Toward visual programming languages for steering scientific computations

摘要

Imagine running a computationally intensive model and being able not only to visualize the data as soon as results start emerging, but also to experiment with and change the underlying calculations in midstream, all with a consistent visual interface. This scenario represents a marriage of these two areas: visual programming languages and steering. We classified these two areas of research in six dimensions to provide insights and to suggest future research directions. The first four measure how thoroughly a system fulfils significant attributes of the requirements of a scenario (which is presented). The fifth and sixth dimensions help clarify the advantages and disadvantages of different approaches by looking at the ways they use various programming paradigms. The dimensions are: the system's steering capabilities; the power and visual extent of the interface; the level of support for preexisting scientific application programs; system generality; the programming paradigm used for scientific programming; and the programming paradigm used for visualization and steering.