Weijun-H commented on PR #25406: URL: https://github.com/apache/datafusion/pull/25406#issuecomment-5713218747
Follow-up for https://github.com/apache/datafusion/pull/25406#discussion_r4035916762: updated in b562850ab. Fully drained output now uses a cursor-only fast path, so it avoids the live-batch mark/remap Vec allocations on the common empty-index path. I kept the liveness check (`row_idx < batch.num_rows()`) in that fast path so exhausted cursors still release their final batch rather than restoring the old retention behavior. The existing builder regression covers both branches: the first partial emit leaves pending indices and uses `retain_live_batches`, while the final drain empties indices and exercises the fast path. -- 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]
