Github user aokolnychyi commented on the issue:
https://github.com/apache/spark/pull/23171
@cloud-fan, yeah, letâs see if this PR is useful.
The original idea wasnât to avoid fixing autoboxing in `InSet`. `In` was
tested on 250 numbers to prove O(1) time complexity on compact values and
outline problems with `InSet`. After this change, `In` will be faster than
`InSet` but this is not the goal. Overall, the intent was to have a tiny and
straightforward change that would optimize `In` expressions even if the number
of elements is less than âspark.sql.optimizer.inSetConversionThresholdâ and
Spark does not use `InSet`.
Once we solve autoboxing issues in `InSet`, we would need to benchmark
against this approach.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
