A relational algebra for SPARQL

The SPARQL query language for RDF provides Semantic Web developers with a powerful tool to extract information from large datasets. This report describes a transformation from SPARQL into the relational algebra, an abstract intermediate language for the expression and analysis of queries. This makes existing work on query planning and optimization available to SPARQL implementors. A further translation into SQL is outlined, and mismatches between SPARQL semantics and the relational approach are discussed. The SPARQL query language is one of the latest additions to the Semantic Web toolbox. It provides powerful facilities to extract information out of large sets of RDF data. This report describes a transformation from SPARQL queries into the relational algebra, an intermediate language for the expression and analysis of queries that is widely used in the database area [2]. An RDF graph can be represented as a “table” with three columns, ?subject, ?predicate and ?object. Each row corresponds to one triple. Similarly, the result of a SPARQL SELECT query is a table of RDF nodes. These tables are RDF relations. The algebra described in this report consists of operators on RDF relations. Some of them are well-known from regular relational algebra, others are slightly modified to reproduce SPARQL semantics. Figure 1 shows a SPARQL query and its translation into a relational operator tree. Expressing SPARQL queries in relational algebra has several benefits: – It makes a large body of work on query planning and optimization available to SPARQL implementers. – It simplifies the problem of supporting SPARQL in database-backed RDF stores by splitting off considerations specific to the store’s database layout. Section 3 outlines a translation into SQL. – While this report assumes a simple relation of triples without further substructure as the basic building block of expressions, it might just as well be a “view” on a more complex source of triples, like, for example, a distributed union of RDF graphs, or an application-specific database mapped into RDF. Substituting their definitions into the operator tree, and subsequent use of expression transformation and query planning could yield efficient execution strategies for those complex sources of RDF data. SELECT ?name ?email WHERE { ?person rdf:type foaf:Person . ?person foaf:name ?name . OPTIONAL { ?person foaf:mbox ?email } }