Reducing Subjectivity in Code Smells Detection: Experimenting with the Long Method

Guidelines for refactoring are meant to improve software systems internal quality and are widely acknowledged as among software´s best practices. However, such guidelines remain mostly qualitative in nature. As a result, judgments on how to conduct refactoring processes remain mostly subjective and therefore non-automatable, prone to errors and unrepeatable. The detection of the Long Method code smell is an example. To address this problem, this paper proposes a technique to detect Long Method objectively and automatically, using a Binary Logistic Regression model calibrated by expert´s knowledge. The results of an experiment illustrating the use of this technique are reported.