Generic Modeling and Mapping Languages for Model Management
Activities in management of schemas and schema mappings are usually solved by special-purpose solutions such as coding wrapper components or manually updating view definitions. The goal of model management is to raise the level of abstraction for metadata-intensive activities by providing a set of high-level operators that automate or semi-automate such tasks. The problems of model management are aggravated by the fact that usually heterogeneous modeling languages, such as the relational data model, XML Schema, or ontologies, are employed within the same organization. Therefore, model management aims at genericness by devising operations that are agnostic about the underlying native metamodels. Current solutions fail to be generic as they are restricted to certain combinations of modeling languages. Therefore, a generic solution for model management problems requires generic languages for modeling and mapping specification as well as algorithms operating on such generic representations. This work solves some of the problems in generic model management. In particular, the work makes the following contributions: 1. A generic metamodel that allows the detailed representation of schemas imported from various native languages. This is required, for instance, by schema matching algorithms which use the knowledge about schemas to produce a mapping between them. 2. The semantics of our generic metamodel serves as the foundation for a formal and generic schema mapping language which allows data exchange and query rewriting between schemas in different modeling languages. Unlike other languages, our mapping language at the same time supports powerful restructuring of data and is closed under composition. 3. Our solutions for schema matching, mapping composition and other model management operations have been integrated into a holistic generic model management prototype system. 4. Our schema mapping language has been used to develop an object-relational mapping tool and a federated data management system that is agnostic about the native metamodels employed by its data sources.
RWTH Aachen University , Aachen , DE.