Github user srowen commented on the issue:
https://github.com/apache/spark/pull/19565
I assume not-selecting a record in a sample is cheaper than just about any
other operation, including filtering on a predicate. All else equal, I'd rather
sample, then evaluate a predicate on only the sample.
The two versions produce the same result though. Either way, every x%
sample of non-empty docs appears with equal probability.
Yes, I doubt one is meaningfully faster than the other overall though. If
that's not the motive though, and there's not a functional change, what's the
purpose here?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
