The hidden injuries of overloading 'ADT'

Abstract

The most commonly stated definition of abstract data type (ADT) is that it is a domain of values and the operations over that domain. So, for example, a language's built-in types, like int are seen to be ADTs. It is our opinion that a pure interpretation of this definition yields a semantics in which using an ADT is the same as using built-in types: the operations are side effect free and there is no concern over alias, shallow copy or synchronization problems. Unfortunately, the term abstract data type has over time been associated with at least three distinct meanings, and those incompatible definitions have often been conflated, causing confusion to students and textbook authors alike. We believe that this has resulted in a loss of appreciation for the value-based semantics of ADTs.