Github user cloud-fan commented on the issue:

    https://github.com/apache/spark/pull/22010
  
    While this saves a shuffle, but the algorithm becomes different. Previously 
we use the shuffe aggregator, which stores data in a `ExternalAppendOnlyMap`. 
Now we use a scala set, which may OOM.


---

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to