580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can dynamic program slicing be adapted for complex programming paradigms such as feature-oriented and object-oriented programs?
This research area focuses on developing dynamic slicing algorithms and intermediate program representations tailored to advanced programming paradigms that introduce complexities beyond traditional procedural programs. These paradigms include feature-oriented programming (FOP), which involves mixin layers and refinements, as well as object-oriented programming (OOP) with mechanisms like inheritance and friend functions. Dynamic slicing in this area matters because it enables precise debugging, testing, and maintenance in software systems with modular features and OO abstractions, overcoming limitations of classical slicing that may yield imprecise or incomplete slices.
Key finding: Proposes the ETBFODS dynamic slicing algorithm specifically designed for feature-oriented programs by constructing a Dynamic Feature Composition Dependence Graph (DFCDG) that encodes unique FOP characteristics such as mixin... Read more
Key finding: Develops dynamic slicing techniques for complex C++ programs by focusing on object-oriented features such as inheritance and friend functions, which affect code reuse and program dependencies. Using program dependence graphs... Read more
2. What are effective methodologies for advanced program slicing combining slicing and program transformation semantics?
This theme investigates formal foundations and practical implementations of program slicing through semantic and transformational approaches using rigorous mathematical models. Program slicing is viewed as program transformations ensuring semantic equivalence or refinement, leading to correctness guarantees. This includes semantic slicing, conditioned slicing, and amorphous slicing—all aiming to produce executable slices that preserve essential program behavior under various criteria and inputs. This focus matters for establishing soundness and enabling advanced slicing functionalities such as slicing with specifications and conditions for use in reverse engineering, program comprehension, and re-engineering.
Key finding: Presents a rigorous formalization of program slicing within the Wide Spectrum Language (WSL) transformation theory, defining slicing as a program transformation that preserves semantic equivalence. Introduces semantic... Read more
Key finding: Introduces ConSUS, a conditioner system for WSL programs enabling conditioned slicing that augments slicing criteria with initial state conditions to prune infeasible paths before static slicing. ConSUS employs symbolic... Read more
Key finding: Outlines the theoretical foundation of WSL transformation theory, emphasizing the importance of formal semantic definitions and correctness proofs in program transformations such as slicing. Demonstrates that program... Read more
Key finding: Applies the improved formalism of WSL-based semantic slicing to a challenging slicing problem, illustrating that minimal semantic slices can be derived for any program with given slicing criteria. The study presents both... Read more
3. How can dynamic slicing address omission bugs to improve fault localization in practice?
This theme investigates the limitations of classical dynamic slicing when confronting omission bugs—faults caused by missing execution of code—and develops approaches for overcoming its dead ends. By conducting empirical studies on bug repositories and analyzing the causes and fixes of omission bugs, this research aims to extend dynamic slicing techniques to localize and recommend relevant suspicious code regions that standard dynamic slicing misses. This matters because omission bugs are prevalent and challenging for automated debugging, requiring refined techniques to enhance fault localization accuracy and developer productivity.
Key finding: Provides an empirical study showing omission bugs comprise 46.4% of bugs in Defects4J and identifies patterns in their causes and fixes. Proposes a neural network model trained on mutated omission bugs to recommend candidate... Read more