com.bigdata.rdf.graph.impl.ram

Class RAMGASEngine.RAMGraphAccessor

java.lang.Object

com.bigdata.rdf.graph.impl.ram.RAMGASEngine.RAMGraphAccessor

All Implemented Interfaces:

IGraphAccessor

Enclosing class:

RAMGASEngine

public static class RAMGASEngine.RAMGraphAccessor
extends Object
implements IGraphAccessor

Constructor Summary

Constructors

Constructor and Description
RAMGASEngine.RAMGraphAccessor(RAMGASEngine.RAMGraph g)

Method Summary

Methods

Modifier and Type	Method and Description
void	advanceView() Hook to advance the view of the graph.
VertexDistribution	getDistribution(Random r) Obtain a weighted distribution of the vertices in the graph from which samples may then be taken.
long	getEdgeCount(IGASContext<?,?,?> ctx, org.openrdf.model.Value u, EdgesEnum edges) Return the #of edges of the specified type for the given vertex.
Iterator<org.openrdf.model.Statement>	getEdges(IGASContext<?,?,?> ctx, org.openrdf.model.Value u, EdgesEnum edges) Return the edges for the given vertex.

Methods inherited from class java.lang.Object

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

Constructor Detail

RAMGASEngine.RAMGraphAccessor

public RAMGASEngine.RAMGraphAccessor(RAMGASEngine.RAMGraph g)

Method Detail

advanceView

public void advanceView()

Description copied from interface: IGraphAccessor

Hook to advance the view of the graph. This is invoked at the end of each GAS computation round for a given IGASProgram.

Specified by:

advanceView in interface IGraphAccessor

getEdgeCount

public long getEdgeCount(IGASContext<?,?,?> ctx,
                org.openrdf.model.Value u,
                EdgesEnum edges)

Description copied from interface: IGraphAccessor

Return the #of edges of the specified type for the given vertex.

Note: This is not always a flyweight operation due to the need to filter for only the observable edge types. If this information is required, it may be best to cache it on the vertex state object for a given IGASProgram.

Specified by:

getEdgeCount in interface IGraphAccessor

Parameters:

ctx - The IGASContext.

u - The vertex.

edges - Typesafe enumeration indicating which edges should be visited.

Returns:

An iterator that will visit the edges for that vertex.

getEdges

public Iterator<org.openrdf.model.Statement> getEdges(IGASContext<?,?,?> ctx,
                                             org.openrdf.model.Value u,
                                             EdgesEnum edges)

Description copied from interface: IGraphAccessor

Return the edges for the given vertex.

Specified by:

getEdges in interface IGraphAccessor

Parameters:

ctx - The IGASContext.

u - The vertex.

edges - Typesafe enumeration indicating which edges should be visited.

Returns:

An iterator that will visit the edges for that vertex.

getDistribution

public VertexDistribution getDistribution(Random r)

Description copied from interface: IGraphAccessor

Obtain a weighted distribution of the vertices in the graph from which samples may then be taken. The weight is the #of in-edges and out-edges in which a given vertex appears. Statements that model property values and statement that model link attributes SHOULD NOT be counted in this sample.

Specified by:

getDistribution in interface IGraphAccessor

Parameters:

r - A random number generator that (a) MAY be used to select the sample distribution (the distribution does not need to be exhaustive so long as it obeys the sampling requirements); and (b) WILL be used to select vertices from that distribution.

Returns:

The distribution. TODO SAMPLING: While the same distribution is used for all SAIL implementations, and while all implementations accept and use a random seed, there is NOT a guarantee the different implementations will sample the same vertices from the same graph. This is because the order in which the vertex sample is collected and the natural order of the vertices currently depends on the implementation objects.

This COULD be fixed if we required the implementation to put the sampled vertices into an order based on their external representation as RDF Values, but still reporting their internal since that is what is used to actually specify a starting vertex of the GAS program. However, even then, implementation specific characteristics might not provide us with the same vertices (e.g., bigdata imposes a canonical mapping on xsd:dateTime).