Github user gatorsmile commented on the issue:
https://github.com/apache/spark/pull/20372
This PR was merged to RC3 of Spark 2.3. For all such fixes, we should not
merge them to Spark 2.3. The performance regression has been witnessed in RC3,
compared with RC2. We did not investigate more. For safety, we have to revert
it from Spark 2.3 and master.
Maybe you can submit a new PR and make it configurable? Does this sound
good to you?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
