]()) - def castOutput(plan: LogicalPlan): LogicalPlan = { - val casted = plan.output.zip(castedTypes).map { - case (e, Some(dt)) if e.dataType != dt => - Alias(Cast(e, dt), e.name)() - case (e, _) => e - } - Project(casted, plan) + // Add extra Project for type promotion if necessary + if (castedTypes.isEmpty) children else children.map(castOutput(_, castedTypes)) + } + + // Add Project if the data types do not match + private def castOutput( + plan: LogicalPlan, + castedTypes: Seq[DataType]): LogicalPlan = { + val casted = plan.output.zip(castedTypes).map { + case (e, dt) if e.dataType != dt => + Alias(Cast(e, dt), e.name)() + case (e, _) => e } + if (casted.exists(_.isInstanceOf[Alias])) Project(casted, plan) else plan + } - if (castedTypes.exists(_.isDefined)) { - (castOutput(left), castOutput(right)) - } else { - (left, right) + // Get the widest type for each attribute in all the children + @tailrec private def getCastedTypes( + children: Seq[LogicalPlan], + attrIndex: Int, + castedTypes: mutable.Queue[DataType]): Seq[DataType] = { + // Return the result after the widen data types have been found for all the children + if (attrIndex >= children.head.output.length) return castedTypes.toSeq + + // For the attrIndex-th attribute, find the widest type + findWiderCommonType(children.map(_.output(attrIndex).dataType)) match { + // If unable to find an appropriate widen type for this column, return an empty Seq + case None => Seq.empty[DataType] + // Otherwise, record the result in the queue and find the type for the next column + case Some(widenType) => + castedTypes.enqueue(widenType) + getCastedTypes (children, attrIndex + 1, castedTypes) --- End diff -- nit: no space before `(`
--- 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