RDD implementation for merge-join that uses a shuffle to partition and sort by keys using an implicit Ordering for K,
and then delegates to an instance of MergeJoin to perform the actual merge logic.
There is an optimization in place to avoid a shuffle in some cases where left or right are guaranteed to be partition-sorted already (ie: via repartitionAndSortWithinPartitions)
:: @DeveloperApi ::
RDD implementation for merge-join that uses a shuffle to partition and sort by keys using an implicit Ordering for
K
, and then delegates to an instance of MergeJoin to perform the actual merge logic.There is an optimization in place to avoid a shuffle in some cases where
left
orright
are guaranteed to be partition-sorted already (ie: viarepartitionAndSortWithinPartitions
)