Comparing Software Architecture Descriptions and Raw Source-Code: A Statistical Analysis of Maintainability Metrics

摘要

The software systems have been exposed to constant changes in a short period of time. It requires high maintainable systems and makes maintainability one of the most important quality attributes. In this work we performed a statistical analysis of maintainability metrics in three mainstream open-source applications, Tomcat (webserver), Jedit (text editor) and Vuze (a peer to peer client). The metrics are applied to source-code and to derived similar architectural metrics using scatter plot, Pearson's correlation coefficient and significance tests. The observations contradict the common assumption that software quality attributes (aka non-functional requirements) are mostly determined at the architectural level and raise new issues for future works in this field.