cloud-fan commented on a change in pull request #24442: [SPARK-27547][SQL] fix
DataFrame self-join problems
URL: https://github.com/apache/spark/pull/24442#discussion_r277758584
##########
File path: sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala
##########
@@ -876,7 +885,30 @@ class Dataset[T] private[sql](
* @since 2.0.0
*/
def join(right: Dataset[_]): DataFrame = withPlan {
- Join(logicalPlan, right.logicalPlan, joinType = Inner, None, JoinHint.NONE)
+ val (joinLeft, joinRight) = prepareJoinPlan(this, right)
+ Join(joinLeft, joinRight, joinType = Inner, None, JoinHint.NONE)
+ }
+
+ // Called by `Dataset#join`, to attach the Dataset id to the logical plan,
so that we
+ // can resolve column reference correctly later. See
`ResolveDatasetColumnReference`.
+ private def createPlanWithDatasetId(): LogicalPlan = {
+ if
(!sparkSession.sessionState.conf.getConf(SQLConf.RESOLVE_DATASET_COLUMN_REFERENCE))
{
+ return logicalPlan
+ }
+
+ // The alias should start with `SubqueryAlias.HIDDEN_ALIAS_PREFIX`, so
that `SubqueryAlias` can
Review comment:
It's possible, but I'm a little worried about changing the constructor of a
widely used plan node like `SubqueryAlais`. It's also a little weird to put the
dataset concept into the catalyst module.
Another idea I've thought about is adding a new no-op plan node to carry the
dataset id. I gave it up because no-op node can be troublesome according to the
past experience with `ResolvedHint` and `AnalysisBarrier`.
----------------------------------------------------------------
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]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
