com.hp.hpl.jena.reasoner.transitiveReasoner
Class TransitiveGraphCache

java.lang.Object
  com.hp.hpl.jena.reasoner.transitiveReasoner.TransitiveGraphCache

All Implemented Interfaces:

Finder

public class TransitiveGraphCache
extends java.lang.Object
implements Finder

Datastructure used to represent a closed transitive reflexive relation. It (mostly) incrementally maintains a transitive reduction and transitive closure of the relationship and so queries should be faster than dynamically computing the closed or reduced relations.

The implementation stores the reduced and closed relations as real graph (objects linked together by pointers). For each graph node we store its direct predecessors and successors and its closed successors. A cost penalty is the storage turnover involved in turning the graph representation back into triples to answer queries. We could avoid this by optionally also storing the manifested triples for the links.

Cycles are currently handled by collapsing strongly connected components. Incremental deletes would be possible but at the price of substanially more storage and code complexity. We compromise by doing the easy cases incrementally but some deletes (those that break strongly connected components) will trigger a fresh rebuild.

TODO Combine this with interval indexes (Agrawal, Borigda and Jagadish 1989) for storing the closure of the predecessor relationship. Typical graphs will be nearly tree shaped so the successor closure is modest (L^2 where L is the depth of the tree branch) but the predecessor closure would be expensive. The interval index would handle predecessor closure nicely.

Version:

$Revision: 1.24 $

Author:

Dave Reynolds

Constructor Summary

TransitiveGraphCache(Node directPredicate, Node closedPredicate)
Constructor - create a new cache to hold the given relation information.

Method Summary

void	addRelation(Triple t) Register a new relation instance in the cache
boolean	cacheAll(Finder graph, Node predicate) Cache all instances of the given predicate which are present in the given Graph.
void	clear() Clear the entire cache contents.
boolean	contains(TriplePattern pattern) Return true if the given pattern occurs somewhere in the find sequence.
TransitiveGraphCache	deepCopy() Create a deep copy of the cache contents.
java.lang.String	dump() Dump a description of the cache to a string for debug.
ExtendedIterator	find(TriplePattern pattern) Basic pattern lookup interface.
ExtendedIterator	findWithContinuation(TriplePattern pattern, Finder continuation) Extended find interface used in situations where the implementator may or may not be able to answer the complete query.
Node	getClosedPredicate() Returns the closedPredicate.
Node	getDirectPredicate() Returns the directPredicate.
boolean	isSubject(Node node) Return true if the given Node is registered as a subject node
ExtendedIterator	listAllSubjects() Return an iterator over all registered subject nodes
void	removeRelation(Triple t) Remove an instance of a relation from the cache.
void	setCaching(boolean enable) Enable/disabling caching of the Triples representing the relationships.

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

TransitiveGraphCache

public TransitiveGraphCache(Node directPredicate,
                            Node closedPredicate)

Constructor - create a new cache to hold the given relation information.

Parameters:

directPredicate - The RDF predicate representing the direct relation

closedPredicate - The RDF predicate representing the closed relation

Method Detail

getClosedPredicate

public Node getClosedPredicate()

Returns the closedPredicate.

Returns:

Node

getDirectPredicate

public Node getDirectPredicate()

Returns the directPredicate.

Returns:

Node

addRelation

public void addRelation(Triple t)

Register a new relation instance in the cache

removeRelation

public void removeRelation(Triple t)

Remove an instance of a relation from the cache.

findWithContinuation

public ExtendedIterator findWithContinuation(TriplePattern pattern,
                                             Finder continuation)

Extended find interface used in situations where the implementator may or may not be able to answer the complete query.

In this case any query on the direct or closed predicates will be assumed complete, any other query will pass on to the continuation.

Specified by:

findWithContinuation in interface Finder

Parameters:

pattern - a TriplePattern to be matched against the data

continuation - either a Finder or a normal Graph which will be asked for additional match results if the implementor may not have completely satisfied the query.

contains

public boolean contains(TriplePattern pattern)

Return true if the given pattern occurs somewhere in the find sequence.

Specified by:

contains in interface Finder

listAllSubjects

public ExtendedIterator listAllSubjects()

Return an iterator over all registered subject nodes

isSubject

public boolean isSubject(Node node)

Return true if the given Node is registered as a subject node

cacheAll

public boolean cacheAll(Finder graph,
                        Node predicate)

Cache all instances of the given predicate which are present in the given Graph.

Parameters:

graph - the searchable set of triples to cache

predicate - the predicate to cache, need not be the registered predicate due to subProperty declarations

Returns:

returns true if new information has been cached

find

public ExtendedIterator find(TriplePattern pattern)

Basic pattern lookup interface.

Specified by:

find in interface Finder

Parameters:

pattern - a TriplePattern to be matched against the data

Returns:

a ExtendedIterator over all Triples in the data set that match the pattern

deepCopy

public TransitiveGraphCache deepCopy()

Create a deep copy of the cache contents. Works by creating a completely new cache and just adding in the direct links.

clear

public void clear()

Clear the entire cache contents.

setCaching

public void setCaching(boolean enable)

Enable/disabling caching of the Triples representing the relationships. If this is enabled then a number of triples quadratic in the graph depth will be stored. If it is disabled then all queries will turn over storage dynamically creating the result triples.

dump

public java.lang.String dump()

Dump a description of the cache to a string for debug.