Connecting Python's RDFLib and Stardog

For a couple of years I have been working with the Python RDFLib library for converting data from various formats to RDF. This library serves this work well but it's sometimes difficult to track down a straightforward, working example of performing a particular operation or task in RDFLib. I have also become interested in learning more about the commercial triple store offerings, which promise better performance and more features than the open source solutions. A colleague has had good experiences with Stardog, a commercial semantic graph database (with a freely licensed community edition) from Clark & Parsia, so I thought I would investigate how to use RDFLib to load data in to Stardog and share my notes.

A "SPARQLStore" and "SPARQLUpdateStore" have been included with Python's RDFLib since version 4.0. These are designed to allow developers to use the RDFLib code as a client to any SPARQL endpoint. Since Stardog supports SPARQL 1.1, developers should be able to connect to Stardog from RDFLib in the similar way they would to other triple stores like Sesame or Fuseki.

Setup Stardog

You will need a working instance of Stardog. Stardog is available under a community license for evaluation after going through a simple registration process. If you haven't setup Stardog before, you might want to checkout Geir Grønmo's triplestores repository where he has Vagrant provisioning scripts for various triple stores. This is how I got up and running with Stardog.

Once Stardog is installed, start the Stardog server with security disabled. This will allow the RDFLib code to connect without a username and password. Obviously you will not want to run Stardog in this way in production but it is convenient for testing.

$./bin/stardog-admin server start --disable-security

Next create a database called "demo" to store our data.

$./bin/stardog-admin db create -n demo

At this point a SPARQL endpoint is available at ready for queries at http://localhost:5820/demo/query.


For this example, we'll add three skos:Concepts to a named graph in the Stardog store.

@prefix rdf: <> .
@prefix rdfs: <> .
@prefix skos: <> .
@prefix xml: <> .
@prefix xsd: <> .

<> a skos:Concept ;
    skos:broader <> ;
    skos:preferredLabel "Baseball" .

<> a skos:Concept ;
    skos:preferredLabel "Sports" .

<> a skos:Concept ;
    skos:preferredLabel "Soccer" .


The complete example code here is available as a Gist.

Setting up the 'store'

We need to initialize a SPARQLUpdateStore as well as a named graph where we will store our assertions.

    from rdflib import Graph, Literal, URIRef
    from rdflib.namespace import RDF, SKOS
    from rdflib.plugins.stores import sparqlstore

    #Define the Stardog store
    endpoint = 'http://localhost:5820/demo/query'
    store = sparqlstore.SPARQLUpdateStore(), endpoint))

    #Identify a named graph where we will be adding our instances.
    default_graph = URIRef('')
    ng = Graph(store, identifier=default_graph)
Loading assertions from a file

We can load our sample turtle file to an in-memory RDFLib graph.

    g = Graph()
    g.parse('./sample-concepts.ttl', format='turtle')

    #Serialize our named graph to make sure we got what we expect.
    print g.serialize(format='turtle')

Since our data is now loaded as an in memory Graph we can add it to Stardog with a SPARQL INSERT DATA operation.

    u'INSERT DATA { %s }' % g.serialize(format='nt')
Use the RDFLib API to inspect the data

Using the RDFLib API, we can list all the Concepts in the Stardog that were just added.

    for subj in ng.subjects(predicate=RDF.type, object=SKOS.Concept):
        print 'Concept: ', subj

And, we can find concepts that are broader than others.

    for ob in ng.objects(predicate=SKOS.broader):
        print 'Broader: ', ob
Use RDFLib to issue SPARQL read queries.

RDFLib allows for binding a prefix to a namespace. This makes our queries easier to read and write.

    store.bind('skos', SKOS)

A SELECT query to get all the skos:preferredLabel for skos:Concepts.

    rq = """
    SELECT ?s ?label
    WHERE {
        ?s a skos:Concept ;
           skos:preferredLabel ?label .
    for s, l in ng.query(rq):
        print s.n3(), l.n3()
Use RDFLib to add assertions.

The RDFLib API can also be used to add new assertions to Stardog.

    soccer = URIRef('')
    ng.add((soccer, SKOS.altLabel, Literal('Football')))

We can now Read statements about soccer using the RDFLib API, which issues the proper SPARQL query to Stardog in the background.

    for s, p, o in ng.triples((soccer, None, None)):
        print s.n3(), p.n3(), o.n3()


With a little setup, we can begin working with Stardog in RDFLib in a similar way that we work with RDFLib and other backends. The sample code here is included in this Gist.

Comments !