suxiaogang223 opened a new pull request, #65581: URL: https://github.com/apache/doris/pull/65581
### What problem does this PR solve? Paimon partition metadata was generated only when runtime partition pruning was enabled. Some count/native/JNI splits therefore reached BE without stable partition values. BE also reused a partition-prune block across splits but inserted replacement columns, corrupting the block layout. ### Changes - Always attach ordered Paimon partition metadata to every data split. - Generate aligned `columns_from_path_keys`, `columns_from_path`, and `columns_from_path_is_null` arrays in FE. - Pass explicit null markers from `FileScanner` to Parquet/ORC readers. - Preserve NULL, empty string, and literal `\N` semantics. - Replace partition columns in the reusable prune block instead of inserting new columns. - Add FE/BE unit coverage and a three-partition static `IN` regression case. This is a focused backport of the relevant #62821 behavior adapted to the pre-#62306 reader architecture. ### Validation - `clang-format` 16 - `git diff --check` - FE checkstyle for touched Java files - Full FE/BE tests will run in CI; local FE UT compilation is blocked by missing generated Thrift Java classes in this worktree. -- 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]
