2026-09-26T00:25:27Zhttps://bura.brunel.ac.uk/oai/request

oai:bura.brunel.ac.uk:2438/45272016-01-27T09:08:31Zcom_2438_23com_2438_58com_2438_8630com_2438_8620col_2438_3672col_2438_210col_2438_8638

Counsell, S Hierons, RM Mubarak, Asma 2010-08-27T10:14:39Z 2010-08-27T10:14:39Z 2010 http://bura.brunel.ac.uk/handle/2438/4527 This thesis was submitted for the degree of Doctor of Philosophy and awarded by Brunel University. Excessive coupling between object-oriented classes in systems is generally acknowledged as harmful and is recognised as a maintenance problem that can result in a higher propensity for faults in systems and a „stored up‟ future problem. Characterisation and understanding coupling at different levels of abstraction is therefore important for both the project manager and developer both of whom have a vested interest in software quality. In this Thesis, coupling trends are empirically investigated over multiple versions of seven Java open-source systems (OSS). The first investigation explores the trends in longitudinal changes to open-source systems given by six coupling metrics. Coupling trends are then explored from the perspective of: the relationship between removed classes and their coupling with other classes in the same package; the relationships between coupling and 'warnings’ in packages and the time interval between versions in Java OSS; the relationship between some of these coupling metrics are also explored. Finally, the existence of an 80/20 rule for the coupling metrics is inspected. Results suggest that developer activity comprises a set of high and low periods (peak and trough‟ effect) evident as a system evolves. Findings also demonstrate that addition of coupling may have beneficial effects on a system, particularly if they are added as new functionality through the package Java feature. The fan-in and fan-out coupling metrics reveal particular features and exhibited a wide range of traits in the classes depending on their high or low values; finally, we revealed that one metric (fan-in) is the only metric that appears consistently to exhibit an 80/20 (Pareto) relationship. Submitted by Carolyn Bailey (carolyn.bailey@brunel.ac.uk) on 2010-08-27T10:14:21Z No. of bitstreams: 1 Fulltext.pdf: 2620047 bytes, checksum: cf55b175071e507c2745415a2261677f (MD5) Approved for entry into archive by Carolyn Bailey(carolyn.bailey@brunel.ac.uk) on 2010-08-27T10:14:39Z (GMT) No. of bitstreams: 1 Fulltext.pdf: 2620047 bytes, checksum: cf55b175071e507c2745415a2261677f (MD5) Made available in DSpace on 2010-08-27T10:14:39Z (GMT). No. of bitstreams: 1 Fulltext.pdf: 2620047 bytes, checksum: cf55b175071e507c2745415a2261677f (MD5) en Brunel University, School of Information Systems, Computing and Mathematics School of Information Systems, Computing and Mathematics http://bura.brunel.ac.uk/bitstream/2438/4527/1/FulltextThesis.pdf Software metrics Evolution Refactoring Power law Warning An empirical study of package coupling in Java open-source Thesis