Pass empty QueryCollectorContext in case of hybrid query to improve l…

…atencies by 20% (#731) * Pass empty QueryCollectorContext in case of hybrid query Signed-off-by: Martin Gaievski <[email protected]>
opensearch-project · May 6, 2024 · 2c556d2 · 2c556d2
1 parent 2dd2e8a
commit 2c556d2
Show file tree

Hide file tree

Showing 3 changed files with 118 additions and 41 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -15,6 +15,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 ## [Unreleased 2.x](https://github.com/opensearch-project/neural-search/compare/2.14...2.x)
 ### Features
 ### Enhancements
+- Pass empty doc collector instead of top docs collector to improve hybrid query latencies by 20% ([#731](https://github.com/opensearch-project/neural-search/pull/731))
 ### Bug Fixes
 - Fix multi node "no such index" error in text chunking processor ([#713](https://github.com/opensearch-project/neural-search/pull/713))
 ### Infrastructure

diff --git a/src/main/java/org/opensearch/neuralsearch/search/query/HybridQueryPhaseSearcher.java b/src/main/java/org/opensearch/neuralsearch/search/query/HybridQueryPhaseSearcher.java
@@ -10,6 +10,8 @@
 import java.util.stream.Collectors;
 
 import com.google.common.annotations.VisibleForTesting;
+import lombok.AccessLevel;
+import lombok.NoArgsConstructor;
 import org.apache.lucene.search.BooleanClause;
 import org.apache.lucene.search.BooleanQuery;
 import org.apache.lucene.search.Query;
@@ -19,8 +21,10 @@
 import org.opensearch.search.aggregations.AggregationProcessor;
 import org.opensearch.search.internal.ContextIndexSearcher;
 import org.opensearch.search.internal.SearchContext;
+import org.opensearch.search.query.ConcurrentQueryPhaseSearcher;
 import org.opensearch.search.query.QueryCollectorContext;
 import org.opensearch.search.query.QueryPhase;
+import org.opensearch.search.query.QueryPhaseSearcher;
 import org.opensearch.search.query.QueryPhaseSearcherWrapper;
 
 import lombok.extern.log4j.Log4j2;
@@ -36,6 +40,14 @@
 @Log4j2
 public class HybridQueryPhaseSearcher extends QueryPhaseSearcherWrapper {
 
+    private final QueryPhaseSearcher defaultQueryPhaseSearcherWithEmptyCollectorContext;
+    private final QueryPhaseSearcher concurrentQueryPhaseSearcherWithEmptyCollectorContext;
+
+    public HybridQueryPhaseSearcher() {
+        this.defaultQueryPhaseSearcherWithEmptyCollectorContext = new DefaultQueryPhaseSearcherWithEmptyQueryCollectorContext();
+        this.concurrentQueryPhaseSearcherWithEmptyCollectorContext = new ConcurrentQueryPhaseSearcherWithEmptyQueryCollectorContext();
+    }
+
     public boolean searchWith(
         final SearchContext searchContext,
         final ContextIndexSearcher searcher,
@@ -49,10 +61,17 @@ public boolean searchWith(
             return super.searchWith(searchContext, searcher, query, collectors, hasFilterCollector, hasTimeout);
         } else {
             Query hybridQuery = extractHybridQuery(searchContext, query);
-            return super.searchWith(searchContext, searcher, hybridQuery, collectors, hasFilterCollector, hasTimeout);
+            QueryPhaseSearcher queryPhaseSearcher = getQueryPhaseSearcher(searchContext);
+            return queryPhaseSearcher.searchWith(searchContext, searcher, hybridQuery, collectors, hasFilterCollector, hasTimeout);
         }
     }
 
+    private QueryPhaseSearcher getQueryPhaseSearcher(final SearchContext searchContext) {
+        return searchContext.shouldUseConcurrentSearch()
+            ? concurrentQueryPhaseSearcherWithEmptyCollectorContext
+            : defaultQueryPhaseSearcherWithEmptyCollectorContext;
+    }
+
     private static boolean isWrappedHybridQuery(final Query query) {
         return query instanceof BooleanQuery
             && ((BooleanQuery) query).clauses().stream().anyMatch(clauseQuery -> clauseQuery.getQuery() instanceof HybridQuery);
@@ -132,4 +151,60 @@ public AggregationProcessor aggregationProcessor(SearchContext searchContext) {
         AggregationProcessor coreAggProcessor = super.aggregationProcessor(searchContext);
         return new HybridAggregationProcessor(coreAggProcessor);
     }
+
+    /**
+     * Class that inherits ConcurrentQueryPhaseSearcher implementation but calls its search with only
+     * empty query collector context
+     */
+    @NoArgsConstructor(access = AccessLevel.PRIVATE)
+    final class ConcurrentQueryPhaseSearcherWithEmptyQueryCollectorContext extends ConcurrentQueryPhaseSearcher {
+
+        @Override
+        protected boolean searchWithCollector(
+            SearchContext searchContext,
+            ContextIndexSearcher searcher,
+            Query query,
+            LinkedList<QueryCollectorContext> collectors,
+            boolean hasFilterCollector,
+            boolean hasTimeout
+        ) throws IOException {
+            return searchWithCollector(
+                searchContext,
+                searcher,
+                query,
+                collectors,
+                QueryCollectorContext.EMPTY_CONTEXT,
+                hasFilterCollector,
+                hasTimeout
+            );
+        }
+    }
+
+    /**
+     * Class that inherits DefaultQueryPhaseSearcher implementation but calls its search with only
+     * empty query collector context
+     */
+    @NoArgsConstructor(access = AccessLevel.PACKAGE)
+    final class DefaultQueryPhaseSearcherWithEmptyQueryCollectorContext extends QueryPhase.DefaultQueryPhaseSearcher {
+
+        @Override
+        protected boolean searchWithCollector(
+            SearchContext searchContext,
+            ContextIndexSearcher searcher,
+            Query query,
+            LinkedList<QueryCollectorContext> collectors,
+            boolean hasFilterCollector,
+            boolean hasTimeout
+        ) throws IOException {
+            return searchWithCollector(
+                searchContext,
+                searcher,
+                query,
+                collectors,
+                QueryCollectorContext.EMPTY_CONTEXT,
+                hasFilterCollector,
+                hasTimeout
+            );
+        }
+    }
 }
diff --git a/src/test/java/org/opensearch/neuralsearch/search/query/HybridQueryPhaseSearcherTests.java b/src/test/java/org/opensearch/neuralsearch/search/query/HybridQueryPhaseSearcherTests.java
@@ -20,13 +20,11 @@
 import static org.opensearch.neuralsearch.search.util.HybridSearchResultFormatUtil.isHybridQueryStartStopElement;
 
 import java.io.IOException;
-import java.util.Arrays;
+import java.util.HashMap;
 import java.util.LinkedList;
-import java.util.List;
 import java.util.Map;
 import java.util.Set;
-import java.util.UUID;
-import java.util.stream.Collectors;
+import java.util.concurrent.ExecutorService;
 
 import org.apache.lucene.document.Document;
 import org.apache.lucene.document.Field;
@@ -39,6 +37,8 @@
 import org.apache.lucene.index.IndexWriter;
 import org.apache.lucene.search.BooleanClause;
 import org.apache.lucene.search.BooleanQuery;
+import org.apache.lucene.search.Collector;
+import org.apache.lucene.search.CollectorManager;
 import org.apache.lucene.search.IndexSearcher;
 import org.apache.lucene.search.Query;
 import org.apache.lucene.search.ScoreDoc;
@@ -58,8 +58,6 @@
 import org.opensearch.index.mapper.MapperService;
 import org.opensearch.index.mapper.TextFieldMapper;
 import org.opensearch.index.query.BoolQueryBuilder;
-import org.opensearch.index.query.MatchAllQueryBuilder;
-import org.opensearch.index.query.QueryBuilder;
 import org.opensearch.index.query.QueryBuilders;
 import org.opensearch.index.query.QueryShardContext;
 import org.opensearch.index.query.TermQueryBuilder;
@@ -78,6 +76,7 @@
 import lombok.SneakyThrows;
 import org.opensearch.search.query.QueryCollectorContext;
 import org.opensearch.search.query.QuerySearchResult;
+import org.opensearch.search.query.ReduceableSearchResult;
 
 public class HybridQueryPhaseSearcherTests extends OpenSearchQueryTestCase {
     private static final String VECTOR_FIELD_NAME = "vectorField";
@@ -88,13 +87,7 @@ public class HybridQueryPhaseSearcherTests extends OpenSearchQueryTestCase {
     private static final String TEST_DOC_TEXT4 = "This is really nice place to be";
     private static final String QUERY_TEXT1 = "hello";
     private static final String QUERY_TEXT2 = "randomkeyword";
-    private static final String QUERY_TEXT3 = "place";
     private static final Index dummyIndex = new Index("dummy", "dummy");
-    private static final String MODEL_ID = "mfgfgdsfgfdgsde";
-    private static final int K = 10;
-    private static final QueryBuilder TEST_FILTER = new MatchAllQueryBuilder();
-    private static final UUID INDEX_UUID = UUID.randomUUID();
-    private static final String TEST_INDEX = "index";
 
     @SneakyThrows
     public void testQueryType_whenQueryIsHybrid_thenCallHybridDocCollector() {
@@ -306,20 +299,22 @@ public void testQueryResult_whenOneSubQueryWithHits_thenHybridResultsAreSet() {
         Query query = queryBuilder.toQuery(mockQueryShardContext);
         when(searchContext.query()).thenReturn(query);
 
+        CollectorManager<? extends Collector, ReduceableSearchResult> collectorManager = HybridCollectorManager
+            .createHybridCollectorManager(searchContext);
+        Map<Class<?>, CollectorManager<? extends Collector, ReduceableSearchResult>> queryCollectorManagers = new HashMap<>();
+        queryCollectorManagers.put(HybridCollectorManager.class, collectorManager);
+        when(searchContext.queryCollectorManagers()).thenReturn(queryCollectorManagers);
+
         hybridQueryPhaseSearcher.searchWith(searchContext, contextIndexSearcher, query, collectors, hasFilterCollector, hasTimeout);
+        hybridQueryPhaseSearcher.aggregationProcessor(searchContext).postProcess(searchContext);
 
         assertNotNull(querySearchResult.topDocs());
         TopDocsAndMaxScore topDocsAndMaxScore = querySearchResult.topDocs();
         TopDocs topDocs = topDocsAndMaxScore.topDocs;
-        assertEquals(1, topDocs.totalHits.value);
+        assertEquals(0, topDocs.totalHits.value);
         ScoreDoc[] scoreDocs = topDocs.scoreDocs;
         assertNotNull(scoreDocs);
-        assertEquals(1, scoreDocs.length);
-        ScoreDoc scoreDoc = scoreDocs[0];
-        assertNotNull(scoreDoc);
-        int actualDocId = Integer.parseInt(reader.document(scoreDoc.doc).getField("id").stringValue());
-        assertEquals(docId1, actualDocId);
-        assertTrue(scoreDoc.score > 0.0f);
+        assertEquals(0, scoreDocs.length);
 
         releaseResources(directory, w, reader);
     }
@@ -340,13 +335,7 @@ public void testQueryResult_whenMultipleTextSubQueriesWithSomeHits_thenHybridRes
         ft.setOmitNorms(random().nextBoolean());
         ft.freeze();
         int docId1 = RandomizedTest.randomInt();
-        int docId2 = RandomizedTest.randomInt();
-        int docId3 = RandomizedTest.randomInt();
-        int docId4 = RandomizedTest.randomInt();
         w.addDocument(getDocument(TEXT_FIELD_NAME, docId1, TEST_DOC_TEXT1, ft));
-        w.addDocument(getDocument(TEXT_FIELD_NAME, docId2, TEST_DOC_TEXT2, ft));
-        w.addDocument(getDocument(TEXT_FIELD_NAME, docId3, TEST_DOC_TEXT3, ft));
-        w.addDocument(getDocument(TEXT_FIELD_NAME, docId4, TEST_DOC_TEXT4, ft));
         w.commit();
 
         IndexReader reader = DirectoryReader.open(w);
@@ -395,18 +384,22 @@ public void testQueryResult_whenMultipleTextSubQueriesWithSomeHits_thenHybridRes
         Query query = queryBuilder.toQuery(mockQueryShardContext);
         when(searchContext.query()).thenReturn(query);
 
+        CollectorManager<? extends Collector, ReduceableSearchResult> collectorManager = HybridCollectorManager
+            .createHybridCollectorManager(searchContext);
+        Map<Class<?>, CollectorManager<? extends Collector, ReduceableSearchResult>> queryCollectorManagers = new HashMap<>();
+        queryCollectorManagers.put(HybridCollectorManager.class, collectorManager);
+        when(searchContext.queryCollectorManagers()).thenReturn(queryCollectorManagers);
+
         hybridQueryPhaseSearcher.searchWith(searchContext, contextIndexSearcher, query, collectors, hasFilterCollector, hasTimeout);
+        hybridQueryPhaseSearcher.aggregationProcessor(searchContext).postProcess(searchContext);
 
         assertNotNull(querySearchResult.topDocs());
         TopDocsAndMaxScore topDocsAndMaxScore = querySearchResult.topDocs();
         TopDocs topDocs = topDocsAndMaxScore.topDocs;
-        assertEquals(4, topDocs.totalHits.value);
+        assertEquals(0, topDocs.totalHits.value);
         ScoreDoc[] scoreDocs = topDocs.scoreDocs;
         assertNotNull(scoreDocs);
-        assertEquals(4, scoreDocs.length);
-        List<Integer> expectedIds = List.of(0, 1, 2, 3);
-        List<Integer> actualDocIds = Arrays.stream(scoreDocs).map(sd -> sd.doc).collect(Collectors.toList());
-        assertEquals(expectedIds, actualDocIds);
+        assertEquals(0, scoreDocs.length);
 
         releaseResources(directory, w, reader);
     }
@@ -705,18 +698,22 @@ public void testWrappedHybridQuery_whenHybridWrappedIntoBoolBecauseOfNested_then
 
         when(searchContext.query()).thenReturn(query);
 
+        CollectorManager<? extends Collector, ReduceableSearchResult> collectorManager = HybridCollectorManager
+            .createHybridCollectorManager(searchContext);
+        Map<Class<?>, CollectorManager<? extends Collector, ReduceableSearchResult>> queryCollectorManagers = new HashMap<>();
+        queryCollectorManagers.put(HybridCollectorManager.class, collectorManager);
+        when(searchContext.queryCollectorManagers()).thenReturn(queryCollectorManagers);
+
         hybridQueryPhaseSearcher.searchWith(searchContext, contextIndexSearcher, query, collectors, hasFilterCollector, hasTimeout);
+        hybridQueryPhaseSearcher.aggregationProcessor(searchContext).postProcess(searchContext);
 
         assertNotNull(querySearchResult.topDocs());
         TopDocsAndMaxScore topDocsAndMaxScore = querySearchResult.topDocs();
         TopDocs topDocs = topDocsAndMaxScore.topDocs;
-        assertTrue(topDocs.totalHits.value > 0);
+        assertEquals(0, topDocs.totalHits.value);
         ScoreDoc[] scoreDocs = topDocs.scoreDocs;
         assertNotNull(scoreDocs);
-        assertEquals(1, scoreDocs.length);
-        ScoreDoc scoreDoc = scoreDocs[0];
-        assertTrue(scoreDoc.score > 0);
-        assertEquals(0, scoreDoc.doc);
+        assertEquals(0, scoreDocs.length);
 
         releaseResources(directory, w, reader);
     }
@@ -979,18 +976,22 @@ public void testAliasWithFilter_whenHybridWrappedIntoBoolBecauseOfIndexAlias_the
         when(searchContext.query()).thenReturn(query);
         when(searchContext.aliasFilter()).thenReturn(termFilter);
 
+        CollectorManager<? extends Collector, ReduceableSearchResult> collectorManager = HybridCollectorManager
+            .createHybridCollectorManager(searchContext);
+        Map<Class<?>, CollectorManager<? extends Collector, ReduceableSearchResult>> queryCollectorManagers = new HashMap<>();
+        queryCollectorManagers.put(HybridCollectorManager.class, collectorManager);
+        when(searchContext.queryCollectorManagers()).thenReturn(queryCollectorManagers);
+
         hybridQueryPhaseSearcher.searchWith(searchContext, contextIndexSearcher, query, collectors, hasFilterCollector, hasTimeout);
+        hybridQueryPhaseSearcher.aggregationProcessor(searchContext).postProcess(searchContext);
 
         assertNotNull(querySearchResult.topDocs());
         TopDocsAndMaxScore topDocsAndMaxScore = querySearchResult.topDocs();
         TopDocs topDocs = topDocsAndMaxScore.topDocs;
-        assertTrue(topDocs.totalHits.value > 0);
+        assertEquals(0, topDocs.totalHits.value);
         ScoreDoc[] scoreDocs = topDocs.scoreDocs;
         assertNotNull(scoreDocs);
-        assertEquals(1, scoreDocs.length);
-        ScoreDoc scoreDoc = scoreDocs[0];
-        assertTrue(scoreDoc.score > 0);
-        assertEquals(0, scoreDoc.doc);
+        assertEquals(0, scoreDocs.length);
 
         releaseResources(directory, w, reader);
     }