Análise de correlação entre métricas de evolução e qualidade de design de software.

Abstract

We have studied the evolution of eight open source projects and five proprietary ones, looking for statistical correlations between complexity and quality measures in terms of bad smells and bugs detected. In all projects, we found strong statistical correlations between complexity (WMC) and quality measures. In all the legacies softwares and five of open sources, the correlation can be considered very strong (r > 0.9). Surprisingly, in three of the open source, the correlation is strong, but negative. This has been attributed to the fact that, in these projects, designers have intentionally controlled the quality measures under study, by applying refactoring strategies. These results suggest that, despite the correlation, there is no necessary cause-effect relation between complexity and quality measures. We conclude that just eliminate bad smells is not a good strategy to be followed if the desired objective is to reduce software design complexity. Then also does not improve software quality in terms associated to software complexity reduction.