Schema Adaptation Under Multi-relation Dependencies

Abstract

In information grid, data sources not only change their data, but also change their schema. On schema changing, the mapping may be left inconsistency. Thus, it is an essential issue to detect and update the inconsistency. We present a novel approach for adapting mappings as schema evolves. We propose a new kind of dependencies, called multi-relation dependencies. Multi-relation dependencies specify the dependent relationship between attributes across relations under a join condition. By the soundness and completeness of reasoning rules of multi-relation dependencies, our approach detects the minimal part of mapping which are inconsistent. And the algorithm adapts mappings on both local schema and global schema changing. Most current approaches do not distinguish the removal of key attributes from non-key attributes. Our study shows that removal of these two kinds of attributes results in different adaptations.