The Smell of Blood: Evaluating Anemia and Bloodshot Symptoms in Web Applications

Abstract

In web applications that adopt layered architecture, Domain Layer is formed by Domain Model. Without any behavior, Anemic Domain Models contain only data. Those behaviors are dispersed into other layers and causing bloodshot symptoms in them. Most empirical studies suggest that symptoms of anemia and bloodshot degrade the maintainability of web applications, but no quantitative research has been done. This paper evaluates intensities of anemia and bloodshot symptoms based on metrics of three Code Smells, i.e. Data Class, Feature Envy and Blob. Furthermore, correlations of the intensities are evaluated using Spearman's rank correlation coefficient. The achieved results of experiments made on multiple versions of open-sourced projects show that over 65% of the applications are affected by anemia and bloodshot symptoms, and their intensities rarely decrease over time. Correlations of the intensities are also discovered within a single version and among multiple versions.