Comparison of Backward Slicing Techniques for Java

Abstract

SUMMARYProgram slicing is an important approach for debugging, program comprehension, impact analysis, etc. There are various program slicing techniques ranging from the lightweight to the more accurate but heavyweight. Comparative analyses are important for selecting the most appropriate technique. This paper presents a comparative study of four backward program slicing techniques for Java. The results show the scalability and precision of these techniques. We develop guidelines that indicate which slicing techniques are appropriate for different situations, based on the results.

Variable Data-Flow Graph for Lightweight Program Slicing and Visualization

“…pointsto analysis and side effect analysis. Although a developer's typical working session time is between 30 minutes and two hours, making SDGs often consumes more than single ses- sion [4]. Second, existing research [5] shows that a straight graph visualization of program slices on an SDG for a typical program is too large and too complicated for a human to understand.…”

Analysis Techniques for Software Maintenance

“…They defined an algorithm to completely unfold object variables without unfolding the same object twice in the same object tree and without loosing dependences based on point-to information. Kashima et al [96] compared four different backward slicing techniques for Java: static execute before (SBE), based in the CFG; context-insensitive slicing (CIS), ignoring the call context; hybrid model (HYB), where the slice is defined as the intersection of SBE and CIS; and improved slicing (IMP), based on the Hammer and Snelting's work [75]. They compared their precision, scalability, and tradeoffs, determining IMP as the more accurate but not applicable to large programs.…”