Abstract
At the crossroad of the fields of Geographic Information and Computing, this research which was carried out comes under the Geomatic field. It deals with the contribution of the model engineering to the design and the development of Geographic Information Systems (GIS). The geomatic application field, a weak capitalization of knowledge during the development process and a poor quality of the captured knowledge during the analysis have led to fix as an aim to develop a supporting tool for the design of Geographic Information Systems adapted to a development process enabling fast prototyping during the analysis phase and ensuring the knowledge capitalization. The dichotomy Geographic Information/Computing required to mobilize knowledge belonging to these two fields. It gives the structure to the research content. The first contribution concerns the GIS modeling. After a quick overview of the background bibliography on the methods and the formalisms used for the design of the GIS, the need to compare the spatial and temporal properties used in these methods and these formalisms has resulted in a terminological study realized with the Set Theory. This study resulted in a general taxonomy of the used terminology. The conceptual proximity between this taxonomy and the UML class diagrams enabled to define a method of derivation of this taxonomy in order to obtain a GIS metamodel and a GIS Design Pattern. The latest will be automatically generated by the transformations defined in model engineering. The terminological study also enabled to specify the relationships between the thematic concepts, the spatial and temporal properties, etc. These relationships contributed to specify the model transformations of the second contribution. The second contribution concerns the model engineering. The objective related to the development process enabling fast prototyping during the analysis phase has led to define the method Continuous Integration Unified Process, which superimposes a fast prototyping cycle in the analysis phase of the Unified Process method. The objective of knowledge capitalization required to design a generalization of the MDA approach called Software Development Process Approach (SDPA). It is based on the statement that the knowledge capitalization is an issue which is met during the whole development process. A multimodel artifact, called Software Development Process Model (SDPM), has been developed to solve this problem. It associates a sub-model to each of the development cycle phases. A diffusion transformation based on the cloning of the concepts enables to transfer, from sub-model to sub-model, the concepts from the analysis sub-model up to the implementation sub-models. A retrodiffusion transformation enables to achieve the reverse transfer. The coherence of the SDPM is ensured by an traceability link architecture which connect all the concepts to their clone and by pre-treatments and/or post-treatments to the transformations of diffusion and of retrodiffusion. Four geomatic transformations enable on the one hand, the capture the spatial and/or temporal properties of the thematic concepts and, on the other hand, the conversion of these properties into UML model elements which can be used by the code generators of the case tool. Finally, three transformations which were carried out on the SQL implementation model fit the diffused model to the code generator of the case tool. The Continuous Integration Unified Process method and the Software Development Process Approach has been implemented in the Objecteering Case Tool.