cloud-fan commented on code in PR #39017:
URL: https://github.com/apache/spark/pull/39017#discussion_r1054345074
##########
connector/connect/server/src/main/scala/org/apache/spark/sql/connect/planner/SparkConnectPlanner.scala:
##########
@@ -132,12 +132,32 @@ class SparkConnectPlanner(session: SparkSession) {
* wrap such fields into proto messages.
*/
private def transformSample(rel: proto.Sample): LogicalPlan = {
+ val input = Dataset.ofRows(session, transformRelation(rel.getInput))
+ val plan = if (rel.getForceStableSort) {
+ // It is possible that the underlying dataframe doesn't guarantee the
ordering of rows in its
+ // constituent partitions each time a split is materialized which could
result in
+ // overlapping splits. To prevent this, we explicitly sort each input
partition to make the
+ // ordering deterministic. Note that MapTypes cannot be sorted and are
explicitly pruned out
+ // from the sort order.
+ val sortOrder = input.logicalPlan.output
+ .filter(attr => RowOrdering.isOrderable(attr.dataType))
+ .map(SortOrder(_, Ascending))
+ if (sortOrder.nonEmpty) {
+ Sort(sortOrder, global = false, input.logicalPlan)
+ } else {
+ input.logicalPlan
+ }
+ } else {
+ input.cache()
Review Comment:
Then we need a new proto message for randomSplit, and can't reuse Sample, to
avoid adding extra cache for `df.sample`.
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
