Skip to main content
CSE 331: Software Design & Implementation

CSE 331: Abstract data types

Programs keep track of data in order to perform their function. For example, a program that keeps track of an office hours queue will have some way to store the students in the queue. Program data often has "rules" that must be followed. For example, the office hours queue should store the students in the order they joined and ensure that each student is present in the queue at most once.

We can use the techniques of testing and reasoning to ensure that the rules are followed. But the question is: how much of the code do we have to test or reason about in order to be sure the rules are followed? Answer: all the code that modifies the data. In a poorly organized program, where many parts of the code all directly modify the data, we need to test and reason about all of them.

A related problem is that we may later decide to change the details of how exactly the data is represented. For example, we may change from a queue based on an array to a queue based on a linked list. Here the question is: how much of the code must be updated to make this change? The answer again is: all the code that modifies the data. In a poorly organized program, that may be a large amount of code.

A well organized program can address both of these problems with an abstract data type. An abstract data type (ADT) is a specification for the operations available to access and modify the underlying data. The specification is written in terms of the abstract state, which is how the client should think about the object. Data abstraction means only accessing a type via its specification, which helps isolate and hide the details of how the data is actually represented from the rest of the program.

An ADT allows the programmer to isolate and minimize the part of the code that directly depends on the details of how the data is represented. To ensure the rules are followed, we only need to test and reason about the ADT itself, not the rest of the program. Similarly, if we change our mind about how to represent the data, only the ADT implementation needs to change, not the rest of the program.