MaxGekk commented on PR #54905: URL: https://github.com/apache/spark/pull/54905#issuecomment-4260605764
> Finding 2: Per-row object allocation in HiveResult hot path > Issue: ClientTypeOps(dt) is called on every (value, dt) pair in the result set, creating a new TimeTypeApiOps instance per call. Additionally, it's called for ALL non-null values of ALL types (not just TimeType), returning None for standard types — adding overhead to every row of every query. This is also my main concern regarding the current approach. In my original PR https://github.com/apache/spark/pull/51467, I tried to minimize impact on existing types. Definitely it will need benchmarking when enabling the framework by default. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
