Reengineering the Modularity of OO Systems

L'Objet, 2001

Modularization has many designations in the object-oriented world although it usually corresponds to system partitioning in groups (clusters) of classes. The problem is that this partitioning can be accomplished in many different ways, with two degrees of freedom: the number of clusters and the way to group classes to build them. We use Cluster Analysis to derive the optimal grouping for a fixed number of clusters. To find the optimal number of clusters we propose a decision criterion based on the definition of a modularization merit factor. We use this theoretical framework in the MOTTO tool and apply it to a reasonably large sample of software systems. The most important results are that (i) the proposed approach is feasible and easy to apply with appropriate tool support and that (ii) analyzed systems are far from fully exploiting the benefits of modularization. RÉSUMÉ: La modularisation peut avoir plusieurs sens dans le monde de l'orienté objet, même si elle correspond habituellement une division d'un système en groupes de classes. Le problème est que cette division peut être accomplie de différentes manières enjouant sur le nombre de groupes et la façon de regrouper les classes. Nous utilisons une méthode de regroupement automatique pour déterminer le regroupement optimal pour un nombre fixe de groupes. Pour trouver le nombre optimal de groupes, nous proposons un critère de décision basé sur la définition d'un facteur de mérite de la modularisation. Nous utilisons ce cadre théorique dans l'outil de MOTTO et nous l'appliquons à un échantillon raisonnablement grand de systèmes logiciels. Les résultats les plus importants sont que (i) l'approche proposée est faisable et est facile à appliquer avec le support d'un outil approprié et que (ii) les systèmes analysés sont loin d'exploiter pleinement les avantages de la modularisation KEY WORDS: Modularity of object-oriented systems, cluster analysis, modularity assessment, MOTTO tool MOTS-CLÉS: Modularité des systèmes à objets, regroupement automatique, évaluation de la modularité, l'outil MOTTO

IEEE Transactions on Software Engineering, 2000

The metrics formulated to date for characterizing the modularization quality of object-oriented software have considered module and class to be synonymous concepts. But, a typical class in object-oriented programming exists at too low a level of granularity in large object-oriented software consisting of millions of lines of code. A typical module (sometimes referred to as a superpackage) in a large object-oriented software system will typically consist of a large number of classes. Even when the access discipline encoded in each class makes for "clean" class-level partitioning of the code, the intermodule dependencies created by associational, inheritance-based, and method invocations may still make it difficult to maintain and extend the software. The goal of this paper is to provide a set of metrics that characterizes large object-oriented software systems with regard to such dependencies. Our metrics characterize the quality of modularization with respect to the APIs of the modules, on the one hand, and, on the other, with respect to such object-oriented intermodule dependencies as caused by inheritance, associational relationships, state access violations, fragile base-class design, etc. Using a two-pronged approach, we validate the metrics by applying them to popular opensource software systems.

Software Metrics Symposium, 1996., …, 1996

This paper describes the results of a study where the impact of Object-Oriented design on software quality characteristics is experimentally evaluated. A suite of metrics for OO design, called MOOD, was adopted to measure the use of OO design mechanisms. Data collected on the development of eight small-sized information management systems based on identical requirements were used to assess the referred impact. Data obtained in this experiment show how OO design mechanisms such as inheritance, polymorphism, information hiding and coupling, can influence quality characteristics like reliability or maintainability. Some predictive models based on OO design metrics are also presented.

Proceedings of TOOLS'96, 1996

The main mechanisms of the Object-Oriented paradigm are supposed to help produce better and cheaper software. As with other things, their usage is more or less intensive, depending mostly on the designer knowledge, experience and ability. This paper reviews a set of metrics called MOOD, suited for evaluating the use of those mechanisms. A MOOD-to-Eiffel binding is introduced. Some code fragments are presented to illustrate the concepts and to clarify the measurement process. A sample of Eiffel libraries is used to collect these metrics. Statistical analysis is performed on the sample and some hypotheses are drawn and discussed. Some preliminary heuristics that can be used during the design process are then derived. Those heuristics can be of some help to designers, mainly if embedded in CASE tools. 1 -It was recognised [Conner94] that, on average, students with some degree of proficiency in another paradigm (e.g. imperative) had more difficulties learning how to design the OO way, than those doing it from the start (i.e. with no previous software development knowledge whatsoever). The explanation was that students had to unlearn habits. 2 -In other related areas such as OO reuse metrics, some interesting experimental validation studies were conducted [Melo95], [Lewis91].

2011

Conclusion: This work contributes with evidence to fill gaps in the body of quantitative results supporting alleged benefits to software modularity brought by AOP languages, namely OT/J.

4th European Conference on Software Maintenance and Reengineering (CSMR'2000), 2000

This paper describes a validation experiment of a quantitative approach to the modularization of object oriented systems. The approach used is based on Cluster Analysis, a statistical technique used in many fields of science to group items. Here, the clusters are modules and the items are classes. A sample of some relatively large object oriented systems was used in this experiment. The calculation of the dissimilarity between classes is based on their relative couplings combined through six different rating schemes. These couplings are classified according to a taxonomy framework where categories were assigned weights. The coupling data were obtained with the MOODKit G2 tool. The results obtained allow to conclude on the applicability of the proposed approach. This work was developed in the realm of the MOOD Project that aims to deliver a quantitative framework to support the design of object oriented systems.

Journal of Object Technology, 2006

The importance of software measurement is increasing leading to development of new measurement techniques. As the development of object-oriented software is rising, more and more metrics are being defined for object-oriented languages. Many metrics have been proposed related to various object-oriented constructs like class, coupling, cohesion, inheritance, information hiding and polymorphism. The applicability of metrics developed by previous researchers is mostly limited to requirement, design and implementation phase. Exception handling is a desirable feature of software that leads to robust design and must be measured. This research addresses this need and introduces a new set of design metrics for object-oriented code. Two metrics are developed that measure the amount of robustness included in the code. The metrics are analytically evaluated against Weyuker's proposed set of nine axioms. These set of metrics are calculated and analyzed for standard projects and accordingly ways in which project managers can utilize these metrics are suggested. SOFTWARE DESIGN METRICS FOR OBJECT-ORIENTED SOFTWARE 122 J OURNAL OF OBJECT TECHNOLOGY V OL. 6, NO. 1

… Metrics Symposium, 1999 …, 1999

As object-oriented (OO) analysis and design techniques become more widely used, the demand on assessing the quality of OO designs increases substantially. Recently, there has been much research effort devoted to developing and empirically validating metrics for OO design quality. Complexity, coupling, and cohesion have received a considerable interest in the field. Despite the rich body of research and practice in developing design quality metrics, there has been less emphasis on dynamic metrics for OO designs. The ...

To cope with the complexity of large object-oriented software systems, developers organize classes into subsystems using the concepts of module or package. Such modular structure helps software systems to evolve when facing new requirements. The organization of classes into packages and/or subsystems represents the software modularization. the software modularization usually follows interrelationships between classes. Ideally, packages should to be loosely coupled and cohesive to a certain extent. However, Studies show that as software evolves to meet requirements and environment changes, the software modularization gradually drifts and looses quality. As a consequence, the software modularization must be maintained. It is thus important to understand, to assess and to optimize the organization of packages and their relationships. Our claim is that the maintenance of large and complex software modularizations needs approaches that help in: (1) understanding package shapes and relationships; (2) assessing the quality of a modularization, as well as the quality of a single package within a given modularization; (3) optimizing the quality of an existing modulariza- tion. In this thesis, we concentrate on three research fields: software visualizations, metrics and algorithms. At first, we define two visualizations that help maintainers: (1) to understand packages structure, usage and relationships; (2) to spot patterns; and (3) to identify misplaced classes and structural anomalies. In addition to visual- izations, we define a suite of metrics that help in assessing the package design quality (i.e., package cohesion and coupling). We also define metrics that assess the quality of a collection of inter-dependent packages from different view points, such as the degree of package coupling and cycles. Finally, we define a search-based algorithm that automatically reduces package coupling and cycles only by moving classes over existing packages. Our optimization approach takes explicitly into account the origi- nal class organization and package structure. It also allows maintainers to control the optimization process by specifying: (1) the maximal number of classes that may change their packages; (2) the classes that are candidate for moving and the classes that should not; (3) the packages that are candidate for restructuring and the packages that should not; and (4) the maximal number of classes that a given package can entail. The approaches presented in this thesis have been applied to real large object- oriented software systems. The results we obtained demonstrate the usefulness of our visualizations and metrics; and the effectiveness of our optimization algorithm.

2008 12th European Conference on Software Maintenance and Reengineering, 2008

Refactoring, in spite of widely acknowledged as one of the best practices of object-oriented design and programming, still lacks quantitative grounds and efficient tools for tasks such as detecting smells, choosing the most appropriate refactoring or validating the goodness of changes. This is a proposal for a method, supported by a tool, for cross-paradigm refactoring (e.g. from OOP to AOP), based on paradigm and formalism-independent modularity assessment.