Reasoning in Data Integration Systems: why LAV and GAV are Siblings

Data integration consists in providing a uniform access to a set of data sources, through a unified representation of the data called global schema; a mapping specifies the relationship between the global schema and the sources. Integrity constraints (ICs) are expressed on the global schema to better represent the domain of interest; in general, ICs are not satisfied by the data at the sources. In this paper we address the problem of query answering in GLAV data integration systems, where tuple-generating dependencies are expressed on the global schema. We solve the problem in an intensional fashion, by presenting a rewriting technique that, taking into account both the ICs and the mapping, allows us to compute the answers to a query, expressed over the global schema, by evaluating the rewritten query directly over the sources. Since the GLAV approach is a generalisation of the basic approaches LAV and GAV, we show that query answering under ICs can be done in the same way in LAV and GAV systems, thus proving that LAV and GAV are siblings, and not opposites.