XiaoHongbo-Hope opened a new pull request, #1025: URL: https://github.com/apache/paimon-rust/pull/1025
## Summary - Determine the final data/filter projection before loading Parquet page indexes, then fetch/decode only the required ColumnIndex and OffsetIndex columns. - Keep predicate page pruning and row selection semantics; missing optional OffsetIndex falls back to coarser reads. - Reuse the footer tail prefetch, scope metadata-cache entries by index coverage, and trace selected index columns and range reads. ## Trade-off This reduces page-index decoding for narrow reads. When index bytes already fit in the footer prefetch, it may not reduce GET requests; full projections may see no benefit. This changes neither the Parquet format nor writer behavior. ## Validation - cargo test -p paimon --lib --quiet: 3604 passed, 6 ignored - cargo clippy -p paimon --all-targets -- -D warnings: passed - cargo fmt --all --check and git diff --check: passed - Added wide, multi-row-group tests for narrow/full projection equivalence, row selection, missing indexes, index coverage, and cache isolation. Current-patch production A/B remains pending. No end-to-end throughput claim is made here. -- 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]
