An Empirical Study On Visualized Assistance For Developer Testing

Software testing is a significant process to guarantee the quality of a software,among which the developer testing plays an important role as a premier method.Developer testing means testing invoked by developers during their developing process,to guarantee the quality and to fix the defects.Such behavior premeates their daily development processes,including unit testing,software debugging,etc.To improve the efficiency and effectiveness,researchers proposed plenty of techniques to aid their behaviors during developer testing,such as coverage support in unit testing and fault localization techniques in debugging process.However,these techniques are not widely used in industry and there lacks empirical study or intuitive method to illustrate the effect and give some explanation.To explore the root cause,we involved two conventional assistant method,coverage testing and fault localization techniques,revealed their results towards developers by visualization,and meanwhile collected their behaviors.We further applied these techniques to a online training platform called Mooctest and conducted an empirical study among it.The experimental results showed that the assistant techniques can indeed facilitate the effectiveness during developer testing,yet hinder the efficiency.To move forward,we further analyzed the developer behavior during unit testing and debugging and mapped them into "First Scan" and "Follow-up Browsing" pattern.Through these patterns,we validate the positive and negative influence aroused by assistant techniques and gave an explanation.