LuciferYang opened a new pull request, #10255: URL: https://github.com/apache/paimon/pull/10255
### Purpose `IcebergDataFileMeta.create` treated a null `statsColumns` as covering the whole Iceberg (table) schema in table order. A null `valueStatsCols` instead means the stats cover the whole write schema, whose column order and set can differ from the table for a partial write. As a result the min/max bounds and null counts drifted onto the wrong Iceberg fields in the manifest, and a strict-subset write could read past the end of the stats row. This passes the file's `writeCols` into `create`. When `statsColumns` is null and `writeCols` is present, it builds the explicit stats-column list from the write columns in their recorded order, mapping a nested leaf path to its top-level field so each stats slot is attributed to the correct column. Full writes keep `writeCols` null and are unaffected. This closes #10253. ### Tests `IcebergDataFileMetaTest.testNullStatsColumnsWithWriteColsAlignByWriteSchema` pins that a partial write of `(b, k)` in write-column order maps the bounds and null counts to `b` and `k` and does not drift onto the unwritten column `a`. `IcebergDataFileMetaTest.testNullStatsColumnsWithNestedWriteColsMapToTopLevel` pins that a nested leaf path attributes its stats to the top-level field rather than to a neighboring column. ### API and Format No. ### Documentation No. -- 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]
