Supervisor

Suitable for

Abstract

Oege de Moor and I wrote a paper called Modeless Structure Editing for Tony Hoare’s retirement symposium. What we had in mind was a family of modeless editors whose underlying model of the document being edited is akin to its abstract syntax tree. It’s clear what this means if the document represents, for example, a program in a programming language or a proof in a proof calculus; but we conjecture that structured prose documents will also be amenable to this approach to a useful extent.

Editing an abstract syntax tree model does not mean that one must necessarily work with a “tree-shaped” view of the document.

The benefits of editing an abstract syntax tree model include:

• The ability to perform semantic checks incrementally and rapidly. Of course semantic checks are realized differently for different formalisms.
• The potential to interact with varieties of views of the document, and the potential to switch views rapidly.
• The potential to mix formalisms within a single document, and to derive a variety of artefacts from it. Imagine, for example, writing lectures about programs in a programming language: one wants (parts of) the program source code to appear in the lecture; and one wants the entire source code of the program to be derived from their “reference text” in the document.

Such an approach has the potential to yield dividends in performance and simplicity of implementation of incremental semantic checks, codegeneration, etc.

The first phase of this project will construct (in Scala) a prototype structure editing toolkit implementing the model explained in this paper. Subsequent phases will provide one or two case studies in its use. The architecture of an editor that could use such a document model already exists – in the shape of our AppleRed editor, whose document model is (in principle) “pluggable”, and that currently uses a linear model (document as a sequence of lines of text). What will be needed is to construct a new “plugin” conforming to the AppleRed document model interface, and to adapt and generalise that interface where necessary.