maropu commented on a change in pull request #29587:
URL: https://github.com/apache/spark/pull/29587#discussion_r482700803
##########
File path:
sql/catalyst/src/main/scala/org/apache/spark/sql/types/StructType.scala
##########
@@ -641,4 +641,30 @@ object StructType extends AbstractDataType {
fields.foreach(s => map.put(s.name, s))
map
}
+
+ /**
+ * Returns a `StructType` that contains missing fields recursively from
`source` to `target`.
+ * Note that this doesn't support looking into array type and map type
recursively.
+ */
+ def findMissingFields(source: StructType, target: StructType, resolver:
Resolver): StructType = {
Review comment:
okay, one nit: a return type `Option[StructType]` for `findXXX` methods
is more natural just like scala collection (e.g., `Seq.find`)?
##########
File path:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/ResolveUnion.scala
##########
@@ -50,18 +122,29 @@ object ResolveUnion extends Rule[LogicalPlan] {
}
}
+ (rightProjectList, aliased)
+ }
+
+ private def unionTwoSides(
+ left: LogicalPlan,
+ right: LogicalPlan,
+ allowMissingCol: Boolean): LogicalPlan = {
+ val rightOutputAttrs = right.output
+
+ // Builds a project list for `right` based on `left` output names
+ val (rightProjectList, aliased) = compareAndAddFields(left, right,
allowMissingCol)
+
// Delegates failure checks to `CheckAnalysis`
- val notFoundAttrs = rightOutputAttrs.diff(rightProjectList)
+ val notFoundAttrs = rightOutputAttrs.diff(rightProjectList ++ aliased)
val rightChild = Project(rightProjectList ++ notFoundAttrs, right)
// Builds a project for `logicalPlan` based on `right` output names, if
allowing
// missing columns.
val leftChild = if (allowMissingCol) {
- val missingAttrs = notFoundAttrs.map { attr =>
- Alias(Literal(null, attr.dataType), attr.name)()
- }
- if (missingAttrs.nonEmpty) {
- Project(leftOutputAttrs ++ missingAttrs, left)
+ // Add missing (nested) fields to left plan.
+ val (leftProjectList, _) = compareAndAddFields(rightChild, left,
allowMissingCol)
+ if (leftProjectList.map(_.toAttribute) != left.output) {
Review comment:
nit:
```
if (leftProjectList.length != left.output.length ||
leftProjectList.map(_.toAttribute) != left.output) {
```
?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
