c21 commented on a change in pull request #29342:
URL: https://github.com/apache/spark/pull/29342#discussion_r465330581
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/joins/ShuffledHashJoinExec.scala
##########
@@ -71,8 +89,137 @@ case class ShuffledHashJoinExec(
val numOutputRows = longMetric("numOutputRows")
streamedPlan.execute().zipPartitions(buildPlan.execute()) { (streamIter,
buildIter) =>
val hashed = buildHashedRelation(buildIter)
- join(streamIter, hashed, numOutputRows)
+ joinType match {
+ case FullOuter => fullOuterJoin(streamIter, hashed, numOutputRows)
+ case _ => join(streamIter, hashed, numOutputRows)
+ }
+ }
+ }
+
+ /**
+ * Full outer shuffled hash join has three steps:
+ * 1. Construct hash relation from build side,
+ * with extra boolean value at the end of row to track look up information
+ * (done in `buildHashedRelation`).
+ * 2. Process rows from stream side by looking up hash relation,
+ * and mark the matched rows from build side be looked up.
+ * 3. Process rows from build side by iterating hash relation,
+ * and filter out rows from build side being looked up already.
+ */
+ private def fullOuterJoin(
+ streamIter: Iterator[InternalRow],
+ hashedRelation: HashedRelation,
+ numOutputRows: SQLMetric): Iterator[InternalRow] = {
+ abstract class HashJoinedRow extends JoinedRow {
+ /** Updates this JoinedRow by updating its stream side row. Returns
itself. */
+ def withStream(newStream: InternalRow): JoinedRow
+
+ /** Updates this JoinedRow by updating its build side row. Returns
itself. */
+ def withBuild(newBuild: InternalRow): JoinedRow
}
+ val joinRow: HashJoinedRow = buildSide match {
+ case BuildLeft =>
+ new HashJoinedRow {
+ override def withStream(newStream: InternalRow): JoinedRow =
withRight(newStream)
+ override def withBuild(newBuild: InternalRow): JoinedRow =
withLeft(newBuild)
+ }
+ case BuildRight =>
+ new HashJoinedRow {
+ override def withStream(newStream: InternalRow): JoinedRow =
withLeft(newStream)
+ override def withBuild(newBuild: InternalRow): JoinedRow =
withRight(newBuild)
+ }
+ }
+ val joinKeys = streamSideKeyGenerator()
+ val buildRowGenerator = UnsafeProjection.create(buildOutput, buildOutput)
+ val buildNullRow = new GenericInternalRow(buildOutput.length)
+ val streamNullRow = new GenericInternalRow(streamedOutput.length)
+
+ def markRowLookedUp(row: UnsafeRow): Unit = {
+ if (!row.getBoolean(row.numFields() - 1)) {
Review comment:
@cloud-fan - updated.
----------------------------------------------------------------
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]
