[hibernate-commits] Hibernate SVN: r12806 - in trunk/HibernateExt/search/src/java/org/hibernate/search: backend/impl/lucene and 2 other directories.

Author: hardy.ferentschik Date: 2007-07-24 05:51:04 -0400 (Tue, 24 Jul 2007) New Revision: 12806 Modified: trunk/HibernateExt/search/src/java/org/hibernate/search/backend/LuceneWork.java trunk/HibernateExt/search/src/java/org/hibernate/search/backend/WorkType.java trunk/HibernateExt/search/src/java/org/hibernate/search/backend/Workspace.java trunk/HibernateExt/search/src/java/org/hibernate/search/backend/impl/lucene/LuceneBackendQueueProcessor.java trunk/HibernateExt/search/src/java/org/hibernate/search/backend/impl/lucene/LuceneWorker.java trunk/HibernateExt/search/src/java/org/hibernate/search/engine/DocumentBuilder.java trunk/HibernateExt/search/src/java/org/hibernate/search/engine/SearchFactoryImplementor.java trunk/HibernateExt/search/src/java/org/hibernate/search/impl/FullTextSessionImpl.java trunk/HibernateExt/search/src/java/org/hibernate/search/impl/SearchFactoryImpl.java Log: HSEARCH-67. Enabled the WorkType.INDEX and chaged all required files to use the lucene indexing parameters in case this new work type is used. Modified: trunk/HibernateExt/search/src/java/org/hibernate/search/backend/LuceneWork.java =================================================================== --- trunk/HibernateExt/search/src/java/org/hibernate/search/backend/LuceneWork.java 2007-07-24 09:48:22 UTC (rev 12805) +++ trunk/HibernateExt/search/src/java/org/hibernate/search/backend/LuceneWork.java 2007-07-24 09:51:04 UTC (rev 12806) @@ -9,12 +9,18 @@ * Represent a Serializable Lucene unit work * * @author Emmanuel Bernard + * @author Hardy Ferentschik */ public abstract class LuceneWork implements Serializable { //TODO set a serial id private Document document; private Class entityClass; private Serializable id; + + /** + * Flag indicating if this lucene work has to be indexed in batch mode. + */ + private boolean batch = false; public LuceneWork(Serializable id, Class entity) { this( id, entity, null ); @@ -26,7 +32,14 @@ this.document = document; } + public boolean isBatch() { + return batch; + } + public void setBatch(boolean batch) { + this.batch = batch; + } + public Document getDocument() { return document; } Modified: trunk/HibernateExt/search/src/java/org/hibernate/search/backend/WorkType.java =================================================================== --- trunk/HibernateExt/search/src/java/org/hibernate/search/backend/WorkType.java 2007-07-24 09:48:22 UTC (rev 12805) +++ trunk/HibernateExt/search/src/java/org/hibernate/search/backend/WorkType.java 2007-07-24 09:51:04 UTC (rev 12806) @@ -2,11 +2,19 @@ package org.hibernate.search.backend; /** + * Enumeration of the different types of Lucene work. This enumeration is used to specify the type + * of index operation to be executed. + * * @author Emmanuel Bernard + * @author Hardy Ferentschik */ public enum WorkType { ADD, UPDATE, - DELETE - //add INDEX at some point to behave differently during the queue process? + DELETE, + + /** + * This type is used for batch indexing. + */ + INDEX } Modified: trunk/HibernateExt/search/src/java/org/hibernate/search/backend/Workspace.java =================================================================== --- trunk/HibernateExt/search/src/java/org/hibernate/search/backend/Workspace.java 2007-07-24 09:48:22 UTC (rev 12805) +++ trunk/HibernateExt/search/src/java/org/hibernate/search/backend/Workspace.java 2007-07-24 09:51:04 UTC (rev 12806) @@ -22,15 +22,24 @@ import org.hibernate.annotations.common.AssertionFailure; /** - * Lucene workspace - * This is not intended to be used in a multithreaded environment + * Lucene workspace. + * <p> + * <b>This is not intended to be used in a multithreaded environment</b>. * <p/> - * One cannot execute modification through an IndexReader when an IndexWriter has been acquired on the same underlying directory - * One cannot get an IndexWriter when an IndexReader have been acquired and modificed on the same underlying directory - * The recommended approach is to execute all the modifications on the IndexReaders, {@link #clean()} }, and acquire the - * index writers + * <ul> + * <li>One cannot execute modification through an IndexReader when an IndexWriter has been acquired + * on the same underlying directory + * </li> + * <li>One cannot get an IndexWriter when an IndexReader have been acquired and modificed on the same + * underlying directory + * </li> + * <li>The recommended approach is to execute all the modifications on the IndexReaders, {@link #clean()}, and acquire the + * index writers + * </li> + * </ul> * * @author Emmanuel Bernard + * @author Hardy Ferentschik */ //TODO introduce the notion of read only IndexReader? We cannot enforce it because Lucene use abstract classes, not interfaces public class Workspace { @@ -41,12 +50,16 @@ private Map<DirectoryProvider, DPStatistics> dpStatistics = new HashMap<DirectoryProvider, DPStatistics>(); private SearchFactoryImplementor searchFactoryImplementor; + /** + * Flag indicating if the current work should be executed the Lucene parameters for batch indexing. + */ + private boolean isBatch = false; + public Workspace(SearchFactoryImplementor searchFactoryImplementor) { this.searchFactoryImplementor = searchFactoryImplementor; } - public DocumentBuilder getDocumentBuilder(Class entity) { return searchFactoryImplementor.getDocumentBuilders().get( entity ); } @@ -115,7 +128,22 @@ Analyzer analyzer = entity != null ? searchFactoryImplementor.getDocumentBuilders().get( entity ).getAnalyzer() : new SimpleAnalyzer(); //never used - writer = new IndexWriter( provider.getDirectory(), analyzer, false ); //have been created at init time + writer = new IndexWriter( provider.getDirectory(), analyzer, false ); //has been created at init time + + LuceneIndexingParameters indexingParams = searchFactoryImplementor.getIndexingParameters(provider); + if(isBatch) + { + writer.setMergeFactor(indexingParams.getBatchMergeFactor()); + writer.setMaxMergeDocs(indexingParams.getBatchMaxMergeDocs()); + writer.setMaxBufferedDocs(indexingParams.getBatchMaxBufferedDocs()); + } + else + { + writer.setMergeFactor(indexingParams.getMergeFactor()); + writer.setMaxMergeDocs(indexingParams.getMaxMergeDocs()); + writer.setMaxBufferedDocs(indexingParams.getMaxBufferedDocs()); + } + writers.put( provider, writer ); } catch (IOException e) { @@ -214,4 +242,11 @@ public long operations; } + public boolean isBatch() { + return isBatch; + } + + public void setBatch(boolean isBatch) { + this.isBatch = isBatch; + } } Modified: trunk/HibernateExt/search/src/java/org/hibernate/search/backend/impl/lucene/LuceneBackendQueueProcessor.java =================================================================== --- trunk/HibernateExt/search/src/java/org/hibernate/search/backend/impl/lucene/LuceneBackendQueueProcessor.java 2007-07-24 09:48:22 UTC (rev 12805) +++ trunk/HibernateExt/search/src/java/org/hibernate/search/backend/impl/lucene/LuceneBackendQueueProcessor.java 2007-07-24 09:51:04 UTC (rev 12806) @@ -5,6 +5,8 @@ import java.util.Comparator; import java.util.List; +import org.apache.commons.logging.Log; +import org.apache.commons.logging.LogFactory; import org.hibernate.search.backend.AddLuceneWork; import org.hibernate.search.backend.LuceneWork; import org.hibernate.search.backend.Workspace; @@ -13,12 +15,18 @@ import org.hibernate.search.store.DirectoryProvider; /** - * Apply the operations to Lucene directories - * avoiding deadlocks + * Apply the operations to Lucene directories avoiding deadlocks. * * @author Emmanuel Bernard + * @author Hardy Ferentschik */ public class LuceneBackendQueueProcessor implements Runnable { + + /** + * Class logger. + */ + private static Log log = LogFactory.getLog( LuceneBackendQueueProcessor.class ); + private List<LuceneWork> queue; private SearchFactoryImplementor searchFactoryImplementor; @@ -33,7 +41,8 @@ workspace = new Workspace( searchFactoryImplementor ); worker = new LuceneWorker( workspace ); try { - deadlockFreeQueue(queue, workspace, searchFactoryImplementor); + deadlockFreeQueue(queue, workspace, searchFactoryImplementor); + checkForBatchIndexing(workspace); for ( LuceneWork luceneWork : queue ) { worker.performWork( luceneWork ); } @@ -44,6 +53,16 @@ } } + private void checkForBatchIndexing(Workspace workspace) { + for ( LuceneWork luceneWork : queue ) { + // if there is at least a single batch index job we put the work space into batch indexing mode. + if(luceneWork.isBatch()){ + log.debug("Setting batch indexing mode."); + workspace.setBatch(true); + } + } + } + /** * one must lock the directory providers in the exact same order to avoid * dead lock between concurrent threads or processes Modified: trunk/HibernateExt/search/src/java/org/hibernate/search/backend/impl/lucene/LuceneWorker.java =================================================================== --- trunk/HibernateExt/search/src/java/org/hibernate/search/backend/impl/lucene/LuceneWorker.java 2007-07-24 09:48:22 UTC (rev 12805) +++ trunk/HibernateExt/search/src/java/org/hibernate/search/backend/impl/lucene/LuceneWorker.java 2007-07-24 09:51:04 UTC (rev 12806) @@ -21,9 +21,10 @@ import org.hibernate.search.engine.DocumentBuilder; /** - * Stateless implementation that perform a work + * Stateless implementation that performs a unit of work. * * @author Emmanuel Bernard + * @author Hardy Ferentschik */ public class LuceneWorker { private Workspace workspace; Modified: trunk/HibernateExt/search/src/java/org/hibernate/search/engine/DocumentBuilder.java =================================================================== --- trunk/HibernateExt/search/src/java/org/hibernate/search/engine/DocumentBuilder.java 2007-07-24 09:48:22 UTC (rev 12805) +++ trunk/HibernateExt/search/src/java/org/hibernate/search/engine/DocumentBuilder.java 2007-07-24 09:51:04 UTC (rev 12806) @@ -54,6 +54,7 @@ * @author Emmanuel Bernard * @author Sylvain Vieujot * @author Richard Hallier + * @author Hardy Ferentschik */ public class DocumentBuilder<T> { private static final Log log = LogFactory.getLog( DocumentBuilder.class ); @@ -393,6 +394,15 @@ queue.add( new AddLuceneWork( id, entityClass, doc ) ); searchForContainers = true; } + else if ( workType == WorkType.INDEX ) { + Document doc = getDocument( entity, id ); + queue.add(new DeleteLuceneWork(id, entityClass) ); + LuceneWork work = new AddLuceneWork( id, entityClass, doc ); + work.setBatch(true); + queue.add(work); + searchForContainers = true; + } + else { throw new AssertionFailure("Unknown WorkType: " + workType); } Modified: trunk/HibernateExt/search/src/java/org/hibernate/search/engine/SearchFactoryImplementor.java =================================================================== --- trunk/HibernateExt/search/src/java/org/hibernate/search/engine/SearchFactoryImplementor.java 2007-07-24 09:48:22 UTC (rev 12805) +++ trunk/HibernateExt/search/src/java/org/hibernate/search/engine/SearchFactoryImplementor.java 2007-07-24 09:51:04 UTC (rev 12806) @@ -1,4 +1,4 @@ -//$Id: $ +// $Id$ package org.hibernate.search.engine; import java.util.Map; @@ -8,10 +8,14 @@ import org.hibernate.search.store.DirectoryProvider; import org.hibernate.search.store.optimization.OptimizerStrategy; import org.hibernate.search.backend.BackendQueueProcessorFactory; +import org.hibernate.search.backend.LuceneIndexingParameters; import org.hibernate.search.backend.Worker; /** + * Interface which gives access to the different directory providers and their configuration. + * * @author Emmanuel Bernard + * @author Hardy Ferentschik */ public interface SearchFactoryImplementor extends SearchFactory { BackendQueueProcessorFactory getBackendQueueProcessorFactory(); @@ -27,4 +31,8 @@ void addOptimizerStrategy(DirectoryProvider<?> provider, OptimizerStrategy optimizerStrategy); public OptimizerStrategy getOptimizerStrategy(DirectoryProvider<?> provider); + + public LuceneIndexingParameters getIndexingParameters(DirectoryProvider<?> provider ); + + void addIndexingParmeters(DirectoryProvider<?> provider, LuceneIndexingParameters indexingParams); } Property changes on: trunk/HibernateExt/search/src/java/org/hibernate/search/engine/SearchFactoryImplementor.java ___________________________________________________________________ Name: svn:keywords + Id Modified: trunk/HibernateExt/search/src/java/org/hibernate/search/impl/FullTextSessionImpl.java =================================================================== --- trunk/HibernateExt/search/src/java/org/hibernate/search/impl/FullTextSessionImpl.java 2007-07-24 09:48:22 UTC (rev 12805) +++ trunk/HibernateExt/search/src/java/org/hibernate/search/impl/FullTextSessionImpl.java 2007-07-24 09:51:04 UTC (rev 12806) @@ -52,7 +52,7 @@ import org.hibernate.type.Type; /** - * Lucene Full text search aware session + * Lucene full text search aware session. * * @author Emmanuel Bernard */ @@ -83,7 +83,7 @@ * Non indexable entities are ignored * The entity must be associated with the session * - * @param entity must not be null + * @param entity The neity to index - must not be <code>null</code>. */ public void index(Object entity) { if (entity == null) return; @@ -94,7 +94,7 @@ DocumentBuilder<Object> builder = searchFactoryImplementor.getDocumentBuilders().get( clazz ); if ( builder != null ) { Serializable id = session.getIdentifier( entity ); - searchFactoryImplementor.getWorker().performWork( entity, id, WorkType.UPDATE, eventSource ); + searchFactoryImplementor.getWorker().performWork( entity, id, WorkType.INDEX, eventSource ); } //TODO //need to add elements in a queue kept at the Session level Modified: trunk/HibernateExt/search/src/java/org/hibernate/search/impl/SearchFactoryImpl.java =================================================================== --- trunk/HibernateExt/search/src/java/org/hibernate/search/impl/SearchFactoryImpl.java 2007-07-24 09:48:22 UTC (rev 12805) +++ trunk/HibernateExt/search/src/java/org/hibernate/search/impl/SearchFactoryImpl.java 2007-07-24 09:51:04 UTC (rev 12806) @@ -1,13 +1,13 @@ //$Id$ package org.hibernate.search.impl; +import java.util.ArrayList; import java.util.HashMap; import java.util.Iterator; +import java.util.List; import java.util.Map; import java.util.Set; import java.util.WeakHashMap; -import java.util.List; -import java.util.ArrayList; import java.util.concurrent.locks.ReentrantLock; import org.apache.lucene.analysis.Analyzer; @@ -22,10 +22,11 @@ import org.hibernate.search.Version; import org.hibernate.search.annotations.Indexed; import org.hibernate.search.backend.BackendQueueProcessorFactory; +import org.hibernate.search.backend.LuceneIndexingParameters; +import org.hibernate.search.backend.LuceneWork; +import org.hibernate.search.backend.OptimizeLuceneWork; import org.hibernate.search.backend.Worker; import org.hibernate.search.backend.WorkerFactory; -import org.hibernate.search.backend.LuceneWork; -import org.hibernate.search.backend.OptimizeLuceneWork; import org.hibernate.search.engine.DocumentBuilder; import org.hibernate.search.engine.SearchFactoryImplementor; import org.hibernate.search.reader.ReaderProvider; @@ -36,6 +37,7 @@ import org.hibernate.util.ReflectHelper; /** + * * @author Emmanuel Bernard */ public class SearchFactoryImpl implements SearchFactoryImplementor { @@ -45,7 +47,7 @@ static { Version.touch(); } - + private Map<Class, DocumentBuilder<Object>> documentBuilders = new HashMap<Class, DocumentBuilder<Object>>(); //keep track of the index modifiers per DirectoryProvider since multiple entity can use the same directory provider private Map<DirectoryProvider, ReentrantLock> lockableDirectoryProviders = @@ -55,7 +57,15 @@ private Worker worker; private ReaderProvider readerProvider; private BackendQueueProcessorFactory backendQueueProcessorFactory; + + + /** + * Each directory provider (index) can have its own performance settings. + */ + private Map<DirectoryProvider, LuceneIndexingParameters> dirProviderIndexingParams = + new HashMap<DirectoryProvider, LuceneIndexingParameters>(); + public BackendQueueProcessorFactory getBackendQueueProcessorFactory() { return backendQueueProcessorFactory; } @@ -64,63 +74,16 @@ this.backendQueueProcessorFactory = backendQueueProcessorFactory; } + @SuppressWarnings( "unchecked" ) public SearchFactoryImpl(Configuration cfg) { //yuk ReflectionManager reflectionManager = getReflectionManager( cfg ); - Class analyzerClass; - String analyzerClassName = cfg.getProperty( Environment.ANALYZER_CLASS ); - if ( analyzerClassName != null ) { - try { - analyzerClass = ReflectHelper.classForName( analyzerClassName ); - } - catch (Exception e) { - throw new SearchException( - "Lucene analyzer class '" + analyzerClassName + "' defined in property '" + Environment.ANALYZER_CLASS + "' could not be found.", - e - ); - } - } - else { - analyzerClass = StandardAnalyzer.class; - } - // Initialize analyzer - Analyzer defaultAnalyzer; - try { - defaultAnalyzer = (Analyzer) analyzerClass.newInstance(); - } - catch (ClassCastException e) { - throw new SearchException( - "Lucene analyzer does not implement " + Analyzer.class.getName() + ": " + analyzerClassName, e - ); - } - catch (Exception e) { - throw new SearchException( "Failed to instantiate lucene analyzer with type " + analyzerClassName, e ); - } - - Iterator iter = cfg.getClassMappings(); - DirectoryProviderFactory factory = new DirectoryProviderFactory(); - while ( iter.hasNext() ) { - PersistentClass clazz = (PersistentClass) iter.next(); - Class<?> mappedClass = clazz.getMappedClass(); - if ( mappedClass != null ) { - XClass mappedXClass = reflectionManager.toXClass( mappedClass ); - if ( mappedXClass != null && mappedXClass.isAnnotationPresent( Indexed.class ) ) { - DirectoryProvider provider = factory.createDirectoryProvider( mappedXClass, cfg, this ); - //TODO move that into DirectoryProviderFactory - if ( !lockableDirectoryProviders.containsKey( provider ) ) { - lockableDirectoryProviders.put( provider, new ReentrantLock() ); - } - final DocumentBuilder<Object> documentBuilder = new DocumentBuilder<Object>( - mappedXClass, defaultAnalyzer, provider, reflectionManager - ); - - documentBuilders.put( mappedClass, documentBuilder ); - } - } - } + Analyzer analyzer = initAnalyzer(cfg); + initDocumentBuilders(cfg, reflectionManager, analyzer); + Set<Class> indexedClasses = documentBuilders.keySet(); - for (DocumentBuilder builder : documentBuilders.values()) { + for (DocumentBuilder builder : documentBuilders.values()) { builder.postInitialize( indexedClasses ); } worker = WorkerFactory.createWorker( cfg, this ); @@ -161,10 +124,18 @@ public void addOptimizerStrategy(DirectoryProvider<?> provider, OptimizerStrategy optimizerStrategy) { dirProviderOptimizerStrategies.put( provider, optimizerStrategy ); } + + public void addIndexingParmeters(DirectoryProvider<?> provider, LuceneIndexingParameters indexingParams) { + dirProviderIndexingParams.put( provider, indexingParams ); + } public OptimizerStrategy getOptimizerStrategy(DirectoryProvider<?> provider) { return dirProviderOptimizerStrategies.get( provider ); } + + public LuceneIndexingParameters getIndexingParameters(DirectoryProvider<?> provider ) { + return dirProviderIndexingParams.get( provider ); + } public ReaderProvider getReaderProvider() { return readerProvider; @@ -207,4 +178,60 @@ queue.add( new OptimizeLuceneWork( entityType ) ); getBackendQueueProcessorFactory().getProcessor( queue ).run(); } + + private void initDocumentBuilders(Configuration cfg, ReflectionManager reflectionManager, Analyzer analyzer) { + Iterator iter = cfg.getClassMappings(); + DirectoryProviderFactory factory = new DirectoryProviderFactory(); + while (iter.hasNext()) { + PersistentClass clazz = (PersistentClass) iter.next(); + Class<?> mappedClass = clazz.getMappedClass(); + if (mappedClass != null) { + XClass mappedXClass = reflectionManager.toXClass(mappedClass); + if (mappedXClass != null && mappedXClass.isAnnotationPresent(Indexed.class)) { + DirectoryProvider provider = factory.createDirectoryProvider(mappedXClass, cfg, this); + // TODO move that into DirectoryProviderFactory + if (!lockableDirectoryProviders.containsKey(provider)) { + lockableDirectoryProviders.put(provider, new ReentrantLock()); + } + final DocumentBuilder<Object> documentBuilder = new DocumentBuilder<Object>(mappedXClass, analyzer, + provider, reflectionManager); + + documentBuilders.put(mappedClass, documentBuilder); + } + } + } + } + + /** + * Initilises the Lucene analyzer to use by reading the analyzer class from the configuration and instantiating it. + * + * @param cfg + * The current configuration. + * @return The Lucene analyzer to use for tokenisation. + */ + private Analyzer initAnalyzer(Configuration cfg) { + Class analyzerClass; + String analyzerClassName = cfg.getProperty(Environment.ANALYZER_CLASS); + if (analyzerClassName != null) { + try { + analyzerClass = ReflectHelper.classForName(analyzerClassName); + } catch (Exception e) { + throw new SearchException("Lucene analyzer class '" + analyzerClassName + "' defined in property '" + + Environment.ANALYZER_CLASS + "' could not be found.", e); + } + } else { + analyzerClass = StandardAnalyzer.class; + } + // Initialize analyzer + Analyzer defaultAnalyzer; + try { + defaultAnalyzer = (Analyzer) analyzerClass.newInstance(); + } catch (ClassCastException e) { + throw new SearchException("Lucene analyzer does not implement " + Analyzer.class.getName() + ": " + + analyzerClassName, e); + } catch (Exception e) { + throw new SearchException("Failed to instantiate lucene analyzer with type " + analyzerClassName, e); + } + return defaultAnalyzer; + } }