Articles

Many of you are probably familiar with basic test techniques like equivalence partitioning and boundary value analysis. In this article, Rex presents an advanced technique for black-box testing called domain analysis. Domain analysis is an analytical way to deal with the interaction of factors or variables within the business logic layer of a program. It is appropriate when you have some number of factors to deal with. These factors might be input fields, output fields, database fields, events, or conditions. They should interact to create two or more situations in which the system will process data differently. Those situations are the domains. In each domain, the value of one or more factors influences the values of other factors, the system's outputs, or the processing performed.

In some cases, the number of possible test cases becomes very large due to the number of variables or factors and the potentially interesting test values or options for each variable or factor. For example, suppose you have 10 integer input fields that accept a number from 0 to 99. There are 10 billion billion valid input combinations.

Equivalence class partitioning and boundary value analysis on each field will reduce but not resolve the problem. You have four boundary values for each field. The illegal values are easy, because you have only 20 tests for those. However, to test each legal combination of fields, you have 1,024 test cases. But do you need to do so? And would testing combinations of boundary values necessarily make for good tests? Are there smarter options for dealing with such combinatorial explosions?

This article was originally published in Quality Matters.