suxiaogang223 opened a new pull request, #65307: URL: https://github.com/apache/doris/pull/65307
### What problem does this PR solve? Fix wrong results when querying Paimon partitioned primary-key tables with partition predicates such as `partition IN (...)` and `enable_runtime_filter_partition_prune=true`. The old file scanner reuses `_runtime_filter_partition_prune_block` across splits but inserted partition columns for every split, which can pollute the block layout and make later split-level partition pruning evaluate against the wrong column state. Paimon split partition metadata also needs to be passed in a stable key order instead of relying on path parsing. ### What is changed? - Replace partition columns in the runtime partition prune block by position instead of inserting new columns. - Add a helper to fill partition columns, including nullable partition values. - Pass ordered Paimon metadata partition values for each split. - Normalize Paimon split partition lists to avoid path parsing for Paimon dummy/native splits. - Add Paimon partition regression assertions comparing JNI and native results under partition `IN` predicates. ### Check List - [x] Regression test - [ ] Unit test - [ ] Manual test Not run locally per investigation plan; expected to be verified by CI/regression pipeline. Related fix: apache/doris#62821 Related Jira: DORIS-26857 -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
