rohityadav1993 commented on PR #19120: URL: https://github.com/apache/pinot/pull/19120#issuecomment-6198835824
@Jackie-Jiang notes 3 and 4 from your review are done in 1ef67c18b6. 3. Finish at the global limit: the merge checks the limit right after emitting a row and leaves through the path that finishes before flushing the final block. It never advances the leader past the last needed row, and every segment is released before a full final block is returned. finish() also drops the cursors' buffered rows, and an exhausted leaf drops its run heap and doc-id copy. Tests: the segment that supplies the last row pulls exactly one block per row it supplied (one extra before), and with a full final block every segment is released and no cursor holds rows when it is returned (both failed before). 4. Response boundary: three tests build makeStreamingInstancePlan with a capturing streamer and assert on the InstanceResponseBlock. They cover operator pairing, rows against the blocking plan, and total docs / segments / docs scanned; AUTO resolving to OFF keeping MinMax through the single-block branch; and a segment rebuilt with valCol as LONG reported as MERGE_RESPONSE with its rows dropped and stats intact. Resource statistics are covered by the P2e test, and the doc-id ownership coverage is in the P1a thread. Notes 1 and 2, P2b, and the AUTO wording are still open; I'll follow up on those separately. -- 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]
