메트릭기반 소프트웨어의 디자인 결점 자동 탐지

Abstract

With the popularity of object-oriented technology, inexperienced developers write low quality codes, which lead to inefficient software and decrease its maintainability and extensibility. It turns out to be an urgent issue regarding how to effectively search and locate software design problems. As for this issue, this thesis designs and implements an automatic tool, which aims to find object-oriented design flaws based on metrics. No measurement, no control. Software measurement is the foundation for software recognition, assessment, prediction and improvement, and thus ensures the quality of software. Since software metrics covers lots of aspects of software engineering, this paper focuses on Java source code metrics quality analysis. As for the selection of metrics, this paper uses traditional ob-ject-oriented metrics, which has been proved to be the “weatherglass” in practice that mostly reflects the design problems. Meanwhile, in order to accommodate the update of Java language, this paper considers the design problems related with new features of Java such as generic abuse. The tool is implemented based on the Eclipse platform and has been released as an Eclipse Plug-in. It realizes the automatic quality analysis of source codes. Compared with other related work, the tool has the following advantages: the upgrade of Java language will not affect the usage of the tool; sources related to the new features of Java can be measured; the report of measured results can be displayed in various formats.