Hi Robert,

Thanks for your suggestion. The fact that I want to do this in SparkSQL is because there is no underlying SQLServer. Well, there might be, but often there isn't. We make use of the possibility of SparkSQL to abstract the underlying data stores (Parquet files, CSV files or SQL data) into a single SQL-alike layer. That is a kind of virtual curation.

In that context, I do not see an option to delegate the UDF to a lower SQL server. Am I overlooking something?

Best regards,

Johan