Evaluación de código mediante múltiples intervalos de métricas

Résumé

Code evolution and maintenance include measurement activities to detect possible flaws and to suggest improvements. An approach to detect anomalies on code entities is based on metric values outside their thresholds. From the 90s, there are plenty of papers that propose general threshold values for a set metrics. One of the criticisms outlined in the literature to this approach is the difficulty of using these results to other contexts. Threshold values may be influenced by multiple variables, including the nature of the problem solved: exception, graphical interface, model and test. This information can be introduced by the user in the measurement process, making it semiautomatic. To verify this hypothesis, this paper describes a case study where the measurement process in guided by the inspector/assessor, who classifies code entities to measure depending on their nature. The case study result suggests some metric thresholds for different kind of problems and relative to a specific organization.