Dependency Injection

RDF4J Repositories and Connections

Package rdf4j.repository and rdf4j.repository.impl provide an implementation of RDF4J Repository and RepositoryConnection API based on an underlying OntopQueryEngine.

Specifically, interface OntopRepository extends interface Repository adding an httpCacheHeaders property(which is used in the endpoint module) and a factory method defaultRepository(). This method takes an OntopSystemConfiguration providing access to settings and GUICE injection facilities, which are used to obtain an OntopQueryEngine and related beans (e.g., RDF4JInputQueryFactory) which are then wrapped by the implementation class OntopVirtualRepository. This class returns RDF4J connections by creating OntopRepositoryConnection instances, which reject RDF4J manipulation methods (repository is read-only) and map RDF4J query methods to calls to a wrapped OntopConnection.

source

RDF4J Queries

RDF4J RepositoryConnections have prepareXXXQuery methods that return Query (and sub-interfaces) objects with evaluate() methods. Query implementation classes for Ontop are provided in package rdf4j.query.impl.

Abstract class AbstractOntopQuery wraps the OntopConnection used to execute the query and implement shared features for controlling query timeouts, bindings, and base IRI, while it does not support controlling datasets and inclusion of inferred statements (always enabled, cannot be changed).

Implementation classes OntopTupleQuery, OntopGraphQuery, and OntopBooleanQuery extend AbstractOntopQuery and provide implementations of RDF4J TupleQuery, GraphQuery and BooleanQuery. Their evaluate() methods obtain an OntopStatement from the wrapped OntopConnection, and then translate (via RDF4JInputQueryFactory) the <query string, parsed query> pair into a matching Ontop query (i.e., SelectQuery, ConstructQuery, DescribeQuery, AskQuery) that is executed via the statement. In case of OntopTupleQuery, the Ontop result -- a TupleResultSet -- is wrapped in a OntopTupleQueryResult that adapts it to the expected RDF4J result, which returns binding sets here implemented by OntopRDF4JBindingSet.

source

Materialization

The rdf4j module provides also an RDF4JMaterializer component used by upper level modules that provides the same functionalities of OntopRDFMaterializer but using an API compliant with the RDF4J one (e.g., Statement instead of Assertion).

Interface RDF4JMaterializer is the main interface, which contains factory methods defaultMaterializer() that instantiate implementation class DefaultRDF4JMaterializer starting from an OntopSystemConfiguration. This configuration is used to obtain the OntopRDFMaterializer to which materialization is delegated to. Materialization is triggered by methods RDF4JMaterializer.materialize() that optionally take a list of concepts (classes, properties) to materialize, and return a MaterializationGraphQuery that allows streaming through materialized triples while monitoring materialization status.

source

Remarks

• method OntopRepository.defaultRepository() should not reference implementation class OntopVirtualRepository (if access to OntopRepositoryConnection.reformulate() is needed, then a connection interface including this method should be defined in the repository package)
• it is safer if OntopVirtualRepository tracks open OntopRepositoryConnections and closes them when it is closed
• suggest adding default namespaces in OntopVirtualRepository and then inheriting/locally manipulate them in OntopRepositoryConnection
• materializing statements in a list in OntopRepositoryConnection.getStatements() will result in out-of-memory when trying to export all / a large amounts of statements
• implementing hasStatement() in terms of getStatements() is hugely inefficient if the latter materializes all results in a list, since it is enough to check whether a non-empty result is returned
• method OntopRepositoryConnection.getStatements() should be implemented in terms of a tuple query to overcome CONSTRUCT limitations with returning named graphs (unless named graphs are not supported at all)
• class OntopGraphQuery materializes all statements in a list with possible out-of-memory errors