reification howto

author: Chris Dollin
version: 1.0
id: $Id: reification.html,v 1.8 2003/08/28 11:29:05 andy_seaborne Exp $

introduction

This document describes the Jena 2 reification API and how to use it. New users of Jena should read this to understand Jena's special support for reification. Since there have been significant changes to this support since Jena 1, users of Jena 1's reification should also read this document to see how to translate their existing code to the new form. As always, consult the Javadoc for interface details.

Reification in RDF and Jena is the ability to treat a Statement as a Resource, and hence to make assertions about that statement. A statement may be reified as many different resources, allowing different manifestations ("statings") of that statement to be treated differently if required.

RDF represents a reified statement as four statements with particular RDF properties and objects: the statement (S, P, O), reified by resource R, is represented by:

We shall call these four such statements a reification quad and the components quadlets. Users of reification in Jena may, by default, simply manipulate reified statements as these quads. However, just as for Bag, Seq, Alt and RDF lists in ordinary models, or ontology classes and individuals in OntModels, Jena has additional support for manipulating reified statements. It also optimises the storage for complete reification quads, avoiding having to store the extra four statements merely to represent one reification.

The interface ReifiedStatement is used to represent a reified statement as a Jena Resource that has direct access to the statement it reifies. The method

returns the Statement that the resource is reifying. All the other Resource methods, of course, may be applied to a ReifiedStatement.

converting resources to reified statements

If a resource R is associated with a reified statement, but might not itself be a ReifiedStatement object, the conversion method RDFNode::as(Class) can be used to find (or create) a ReifiedStatement: For example, a model that has been read in from an RDF/XML file may have reified statements: knowing the name of the resource allows a ReifiedStatement object to be constructed without knowing the statement itself.

If there is no such associated reified statement, a CannotReifyException is thrown. To find out in advance if the conversion is possible, use the predicate RDFNode::canAs(ReifiedStatement.class). (Jena only counts as "an associated reified statement" a resource with exactly one rdf:subject, rdf:predicate, and rdf:object which has rdf:type rdf:Statement. It can of course have other properties.)

Once the ReifiedStatement has been constructed, it retains its Statement even if some (or all) of the original quadlets are removed from the model. This is a feature of the current implementation that might go away; do not rely on it.

testing statements for reification

You may wish to know if some Statement is reified. The methods Statement::isReified() and Model::isreified(Statement) return true if (and only if) the statement has been reified in the model. Note that the Statement method tests to see if the statement is reified in its own model, and the model method tests to see if the Statement is reified in that model; there is no test to see if a Statement is reified in any other models.

listing reified statements

Just as listStatements is used to find the statements present in some model, there are methods for finding the reified statements of a model. Each of them returns a RSIterator object, which is an iterator each of who's elements are ReifiedStatements and for which the convenience method nextRS() will deliver a suitably-cast reified statement.

creating reified statements directly

You do not have to create reified statements by asserting their quads into a Model; they can be created directly from their Statements using one of the methods: Each of these returns a ReifiedStatement who's getStatement() method delivers the original statement (actually, a .equals() statement; it may not be the identical statement). If the creation method passed in a (non-null) String, the ReifiedStatement is a named resource and that string is its URI. Otherwise it is a newly-minted bnode. The methods on Statement create a reified statement in that statements model; those on Model create a reified statement in that model.

It is not permitted for two different (non-equals) statements to be reified onto the same resource. An attempt to do so will generate an AlreadyReifiedException.

The additional method Model::getAnyReifiedStatement(Statement) returns some reification of the supplied Statement; an existing one if possible, otherwise a fresh one (reified by a fresh bnode).

reification and Model::add(Model)

When one model is added to another, as well as the ordinary statements of the model being added, the reified statements are copied across. If this is not desired, there is a two-argument form: If suppress is true, then the reified statements are not copied. (This choice arose from comments on earlier versions of the Jena 2 API; users expected the reified statements to be copied.)

removing reified statements

There are two methods which remove all the reifications of a Statement in some Model: All the reified statements in the model that reify the given statement are removed, whatever their reifying resource. To remove a particular reified statement only, use Similarly to Model::add(Model), the method model.remove(Model m) will remove all the reified statements of m from model, and the two-argument form model.remove(m,true) will not.

reification styles

By default and as you might expect, Jena models allow reification quads to be manifested as ReifiedStatements. Similarly, explicitly created ReifiedStatements are visible as statement quads.

Sometimes this is not desirable. For example, in an application that reifies large numbers of statements in the same model as those statements, most of the results from listStatements() will be quadlets; this is inefficient and confusing. One choice is to reify the statements in a different model. Another is to take advantage of reification styles.

Each model has a reification style, described by constants in ModelFactory. The default style is called Standard because it behaves mostly closely to the RDF standard. There are two other reification styles to choose from:

The method ModelFactory.createDefaultModel() takes an optional Style argument, which defaults to Standard. Similarly, createFileModelMaker() and createMemModelMaker() can take Style arguments which are applied to every model they create.

To take a model with hidden reification quads and expose them as statements, the method ModelFactory.withHiddenStatements(Model m) produces a new model which does just that.