XiaoHongbo-Hope commented on PR #614: URL: https://github.com/apache/paimon-rust/pull/614#issuecomment-5083908609
Thanks for the thorough review. Addressing the two P1s: - **P1 (full_text_search)** — Fixed in this PR (commit 9715860). `full_text.rs` had its own `SearchResult::top_k` with the identical `len <= k → clone` bug; it now sorts best-first on both paths, and the test that asserted the old `[1,4]/[0.1,0.2]` order is updated to `[4,1]/[0.2,0.1]`. - **P1 (DataFusion `vector_search` loses ranking end-to-end)** — Handled by the companion PR #613, which stops the TVF from re-reading rows in file order (`to_row_ranges` + physical scan) and instead gathers them back into the index's rank. That layer is out of scope for this primitive-level fix; the two PRs together close the loop for all `limit` values. (The `extract_ids` sort you flagged is why #613 adds an order-preserving `extract_ids_in_order` assertion.) So this PR now covers both `SearchResult::top_k` primitives (vector + full_text); #613 covers the DataFusion scan. -- 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]
