Querying RDF Databases with Sub-CONSTRUCTs

TL;DR

This paper introduces GrAL, a core graph query language focused on CONSTRUCT queries in RDF databases, defining semantics and recursive pattern evaluation, including sub-CONSTRUCTs and aggregations.

Contribution

It presents a formal semantics for CONSTRUCT queries with sub-patterns in RDF, extending RDF graphs with isolated nodes and defining a recursive evaluation process.

Findings

Formal semantics for GrAL with sub-CONSTRUCTs

Extension of RDF graphs with isolated nodes

Original local aggregation processing

Abstract

Graph query languages feature mainly two kinds of queries when applied to a graph database: those inspired by relational databases which return tables such as SELECT queries and those which return graphs such as CONSTRUCT queries in SPARQL. The latter are object of study in the present paper. For this purpose, a core graph query language GrAL is defined with focus on CONSTRUCT queries. Queries in GrAL form the final step of a recursive process involving so-called GrAL patterns. By evaluating a query over a graph one gets a graph, while by evaluating a pattern over a graph one gets a set of matches which involves both a graph and a table. CONSTRUCT queries are based on CONSTRUCT patterns, and sub-CONSTRUCT patterns come for free from the recursive definition of patterns. The semantics of GrAL is based on RDF graphs with a slight modification which consists in accepting isolated nodes. Such an extension of RDF graphs eases the definition of the evaluation semantics, which is mainly captured by a unique operation called Merge. Besides, we define aggregations as part of GrAL expressions, which leads to an original local processing of aggregations.