Kaynak Kodlardaki Kötü Kokuların Otomatik Tespiti İçin Eclıpse Eklentisi

Abstract

Code smells in source codes are code fragments that do not prevent the functionality of the developed application, but which reduce code quality, make code maintenance and understandability difficult and require refactoring. Those types of smells could be found in a class as a whole or in a specific method of a class. Detecting those code smells by manual reviewing is a process that could increase the probability of unintentional omission in terms of the requirement of time, budget, and manpower as the project grows. Code smells can be caused by errors in the design phase as well as by the developer’s preferences in the design to code conversion phase. In this article, we will introduce an Eclipse plugin that enables automatic detection of code smells in Java source code and presents the detected code smells to developers and maintainers. In this way, the software developers and maintainers can continuously evaluate the quality of the software with realistic values, recognize and refactor the modules that could cause a bug. This provides better quality, easier maintainability and effective testability in software. The developed plugin is tested on the data sets used in fault estimation, and statistical correlation between software fault and code smells is presented. According to results, existence of code smells is unrelated with the software faults. However, existing faults are statistically related with code smells.