We’re happy to announce the release of Stardog 4.2, our first major release since closing our first funding round in July. Complete release notes are available. Let’s review the highlights.
Our mission is to unify all enterprise data, which includes, crucially, the unstructured stuff. In 4.2 we’ve added new capabilities to store, search, index, retrieve, and unify the unstructured data of documents and files. User-defined NLP, text mining, or machine learning algorithms can (optionally) be called at document ingest time to extract structured data, i.e., nodes and edges, from unstructured data.
To be honest, this should have been Stardog 5.0 release based on this new set of capabilities, but we have some backward-incompatible API changes coming up which we will isolate in the 5.0 release.
We’ve also added the ability to give queries a name (either per database or per server) and then use that name any place where a query is legal. We serialize the query in Stardog’s system catalog and then look it up later. This paves the way for a more general stored procedure-like capability in a future release; but, note, that Stardog’s declarative rule capability, which we’ve shipped forever, is already quite sproc-like.
We realized that update queries with auto transactions were causing cluster nodes to be expelled; that’s no good so we fixed it. And now cluster session and connection timeouts are easily configured, which makes the cluster more stable if network latency is high.
Aggregate Queries Resilience
Off-heap support for query evaluation is extended in Stardog 4.2 to cover SPARQL 1.1 aggregates. This change makes Stardog more resilient to OOM errors on hard queries that include aggregation. In low memory scenarious Stardog will move partially computed aggregates off heap and then, if needed, spill them to disk.
This is especially important for queries that contain aggregates with
COUNT(DISTINCT *) or
SUM(DISTINCT ?x)—which can require large
state during computation.
Stardog 4.2 is our best release ever and moves us closer to our goal of unifying all enterprise data. Check it out: free 30-day evals are waiting for you at http://stardog.com/.