Modeling Design Patterns for Semi-Automatic Reuse in System Design

Introduction

Patterns are types of themes that specify recurring processes, events, or elements. Commonly used in different engineering fields, patterns can be classified into analysis patterns, design patterns, organization patterns, process patterns, and domain-specific patterns. Batra (2005) claimed that pattern recognition can be considered as a conceptual modeling technique. He further stated that the main challenge with patterns is “to provide a limited number of patterns at a fairly high level of abstraction and an appropriate level of granularity…” (Batra, 2005, p. 87). Identifying locations in which patterns might be applied can be a tedious task (Eden, 2002). Purao, Storey and Han (2003) described a prototype for automatically generating a conceptual design based on analysis patterns. Their research, which concerns high level analysis and design, aims at producing models automatically from textual requirements. Blomqvist (2005) presented an approach for creating ontology patterns semi-automatically by utilizing knowledge organized as patterns from other areas, like data modeling, knowledge reuse, software analysis, and software design. While these approaches are appropriate for high-level conceptual modeling activities, they are less applicable at the detailed design stage. As the level of design becomes more specific, different parameters, such as code optimization, system limitations, and other non-functional requirements, have to be taken into consideration. These require careful and detailed descriptions of the patterns, their essence, and their exact usages.

In this work, we concentrate on design patterns, which describe reusable solutions for recurring design problems in given contexts (Gamma, Helm, Johnson & Vlissides, 1994). While design patterns may help produce better design and implementation of applications (Prechelt, Unger, Philipssen & Tichy, 2002; Schmidt, 1995), their appropriate use is often hindered due to the inherent ambiguity in the existing ways of description and representation (Taibi & Ngo, 2003). This may impede effective use of design patterns in particular applications, as designers may not be able to appreciate benefits or predict shortcomings associated with their proper use (Dong, Yang & Zhang, 2007). Misuse of design patterns typically results from failure to understand the rationale behind the patterns (Wendorff, 2001) or from difficulties in incorporating the patterns into a specific system design (Abdul Jalil & Azman Mohd Noah, 2007).

In view of these observations, we have developed an approach, accompanied by a tool, for clearly modeling design patterns and applying them semi-automatically to system models. In doing so, we utilize knowledge we have gained from modeling and categorizing design patterns. Our approach supports modeling of the different aspects of design patterns, including their problem and solution specifications, their essence, and their correct usage in particular systems. The design pattern models are specified using Object-Process Methodology (OPM) (Dori, 2002), which enables describing structural and behavioral aspects of design patterns at different granularity and abstraction levels. These design pattern models provide the basis for an algorithm for searching design problems in a given system design and suggesting corrections for them in the form of design patterns. Evaluating our approach and the feasibility of its algorithm using a case study research methodology, we found that the suggested approach successfully locates the main design problems modeled by the selected design patterns.

The structure of the rest of the paper is as follows. First, the relevant literature is reviewed, followed by a section that introduces the suggested design patterns representation method. Then, the algorithm for improving a system’s design via the use of design patterns is presented, followed by a report on the evaluation of the approach. Finally, we summarize and propose possible future research directions.