Github user srowen commented on the issue:
https://github.com/apache/spark/pull/19406
It's probably OK, though kind of special cases the first quantile. This
expression doesn't incorporate relativeError. I'm actually not sure why it
needs to account for relativeError here; it's an extra level of approximation.
But the asymmetry seems a bit wrong.
It is after all approximate? the loop below actually accounts for the same
case, but incorporates relativeError.
When in doubt ask the author: @clockfly
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
