Programming in several dimensions

Writing a program to solve a problem is a process that can be divided into two phases: first, we invent a mental model of the solution; secondly, we map the mental model onto a physical representation. The mental model is multidimensional and syntax-free; in today’s textual programming languages, the physical representation is singledimensional and syntax-burdened. In fact, it hasn’t changed greatly since Algol 60. Mapping from one representation to the other has remained a painstaking and error-prone task, in spite of the ready availability of immensely faster computers, massive amounts of memory, high-resolution graphics displays, and powerful graphic input mechanisms. The Hyperprogramming paradigm exploits these capabilities. A hyperprogramming language employs different visualisations for different program components for example one visual syntax is suitable for visualising algorithms and another is suitable for visualising subroutine nesting. Each visualisation is designed for minimal overlap with the others, and where overlap is essential, hyperlinks between the views are automatically provided to allow easy navigation between them, and automatic updating of shared information. HyperPascal was developed as a vehicle for exploring this idea. In creating a program, a HyperPascal programmer edits information in three main visualisations: • the action window visualisation, which represents algorithms using a visual language based on structure diagrams; • the data structure templates visual component, which represents dynamic data structure algorithms using beforeand-after pictures • the scope window visualisation, which represents declarations as a nested visualisation analogous to conventional subroutine nesting .