Github user srowen commented on the issue: https://github.com/apache/spark/pull/19406 (What is the answer in your test case before the change?) Yea, I guess I am not sure why this method has to use relativeError at all, but I didn't think about it much. If it didn't I think it already does the same thing you've done. Which then leads me to ask why there's an approximation in this part of the process at all.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org