A comparison of CK and Martin's package metric suites in predicting reusability in open source object-oriented software

Packages are units that organize source code in large object-oriented systems. Metrics used at the package granularity level mostly characterize attributes such as complexity, size, cohesion and coupling. Many of these characterized attributes have direct relationships with the quality of the software system being produced. Empirical evidence is required to support the collection of measures for such metrics; hence these metrics are used as early indicators of such important external quality attributes. This research compared the CK and Martin’s package metric suites in order to characterize the package reusability level in object-oriented software. Comparing the package level of metrics suites as they measure an external software quality attribute is supposed to help a developer knows which metric suite can be used to effectively predict the software quality attribute at package level. In this research two open source Java applications, namely; jEdit and BlueJ were used in the evaluation of two package metrics suites, and were compared empirically to predict the package reusability level. The metric measures were also used to compare the effectiveness of the metrics in these package metrics suites in evaluating the reusability at the package granularity level. Thereafter metric measures of each package were normalized to allow for the comparison of the package reusability level among packages within an application. The Bansiya reusability model equation was adapted as a reusability reference quality model in this research work. Correlation analysis was performed to help compare the metrics within package metrics suites. Through the ranking of the package reusability level, results show that the jEdit application has 30% of its packages ranked with a very high reusability level, thus conformed to the Pareto rule (80:20). This means that the jEdit application has packages that are more reusable than packages in the BlueJ application. Empirically, the Martin’s package coupling metric Ce with an r value of 0.68, is ranked as having a positive strong correlation with RL, and this has distinguished the Martin’s package metrics suite as an effective predictor of package reusability level from the CK package metrics suite.

Name dc:type.qualificationname

mphil

Level dc:type.qualificationlevel

masters

Grantor dc:publisher.institution

Universiti Tun Hussein Onn Malaysia

Year dc:date.issued

2016

Author dc:creator

• Alhadi Meftah, Khaled

• QA76.75-76.765 Computer software

Language dc:language

en