nat-openai opened a new issue, #26342: URL: https://github.com/apache/datafusion/issues/26342
### Describe the bug The native `array_repeat` kernel in DataFusion can reject valid array inputs whose elements are nonnullable. It reconstructs the inner list with a nullable `item` field, while the outer list still expects the original type. ### To Reproduce Run this query with native execution, preserving `array(1)` as an Arrow list with a nonnullable `Int32` element field (`containsNull=false`): ```sql SELECT id, array_repeat(array(1), id) AS repeated FROM VALUES (1), (2) AS t(id); ``` (Input paths that widen the element field to nullable can mask the bug, so this query does not fail on every Comet build.) Execution fails with an Arrow type mismatch: ```text ListArray expected data type List(non-null Int32) got List(Int32) ``` ### Expected behavior In Spark, the query succeeds: ```text id repeated 1 [[1]] 2 [[1], [1]] ``` The native kernel should preserve the input list's element field, including nullability, name, and metadata, rather than replacing it with a new field. ### Additional context _No response_ ### Willingness to contribute I would like to contribute a fix for this bug, but request guidance from the DataFusion community -- 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]
