Scalable RDF for Java
Milestone number 2 of the upcoming 6.0.0 release of RDF4J is now available for download.
RDF4J 6.0.0 is a major release of the RDF4J framework. This second milestone continues the work from Milestone 1 with RDF 1.2 and SPARQL 1.2 support, new safeguards for long-running queries, query-planning improvements, FedX optimizations, and deployment updates.
Please note that RDF4J 6.0.0-M2 requires Java 25.
Some of the highlights covered in this second milestone:
This milestone build is not yet feature-complete, but we are putting it out to receive early feedback on all the improvements we have put in.
RDF4J 6.0.0-M2 adds broad support for RDF 1.2 and SPARQL 1.2 (GH-5327). This includes RDF 1.2 parser and writer updates for Turtle, TriG, N-Triples, N-Quads, and RDF/XML, SPARQL 1.2 grammar and evaluation changes, and conformance-test coverage for the new syntax and behavior.
The model API now uses RDF 1.2 triple-term terminology. Legacy RDF-star parser/writer formats and names have been replaced by the corresponding triple-term APIs and formats. Triple terms are supported in object position, matching RDF 1.2 semantics.
Support is not yet complete for every store. The SPARQL 1.2 conformance suite is enabled for MemoryStore, while NativeStore and LMDBStore still have known gaps for triple-term and base-direction support.
RDF4J now includes query circuit breakers for server and workbench query execution (GH-5809). These monitor JVM memory pressure, reject or abort query work when pressure becomes unsafe, release in-flight query resources, and expose breaker status for operators.
Sail-backed repositories can also be configured to log slow queries (GH-5819). The implementation tracks query context through Sail evaluation and logs queries that exceed a configured threshold before producing their first result.
The query optimizer has received new sketch-based, unified, and cascade-style planning work (GH-5720, PR-5785, PR-5796), including runtime telemetry and plan-comparison tooling. LMDB cardinality estimation has also been improved with a page-walking estimator that reads LMDB data files directly (GH-5686).
FedX source selection can now group remote source checks per endpoint
(GH-5813). For larger federations, this reduces source-selection
round trips from one ASK query per statement-pattern/endpoint pair to grouped SELECT queries using
BIND(EXISTS { ... }).
The optimization is enabled by default and can be disabled through FedXConfig#withEnableGroupedSourceSelection(false).
The FedX module also received cleanup for deprecated and unwired 6.0-era code paths (GH-5828).
The Docker setup now targets Java 25 application-server images: Jetty 12 with EE11 modules and Tomcat 11
(GH-5801). The developer testing guide and Server/Workbench
deployment guide now document the docker/run.sh workflow and Docker Compose usage.
RDF4J now ships additional vocabulary modules for EU data-portal and dataspace vocabularies, annotation vocabularies, and bibliographic vocabularies (GH-5709). The PROV-O vocabulary also includes the suggested inverse properties listed by PROV-O (GH-5706).
Milestone number 1 of the upcoming 6.0.0 release of RDF4J is now available for download.
RDF4J 6.0.0 is a major release of the RDF4J framework, focusing on dependency upgrades, LMDB improvements, and HTTP client improvements.
Please note that RDF4J 6.0.0-M1 requires Java 25.
Some of the highlights covered in this first milestone:
This milestone build is not yet feature-complete, but we are putting it out to receive early feedback on all the improvements we have put in.
RDF4J 5.3.1 is now available. This is a patch release fixing 2 bugs.
For more details, have a look at the release notes.
Eclipse RDF4J™ is a powerful Java framework for processing and handling RDF data. This includes creating, parsing, scalable storage, reasoning and querying with RDF and Linked Data. It offers an easy-to-use API that can be connected to all leading RDF database solutions. It allows you to connect with SPARQL endpoints and create applications that leverage the power of linked data and Semantic Web.