viirya commented on PR #6616: URL: https://github.com/apache/arrow-rs/pull/6616#issuecomment-2436119691
Yea, these are the only cases and I believe that are not "many". Not to mention that they are reasonable cases as these arrays are designed to be shared/reused. I already mentioned this in previous comment. That's why I don't understand that you claim that performance regression will be a problem for such a change. Except for above cases, the only cases reusing is happening on null buffer in `take` kernel. I doubt it will bring much extra cost on copying them. With a negligible cost, making `take` behaves more close to the common semantics of "take" (philosophical? 🙂 ). I don't get why this patch gets string objection. Anyway, as the workaround was already added to bypass the issue. I respect your decision here. -- 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]
