alamb opened a new issue, #7749: URL: https://github.com/apache/arrow-datafusion/issues/7749
### Is your feature request related to a problem or challenge? After https://github.com/apache/arrow-datafusion/pull/7721 a `SortExec` with a limit will use a special `TopK` . We have basic unit tests, but I think the coverage could be improved, specifically with Fuzz testing ### Describe the solution you'd like What I would like is a new fuzz test to be added to the the existing fuzz cases: https://github.com/apache/arrow-datafusion/tree/main/datafusion/core/tests/fuzz_cases The structure of `SortTest` in https://github.com/apache/arrow-datafusion/blob/e95a24b5a260e0e2f603d52682d36cce192676f8/datafusion/core/tests/fuzz_cases/sort_fuzz.rs#L111 might be a good one to follow The basic outline would be: 1. Create an input with several columns (integers, strings, floats) 2. Reorder the input randomly 3. Divide the input up multiple batches using [`make_staggered_batches`](https://github.com/apache/arrow-datafusion/blob/0d82620bb45db35ce30ec170cb68e953fbc94366/datafusion/core/tests/fuzz_cases/aggregate_fuzz.rs#L172) 4. Run a query like `SELECT * FROM t ORDER BY <col(s)> LIMIT <N>` and collect the output 5. Compute the expected result programmatically (e.g. by sort the data, prior to creating RecordBatches) 6. Ensure the output matches the expected result Input size: 1000 rows Parameters to vary 1. sort cols: (int), (string), (float), (int, string), (string, int), etc. 2. N: 1, 10, 100, 300 (aka how many are kept) Bonus points make it easy to add new columns / types (e.g. like string dictionary) ### Describe alternatives you've considered _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]
