Stop-list Slicing

Abstract

Traditional program slicing requires two parameters: a program location and a variable, or perhaps a set of variables, of interest. Stop-list slicing adds a third parameter to the slicing criterion: those variables that are not of interest. This third parameter is called the stoplist. When a variable in the stop-list is encountered, the data-flow dependence analysis of slicing is terminated for that variable. Stop-list slicing further focuses on the computation of interest, while ignoring computations known or determined to be uninteresting. This has the potential to reduce slice size when compared to traditional forms of slicing. In order to assess the size of the reduction obtained via stop-list slicing, the paper reports the results of three empirical evaluations: a large scale empirical study into the maximum slice size reduction that can be achieved when all program variables are on the stop-list; a study on a real program, to determine the reductions that could be obtained in a typical application; and qualitative case-based studies to illustrate stop-list slicing in the small. The large-scale study concerned a suite of 42 programs of approximately 800KLoc in total. Over 600K slices were computed. Using the maximal stoplist reduced the size of the computed slices by about one third on average. The typical program showed a slice size reduction of about one-quarter. The casebased studies indicate that the comprehension effects are worth further consideration.