wangyong9999 opened a new issue, #288: URL: https://github.com/apache/paimon-cpp/issues/288
### Search before asking - [x] I searched the existing issues and found no duplicate. ### Paimon-cpp version Upstream main at `ead92078` (before PR #284). ### Minimal reproduce step 1. Write a fixed-bucket primary-key table with two buckets. 2. Read the same snapshot with the current `bucket` option set to four and an equality predicate covering the bucket key. Choose a key whose computed bucket differs between the two counts. Do not set an explicit bucket filter. 3. Compare the returned rows with the same lookup using the original bucket option. The existing KV scan derives one bucket from the current option and applies it to historical files. A matching file can be discarded because its stored bucket was calculated with a different count. For an append table, write keys across several fixed buckets with overlapping file statistics, then scan with an equality predicate on the complete bucket key. The scan retains files from other buckets even though the predicate determines which bucket can contain the key. ### What doesn't meet your expectations? Equality-based pruning should work for both append and KV scans, and changing the current bucket option must not hide matching historical data. The bucket comparison needs each manifest entry's recorded total bucket count. Historical schemas with compatible bucket-key identities, types and hash functions should remain eligible. ### Anything else? PR #284 implements a shared selector in `FileStoreScan`, preserves explicit-filter semantics, and keeps inferred manifest-cache candidates safe across historical bucket counts and schemas. It includes unit and integration coverage for rescaled options, schema compatibility and repeated cached lookups. ### Are you willing to submit a PR? - [x] I'm willing to submit a PR! See #284. -- 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]
