Generic constraint logic programming and incompleteness in the analysis of software

Abstract

Current logic programming systems, as typified by Prolog, contain limitations which restrict their usefulness during the specification, design, and testing of software. A major limitation is the inability to perform analysis in the presence of incomplete information. Three sources of incompleteness are discussed. The result of this analysis suggests an efficient algorithm for a special case. Furthermore, it partitions the problem for testing purposes into three classes: (1) those points which satisfy the special case; (2) those which satisfy the general case but not the special case; and (3) those points which do not satisfy either. Thus, the analysis has uncovered possible structure within the implementation including the case in which the implementation has failed to address class 2 correctly (error of omission) but handle classes 1 and 3 correctly.<>