Github user sddyljsx commented on the issue: https://github.com/apache/spark/pull/21859 'The ShuffleWriter should treat RangePartitioner specially and consume the sampled data in RangePartitioner instead of the input iterator.' This idea is good, maybe we can cache both the K and V when doing sample. I will have a try on this idea.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org