cloud-fan commented on issue #25902: [SPARK-29213][SQL] Make it consistent when get notnull output and generate null checks in FilterExec URL: https://github.com/apache/spark/pull/25902#issuecomment-534830584 @viirya your proposal works, but is a little against the original design goal. Think about `IsNotNull(a + b + c + d ..)` and `Substring(a)`. When codegen `Substring(a)`, with your proposal, we will codegen `IsNotNull(a + b + c + d ..)` first, which loads more columns and may cause perf regression. I have a slightly different proposal: when codegen `Substring(a)`, check if there is a `IsNotNull(a)` predicate. If there is, codegen it first. If not, and `a` is a not-nullable attribute, then generate a `IsNotNull(a)` and codegen it first.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
