[hibernate-commits] Hibernate SVN: r16755 - in search/trunk/src: main/java/org/hibernate/search/filter and 4 other directories.

Author: epbernard Date: 2009-06-10 20:43:09 -0400 (Wed, 10 Jun 2009) New Revision: 16755 Added: search/trunk/src/main/java/org/hibernate/search/filter/FullTextFilterImplementor.java search/trunk/src/main/java/org/hibernate/search/filter/ShardSensitiveOnlyFilter.java search/trunk/src/test/java/org/hibernate/search/test/shards/CustomerShardingStrategy.java search/trunk/src/test/java/org/hibernate/search/test/shards/CustomerShardingStrategyTest.java Modified: search/trunk/src/main/docbook/en-US/modules/configuration.xml search/trunk/src/main/docbook/en-US/modules/query.xml search/trunk/src/main/java/org/hibernate/search/filter/ChainedFilter.java search/trunk/src/main/java/org/hibernate/search/query/FullTextFilterImpl.java search/trunk/src/main/java/org/hibernate/search/query/FullTextQueryImpl.java search/trunk/src/main/java/org/hibernate/search/store/IdHashShardingStrategy.java search/trunk/src/main/java/org/hibernate/search/store/IndexShardingStrategy.java search/trunk/src/main/java/org/hibernate/search/store/NotShardedStrategy.java search/trunk/src/test/java/org/hibernate/search/test/configuration/UselessShardingStrategy.java Log: HSEARCH-251 Query on a shard subset based on a filter activation Modified: search/trunk/src/main/docbook/en-US/modules/configuration.xml =================================================================== --- search/trunk/src/main/docbook/en-US/modules/configuration.xml 2009-06-10 21:48:21 UTC (rev 16754) +++ search/trunk/src/main/docbook/en-US/modules/configuration.xml 2009-06-11 00:43:09 UTC (rev 16755) @@ -206,21 +206,33 @@ <section id="search-configuration-directory-sharding" revision="1"> <title>Sharding indexes</title> - <para>In some extreme cases involving huge indexes (in size), it is - necessary to split (shard) the indexing data of a given entity type into - several Lucene indexes. This solution is not recommended until you reach - significant index sizes and index update times are slowing the application - down. The main drawback of index sharding is that searches will end up - being slower since more files have to be opened for a single search. In + <para>In some cases, it is necessary to split (shard) the indexing data of + a given entity type into several Lucene indexes. This solution is not + recommended unless there is a pressing need because by default, searches + will be slower as all shards have to be opened for a single search. In other words don't do it until you have problems :)</para> - <para>Despite this strong warning, Hibernate Search allows you to index a - given entity type into several sub indexes. Data is sharded into the - different sub indexes thanks to an - <classname>IndexShardingStrategy</classname>. By default, no sharding - strategy is enabled, unless the number of shards is configured. To - configure the number of shards use the following property</para> + <para>For example, sharding may be desirable if:</para> + <itemizedlist> + <listitem> + <para>A single index is so huge that index update times are slowing + the application down.</para> + </listitem> + + <listitem> + <para>A typical search will only hit a sub-set of the index, such as + when data is naturally segmented by customer, region or + application.</para> + </listitem> + </itemizedlist> + + <para>Hibernate Search allows you to index a given entity type into + several sub indexes. Data is sharded into the different sub indexes thanks + to an <classname>IndexShardingStrategy</classname>. By default, no + sharding strategy is enabled, unless the number of shards is configured. + To configure the number of shards use the following property</para> + <example> <title>Enabling index sharding by specifying nbr_of_shards for a specific index</title> @@ -243,10 +255,21 @@ <programlisting>hibernate.search.&lt;indexName&gt;.sharding_strategy my.shardingstrategy.Implementation</programlisting> </example> + <para>Using a custom <classname>IndexShardingStrategy</classname> + implementation, it's possible to define what shard a given entity is + indexed to. </para> + + <para>It also allows for optimizing searches by selecting which shard to + run the query onto. By activating a filter (see <xref + linkend="query-filter-shard" />), a sharding strategy can select a subset + of the shards used to answer a query + (<classname>IndexShardingStrategy.getDirectoryProvidersForQuery</classname>) + and thus speed up the query execution.</para> + <para>Each shard has an independent directory provider configuration as described in <xref linkend="search-configuration-directory" />. The - DirectoryProvider default name for the previous example are - <literal>&lt;indexName&gt;.0</literal> to + <classname>DirectoryProvider</classname> default name for the previous + example are <literal>&lt;indexName&gt;.0</literal> to <literal>&lt;indexName&gt;.4</literal>. In other words, each shard has the name of it's owning index followed by <constant>.</constant> (dot) and its index number.</para> @@ -367,14 +390,15 @@ <entry>Out of the box support for the Apache Lucene back end and the JMS back end. Default to <literal>lucene</literal>. Supports - also <literal>jms</literal> and <literal>blackhole</literal>.</entry> + also <literal>jms</literal> and + <literal>blackhole</literal>.</entry> </row> <row> <entry><literal>hibernate.search.worker.execution</literal></entry> - <entry>Supports synchronous and asynchronous execution. Default - to <literal><literal>sync</literal></literal>. Supports also + <entry>Supports synchronous and asynchronous execution. Default to + <literal><literal>sync</literal></literal>. Supports also <literal>async</literal>.</entry> </row> @@ -445,8 +469,8 @@ <section> <title>Slave nodes</title> - <para>Every index update operation is sent to a JMS queue. Index querying - operations are executed on a local index copy.</para> + <para>Every index update operation is sent to a JMS queue. Index + querying operations are executed on a local index copy.</para> <example> <title>JMS Slave configuration</title> @@ -605,8 +629,9 @@ <para>To enable Hibernate Search in Hibernate Core (ie. if you don't use Hibernate Annotations), add the <literal>FullTextIndexEventListener</literal> for the following six - Hibernate events and also add it after the default - <literal>DefaultFlushEventListener</literal>, as in the following example.</para> + Hibernate events and also add it after the default + <literal>DefaultFlushEventListener</literal>, as in the following + example.</para> <example> <title>Explicitly enabling Hibernate Search by configuring the @@ -768,13 +793,13 @@ terms.</para> <para>This silently truncates large documents, excluding from the index all terms that occur further in the document. If you know your source documents are large, be sure to - set this value high enough to accommodate the expected size. If you - set it to Integer.MAX_VALUE, then the only limit is your memory, - but you should anticipate an OutOfMemoryError. </para> <para>If - setting this value in <literal>batch</literal> differently than in - <literal>transaction</literal> you may get different data (and - results) in your index depending on the indexing - mode.</para></entry> + set this value high enough to accommodate the expected size. If + you set it to Integer.MAX_VALUE, then the only limit is your + memory, but you should anticipate an OutOfMemoryError. </para> + <para>If setting this value in <literal>batch</literal> + differently than in <literal>transaction</literal> you may get + different data (and results) in your index depending on the + indexing mode.</para></entry> <entry>10000</entry> </row> @@ -852,24 +877,26 @@ </tbody> </tgroup> </table> - - <para>To tune the indexing speed it might be useful to time the - object loading from database in isolation from the writes to the index. - To achieve this set the <literal>blackhole</literal> as worker backend and start - you indexing routines. - This backend does not disable Hibernate Search: it will still generate the needed - changesets to the index, but will discard them instead of flushing them to the index. - As opposite to setting the <literal>hibernate.search.indexing_strategy</literal> - to <literal>manual</literal> when using <literal>blackhole</literal> it will possibly load - more data to rebuild the index from associated entities.</para> - + + <para>To tune the indexing speed it might be useful to time the object + loading from database in isolation from the writes to the index. To + achieve this set the <literal>blackhole</literal> as worker backend and + start you indexing routines. This backend does not disable Hibernate + Search: it will still generate the needed changesets to the index, but + will discard them instead of flushing them to the index. As opposite to + setting the <literal>hibernate.search.indexing_strategy</literal> to + <literal>manual</literal> when using <literal>blackhole</literal> it will + possibly load more data to rebuild the index from associated + entities.</para> + <programlisting>hibernate.search.worker.backend blackhole</programlisting> - - <para>The recommended approach is to focus first on optimizing the object loading, and then - use the timings you achieve as a baseline to tune the indexing process.</para> - <para>The <literal>blackhole</literal> backend is not meant to be used in production, only - as a tool to identify indexing bottlenecks.</para> - + + <para>The recommended approach is to focus first on optimizing the object + loading, and then use the timings you achieve as a baseline to tune the + indexing process.</para> + + <para>The <literal>blackhole</literal> backend is not meant to be used in + production, only as a tool to identify indexing bottlenecks.</para> </section> <section id="search-configuration-directory-lockfactories" revision="1"> @@ -883,6 +910,8 @@ for most cases, but it's possible to specify for each index managed by Hibernate Search which LockingFactory you want to use.</para> + + <para>Some of these locking strategies require a filesystem level lock and may be used even on RAM based indexes, but this is not recommended and of no practical use.</para> @@ -976,7 +1005,7 @@ </tgroup> </table></para> - Configuration example: + Configuration example: <programlisting>hibernate.search.default.locking_strategy simple hibernate.search.Animals.locking_strategy native @@ -988,4 +1017,4 @@ </section> -</chapter> +</chapter> \ No newline at end of file Modified: search/trunk/src/main/docbook/en-US/modules/query.xml =================================================================== --- search/trunk/src/main/docbook/en-US/modules/query.xml 2009-06-10 21:48:21 UTC (rev 16754) +++ search/trunk/src/main/docbook/en-US/modules/query.xml 2009-06-11 00:43:09 UTC (rev 16755) @@ -345,8 +345,8 @@ </listitem> <listitem> - <para>FullTextQuery.OBJECT_CLASS: returns the class of the - indexed entity.</para> + <para>FullTextQuery.OBJECT_CLASS: returns the class of the indexed + entity.</para> </listitem> <listitem> @@ -545,7 +545,7 @@ </section> </section> - <section> + <section id="query-filter"> <title>Filters</title> <para>Apache Lucene has a powerful feature that allows to filter query @@ -833,6 +833,105 @@ time spent to execute the query)</para> </listitem> </itemizedlist> + + <section id="query-filter-shard"> + <title>Using filters in a sharded environment</title> + + <para>It is possible, in a sharded environment to execute queries on a + subset of the available shards. This can be done in two steps:</para> + + <itemizedlist> + <listitem> + <para>create a sharding strategy that does select a subset of + <classname>DirectoryProvider</classname>s depending on sone filter + configuration</para> + </listitem> + + <listitem> + <para>activate the proper filter at query time</para> + </listitem> + </itemizedlist> + + <para>Let's first look at an example of sharding strategy that query on + a specific customer shard if the customer filter is activated.</para> + + <programlisting>public class CustomerShardingStrategy implements IndexShardingStrategy { + + // stored DirectoryProviders in a array indexed by customerID + private DirectoryProvider&lt;?&gt;[] providers; + + public void initialize(Properties properties, DirectoryProvider&lt;?&gt;[] providers) { + this.providers = providers; + } + + public DirectoryProvider&lt;?&gt;[] getDirectoryProvidersForAllShards() { + return providers; + } + + public DirectoryProvider&lt;?&gt; getDirectoryProviderForAddition(Class&lt;?&gt; entity, Serializable id, String idInString, Document document) { + Integer customerID = Integer.parseInt(document.getField("customerID").stringValue()); + return providers[customerID]; + } + + public DirectoryProvider&lt;?&gt;[] getDirectoryProvidersForDeletion(Class&lt;?&gt; entity, Serializable id, String idInString) { + return getDirectoryProvidersForAllShards(); + } + +<emphasis role="bold"> /** + * Optimization; don't search ALL shards and union the results; in this case, we + * can be certain that all the data for a particular customer Filter is in a single + * shard; simply return that shard by customerID. + */ + public DirectoryProvider&lt;?&gt;[] getDirectoryProvidersForQuery(FullTextFilterImplementor[] filters) { + FFullTextFilter filter = getCustomerFilter(filters, "customer"); + if (filter == null) { + return getDirectoryProvidersForAllShards(); + } + else { + return new DirectoryProvider[] { providers[Integer.parseInt(filter.getParameter("customerID").toString())] }; + } + } + + private FullTextFilter getFilter(FullTextFilterImplementor[] filters, String name) { + for (FullTextFilterImplementor filter: filters) { + if (filter.getName().equals(name)) return filter; + } + return null; + }</emphasis> + +}</programlisting> + + <para>In this example, if the filter named <literal>customer</literal> + is present, we make sure to only use the shard dedicated to this + customer. Otherwise, we return all shards. A given Sharding strategy can + react to one or more filters and depends on their parameters.</para> + + <para>The second step is simply to activate the filter at query time. + While the filter can be a regular filter (as defined in <xref + linkend="query-filter" />) which also filters Lucene results after the + query, you can make use of a special filter that will only be passed to + the sharding strategy and otherwise ignored for the rest of the query. + Simply use the <classname>ShardSensitiveOnlyFilter</classname> class + when declaring your filter.</para> + + <programlisting>@Entity @Indexed +<emphasis role="bold">@FullTextFilterDef(name="customer", impl=ShardSensitiveOnlyFilter.class)</emphasis> +public class Customer { + ... +} + + +FullTextQuery query = ftEm.createFullTextQuery(luceneQuery, Customer.class); +<emphasis role="bold">query.enableFulltextFilter("customer").setParameter("CustomerID", 5);</emphasis> +@SuppressWarnings("unchecked") +List&lt;Customer&gt; results = query.getResultList();</programlisting> + + <para>Note that by using the + <classname>ShardSensitiveOnlyFilter</classname>, you do not have to + implement any Lucene filter. Using filters and sharding strategy + reacting to these filters is recommended to speed up queries in a + sharded environment.</para> + </section> </section> <section> @@ -866,4 +965,4 @@ run Lucene specific queries. Check <xref linkend="search-lucene-native" /> for more information.</para> </section> -</chapter> +</chapter> \ No newline at end of file Modified: search/trunk/src/main/java/org/hibernate/search/filter/ChainedFilter.java =================================================================== --- search/trunk/src/main/java/org/hibernate/search/filter/ChainedFilter.java 2009-06-10 21:48:21 UTC (rev 16754) +++ search/trunk/src/main/java/org/hibernate/search/filter/ChainedFilter.java 2009-06-11 00:43:09 UTC (rev 16755) @@ -30,6 +30,10 @@ this.chainedFilters.add( filter ); } + public boolean isEmpty() { + return chainedFilters.size() == 0; + } + public BitSet bits(IndexReader reader) throws IOException { throw new UnsupportedOperationException(); } Added: search/trunk/src/main/java/org/hibernate/search/filter/FullTextFilterImplementor.java =================================================================== --- search/trunk/src/main/java/org/hibernate/search/filter/FullTextFilterImplementor.java (rev 0) +++ search/trunk/src/main/java/org/hibernate/search/filter/FullTextFilterImplementor.java 2009-06-11 00:43:09 UTC (rev 16755) @@ -0,0 +1,15 @@ +package org.hibernate.search.filter; + +import org.hibernate.search.FullTextFilter; + +/** + * @author Emmanuel Bernard + */ +public interface FullTextFilterImplementor extends FullTextFilter { + /** + * Returns the Filter name + */ + String getName(); + + //TODO should we expose Map<String, Object> getParameters() +} Added: search/trunk/src/main/java/org/hibernate/search/filter/ShardSensitiveOnlyFilter.java =================================================================== --- search/trunk/src/main/java/org/hibernate/search/filter/ShardSensitiveOnlyFilter.java (rev 0) +++ search/trunk/src/main/java/org/hibernate/search/filter/ShardSensitiveOnlyFilter.java 2009-06-11 00:43:09 UTC (rev 16755) @@ -0,0 +1,12 @@ +package org.hibernate.search.filter; + +/** + * When using this class in @FullTextFilterDef.impl, Hibernate Search + * considers the filter to be only influencing the sharding strategy. + * + * This filter is not applied on the results of the Lucene query. + * + * @author Emmanuel Bernard + */ +public interface ShardSensitiveOnlyFilter { +} Modified: search/trunk/src/main/java/org/hibernate/search/query/FullTextFilterImpl.java =================================================================== --- search/trunk/src/main/java/org/hibernate/search/query/FullTextFilterImpl.java 2009-06-10 21:48:21 UTC (rev 16754) +++ search/trunk/src/main/java/org/hibernate/search/query/FullTextFilterImpl.java 2009-06-11 00:43:09 UTC (rev 16755) @@ -5,11 +5,12 @@ import java.util.Map; import org.hibernate.search.FullTextFilter; +import org.hibernate.search.filter.FullTextFilterImplementor; /** * @author Emmanuel Bernard */ -public class FullTextFilterImpl implements FullTextFilter { +public class FullTextFilterImpl implements FullTextFilterImplementor { private final Map<String, Object> parameters = new HashMap<String, Object>(); private String name; Modified: search/trunk/src/main/java/org/hibernate/search/query/FullTextQueryImpl.java =================================================================== --- search/trunk/src/main/java/org/hibernate/search/query/FullTextQueryImpl.java 2009-06-10 21:48:21 UTC (rev 16754) +++ search/trunk/src/main/java/org/hibernate/search/query/FullTextQueryImpl.java 2009-06-11 00:43:09 UTC (rev 16755) @@ -54,9 +54,12 @@ import org.hibernate.search.filter.ChainedFilter; import org.hibernate.search.filter.FilterKey; import org.hibernate.search.filter.StandardFilterKey; +import org.hibernate.search.filter.FullTextFilterImplementor; +import org.hibernate.search.filter.ShardSensitiveOnlyFilter; import org.hibernate.search.reader.ReaderProvider; import static org.hibernate.search.reader.ReaderProviderHelper.getIndexReaders; import org.hibernate.search.store.DirectoryProvider; +import org.hibernate.search.store.IndexShardingStrategy; import org.hibernate.search.util.ContextHelper; import static org.hibernate.search.util.FilterCacheModeTypeHelper.cacheInstance; import static org.hibernate.search.util.FilterCacheModeTypeHelper.cacheResults; @@ -84,6 +87,7 @@ private Integer resultSize; private Sort sort; private Filter filter; + private Filter userFilter; private Criteria criteria; private String[] indexProjection; private Set<String> idFieldNames; @@ -92,7 +96,9 @@ private SearchFactoryImplementor searchFactoryImplementor; private Map<String, FullTextFilterImpl> filterDefinitions; private int fetchSize = 1; + private static final FullTextFilterImplementor[] EMPTY_FULL_TEXT_FILTER_IMPLEMENTOR = new FullTextFilterImplementor[0]; + /** * Constructs a <code>FullTextQueryImpl</code> instance. * @@ -127,7 +133,7 @@ * {@inheritDoc} */ public FullTextQuery setFilter(Filter filter) { - this.filter = filter; + this.userFilter = filter; return this; } @@ -399,20 +405,27 @@ } private void buildFilters() { - if ( filterDefinitions == null || filterDefinitions.size() == 0 ) { - return; // there is nothing to do if we don't have any filter definitions + ChainedFilter chainedFilter = null; + if ( ! ( filterDefinitions == null || filterDefinitions.size() == 0 ) ) { + chainedFilter = new ChainedFilter(); + for ( FullTextFilterImpl fullTextFilter : filterDefinitions.values() ) { + Filter filter = buildLuceneFilter( fullTextFilter ); + if (filter != null) chainedFilter.addFilter( filter ); + } } - ChainedFilter chainedFilter = new ChainedFilter(); - for ( FullTextFilterImpl fullTextFilter : filterDefinitions.values() ) { - Filter filter = buildLuceneFilter( fullTextFilter ); - chainedFilter.addFilter( filter ); + if ( userFilter != null ) { + //chainedFilter is not always necessary here but the code is easier to read + if (chainedFilter == null) chainedFilter = new ChainedFilter(); + chainedFilter.addFilter( userFilter ); } - if ( filter != null ) { - chainedFilter.addFilter( filter ); + if ( chainedFilter == null || chainedFilter.isEmpty() ) { + filter = null; } - filter = chainedFilter; + else { + filter = chainedFilter; + } } /** @@ -430,6 +443,10 @@ * as FilterCachingStrategy ensure a memory barrier between concurrent thread calls */ FilterDef def = searchFactoryImplementor.getFilterDefinition( fullTextFilter.getName() ); + //def can never be null, ti's guarded by enableFullTextFilter(String) + + if ( isPreQueryFilterOnly(def) ) return null; + Object instance = createFilterInstance( fullTextFilter, def ); FilterKey key = createFilterKey( def, instance ); @@ -449,6 +466,10 @@ return filter; } + private boolean isPreQueryFilterOnly(FilterDef def) { + return def.getImpl().equals( ShardSensitiveOnlyFilter.class ); + } + private Filter createFilter(FilterDef def, Object instance) { Filter filter; if ( def.getFactoryMethod() != null ) { @@ -633,7 +654,7 @@ */ private IndexSearcher buildSearcher(SearchFactoryImplementor searchFactoryImplementor) { Map<Class<?>, DocumentBuilderIndexedEntity<?>> builders = searchFactoryImplementor.getDocumentBuildersIndexedEntities(); - List<DirectoryProvider> directories = new ArrayList<DirectoryProvider>(); + List<DirectoryProvider> targetedDirectories = new ArrayList<DirectoryProvider>(); Set<String> idFieldNames = new HashSet<String>(); Similarity searcherSimilarity = null; @@ -653,9 +674,7 @@ idFieldNames.add( builder.getIdKeywordName() ); allowFieldSelectionInProjection = allowFieldSelectionInProjection && builder.allowFieldSelectionInProjection(); } - final DirectoryProvider[] directoryProviders = builder.getDirectoryProviderSelectionStrategy() - .getDirectoryProvidersForAllShards(); - populateDirectories( directories, directoryProviders ); + populateDirectories( targetedDirectories, builder ); } classesAndSubclasses = null; } @@ -679,10 +698,8 @@ idFieldNames.add( builder.getIdKeywordName() ); allowFieldSelectionInProjection = allowFieldSelectionInProjection && builder.allowFieldSelectionInProjection(); } - final DirectoryProvider[] directoryProviders = builder.getDirectoryProviderSelectionStrategy() - .getDirectoryProvidersForAllShards(); searcherSimilarity = checkSimilarity( searcherSimilarity, builder ); - populateDirectories( directories, directoryProviders ); + populateDirectories( targetedDirectories, builder ); } this.classesAndSubclasses = involvedClasses; } @@ -691,7 +708,7 @@ //compute optimization needClassFilterClause //if at least one DP contains one class that is not part of the targeted classesAndSubclasses we can't optimize if ( classesAndSubclasses != null ) { - for ( DirectoryProvider dp : directories ) { + for ( DirectoryProvider dp : targetedDirectories ) { final Set<Class<?>> classesInDirectoryProvider = searchFactoryImplementor.getClassesInDirectoryProvider( dp ); @@ -712,7 +729,7 @@ } //set up the searcher - final DirectoryProvider[] directoryProviders = directories.toArray( new DirectoryProvider[directories.size()] ); + final DirectoryProvider[] directoryProviders = targetedDirectories.toArray( new DirectoryProvider[targetedDirectories.size()] ); IndexSearcher is = new IndexSearcher( searchFactoryImplementor.getReaderProvider().openReader( directoryProviders @@ -722,7 +739,19 @@ return is; } - private void populateDirectories(List<DirectoryProvider> directories, DirectoryProvider[] directoryProviders) { + private void populateDirectories(List<DirectoryProvider> directories, DocumentBuilderIndexedEntity builder) { + final IndexShardingStrategy indexShardingStrategy = builder.getDirectoryProviderSelectionStrategy(); + final DirectoryProvider[] directoryProviders; + if ( filterDefinitions != null && !filterDefinitions.isEmpty() ) { + directoryProviders = indexShardingStrategy.getDirectoryProvidersForQuery( + filterDefinitions.values().toArray( new FullTextFilterImplementor[filterDefinitions.size()] ) + ); + } + else { + //no filter get all shards + directoryProviders = indexShardingStrategy.getDirectoryProvidersForQuery( EMPTY_FULL_TEXT_FILTER_IMPLEMENTOR ); + } + for ( DirectoryProvider provider : directoryProviders ) { if ( !directories.contains( provider ) ) { directories.add( provider ); Modified: search/trunk/src/main/java/org/hibernate/search/store/IdHashShardingStrategy.java =================================================================== --- search/trunk/src/main/java/org/hibernate/search/store/IdHashShardingStrategy.java 2009-06-10 21:48:21 UTC (rev 16754) +++ search/trunk/src/main/java/org/hibernate/search/store/IdHashShardingStrategy.java 2009-06-11 00:43:09 UTC (rev 16755) @@ -6,6 +6,8 @@ import org.apache.lucene.document.Document; +import org.hibernate.search.filter.FullTextFilterImplementor; + /** * This implementation use idInString as the hashKey. * @@ -31,6 +33,10 @@ return new DirectoryProvider[] { providers[hashKey( idInString )] }; } + public DirectoryProvider<?>[] getDirectoryProvidersForQuery(FullTextFilterImplementor[] fullTextFilters) { + return getDirectoryProvidersForAllShards(); + } + private int hashKey(String key) { // reproduce the hashCode implementation of String as documented in the javadoc // to be safe cross Java version (in case it changes some day) Modified: search/trunk/src/main/java/org/hibernate/search/store/IndexShardingStrategy.java =================================================================== --- search/trunk/src/main/java/org/hibernate/search/store/IndexShardingStrategy.java 2009-06-10 21:48:21 UTC (rev 16754) +++ search/trunk/src/main/java/org/hibernate/search/store/IndexShardingStrategy.java 2009-06-11 00:43:09 UTC (rev 16755) @@ -6,6 +6,8 @@ import org.apache.lucene.document.Document; +import org.hibernate.search.filter.FullTextFilterImplementor; + /** * Defines how a given virtual index shards data into different DirectoryProviders * @@ -32,4 +34,13 @@ * id and idInString can be null. If null, all the directory providers containing entity types should be returned */ DirectoryProvider<?>[] getDirectoryProvidersForDeletion(Class<?> entity, Serializable id, String idInString); + + /** + * return the set of DirectoryProvider(s) where the entities matching the filters are stored + * this optional optimization allows queries to hit a subset of all shards, which may be useful for some datasets + * if this optimization is not needed, return getDirectoryProvidersForAllShards() + * + * fullTextFilters can be empty if no filter is applied + */ + DirectoryProvider<?>[] getDirectoryProvidersForQuery(FullTextFilterImplementor[] fullTextFilters); } Modified: search/trunk/src/main/java/org/hibernate/search/store/NotShardedStrategy.java =================================================================== --- search/trunk/src/main/java/org/hibernate/search/store/NotShardedStrategy.java 2009-06-10 21:48:21 UTC (rev 16754) +++ search/trunk/src/main/java/org/hibernate/search/store/NotShardedStrategy.java 2009-06-11 00:43:09 UTC (rev 16755) @@ -6,6 +6,7 @@ import org.apache.lucene.document.Document; import org.hibernate.annotations.common.AssertionFailure; +import org.hibernate.search.filter.FullTextFilterImplementor; /** * @author Emmanuel Bernard @@ -31,4 +32,8 @@ return directoryProvider; } + public DirectoryProvider<?>[] getDirectoryProvidersForQuery(FullTextFilterImplementor[] fullTextFilters) { + return directoryProvider; + } + } Modified: search/trunk/src/test/java/org/hibernate/search/test/configuration/UselessShardingStrategy.java =================================================================== --- search/trunk/src/test/java/org/hibernate/search/test/configuration/UselessShardingStrategy.java 2009-06-10 21:48:21 UTC (rev 16754) +++ search/trunk/src/test/java/org/hibernate/search/test/configuration/UselessShardingStrategy.java 2009-06-11 00:43:09 UTC (rev 16755) @@ -8,6 +8,7 @@ import org.apache.lucene.document.Document; import org.hibernate.search.store.DirectoryProvider; import org.hibernate.search.store.IndexShardingStrategy; +import org.hibernate.search.filter.FullTextFilterImplementor; /** * Used to test the configuration of a third-party strategy @@ -27,6 +28,10 @@ return null; } + public DirectoryProvider<?>[] getDirectoryProvidersForQuery(FullTextFilterImplementor[] fullTextFilters) { + return null; + } + public void initialize(Properties properties, DirectoryProvider<?>[] providers) { Enumeration<?> propertyNames = properties.propertyNames(); int counter; Added: search/trunk/src/test/java/org/hibernate/search/test/shards/CustomerShardingStrategy.java =================================================================== --- search/trunk/src/test/java/org/hibernate/search/test/shards/CustomerShardingStrategy.java (rev 0) +++ search/trunk/src/test/java/org/hibernate/search/test/shards/CustomerShardingStrategy.java 2009-06-11 00:43:09 UTC (rev 16755) @@ -0,0 +1,65 @@ +package org.hibernate.search.test.shards; + +import java.io.Serializable; +import java.util.Properties; + +import org.apache.lucene.document.Document; + +import org.hibernate.search.FullTextFilter; +import org.hibernate.search.filter.FullTextFilterImplementor; +import org.hibernate.search.store.DirectoryProvider; +import org.hibernate.search.store.IndexShardingStrategy; + +/** + * Shards an index containing data for multiple customers by customerID. customerID is + * provided as a property on all indexes entities, and is also defined as a Filter. + * + * The number of shards should be configured to be MAX(customerID). + * + * @author Chase Seibert + */ +public class CustomerShardingStrategy implements IndexShardingStrategy { + + // stored DirectoryProviders in a array indexed by customerID + private DirectoryProvider<?>[] providers; + + public void initialize(Properties properties, DirectoryProvider<?>[] providers) { + this.providers = providers; + } + + public DirectoryProvider<?>[] getDirectoryProvidersForAllShards() { + return providers; + } + + public DirectoryProvider<?> getDirectoryProviderForAddition(Class<?> entity, Serializable id, String idInString, Document document) { + Integer customerID = Integer.parseInt(document.getField("customerID").stringValue()); + return providers[customerID]; + } + + public DirectoryProvider<?>[] getDirectoryProvidersForDeletion(Class<?> entity, Serializable id, String idInString) { + return getDirectoryProvidersForAllShards(); + } + + /** + * Optimization; don't search ALL shards and union the results; in this case, we + * can be certain that all the data for a particular customer Filter is in a single + * shard; simply return that shard by customerID. + */ + public DirectoryProvider<?>[] getDirectoryProvidersForQuery(FullTextFilterImplementor[] filters) { + FullTextFilter filter = getCustomerFilter(filters, "customer"); + if (filter == null) { + return getDirectoryProvidersForAllShards(); + } + else { + return new DirectoryProvider[] { providers[Integer.parseInt(filter.getParameter("customerID").toString())] }; + } + } + + private FullTextFilter getCustomerFilter(FullTextFilterImplementor[] filters, String name) { + for (FullTextFilterImplementor filter: filters) { + if (filter.getName().equals(name)) return filter; + } + return null; + } + +} Added: search/trunk/src/test/java/org/hibernate/search/test/shards/CustomerShardingStrategyTest.java =================================================================== --- search/trunk/src/test/java/org/hibernate/search/test/shards/CustomerShardingStrategyTest.java (rev 0) +++ search/trunk/src/test/java/org/hibernate/search/test/shards/CustomerShardingStrategyTest.java 2009-06-11 00:43:09 UTC (rev 16755) @@ -0,0 +1,56 @@ +package org.hibernate.search.test.shards; + +import junit.framework.TestCase; +import org.apache.lucene.document.Document; +import org.apache.lucene.document.Field; + +import org.hibernate.search.query.FullTextFilterImpl; +import org.hibernate.search.store.DirectoryProvider; +import org.hibernate.search.store.RAMDirectoryProvider; + +/** + * @author Chase Seibert + */ +public class CustomerShardingStrategyTest extends TestCase { + + private CustomerShardingStrategy shardStrategy; + + protected void setUp() throws Exception { + shardStrategy = new CustomerShardingStrategy(); + + // initilaize w/ 10 shards + shardStrategy.initialize( null, new DirectoryProvider[] { + new RAMDirectoryProvider(), + new RAMDirectoryProvider(), + new RAMDirectoryProvider(), + new RAMDirectoryProvider(), + new RAMDirectoryProvider(), + new RAMDirectoryProvider(), + new RAMDirectoryProvider(), + new RAMDirectoryProvider(), + new RAMDirectoryProvider(), + new RAMDirectoryProvider() + } ); + } + + public void testGetDirectoryProvidersForQuery() { + + FullTextFilterImpl filter = new FullTextFilterImpl(); + filter.setName("customer"); + filter.setParameter("customerID", 5); + + // customerID == 5 should correspond to just a single shard instance + DirectoryProvider[] providers = shardStrategy.getDirectoryProvidersForQuery(new FullTextFilterImpl[] { filter }); + assertTrue(providers.length == 1); + + // create a dummy document for the same customerID, and make sure the shard it would be + // indexed on matches the shard returned by getDirectoryProvidersForQuery() + Document document = new Document(); + document.add(new Field("customerID", "5", Field.Store.NO, Field.Index.UN_TOKENIZED)); + + assertTrue(providers[0].equals( + shardStrategy.getDirectoryProviderForAddition(null, null, null, document) + )); + } + +} \ No newline at end of file