PROGRAMMING WITH CONSTRAINTS IN AN OBJECT-ORIENTED GIS 
FRANCOIS BOUILLE 
5 AFIN DE 
LIST f j ité P.M. j ; 
FORMATION DE S.T., Boîte 1000, Université P.M.Curie, 4 Place Jussieu 
75252 PARIS cédex 05, FRANCE 
ABSTRACT 
All geographical phenomenons may be easily structured in an object-oriented GIS. The HBDS model is one of 
the possibilities for doing that. Based upon six basical persistent abstract data types and their extensions, it allows 
to build a class-based integrated platform as a kernel for GIS design and development. It includes an expert 
system managing fuzzy rules and facts, and a neural engine allowing automated learning; but for expressing most 
of constraints in a GIS, the rules are not the universal tool, and the concept of "constraint" must be introduced as a 
new component which can be defined, handled and applied. It requires a special architecture and a specific 
vocabulary. This new tool must be used in cooperation with the others, such as the classical "condition" carried by 
any component, such as more complex tools i.e. the "rule" , the "process" and the "neuron", all concepts making a 
GIS fully dynamical. These constraints are particularly usefull when simultaneously dealing with several types; 
they may be strict or fuzzy; we may distinguish too constraints relevant to conditions which must be absolutely 
verified, and that relevant to conditions which must never happen. As all the components of the HBDS model, 
constraints have an intrinsic graphical representation, are organized into classes, present their own properties and 
relationships with other components and between contraints themselves. This new tool introduced in GIS provides 
a new way of expressing the constraints but requires to take care in programming. 
| tardive du 
RESUME 
Tout phénomène géographique est facilement structuré dans un SIG orienté objet. Le modèle HBDS représente 
l'une de ces possibilités. Fondé sur six types abstraits de données persistants et sur leurs extensions, il permet de 
construire une plateforme intégrée orientée objet, constituant un noyau pour la conception et le développement de 
SIG. Il comprend un systéme expert gérant des faits et régles flous, ainsi qu'un moteur neuronal permettant un 
apprentissage automatique; mais pour exprimer la plupart des contraintes dans un SIG, les régles ne constituent 
pas l'outil universel, et le concept de "contrainte" doit étre introduit comme un nouveau composant qui peut étre 
défini, manipulé et appliqué. Ceci demande une architecture spéciale et un vocabulaire spécifique. Ce nouvel outil 
doit étre utilisé en coopération avec les autres, tels que la classique "condition" qui peut étre portée par n'importe 
quel composant, tels que des outils plus complexes comme la "règle", le "processus" et le "neurone", tous ces 
composants donnant au SIG son aspect dynamique. Ces contraintes sont particuliérement utiles lorsque traitant 
simultanément de plusieurs types; elles peuvent être strictes ou floues,; l'on peut distinguer aussi les contraintes 
relevant de conditions qui doivent étre absoluement vérifiées et celles qui traitent de circonstances qui ne doivent 
jamais arriver. Comme tous les composants du modéle HBDS, les contraintes ont une représentation graphique 
intrinsèque, sont organisées en classes, presentent leurs propres propriétés et relations, soit avec les autres 
composants, soit entre contraintes elles mêmes. 
KEY WORDS: constraint programming, knowledge base, neural model, object-orientation. 
1. INTRODUCTION geographical model. That is why we propose to 
introduce in the GIS area a new way of 
Most of components in a GIS, for instance dealing programming by using constraints. 
With urban networks, are easily structured, captured, 
stored and processed in an object-oriented 2. BRIEF RECALL OF THE HBDS MODEL 
environment. Such a methodology allows to express 
most of properties and relationships between the HBDS means "Hypergraph-Based Data Structure; 
components better than a relational approach could early designed in year 1977 (Bouillé 77a, 77b, 79), 
do. Nevertheless, most of o.-o. models have some it was progressively developped for providing all the 
difficulties in expressing the constraints between the interconnected tools a GIS user may require. We 
geographical elements of a structure. Though the use may distinguish three things: ~the model which is 
of an expert system is an interesting way, rules do computer-independent, -the complete system, —and 
not provide a well-suited methodology for the applications, with two major fields: -the oil 
expressing and applying constraints on a exploration and production, -the Geographic 
39