29 April 2010
Mining design intensions with FCA (Formal Concept Analysis)
Meeting Room 10, 2nd Floor, JLB
12:30pm - 13:45pm
During the maintenance phase, developers must understand the constraints and contracts between the components of the system, to avoid architectural degradation.
However, many of these constraints are implicit or their documentation outdated.
To overcome this problem, we propose an approach to identify implicit regularities in the source code (e.g. naming conventions, design patterns, idioms, protocols of usage, etc.). This approach is based on Formal Concept Analysis and association rules. We extract from a static analysis of the source-code a set of facts that are classified using FCA. Based on this classification, we can detect related properties and exceptions using association rules. By relating properties, we are closer to make explicit the intension (as the specification of a set, not what the developer meant) of the design rules. The uses of regularity mining include documentation, guidance while modifying the code, and warnings when creating irregularities.
Angela Lozano is a post-doc researcher at the Université catholique de Louvain (Belgium). Currently, she is working on mining source code to detect and improve variability. She obtained her PhD at the Open University in 2009. Her PhD thesis proposes a method to assess the effect of source code characteristics on the changeability of source code entities; in particular, the effect of clones on methods. She obtained her master's degree from Vrije Universiteit of Brussels in the context of the European Master on Object Oriented Software Engineering program at 2003. Her master thesis proposed how to transform java source code into directed graphs. Her interests include source code analysis, refactoring, evolution, software metrics, bad smells, and software design principles.
Save to your Calendar