Program slicing to understand software generators

Many large mainframe based business applications were developed with proprietary macro-based generator systems. To maintain these applications the legacy generators employed to generate them have to be maintained as well. This is a problem because of the generator code that is often hard to understand. A solution would be to replace the old generator technology with a new one. However, this requires a deep understanding of the old generators and the features of the programs they generate. This understanding is currently not available. We propose program slicing as a technique to support program understanding of generators and show its application in four use cases in that domain. We show the feasibility of the technique using real-world generators.