Search Results

We propose an extension of the graph query language SPARQL to support temporal data. The proposal follows the approach of temporal databases, where each tuple is annotated with a finite set of intervals, and joins and other operators implicitly take account of these annotations. We adapt this model to the specifics of SPARQL query evaluation, which is defined in terms of compatibility of partial solution mappings. Furthermore, unlike existing proposals, our model allows predicates over intervals (e.g., Allen’s relations) in temporal atoms of filter expressions, while ensuring that query evaluation remains tractable in data complexity. Our semantics comes in three equivalent flavours, with different levels of abstraction. The basic semantics provides a snapshot view of data (it is, however, applicable only to queries without temporal atoms). In the second semantics, induced by an SPM-semiring, solution mappings are annotated with finite sets of intervals. This semantics covers temporal atoms and coincides with the snapshot semantics when the latter is defined. The third, most concrete semantics is equivalent to the second, but based on single intervals, which naturally leads to an implementation over a triple store or a relational database, with tables featuring columns for interval endpoints. As a proof of concept, we evaluate a database storage prototype and show that it addresses previously identified performance limitations of triple stores for temporal data.