Design pattern decay: a study of design pattern grime and its impact on quality and technical debt
Griffith, Isaac Daniel
MetadataShow full item record
Technical debt is a financial metaphor describing the trade-off between the short-term benefits gained and long-term consequences of design and implementation shortcuts taken over the evolution of a software product. These shortcuts typically manifest as design disharmonies such as code smells, anti-patterns, or design pattern grime. Design pattern grime, which manifests as the accumulation of unnecessary or unrelated software artifacts within design pattern instance classes is of serious concern. Design patterns represent agreed upon methods to solve common problems and are based upon sound principles of good design; thus, these pattern instances' decay implies an evolution away from good design. Currently, little is known about the causal nature of design pattern grime on technical debt and quality or how these three issues interrelate. What is the nature of the relationships between structural design pattern grime, software maintainability, and technical debt measurement? To better understand design pattern grime, we have extended the structural grime taxonomy. We developed an approach to generate design pattern grime instances and inject them with design pattern grime. Using this approach, we conducted 7 experiments evaluating the effects of 26 forms of grime, at 6 severity levels within 16 design pattern types, on software maintainability and technical debt. The results showed that depending upon grime type, grime severity, and pattern type, grime does significantly affect both maintainability and technical debt. We also conducted a verification study on pairs of pattern instances from open-source software systems to evaluate how well the injection process represents the real effects of grime and to verify the results of the experiments. The results of this study showed that there is a disconnect between the injection process and reality, indicating that refinements are still needed. However, the verification study worked as expected in indicating where issues may exist in the process.