Augmenting SQL with Dynamic Restructuring to Support Interoperability in a Relational Federation

In this work, we consider augmenting SQL with constructs allowing dynamic restructuring. Dynamic restructuring occurs naturally in a framework including second-order capabilities for integrating metadata and data within single queries. The context and motivating example for this extension is a federation of relational databases, containing semantically similar information in schematically disparate formats. In ordinary SQL, the output schema for a given query is always known. In order to support interoperability within such a federation, it is desirable to add to SQL the ability to dynamically restructure information among databases, based on the data and metadata of the input relations. Based on a motivating example, we define a hierarchy for characterizing dynamic restructuring capabilities. We then describe a language kernel, the Meta-Query Language (MQL), for dynamically restructuring relational data. MQL includes only simple extensions of SQL but captures almost all of the hierarchy of dynamic restructuring capabilities. We then investigate two extensions of this kernel for obtaining complete data/metadata integration. One extension (MQL+ON) remains LOGSPACE while the other (MQL+JOINALL) captures the PSPACE queries. We conclude by comparing these two extensions and stating the open problem of developing a declarative extension of MQL that captures PTIME and retains completeness with respect to our dynamic restructuring hierarchy.