UML activities at runtime: Experiences of using interpreters and running generated code

Abstract

The execution semantics of activities in UML is based on a token flow concept. As flows from a source to a target may contain control nodes and thus tokens may flow to different targets depending on other concurrent flows or on guards annotated to edges, the computation of possible flows is complex. Rules defining when tokens may traverse an edge can be (and most often are) implemented in interpreters. Generating code is possible, too, but it is rarely seen in academic as well as in commercial tools. However, the compilation of activities to code may speed up the execution of activities. In this paper, we present an interpreter for activities, an enhanced interpreter using static analysis of activities before executing them as well as a code generation approach. We compare these different techniques with regard to runtime behavior and consumption of resources. © 2010 Springer-Verlag.