PetarVasiljevic-DB commented on code in PR #50921:
URL: https://github.com/apache/spark/pull/50921#discussion_r2192283750
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/v2/V2ScanRelationPushDown.scala:
##########
@@ -98,6 +107,138 @@ object V2ScanRelationPushDown extends Rule[LogicalPlan]
with PredicateHelper {
filterCondition.map(Filter(_, sHolder)).getOrElse(sHolder)
}
+ def pushDownJoin(plan: LogicalPlan): LogicalPlan = plan.transformUp {
+ // Join can be attempted to be pushed down only if left and right side of
join are
+ // compatible (same data source, for example). Also, another requirement
is that if
+ // there are projections between Join and ScanBuilderHolder, these
projections need to be
+ // AttributeReferences. We could probably support Alias as well, but this
should be on
+ // TODO list.
+ // Alias can exist between Join and sHolder node because the query below
is not valid:
+ // SELECT * FROM
+ // (SELECT * FROM tbl t1 JOIN tbl2 t2) p
+ // JOIN
+ // (SELECT * FROM tbl t3 JOIN tbl3 t4) q
+ // ON p.t1.col = q.t3.col (this is not possible)
+ // It's because there are 2 same tables in both sides of top level join
and it not possible
+ // to fully qualified the column names in condition. Therefore, query
should be rewritten so
+ // that each of the outputs of child joins are aliased, so there would be
a projection
+ // with aliases between top level join and scanBuilderHolder (that has
pushed child joins).
Review Comment:
No, this explains why we don't support Aliases in Projections between Join
and Scan nodes. This problem is not approach-specific, it's more like
reasoning behind design decision
--
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.
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
