Validating New Method for Measuring Cohesion in Object-Oriented Projects

Abstract

Class cohesion, understood as the degree of how tightly are bound or related its internal elements to one another, is often a critically important for quality assurance (QA) of object-oriented software. Typically, a low cohesive class contains disparate and/or isolated members; therefore, cohesion level is useful for detecting of poorly designed classes and ensuring faster and better system reconfiguration. There are over thirty different metrics to measure cohesion, based on class member analysis in terms of number and structure of attributes and methods. Utilizing class cohesion metrics can promote Java code static analysis quality, improve object-oriented programming practices, and suggest advanced and efficient ones. The jPeek tool with its library for inspecting objects in Java was developed by Huawei Technologies Co. Ltd. to achieve these goals [1]. The idea of code quality assessment is well known for a long time; class connectivity metrics were proposed by community several years ago and have not become generally applicable practice in industrial programming. The main goal of this study is to make a methodology for using cohesion metrics to analyze code that allows users to estimate their projects in the most appropriate way. During the study, each of the metrics was studied in detail, the influence of programming patterns on the result of the metrics was evaluated, and as a result a methodology for using the metrics to obtain the most reliable estimate was formulated.