On the Relevance of Code Anomalies for Identifying Architecture Degradation Symptoms

The longevity of evolving software systems largely depends on their resilience to architectural design degradation. It is often assumed that code anomalies are always key indicators of architecture degradation symptoms. The problem is that there is still limited knowledge about the circumstances under which code anomalies represent architectural problems. Without this knowledge, developers are not able to implement architecturally-relevant strategies for code refactoring. This paper presents an empirical study about the influence of code anomalies on architecture degradation symptoms. To this end, we studied the relationship between code anomalies and architecture problems in 6 software systems, which were intended to adhere different architectural decompositions. A total of 40 versions and 2056 code anomalies were analyzed. Our study revealed that 78% of all architecture problems in the programs were related to code anomalies. In particular, more than 33% of all architecture problems were unexpectedly induced by anomalous code elements in the systems´ first version. We also found that the refactoring strategies, even when frequently applied in those systems, did not significantly contribute to remove architecturally-relevant code anomalies.