Got bored by the relational database?
Switch to a RDF store!
Fabrizio Giudici
Tidalwave s.a.s.

Who I am
● Java consultant since 1996
● Senior architect
● Java instructor for Sun since 1998
● Member of the NetBeans Dream Team
● Technical Writer, Blogger at Java.Net, DZone
● http://weblogs.java.net/blog/fabriziogiudici/
● http://www.tidalwave.it/people

Where I am using RDF stores
http://bluemarine.tidalwave.it

Agenda
● Why the RDBMs?
● RDBMs issues
● The Semantic Model
● OpenSesame and Elmo
● A few code samples
● Conclusion

RDBMs are everywhere

What do we expect from a RDBM?
● Persistence
● Reliability
● Transactions
● Integrability
● Manageability

Lack of cohesion
● Do we really need a RDBM for those things?
● No, we don't
– Persistence and transactions are good
– The specific relational schema is evil
● RDBMs sell those stuff in a single package

ER-OO impedance
● Entity-Relationship is different than OO
– Primary keys
– No inheritance
– No behaviour
– Normalization rules
– Relationship through external keys

ORMs
● Tools to minimize the ER-OO impedance
● Java has got a standard API: JPA
– Hibernate, TopLink, EclipseLink, OpenJPA
– Tries to abstract the database à la Java
● Good, but the RDBM has still to be designed
● And maintained

Can we get rid of the relational database?

The Semantic Model
● Semantic Technology != Semantic Web
● RDF: Resource Description Framework
● “Triples” are the atomic information item
● Subject / predicate / object
– Java / is-a / programming-language
– Fabrizio / is-member-of / NetBeans Dream Team
– Verona / is-part-of / Veneto
– Verona / has-plate / “VR”

The Semantic Model
● The subject is a resource
● The predicate is a property
● The object is a value
● A value is a resource or a primitive type
● Resources, properties identified by URL/URN
– Just a naming scheme
– Not necessarily web-related

Formal representation
● RDF is not related to XML
● XML is just one of the way to represent RDF
– XML-RDF, unfortunately referred to as RDF
● Notation 3 (N3), another popular representation
– Much more human-readable
● Other formats exist
● RDF representation is often referred to as
“serialization”

(XML-)RDF is near to you
● RSS/RDF
● Dublin Core
● XMP by Adobe

Compared to RDBMs
● There's no fixed schema
– Everything is a triple
– “AAA slogan”: Anyone can say Anything
about Any topic
● Adding new data types is adding triples
– No need to add / alter tables
– Maintainance is just updating data
● Databases can be distributed (federations)
– Can be merged by just copying triples together

What about performance?
● Not as optimized as SQL
● There's no spread knowledge about tuning as
for SQL
● Some missing parts
– E.g. Sesame still misses select count(*)

OpenSesame, Elmo
● Popular Java infrastructure for RDF
– FLOSS
– http://www.openrdf.org
● Elmo providers JPA-like operations
– Annotations
– Specific API or even subset of JPA

A simple code example
import org.openrdf.elmo.annotations.rdf;
@rdf(GeoVocabulary.URI_GEO_LOCATION)
public class GeoLocation
{
@rdf("http://www.w3.org/2003/01/geo/wgs84_pos#lat")
private Double latitude;
@rdf("http://www.w3.org/2003/01/geo/wgs84_pos#long")
private Double longitude;
@rdf("http://www.w3.org/2003/01/geo/wgs84_pos#alt")
private Double altitude;
@rdf("http://www.tidalwave.it/rdf/geo/2009/02/22#code")
private String code;
}
● Note the use of standard ontologies

A simple code example
● Declare persistent classes in
META-INF/org.openrdf.elmo.concepts
● Choose a store
– Memory
– Memory backed by file
– Database (transactional)
Repository repository = new SailRepository(
new MemoryStore(new File("/tmp/RDFStore")));
repository.initialize();
ElmoModule module = new ElmoModule();
SesameManagerFactory factory = new SesameManagerFactory(module, repository);
SesameManager em = factory.createElmoManager();

Use as JPA EntityManager
em.getTransaction().begin();
GeoLocation genova = new GeoLocation();
genova.setLatitude(45.0);
genova.setLongitude(9.0);
genova.setCode("GE");
em.persist(genova);
em.getTransaction().commit();

Queries
● There are specific query languages
● SPARQL is one of the most popular
● Similar to SQL, but triples in place of tables
PREFIX wgs84: <http://www.w3.org/2003/01/geo/wgs84_pos#>
SELECT ?location WHERE
{
?location wgs84:lat ?lat
}

Running a query
em.getTransaction().begin();
String queryString = "PREFIX wgs84: <http://www.w3.org/2003/01/geo/wgs84_pos#>n" +
"SELECT ?location WHERE n" +
" {n" +
" ?location a ?type.n" +
" ?location wgs84:lat ?latn" +
" }";
final ElmoQuery query = em.createQuery(queryString).
setType("type", GeoLocation.class).
setParameter("lat", 45.0);
final List<GeoLocation> result = query.getResultList();
for (GeoLocation l : result)
{
System.err.println(l);
}
em.getTransaction().commit();

Scratching the surface
● Elmo is powerful
● Supports advanced constructs
– Objects with “multiple personality”
– Mixins

Open issues
● OpenSesame doesn't support all databases
● Lack of experience
– Programming skills
– Maintainance
– Tuning
– Managerial culture
● Not widespread
● Performance?

Conclusion
● RDBMs are mainstream, but old
● They lead to rigid schemata, don't fit the OO
● It's possible to use something different
● RDF stores can be a viable alternative
Fabrizio.Giudici@tidalwave.it