Applying evolutionary approaches to data flow testing at unit level

Abstract

Data flow testing is a white box testing approach that uses the dataflow relations in a program for the selection of test cases. Evolutionary testing uses the evolutionary approaches for the generation and selection of test data. This paper presents a novel approach applying evolutionary algorithms for the automatic generation of test paths using data flow relations in a program. Our approach starts with a random initial population of test paths and then based on the selected testing criteria new paths are generated by applying a genetic algorithm. A fitness function evaluates each chromosome (path) based on the selected data flow testing criteria and computes its fitness. We have applied one point crossover and mutation operators for the generation of new population. The approach has been implemented in Java by a prototype tool called ETODF for validation. In experiments with this prototype, our approach has much better results as compared to random testing. © 2011 Springer-Verlag.