viirya opened a new issue, #807:
URL: https://github.com/apache/datafusion-comet/issues/807
### Describe the bug
In #651, Comet encountered some errors on Spark test like following. It is
because DataFusion SortExec calls `lexsort_to_indices` and the function
fallbacks to `sort_to_indices` for single column case. However,
`sort_to_indices` doesn't support all data types, e.g., struct. For unsupported
data types, `lexsort_to_indices` should not fallback to `sort_to_indices`.
```
[info] - vectorized reader: struct *** FAILED *** (515 milliseconds)
[info] org.apache.spark.SparkException: Job aborted due to stage failure:
Task 0 in stage 644.0 failed 1 times, most recent failure: Lost task 0.0 in
stage 644.0 (TID 862) (192.168.86.44 executor driver): org.apache.comet.CometNa
tiveException: Compute error: Sort not supported for data type Struct([Field
{ name: "_1", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered:
false, metadata: {} }, Field { name: "_2", data_type: Utf8, nullable: true,
dict_id: 0, dict_is_ordered: false, metadata: {} }])
```
I'm going to fix it at arrow-rs side:
https://github.com/apache/arrow-rs/issues/6226. But before we have new
releases, we need to fallback to Spark for unsupported cases like struct.
### Steps to reproduce
_No response_
### Expected behavior
_No response_
### Additional context
_No response_
--
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]
