On Test Data Generation of Object-Oriented Software

Abstract

Nowadays, Object-Oriented (OO)

Integrating Evolutionary Testing with Reinforcement Learning for Automated Test Generation of Object‐Oriented Software

“…The former indicates how close the execution trace of an individual is to reaching the target branch, whereas the latter measures how far the execution trace of the individual takes the desired alternative branch at the first mismatching predicate. Quantitatively, the fitness function can be formulated by the approach level plus normalized branch distance [11] : fitness = approach level + norm(branch distance) (1) where the branch distance is normalized with the formula:…”

“…More specifically, in a GP T ree individual, three mutation strategies can be employed. 1 Promotion strategy moves a subtree of a parent to one descendant node with compatible return type and access scope. 2 Demotion strategy moves a subtree of a parent to one ancestor node with compatible return type and access scope.…”

“…Automated test generation of Object-oriented (OO) software has been a sustained topic of interest due to its benefit in terms of quality improvement and cost saving [1] . It aims at automatically creating a suite of test cases that achieves a specified coverage criterion, such as branch coverage for the Software under test (SUT), each test case consisting of a method call sequence that cover (or reach) a target branch.…”

Object-Oriented Evolutionary Testing

“…Even though the majority of the OO literature is devoted to the study of either GAs or GP, there are several studies that focus on distinct evolutionary techniques. In fact, as stated in (Arcuri & Yao, 2007c), other metaheuristic techniques have the potential to achieve promising results in this area.…”

“…An analysis of the current literature on OOET allows making some observations: nearly all studies have been developed with basis on the program's structure, with the objective being that of attaining a coverage criterion (usually statement or branch coverage). Java is clearly the programming language of choice for the purposes of implementation and experimentation; and even though several test objects have considered for experimentation purposes, nearly all works (and, in particular, those that do not use custom-made classes) employ container classes (e.g., Stack, BitSet, Vector, TreeMap) as a basis for their studies, mostly due to the lack of a common benchmark which can be used by researchers to test and compare their techniques (Arcuri & Yao, 2007c).…”