In Spark, there are transformation methods like map, reduce, filter. These methods are specific to Spark and are called Spark transformations. Howcome these methods are present in Scala programming language (without Spark).
I think the answer to the question you said as the better question to ask is that since Spark requires these methods to do the work they are intended to on a cluster of machines so they require a different version of these methods. That is what I think. I may be wrong.
Spark got inspired for it's API from Scala's collection API that has those methods (filter, map, flatMap, reduce, etc). When invoked in a Scala collection, they run locally, returning a new collection, when using Spark, the methods will be invoked on the RDD API, and will return a transformed RDD that will run in the Spark cluster.