Yes, the optimization you mentioned is related to storage (so it speeds up loading from storage only before any transformations are made you need to manipulate portions which are create on cluster after transform() is made)


My blog: https://databrickster.medium.com/