Visualization of automated program repair focusing on suspiciousness values

Automated program repair (APR) can realize efficient debugging in software development. Automated program corrections using genetic algorithms (GA) can repair programs, including those with multiple bugs, but the repair process of GA-based APR is difficult to understand using logs because many modification program codes are generated. Consequently, Matsumoto et al. implemented a methodology for visualizing the process. Their proposed methodology provides an intuitive understanding of the conformance values (test case pass rates), generations, states, and operations performed to generate each variant; however, it lacks sufficient information to analyze whether defect localization is appropriate in APR. Herein we propose a new methodology to visualize the impact of fault localization on program evolution in GA-based APR and create a new tool. Additionally, a case study demonstrates the effectiveness of the proposed methodology and future works are considered.