RETRO: A Framework for Semantics Preserving SQL-to-SPARQL Translation

There have been several attempts to make RDBMS and RDF stores inter-operate. The most popular one, D2RQ, has explored one direction i.e. to look at RDBMS through RDF lenses. In this paper we present RETRO, which investigates the reverse direction i.e. to look at RDF through Relational lenses. RETRO generates a relational schema from an RDF store, enabling a user to query RDF data using SQL. A significant advantage of this direction in-addition to interoperability is that it makes numerous relational tools developed over past several decades, available to the RDF stores. In order to provide interoperability between these two DB systems one needs to resolve the heterogeneity between their respective data models and include schema mapping, data mapping and query mapping in the transformation process [1]. However, like D2RQ, RETRO chooses not to physically transform the data and deals only with schema mapping and query mapping. RETRO’s schema mapping derives a domain specific relational schema from RDF data and its query mapping transforms an SQL query over the schema into a provably equivalent SPARQL query, which in-turn is executed upon the RDF store. Since RETRO is a read-only framework, its query mapping uses only a relevant and relationally complete subset of SQL. A proof of correctness of this transformation is given based on compositional semantics of the two query languages.