zml1206 commented on PR #12231: URL: https://github.com/apache/gluten/pull/12231#issuecomment-4624108452
> > However, LocalPartitionWriter::mergeSpills() later merges spill files by scanning pids in ascending order, and Spill::nextPayload(pid) only consumes the queue head. > > Could `evictPartitionBuffersMinSize` be executed multiple times, causing the payloads of spiller to resemble `[1, 2, 3, 1, 2, 3]`? Yes, multiple calls can produce a sequence like [1, 2, 3, 1, 2, 3] across spill rounds. That is OK because each call writes its selected partitions in pid order, and different calls become separate spill batches/files. The correctness requirement here is that a single spill batch is pid-ordered; later merge reads payloads by partition across spill batches. -- 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]
