Classifying Dead Code in Software Development

Abstract

Dead code pervades as an issue in the world of software development as a source of many famous software disasters such as the ARIANE 5 rocket failure and chemical bank withdrawal error. Defining dead code on narrow levels of granularity has not been fully explored, yet is crucial to better our understanding of dead code. Here we will be starting a discussion on how to approach classifying dead code via comparing dead code research specific to an industry segment. Research will be compared primarily by methodology and limitations. Dead code subtype classifications are gleaned from research comparisons and can serve as a methodology for identifying new dead code subtype classifications. Being able to identify dead code is a step in the right direction, but our approach in removing dead code must also be sound. In general, developers do not have enough resources to focus on eliminating dead code which is why automation is utilized instead. Automation in the form of static and dynamic code analysis, which are currently limited in dead code removal capability due to dead code’s abstractness. However, if we utilize our dead code subtype classifications, we can target specific types of dead code with static/dynamic code analysis that are more quantifiable (hence detectable by automation) than general dead code. In conclusion, if we classify dead code and specify what type of dead code we want to remove, we could find new ways to utilize static and dynamic code analysis in dead code elimination.