Github user mateiz commented on the pull request: https://github.com/apache/spark/pull/1499#issuecomment-49669025 I've now updated it with I believe all of @aarondav's fixes and an optimization he suggested for in-memory-only data (don't bother sorting by key or merge-sorting). I also tried running this on all our test suites by making sort-based shuffle the default, and it almost works, except there's a test with custom Product2 types in ShuffledRDD that breaks. However upon further investigation I'm not sure that feature was fully supported even in the current version (our current spilling code would also not return the right type), so we might remove that.
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---