In Spark, operations are either actions or transformations. How would you categorize caching and persistence? Would you consider them actions or transformations?

In Spark, cache() and persist() are lazy transformations that mark an RDD or DataFrame to be stored in memory or on disk with a specified storage level. They do not trigger a job by themselves; the actual caching occurs when an action forces evaluation of the RDD. Because they are applied to an RDD and return a new RDD with a persistence flag, they are classified as transformations. They are useful for speeding up subsequent actions that reuse the same data, but they do not produce output like actions do.