Spark AS lookup implementation

Question

I want to replace our UDF implementation of AS lookup with native spark functions.

AS lookup table has circa 500 000 records which look like this:

+---