Currently in alpha state
https://issues.apache.org/jira/browse/SPARK-23818 compare with transport
- Better primitives,
- Better collections API
- Better expressions API for
- Type safety
- no per record instanceof checks and per record figuring out which data type it is
- apache/spark#5154
- useful for ML while keeping the generic nature of API
- without stddata everywhere
- https://github.com/lesbroot/typedudf/tree/master/src/main/scala/typedudf