yadavay-amzn commented on PR #56243: URL: https://github.com/apache/spark/pull/56243#issuecomment-4697187362
Thanks for the detailed trace, @cloud-fan — you'''re right. I worked through the coalesce math again: the single-partition case only occurs when `totalSize <= advisory`, which this floor explicitly skips, and where the floor engages (`totalSize > advisory`) the baseline already yields `>= ceil(totalSize/advisory) >= 2` partitions. The only residual effect is shrinking `targetSize` below advisory, which just produces more sub-advisory partitions — not a real improvement, and it mis-models a join'''s cost as its input size as you noted. I couldn'''t produce a master-failing test that reflects the stated goal. Closing this; appreciate the careful review. -- 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]
