Automatic layout of UML class diagrams : a topology-shape-metrics approach

Class diagrams are among the most popular tools for the analysis and the modeling of object-oriented software systems. They are ubiquotous in today's software development processes. The Unified Modeling Language (UML), which replaced earlier modeling languages, is the established standard notation for class diagrams. A class diagram describes the static relationships between classes and objects in a software system. There are three different relationships between classes in object-oriented systems: dependencies, associations and generalizations. The graphical representation of each of them follows its own conventions. Particularly generalizations play a special role and are represented generally as upward pointing polyline curves. In some scenarios the contents of a class diagram is known, but not its graphical representation, especially not the placement of the elements in the diagram. This is the case when the diagram hasn't been created interactively but comes from another source. Possible sources are applications for the automatic documentation of software or reverse-engineering tools. If a human beholder wants to gain insight from such a diagram, it has to be drawn in a readable fashion. In this work we will present new algorithms for the automatic and interactive layout of class diagrams. Since class diagrams can be modeled by graphs, we will use methods and techniques from the field of graph drawing. Especially we will use a technique known as the "topology-shape-metrics" approach. For this purpose we enhance the topology-shape-metrics approach to a practical tool which can consider the conventions for drawing class diagrams and extend all three phases of it. For each phase we present new algorithms along with an analysis of their performance. We show in an experimental evaluation, that the new algorithms outperform significantly previous approaches for the automatic layout of class diagrams. All extensions to the topology-shape-metrics approach are of a very general nature and apply also to other application domains than class diagrams. With these new algorithms we extend the number of application domains of the topology-shape-metrics approach dramatically.

Author

• Eiglsperger, Markus

Identifier

hdl:10900/48535

OAI record oai:identifier

https://publikationen.uni-tuebingen.de/oai/request?verb=GetRecord&metadataPrefix=mets&identifier=oai%3Apublikationen.uni-tuebingen.de%3A10900%2F48535