Class SourceSelection

java.lang.Object

org.eclipse.rdf4j.federated.optimizer.SourceSelection

public class SourceSelection
extends Object

Performs source selection during query optimization by determining, for each statement pattern in a BGP, which federation members (endpoints) can contribute results.

Algorithm

For each statement pattern and each endpoint the SourceSelectionCache is consulted first. Patterns whose source membership is already known from a previous query are resolved without any remote communication. Only patterns with a POSSIBLY_HAS_STATEMENTS assurance require a remote check. All remote checks are executed in parallel using the FedX worker-thread infrastructure (SourceSelection.SourceSelectionExecutorWithLatch), and the calling thread blocks until every check has completed or the query timeout is reached.

Remote check strategies

Two strategies are supported for performing the remote checks, selected per endpoint based on the configuration:

Grouped source selection (default, enableGroupedSourceSelection=true)

All statement patterns that require a remote check for a given endpoint are batched into a single SPARQL SELECT query of the form:

SELECT * WHERE {
  BIND(EXISTS { <pattern_0> } AS ?stmt_0)
  BIND(EXISTS { <pattern_1> } AS ?stmt_1)
  ...
}

This reduces the number of remote requests from O(S × M) to O(M), where S is the number of statement patterns and M the number of federation members. It is particularly effective in high-latency settings. Grouped checks are only used when more than two patterns require a check for the same endpoint; otherwise, individual ASK queries are sent. The grouped check is implemented in SourceSelection.ParallelGroupedCheckTask.

Individual ASK queries (classic FedX, enableGroupedSourceSelection=false)

One ASK query is sent per statement pattern and endpoint, yielding up to S × M remote requests. Implemented in SourceSelection.ParallelCheckTask.

Configuration

The source selection strategy is controlled via FedXConfig:

• withEnableGroupedSourceSelection(boolean) — enables or disables grouped source selection (default: true)
• withSourceSelectionCacheSpec(String) — configures the Guava CacheBuilderSpec for the source selection cache
• withSourceSelectionCacheFactory(...) — supplies a custom SourceSelectionCacheFactory

Author:

Andreas Schwarte

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
protected class	SourceSelection.CheckTaskPair
protected static class	SourceSelection.ParallelCheckTask	Task for sending an ASK request to the endpoints (for source selection)
protected static class	SourceSelection.ParallelGroupedCheckTask
protected static class	SourceSelection.ParallelSourceSelectionTask
protected static class	SourceSelection.SourceSelectionExecutorWithLatch

Field Summary

Fields

Modifier and Type	Field	Description
protected final SourceSelectionCache	cache
protected final List<Endpoint>	endpoints
protected final FederationContext	federationContext
protected final QueryInfo	queryInfo
protected Map<StatementPattern, List<StatementSource>>	stmtToSources	Map statements to their sources.

Constructor Summary

Constructors

Constructor	Description
SourceSelection(List<Endpoint> endpoints, SourceSelectionCache cache, FederationContext federationContext, QueryInfo queryInfo)

Method Summary

Modifier and Type	Method	Description
protected void	addSource(StatementPattern stmt, StatementSource source)	Add a source to the given statement in the map (synchronized through map)
void	doSourceSelection(List<StatementPattern> stmts)	Perform source selection for the provided statements using cache or remote ASK queries.
Set<Endpoint>	getRelevantSources()	Retrieve a set of relevant sources for this query.

Methods inherited from class Object

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

Field Details

endpoints

protected final List<Endpoint> endpoints

cache

protected final SourceSelectionCache cache

federationContext

protected final FederationContext federationContext

queryInfo

protected final QueryInfo queryInfo

stmtToSources

protected Map<StatementPattern, List<StatementSource>> stmtToSources

Map statements to their sources. Use synchronized access!

Constructor Details

SourceSelection

public SourceSelection(List<Endpoint> endpoints,
 SourceSelectionCache cache,
 FederationContext federationContext,
 QueryInfo queryInfo)

Method Details

doSourceSelection

public void doSourceSelection(List<StatementPattern> stmts)

Perform source selection for the provided statements using cache or remote ASK queries. Remote ASK queries are evaluated in parallel using the concurrency infrastructure of FedX. Note, that this method is blocking until every source is resolved. The statement patterns are replaced by appropriate annotations in this optimization.

Parameters:

stmts -

getRelevantSources

public Set<Endpoint> getRelevantSources()

Retrieve a set of relevant sources for this query.

Returns:

the relevant sources

addSource

protected void addSource(StatementPattern stmt,
 StatementSource source)

Add a source to the given statement in the map (synchronized through map)

Parameters:

stmt -

source -