viirya commented on issue #8866: URL: https://github.com/apache/arrow-datafusion/issues/8866#issuecomment-1894273366
I feel that it is not so reasonable to bind the nature of a volatile scalar function to some special way on expanding its arguments or how DataFusion treats its output. The definition of volatile scalar function doesn't include such spec so by doing this we might create something weird to understand by outside DataFusion. The special treatment of argument/output is specified to `make_scalar_function` which is useful for special cases of built-in scalar function to save time on implementation. However this assumption doesn't apply generally to all use cases and that's the problem as users are easily to misuse it. The good news is we have `ScalarUDFImpl` now and we are moving to deprecate previous function definition of scalar UDF. I think for this issue we just need to deprecate `make_scalar_function` and make its doc more clear as I proposed in the #8878. -- 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]
