Abstract

Over the last decade, the technical debt metaphor has gained in popularity, and many tools exist today that can calculate the debt associated with a miscellany of source code. However, no corpus of studies has investigated the effects that creation and refactoring of conceptual models have on technical debt of corresponding code. Our work addresses this fundamental gap by first providing a map of correspondences between recognized model smells of UML Class Diagrams and Java source code issues. We then describe a set of empirical studies to calculate the technical debt of generated source code as a result of refactorings performed on their corresponding models. Our results reveal a significant disconnect between model smells and technical debt values of resultant generated source code, and little effect of model refactorings on reducing these values. However, once correspondences between model smells and code issues are defined, model refactoring proves helpful in preventing technical debt from a high abstraction level. We exemplify this scenario by providing an in-depth example, and conclude with a discussion of results.