What is the difference between map and flatMap?

In Spark, map applies a function to each element of an RDD/DataFrame and returns a new element for each input element, preserving the one-to-one relationship. flatMap also applies a function, but that function can return zero, one, or many elements; the results are then flattened into a single RDD/DataFrame. For example, map on ["a", "b"] with a function that returns the length would give [1,1], whereas flatMap with a function that splits each string into characters would give ["a","b"].