From one single full raw data, I have to extract some data and dump multiple type of data.

In Simple Example

case class A (id: Int, age: Int, name: String)

case class B (id: Int, age: Int, married: Boolean)

case class C (id: Int, age: Int, gender: String)

Dumping it sequentially might be naive solution.
But since those case classes have multiple common fields, I think it is quite inefficient.

What I though was making some parent trait or abstract class and extends it.
And then add category field in case class.
Finally use partitionBy(category) to dump in different path.

In example

trait P(category: String, id: Int, age: Int)

case class A (category: String, id: Int, age: Int, name: String)

case class B (category: String, id: Int, age: Int, married: Boolean)

case class C (category: String, id: Int, age: Int, gender: String)

On my IDE, it looks like making RDD[P] works, but transforming it in to `DataSet(through .toDS() method) could not work.

Is there any way to solve it?
(Not making class with complex methods like compare or etc…)