JingsongLi commented on PR #858: URL: https://github.com/apache/paimon-rust/pull/858#issuecomment-5771875634
Requirement fit: **SUPPORTED**. Implementation: **CLEAN in this follow-up review** of `72d9300cc080ef486a6b4284aeebaf36cf94baad`. The latest commit addresses my partition-predicate finding. The selected `Eq`/`In` candidates no longer discard the remaining conjuncts, and the round-trip check preserves full predicate evaluation when a literal cannot be represented exactly (including timestamp precision and NaN). The added SQL and native-count regressions cover mixed `IN`/range conditions and contradictory conditions. The Avro change also addresses the record-walk concern: full, filtered, and slim manifest-entry decoding now share the top-level traversal, and the slim OCF loops share one helper. I checked the nullable, reordered, unknown-field, and missing-count cases against the differential tests. Local verification on this head: - `cargo +1.94.0 test -p paimon --lib partition_filter` — 16 passed. - `cargo +1.94.0 test -p paimon --lib partition_row_count` — 13 passed. - `cargo +1.94.0 test -p paimon-datafusion --test partition_count_pushdown --test count_pushdown` — 14 + 9 passed, including snapshot pinning, schema changes, deletion vectors, and execution-time fallback. The metadata-only grouped-count path has direct end-to-end value for partition-only aggregate queries. I found no new actionable regression in this follow-up. This is a focused re-review of the changes since my previous comments, rather than a blanket approval of every earlier change. -- 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]
