com.bigdata.relation.rule.eval.IJoinNexus Maven / Gradle / Ivy
/*
Copyright (C) SYSTAP, LLC DBA Blazegraph 2006-2016. All rights reserved.
Contact:
SYSTAP, LLC DBA Blazegraph
2501 Calvert ST NW #106
Washington, DC 20008
[email protected]
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
/*
* Created on Jun 25, 2008
*/
package com.bigdata.relation.rule.eval;
import java.io.Serializable;
import java.util.Iterator;
import java.util.Properties;
import java.util.concurrent.ExecutorService;
import com.bigdata.bop.BOp;
import com.bigdata.bop.IBindingSet;
import com.bigdata.bop.IConstant;
import com.bigdata.bop.IConstraint;
import com.bigdata.bop.IPredicate;
import com.bigdata.bop.Var;
import com.bigdata.bop.joinGraph.IEvaluationPlan;
import com.bigdata.bop.joinGraph.IEvaluationPlanFactory;
import com.bigdata.bop.joinGraph.IRangeCountFactory;
import com.bigdata.btree.keys.ISortKeyBuilder;
import com.bigdata.config.Configuration;
import com.bigdata.config.IValidator;
import com.bigdata.io.IStreamSerializer;
import com.bigdata.journal.AbstractTask;
import com.bigdata.journal.ConcurrencyManager;
import com.bigdata.journal.IIndexManager;
import com.bigdata.mdi.IMetadataIndex;
import com.bigdata.mdi.PartitionLocator;
import com.bigdata.relation.AbstractResource;
import com.bigdata.relation.IMutableRelation;
import com.bigdata.relation.IRelation;
import com.bigdata.relation.RelationFusedView;
import com.bigdata.relation.accesspath.IAccessPath;
import com.bigdata.relation.accesspath.IAsynchronousIterator;
import com.bigdata.relation.accesspath.IBlockingBuffer;
import com.bigdata.relation.accesspath.IBuffer;
import com.bigdata.relation.accesspath.IElementFilter;
import com.bigdata.relation.accesspath.UnsynchronizedArrayBuffer;
import com.bigdata.relation.locator.IResourceLocator;
import com.bigdata.relation.rule.IProgram;
import com.bigdata.relation.rule.IRule;
import com.bigdata.relation.rule.IStep;
import com.bigdata.relation.rule.Rule;
import com.bigdata.relation.rule.eval.pipeline.JoinTask;
import com.bigdata.service.AbstractScaleOutFederation;
import com.bigdata.service.IDataService;
import com.bigdata.service.IDataServiceCallable;
import com.bigdata.striterator.IChunkedOrderedIterator;
import com.bigdata.striterator.IKeyOrder;
/**
* Interface provides an interoperability nexus for the {@link IPredicate}s,
* {@link IBindingSet}s, and {@link ISolution}s for the evaluation of an
* {@link IRule} and is responsible for resolving the relation symbol to the
* {@link IRelation} object. Instances of this interface may be type-specific
* and allow you to control various implementation classes used during
* {@link IRule} execution.
*
* Note: This interface is NOT {@link Serializable}. Use an
* {@link IJoinNexusFactory} to create instances of this interface.
*
* @author Bryan Thompson
* @version $Id$
*/
public interface IJoinNexus {
/**
* The factory for rule statistics objects.
*/
IRuleStatisticsFactory getRuleStatisticsFactory();
/**
* The factory object is used to materialize appropriate {@link IJoinNexus}
* instances when the rule execution crosses an RMI boundary.
*/
IJoinNexusFactory getJoinNexusFactory();
/**
* The factory object for range counts used by {@link IEvaluationPlan}s.
*/
IRangeCountFactory getRangeCountFactory();
/**
* The kind of operation that is being executed (Query, Insert, or Delete).
*/
ActionEnum getAction();
/**
* When true, rule level parallelism is disabled and the
* {@link ISolution} buffers are flushed after after every {@link IStep}.
* This can be enabled if you are exploring apparent concurrency problems
* with the rules. It should normally be false for better
* performance.
* @deprecated by bop annotations.
*/
boolean forceSerialExecution();
/**
* The maximum #of subqueries for the first join dimension that will be
* issued in parallel. Use ZERO(0) to avoid submitting tasks to the
* {@link ExecutorService} entirely and ONE (1) to submit a single task at a
* time to the {@link ExecutorService}.
* @deprecated by bop annotations.
*/
int getMaxParallelSubqueries();
/**
* The #of elements in a chunk for query or mutation. This is normally a
* relatively large value on the order of 10,000 or better.
* @deprecated by bop annotations.
*/
int getChunkCapacity();
/**
* The #of chunks that can be held by an {@link IBuffer} that is the target
* or one or more {@link UnsynchronizedArrayBuffer}s. This is generally a
* small value on the order of the #of parallel producers that might be
* writing on the {@link IBuffer} since the capacity of the
* {@link UnsynchronizedArrayBuffer}s is already quite large (10k or better
* elements, defining a single "chunk" from a single producer).
*
* @see #getMaxParallelSubqueries()
* @see #getChunkCapacity()
* @deprecated by bop annotations.
*/
int getChunkOfChunksCapacity();
/**
* The #of elements that will be materialized in a fully buffered read by an
* {@link IAccessPath}. When this threshold is exceeded the
* {@link IAccessPath} will use an {@link IAsynchronousIterator} instead.
* This value should on the order of {@link #getChunkCapacity()}.
*
* @see IAccessPath#iterator(long,long, int)
* @see AbstractResource.Options#FULLY_BUFFERED_READ_THRESHOLD
* @deprecated by bop annotations.
*/
int getFullyBufferedReadThreshold();
/**
* Resolve the property value using the {@link IIndexManager}, the namespace
* of the resource, and the {@link Properties} instance to be tested as
* hidden parameters.
*
* @param name
* The property name.
* @param defaultValue
* The default.
*
* @return The resolved property value.
*
* @see Configuration
*/
String getProperty(final String name, final String defaultValue);
/**
* Resolves, parses, and validates the property value.
*
* @param name
* The property name.
* @param defaultValue
* The default value.
* @return
*/
T getProperty(final String name, final String defaultValue,
IValidator validator);
/**
* Binds variables from a visited element.
*
* Note: The bindings are propagated before the constraints are verified so
* this method will have a side-effect on the bindings even if the
* constraints were not satisfied. Therefore you should clone the bindings
* before calling this method.
*
* @param rule
* The rule.
* @param index
* The index of the {@link IPredicate} in the body of the
* {@link Rule}.
* @param e
* An element materialized by the {@link IAccessPath} for that
* {@link IPredicate}.
* @param bindingSet
* the bindings to which new bindings from the element will be
* applied.
*
* @return true unless the new bindings would violate any of
* the {@link IConstraint}s declared for the {@link Rule}).
*
* @throws NullPointerException
* if an argument is null.
* @throws IndexOutOfBoundsException
* if the index is out of bounds.
*
* @deprecated by {@link #bind(IPredicate, Object, IBindingSet)}
*/
boolean bind(final IRule rule, final int index, final Object e,
final IBindingSet bindings);
/**
* Binds variables from a visited element.
*
* Note: The bindings are propagated before the constraints are verified so
* this method will have a side-effect on the bindings even if the
* constraints were not satisfied. Therefore you should clone the bindings
* before calling this method.
*
* @param pred
* The {@link IPredicate} from which the element was read.
* @param constraints
* An array of constraints which must be satisfied (optional).
* @param e
* An element materialized by the {@link IAccessPath} for that
* {@link IPredicate}.
* @param bindingSet
* the bindings to which new bindings from the element will be
* applied.
*
* @return true unless the new bindings would violate any of
* the optional {@link IConstraint}.
*
* @throws NullPointerException
* if an argument is null.
*/
boolean bind(IPredicate pred, IConstraint[] constraints, Object e,
IBindingSet bindings);
/**
* Return a 'fake' binding for the given variable in the specified
* predicate. The binding should be such that it is of a legal type for the
* slot in the predicate associated with that variable. This is used to
* discovery the {@link IKeyOrder} associated with the {@link IAccessPath}
* that will be used to evaluate the predicate when it appears in the tail
* of an {@link IRule} for a given {@link IEvaluationPlan}.
*
* @param predicate
* The predicate.
* @param var
* A variable appearing in that predicate.
*
* @return The 'fake' binding.
*/
IConstant fakeBinding(IPredicate predicate, Var var);
/**
* Create a new {@link ISolution}. The behavior of this method generally
* depends on bit flags specified when the {@link IJoinNexus} was created.
*
* Note: For many purposes, it is only the computed {@link #ELEMENT}s that
* are of interest. For high-level query, you will generally specify only
* the {@link #BINDINGS}. The {@link #BINDINGS} are also useful for some
* truth maintenance applications. The {@link #RULE} is generally only of
* interest for inspecting the behavior of some rule set.
*
* @param rule
* The rule.
* @param bindingSet
* The bindings (the implementation MUST clone the bindings if
* they will be saved with the {@link ISolution}).
*
* @return The new {@link ISolution}.
*
* @throws IllegalArgumentException
* if any parameter is null.
*
* @see #ELEMENT
* @see #BINDINGS
* @see #RULE
* @see #solutionFlags()
* @see Solution
*/
ISolution newSolution(IRule rule, IBindingSet bindingSet);
/**
* Return an {@link ISortKeyBuilder} for an {@link IBindingSet}. The sort
* key may be used to SORT {@link IBindingSet}s or to impose a DISTINCT
* filter on {@link ISolution}s, etc.
*
* @param rule
* The rule that will determine the order imposed amoung the
* bound variables (which variable is 1st, 2nd, 3rd, etc.).
*
* @return The sort key builder (NOT thread-safe).
*/
ISortKeyBuilder newBindingSetSortKeyBuilder(IRule rule);
/**
* The flags that effect the behavior of {@link #newSolution(IRule, IBindingSet)}.
*/
int solutionFlags();
/**
* Bit flag indicating that {@link #newSolution(IRule, IBindingSet)} should
* materialize an element from the {@link IRule} and {@link IBindingSet} and
* make it available via {@link ISolution#get()}.
*/
final int ELEMENT = 1 << 0;
/**
* Bit flag indicating that {@link #newSolution(IRule, IBindingSet)} should
* clone the {@link IBindingSet} and make it available via
* {@link ISolution#getBindingSet()}.
*/
final int BINDINGS = 1 << 1;
/**
* Bit flag indicating that {@link #newSolution(IRule, IBindingSet)} make
* the {@link IRule} that generated the {@link ISolution} available via
* {@link ISolution#getRule()}.
*/
final int RULE = 1 << 2;
/**
* {@link #ELEMENT} and {@link #BINDINGS} and {@link #RULE}.
*/
final int ALL = ELEMENT|BINDINGS|RULE;
/**
* Factory for {@link IBindingSet} implementations.
*
* Note: The factory MUST apply any bound
* {@link IRule#getConstants() constants} for the {@link IRule} before
* returning the {@link IBindingSet}.
*
* @param rule
* The rule whose bindings will be stored in the binding set.
*
* @return A new binding set suitable for that rule.
*/
IBindingSet newBindingSet(IRule rule);
/**
* Return the effective {@link IRuleTaskFactory} for the rule. When the rule
* is a step of a sequential program writing on one or more
* {@link IMutableRelation}s, then the returned {@link IStepTask} must
* automatically flush the buffer after the rule executes in order to ensure
* that the state of the {@link IMutableRelation}(s) are updated before the
* next {@link IRule} is executed.
*
* @param parallel
* true unless the rule is a step is a sequential
* {@link IProgram}. Note that a sequential step MUST flush its
* buffer since steps are run in sequence precisely because they
* have a dependency!
* @param rule
* A rule that is a step in some program. If the program is just
* a rule then the value of parallel does not matter. The
* buffer will is cleared when it flushed so a re-flushed is
* always a NOP.
*
* @return The {@link IStepTask} to execute for that rule.
*
* @see RunRuleAndFlushBufferTaskFactory
* @see RunRuleAndFlushBufferTask
*
* @deprecated by annotations on individual {@link BOp}s which specify which
* join operator is to be used on a join by join basis.
*/
IRuleTaskFactory getRuleTaskFactory(boolean parallel, IRule rule);
/**
* Return the factory for {@link IEvaluationPlan}s.
*
* @return The factory.
*/
IEvaluationPlanFactory getPlanFactory();
/**
* Equivalent to {@link IJoinNexusFactory#getWriteTimestamp()}.
*/
long getWriteTimestamp();
/**
* Equivalent to {@link IJoinNexusFactory#getReadTimestamp()}.
*/
long getReadTimestamp();
/**
* Locate and return the view of the relation identified by the
* {@link IPredicate}. The implementation must choose a view that will
* accept writes iff this is a mutation operation and which is associated
* with an appropriate timestamp.
*
* @param pred
* The {@link IPredicate}, which MUST be the head of some
* {@link IRule}.
*
* @return The {@link IRelation}, which will never be a fused view and
* which will accept writes iff the rules are being executed as a
* mutation operation.
*/
IRelation getHeadRelationView(IPredicate pred);
/**
* Locate and return the view of the relation(s) identified by the
* {@link IPredicate}.
*
* Note: This method is responsible for returning a fused view when more
* than one relation name was specified for the {@link IPredicate}. It
* SHOULD be used whenever the {@link IRelation} is selected based on a
* predicate in the tail of an {@link IRule} and could therefore be a fused
* view of more than one relation instance. (The head of the {@link IRule}
* must be a simple {@link IRelation} and not a view.)
*
* Note: The implementation should choose the read timestamp for each
* relation in the view using {@link #getReadTimestamp(String)}.
*
* @param pred
* The {@link IPredicate}, which MUST be a tail from some {@link IRule}.
*
* @return The {@link IRelation}, which might be a
* {@link RelationFusedView}.
*/
IRelation getTailRelationView(IPredicate pred);
// /**
// * Obtain an access path reading from the view for the relation associated
// * with the specified predicate (from the tail of some rule).
// *
// * @param pred
// * The predicate.
// *
// * @return The access path.
// *
// * @deprecated by {@link #getTailAccessPath(IRelation, IPredicate)} which
// * factors out obtaining the {@link IRelation} view into the
// * caller, which turns out to be a significant cost savings.
// */
// IAccessPath getTailAccessPath(IPredicate pred);
/**
* Obtain an access path reading from relation for the specified predicate
* (from the tail of some rule).
*
* Note that passing in the {@link IRelation} is important since it
* otherwise must be discovered using the {@link IResourceLocator}. By
* requiring the caller to resolve it before hand and pass it into this
* method the contention and demand on the {@link IResourceLocator} cache is
* reduced.
*
* @param relation
* The relation.
* @param pred
* The predicate. When {@link IPredicate#getPartitionId()} is
* set, the returned {@link IAccessPath} MUST read on the
* identified local index partition (directly, not via RMI).
*
* @return The access path.
*/
IAccessPath getTailAccessPath(IRelation relation, IPredicate pred);
/**
* Return an iterator visiting the {@link PartitionLocator} for the index
* partitions from which an {@link IAccessPath} must read in order to
* materialize all elements which would be visited for that predicate.
*
* Note: You can use an {@link IDataServiceCallable} to obtain the
* reference of the {@link IDataService} and pass that into your
* {@link AbstractTask} in order to have the federation reference available
* when running under the {@link ConcurrencyManager}.
*
* @param predicate
* The predicate, with whatever bindings already applied.
* @param fed
* The federation, which is required in order to access the
* {@link IMetadataIndex} for a scale-out index.
* @param joinNexus
* The {@link IJoinNexus}.
*
* @return The iterator.
*/
Iterator locatorScan(
final AbstractScaleOutFederation fed, final IPredicate predicate);
/**
* Used to locate indices, relations and relation containers.
*/
IIndexManager getIndexManager();
/**
* Run as a query.
*
* @param step
* The {@link IRule} or {@link IProgram}.
*
* @return An iterator from which you can read the solutions.
*
* @throws IllegalStateException
* unless this is an {@link ActionEnum#Query}.
* @throws IllegalArgumentException
* if either argument is null.
*
* @todo ISolution is only used when constructing entailments for which we
* also need to know the rule which licensed the entailment in order
* to generate the justification chain. This should all be done by a
* CONSTRUCT operator. The {@link #solutionFlags()} should also go
* away, including {@link #BINDINGS}, {@link #ELEMENT}, {@link #RULE},
* and {@link #ALL}.
*/
IChunkedOrderedIterator runQuery(IStep step) throws Exception;
/**
* Run as mutation operation (it will write any solutions onto the relations
* named in the head of the various {@link IRule}s).
*
* @param step
* The {@link IRule} or {@link IProgram}.
*
* @return The mutation count (#of distinct elements modified in the
* relation(s)).
*
* @throws IllegalArgumentException
* unless {@link ActionEnum#isMutation()} is true.
* @throws IllegalArgumentException
* if either argument is null.
*/
long runMutation(IStep step) throws Exception;
/**
* Return a thread-safe buffer onto which chunks of computed
* {@link ISolution}s will be written. The client will drain
* {@link ISolution}s from buffer using {@link IBlockingBuffer#iterator()}.
*/
IBlockingBuffer newQueryBuffer();
/**
* Return a thread-safe buffer onto which chunks of computed
* {@link ISolution}s will be written. When the buffer is
* {@link IBuffer#flush() flushed} the chunked {@link ISolution}s will be
* inserted into the {@link IMutableRelation}.
*
* @param relation
* The relation.
*
* @return The buffer.
*/
IBuffer newInsertBuffer(IMutableRelation relation);
/**
* Return a thread-safe buffer onto which chunks of computed
* {@link ISolution}s will be written. When the buffer is
* {@link IBuffer#flush() flushed} the chunks of {@link ISolution}s will be
* deleted from the {@link IMutableRelation}.
*
* @param relation
* The relation.
*
* @return The buffer.
*/
IBuffer newDeleteBuffer(IMutableRelation relation);
/**
* Return a buffer suitable for a single-threaded writer that flushes onto
* the specified targetBuffer.
*
* The returned buffer MUST apply the optional filter value returned by
* {@link #getSolutionFilter()} in order to keep individual
* {@link ISolution}s out of the buffer. Filtering is done at this level
* since the targetBuffer contains chunks of solutions.
*
* @param targetBuffer
* A thread-safe buffer for chunks of {@link ISolution}s that
* was allocated with {@link #newQueryBuffer()},
* {@link #newInsertBuffer(IMutableRelation)}, or
* {@link #newDeleteBuffer(IMutableRelation)}.
* @param chunkCapacity
* The capacity of the new buffer. This should be maximum chunk
* size that will be produced or {@link #getChunkCapacity()} if
* you do not have better information.
*
* @return A non-thread-safe buffer.
*/
IBuffer newUnsynchronizedBuffer(
IBuffer targetBuffer, int chunkCapacity);
/**
* Return the {@link IElementFilter} that will be used to reject solutions
* based on the bindings for the head of the rule -or- null
* if no filter will be imposed. This may be used for query or mutation.
*
* @return The optional filter.
*
* @todo Normally the {@link IElementFilter} is specified in terms of the
* elements of some {@link IRelation} and then wrapped as a
* {@link SolutionFilter}.
*
* The return type here should be strongly typed. The generic argument
* is not a sufficient guarantee and the compiler can be fooled by
* passing an {@link IElementFilter} that is already defined in terms
* of an {@link ISolution} rather than element of an {@link IRelation}.
*/
IElementFilter getSolutionFilter();
/**
* Return the object that is responsible for (de-)serializing chunks of
* {@link ISolution}s. This is used by high-level query to transfer
* {@link ISolution}s back to the client.
*/
IStreamSerializer getSolutionSerializer();
/**
* Return the object that is responsible for (de-)serializing chunks of
* {@link IBindingSet}s. This is used by {@link JoinTask}s to transfer
* intermediate {@link IBindingSet}s from one join dimension to the next.
*/
IStreamSerializer getBindingSetSerializer();
}