Schema translation: A three-level semantic abstraction approach

Abstract

An intermediate mapping (IM) data model is proposed for translating schema, instances and values of a logical data base (LDB). The schema translation process is divided into four phases: REWRITE, DERIVE, FILTER and ACTION. In the REWRITE phase, the data structures and constraints of the source LDB are written into the IM-schema. The IM-schema is then used to derive the schema of the target LDB in the DERIVE phase. Several translation operators are invoked in these two phases, and the redundant ones are filtered out in the FILTER phase using dominance among operators. The operators, retained after the third phase, are executed in the final ACTION phase to get the target LDB. The correctness of the translation is established by proving the semantic content preservation properties of the translation operators. At the end, we provide an example of the translation process using a relational LDB as a source and a hierarchical LDB as a target.