Targeted adequacy criteria for search-based test data generation

Abstract

In test data generation approaches, structural coverage criteria are popular and considered cost effective mainly because they are relatively easy to implement. However, we cannot establish a direct link between achieving good structural coverage and test goals such as finding errors. On the other hand, specific criteria can explicitly target such goals but they may require a full formal specification of the program, which makes them expensive and hard to use. In this paper, we propose a new approach to describe targeted test criteria using few structural information of the system under test. The new criteria can easily describe specific situations such as errors and can be transformed into a fitness function. A search technique, in this case ACO (Ant Colony Optimization), can then be used to automatically generate adequate test data. A case study is presented to illustrate the applicably and the usefulness of the approach.

Test data generation from Hibernate constraints

“…There are plethora of metaheuristic algorithms have been used in many researches to generate test data, genetic algorithm (GA) [5] , ant colony optimization (ACO) [6] and particle swarm optimization (PSO) [16]. As will be explained below, these algorithms are most widely used in many researches.…”

“…In ant colony optimization [6], the authors proposed an approach to describe targeted test criteria using constraints inside the program code. Testers are assumed that they have sufficient knowledge of the SUT structure.…”

“…Fortunately, Hibernate validator can be used for checking whether each data of the fields of an object satisfy the Hibernate constraints or not. At present, a lot of researches [5] [6] showed that test data can be generated from constraints with search techniques.…”