Visual Language Editor Toolkit

Researchers: John Reekie---Postdoctoral Researcher
Advisor:Edward A. Lee
Sponsor:
This work aims to make it easy to generate visual language editors. We treat a visual program as a ``picture.'' Pictures are hierarchically composed of simpler pictures using picture combinators: examples includes aggregation of pictures, containment of one picture by another, and connection by arrowed lines.

These combinators can be represented by visual ``meta-syntax'' that glues together smaller bits of visual syntax. This is analogous to the use of meta-syntax like [...] and {...} in BNF grammars. We plan to use structural inference rules (with pictures) to specify visual syntaxes.

Our approach to implementing a visual language editor is directly implement each picture combinator as a class. Together with a set of classes for ``primitive'' pictures (lines, boxes, labels etc), these classes form a framework that will (we hope) make it quite easy to construct an editor once given its syntax in terms of pictures and picture combinators. I'm currently working on a prototype visual language editing framework in [incr Tcl]/Tk.

Unfortunately, the syntactic approach is no help for specifying user interaction. Our first stab was based on the Tk event model, but this is quite low-level. We are now using an approach based on Brad Meyer's interactors, and have found that this cleans up interaction code considerably, making it both modular and reusable.