A Feature Envy Detection Method Based on Dataflow Analysis

摘要

Feature Envy is a code smell indicating that a particular class is showing too much interest in the methods/attributes of another class. Several feature-envy detection approaches have been proposed. However, these approaches consider an entire method as a unit for detection. When a method is lengthy, the exact location of the problematic statement may not be immediately obvious, and when a method mixes several kinds of behaviors, these approaches could be easily fooled. This paper proposes a characterization of feature envy and a detection approach. A tool, called FEED (FEature Envy Detector), based on dataflow analysis is developed to perform feature-envy detection. In comparison to previous approaches, the proposed approach offers a better detection granularity and also provides a better detection accuracy.