Instances in a dataset are called Records. Each record has a key, unique within a dataset, and zero or more cells. Record instances held within a given dataset are immutable. Changing one or more values in a record creates a new instance of that record which replaces the old instance.