A framework for component mining of Java applications via dynamic slicing

Abstract

This paper explores the use of program slicing as a tool for "component mining" of Java/spl trade/ source code. We define component mining to be the extraction of an executable slice from source code, which satisfies a specific use case (or set of use cases) and provides a standard component interface for its use. However, before a component can be generated, the desired features themselves must be isolated from the source code. Traditionally, software slicing has concerned itself with the value of a variable of interest (or set of variables) at a specific point of execution. This severely limits the usefulness of the traditional definition of a slicing criterion as a tool for feature isolation. We propose the repurposing of software "unit-tests" to aid in the isolation of features of interest within source code. By executing a target application in our JPDA (Java platform debugging architecture) based slicer, a selected unit-test may then serve as the slicing criterion.