GVU Technical Report Number:
GIT-GVU-97-25
Title:
Automatic Generation of Interactive Systems from Declarative Models
Authors:
Kurt Stirewalt
Abstract:
This dissertation applies formal methods to the automatic generation of
interactive systems from multiple declarative models. We are interested in
two kinds of models: user-task and presentation.
The MASTERMIND Dialogue Language (MDL) is presented.
MDL is a notation for describing interactive system behavior in terms
of user tasks. MDL task models are expressed independently of
other models, like presentation, but are later composed with the behavior
of these other models. Such separation is important for preserving the
integrity of models over the lifetime of a system. The technical challenge
in this approach is to generate code that combines the functionality of
task and presentation models without violating that integrity.
To meet this challenge we implement MDL task models as run-time
dialogue constraints that synchronize with presentation
components. The constraint engine is implemented as a virtual machine that
simulates the execution of tasks and resolves the dependencies that arise
as a result of task and presentation model composition.
To simplify the generation process, a toolkit of reusable run-time
components is provided. Each component in this toolkit implements an MDL
operator, and components aggregate into trees whose structure corresponds
one-to-one with the abstract syntax tree of a corresponding MDL model.
Thus implementations can be generated through a simple syntactic
transformation of the MDL source code. The design correctness of these
components is validated by a novel application of symbolic model checking.
The run-time attributes of systems generated using this approach are
measured, and we conclude that this strategy of model composition is
feasible for use with real interactive systems.
Keywords:
Interactive systems, declarative models, MASTERMIND
You can access this technical report via:
PDF
Postscript
|
