Detecting architecturally-relevant code smells in evolving software systems

Refactoring tends to avoid the early deviation of a program from its intended architecture design. However, there is little knowledge about whether the manifestation of code smells in evolving software is indicator of architectural deviations. A fundamental difficulty in this process is that developers are only equipped with static analysis techniques for the source code, which do not exploit traceable architectural information. This work addresses this problem by: (1) identifying a family of architecturally-relevant code smells; (2) providing empirical evidence about the correlation of code smell patterns and architectural degeneration; (3) proposing a set of metrics and detection strategies and that exploit traceable architectural information in smell detection; and (4) conceiving a technique to support the early identification of architecture degeneration symptoms by reasoning about code smell patterns.