Reasonable abstractions: Semantics for dynamic data visualization

Chi showed how to treat visualization programing models abstractly. This provided a firm theoretical basis for the data-state model of visualization. However, Chi´s models did not look deeper into fine-grained program properties, such as execution semantics. We present conditionally deterministic and resource bounded semantics for the data flow model of visualization based on E-FRP. These semantics are used in the Stencil system to move between data state and data flow execution, build task-based parallelism, and build complex analysis chains for dynamic data. This initial work also shows promise for other complex operators, compilation techniques to enable efficient use of time and space, and mixing task and data parallelism.