Automatic Application of the Data-State Model in Data-Flow Contexts

The data-state and data-flow models of information visualization are known to be expressively equivalent. Each model is most effective for different combinations of analysis processes and data characteristics. Visualization frameworks tend to either (1) work within a single model or (2) permit either model in separate sub-frameworks. In either case, converting between the two models falls entirely to the programmer. The theoretical basis for automatic translation between the two models was established by Chi. However, that process is insufficiently specified to be directly implemented. This paper characterizes the practical advantages of the data-state model. This is used to identify when such a transformation is beneficial. It then expands on Chi's theoretical framework to provide the tools for translating visualization program fragments from the data-flow to the data-state model. A partial implementation of the expanded theory is described for the Stencil visualization environment.