A simple static model for understanding the dynamic behavior of programs

Abstract

To maintain software and to adapt it to changing requirements, one needs to have a solid understanding of both its structure and behavior. While there are a number of reverse engineering tools that aid in understanding the runtime behavior of programs, these are mostly based on variants of UML sequence diagrams or statechart diagrams. We propose a new model for understanding runtime behavior that presents several advantages over the more traditional models: it has a simple syntax (a very small subset of UML). Second one can tailor the same type of model to different abstraction levels while maintaining traceability. Third, they have a clearly defined semantics that makes them executable. Finally they capture both static and dynamic aspects of a system. We present three scenarios where our model - named EOP-model - aids with program comprehension: (1) high-level debugging, with which one can observe the execution of the program at a higher level of abstraction; (2) high-level program slicing, which allows to identify the data items in our model that may influence an operation, regardless of the abstraction level, and (3) generating abstract views at varying levels of detail while maintaining traceability between model elements and the code.