zhuqi-lucas opened a new pull request, #21756: URL: https://github.com/apache/datafusion/pull/21756
## Which issue does this PR close? Closes #21755 ## Rationale When a hive-partitioned listing table has files in the root directory (not inside any `partition_col=value/` path), queries that reference partition columns crash with `Unable to get field named "partition_col"`. This is a common scenario when a table transitions from non-partitioned to hive-partitioned storage — the original root file may still exist alongside the new partition directories. ## What changes are included in this PR? `try_into_partitioned_file` now returns `Ok(None)` for files that don't match the partition structure (where `parse_partitions_for_path` returns `None`). The caller skips them via `try_filter_map`. Previously, `None` from `parse_partitions_for_path` was converted to empty `partition_values` via `.into_iter().flatten()`, causing downstream errors. ## Are these changes tested? Yes, 5 unit tests added: - `test_try_into_partitioned_file_valid_partition` — normal case - `test_try_into_partitioned_file_root_file_skipped` — root file skipped - `test_try_into_partitioned_file_wrong_partition_name` — wrong partition col name - `test_try_into_partitioned_file_multiple_partitions` — multi-level partitions - `test_try_into_partitioned_file_partial_partition_skipped` — incomplete partition path ## Are there any user-facing changes? Files outside the hive partition structure are now silently skipped instead of causing query failures. This may change `COUNT(*)` results if such files previously contributed rows (with empty partition values). -- 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]
