Extending dynamic aspect mining with static information

Abstract

Aspect mining tries to identify crosscutting concerns in legacy systems and thus supports the refactoring into an aspect-oriented design. We briefly introduce DynAMiT, a dynamic aspect mining tool that detects crosscutting concerns based on tracing method executions. While the approach is generally fairly precise, further analysis revealed that some false positives were systematically caused by dynamic binding. Furthermore, some aspect candidates were blurred or not detected due to not-sufficient tracing mechanisms of method executions when using AspectJ's execution pointcuts for the trace generation. We enhanced the mining capabilities of DynAMiT by taking additional static type information into account and generating the traces using call pointcuts instead. In an initial case study with AnChoVis, a 1300 LOC Java program, the number of mined aspect candidates increased by a factor of three, while the number of false positives remained zero.