lxy-9602 commented on PR #323: URL: https://github.com/apache/paimon-cpp/pull/323#issuecomment-5630380548
Thanks for the contribution and the performance improvement. Could we refactor this before merging? Java’s `ManifestAvroReader` only uses the top-level `bucket` and `totalBuckets` fields for filtering. I’m not sure why `_FILE._SCHEMA_ID` is needed here, since `bucket-key` is immutable. Schema ID also seems too broad as a bucket-layout identifier. Could we reuse or extend the existing late-materialization framework instead of implementing another probe/bitmap/reread flow in `ManifestFile`? The selector is roughly: ```text _BUCKET = target OR _TOTAL_BUCKETS IS NULL OR _TOTAL_BUCKETS != expectedTotalBuckets ``` `_VERSION = 2` may still be validated for every row. This would also keep Manifest-specific logic out of the generic Avro reader. -- 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]
