A Genetic Programming Approach to Automated Test Generation for Object-Oriented Software

Abstract

This article proposes a new method for creating testsoftware for object-oriented systems using a geneticprogramming approach. It is believed that this approachis advantageous over the more established search-basedtest-case generation approaches because the testsoftware is represented and altered as a fullyfunctional computer program. Genetic programming (GP)uses a tree-shaped data structure which is moredirectly comparable and suitable for being mappedinstantly to abstract syntax trees commonly used incomputer languages and compilers. These structures canbe manipulated and executed directly, bypassingintricate and error prone conversion procedures betweendifferent representations. In addition, tree structuresmake more operations possible, which keep the structureand semantics of the evolving test software betterintact during program evolution, compared to linearstructures. This speeds up the evolutionary programgeneration process because the loss of evolvedstructures due to mutations and crossover is preventedmore effectively.