LUCENE-6850: Optimize BooleanScorer for sparse clauses.

git-svn-id: https://svn.apache.org/repos/asf/lucene/dev/trunk@1710591 13f79535-47bb-0310-9956-ffa450edef68

lucene/CHANGES.txt

@@ -156,6 +156,8 @@ Optimizations
   strings larger than 64kb by an amount equal to string's utf8 size.
   (Dawid Weiss, Robert Muir, shalin)
 
+* LUCENE-6850: Optimize BooleanScorer for sparse clauses. (Adrien Grand)
+
 Bug Fixes
 
 * LUCENE-6817: ComplexPhraseQueryParser.ComplexPhraseQuery does not display 

lucene/core/src/java/org/apache/lucene/search/BooleanScorer.java

@@ -83,11 +83,11 @@ final class BooleanScorer extends BulkScorer {
     }
 
     void advance(int min) throws IOException {
-      score(null, min, min);
+      score(orCollector, null, min, min);
     }
 
-    void score(Bits acceptDocs, int min, int max) throws IOException {
+    void score(LeafCollector collector, Bits acceptDocs, int min, int max) throws IOException {
-      next = scorer.score(orCollector, acceptDocs, min, max);
+      next = scorer.score(collector, acceptDocs, min, max);
     }
   }
 
@@ -179,6 +179,9 @@ final class BooleanScorer extends BulkScorer {
     if (minShouldMatch < 1 || minShouldMatch > scorers.size()) {
       throw new IllegalArgumentException("minShouldMatch should be within 1..num_scorers. Got " + minShouldMatch);
     }
+    if (scorers.size() <= 1) {
+      throw new IllegalArgumentException("This scorer can only be used with two scorers or more, got " + scorers.size());
+    }
     for (int i = 0; i < buckets.length; i++) {
       buckets[i] = new Bucket();
     }
@@ -237,12 +240,12 @@ final class BooleanScorer extends BulkScorer {
     }
   }
 
-  private void scoreWindow(LeafCollector collector, Bits acceptDocs, int base, int min, int max,
+  private void scoreWindowIntoBitSetAndReplay(LeafCollector collector, Bits acceptDocs,
-      BulkScorerAndDoc[] scorers, int numScorers) throws IOException {
+      int base, int min, int max, BulkScorerAndDoc[] scorers, int numScorers) throws IOException {
     for (int i = 0; i < numScorers; ++i) {
       final BulkScorerAndDoc scorer = scorers[i];
       assert scorer.next < max;
-      scorer.score(acceptDocs, min, max);
+      scorer.score(orCollector, acceptDocs, min, max);
     }
 
     scoreMatches(collector, base);
@@ -270,14 +273,7 @@ final class BooleanScorer extends BulkScorer {
     return headTop;
   }
 
-  private void scoreWindow(LeafCollector collector, Bits acceptDocs, int windowBase, int windowMin, int windowMax) throws IOException {
+  private void scoreWindowMultipleScorers(LeafCollector collector, Bits acceptDocs, int windowBase, int windowMin, int windowMax, int maxFreq) throws IOException {
-    // Fill 'leads' with all scorers from 'head' that are in the right window
-    leads[0] = head.pop();
-    int maxFreq = 1;
-    while (head.size() > 0 && head.top().next < windowMax) {
-      leads[maxFreq++] = head.pop();
-    }
-
     while (maxFreq < minShouldMatch && maxFreq + tail.size() >= minShouldMatch) {
       // a match is still possible
       final BulkScorerAndDoc candidate = tail.pop();
@@ -296,7 +292,7 @@ final class BooleanScorer extends BulkScorer {
       }
       tail.clear();
 
-      scoreWindow(collector, acceptDocs, windowBase, windowMin, windowMax, leads, maxFreq);
+      scoreWindowIntoBitSetAndReplay(collector, acceptDocs, windowBase, windowMin, windowMax, leads, maxFreq);
     }
 
     // Push back scorers into head and tail
@@ -308,21 +304,64 @@ final class BooleanScorer extends BulkScorer {
     }
   }
 
+  private void scoreWindowSingleScorer(BulkScorerAndDoc bulkScorer, LeafCollector collector,
+      Bits acceptDocs, int windowMin, int windowMax, int max) throws IOException {
+    assert tail.size() == 0;
+    final int nextWindowBase = head.top().next & ~MASK;
+    final int end = Math.max(windowMax, Math.min(max, nextWindowBase));
+    
+    bulkScorer.score(collector, acceptDocs, windowMin, end);
+    
+    // reset the scorer that should be used for the general case
+    collector.setScorer(fakeScorer);
+  }
+
+  private BulkScorerAndDoc scoreWindow(BulkScorerAndDoc top, LeafCollector collector,
+      LeafCollector singleClauseCollector, Bits acceptDocs, int min, int max) throws IOException {
+    final int windowBase = top.next & ~MASK; // find the window that the next match belongs to
+    final int windowMin = Math.max(min, windowBase);
+    final int windowMax = Math.min(max, windowBase + SIZE);
+
+    // Fill 'leads' with all scorers from 'head' that are in the right window
+    leads[0] = head.pop();
+    int maxFreq = 1;
+    while (head.size() > 0 && head.top().next < windowMax) {
+      leads[maxFreq++] = head.pop();
+    }
+
+    if (minShouldMatch == 1 && maxFreq == 1) {
+      // special case: only one scorer can match in the current window,
+      // we can collect directly
+      final BulkScorerAndDoc bulkScorer = leads[0];
+      scoreWindowSingleScorer(bulkScorer, singleClauseCollector, acceptDocs, windowMin, windowMax, max);
+      return head.add(bulkScorer);
+    } else {
+      // general case, collect through a bit set first and then replay
+      scoreWindowMultipleScorers(collector, acceptDocs, windowBase, windowMin, windowMax, maxFreq);
+      return head.top();
+    }
+  }
+
   @Override
   public int score(LeafCollector collector, Bits acceptDocs, int min, int max) throws IOException {
     fakeScorer.doc = -1;
     collector.setScorer(fakeScorer);
 
+    final LeafCollector singleClauseCollector;
+    if (coordFactors[1] == 1f) {
+      singleClauseCollector = collector;
+    } else {
+      singleClauseCollector = new FilterLeafCollector(collector) {
+        @Override
+        public void setScorer(Scorer scorer) throws IOException {
+          super.setScorer(new BooleanTopLevelScorers.BoostedScorer(scorer, coordFactors[1]));
+        }
+      };
+    }
+
     BulkScorerAndDoc top = advance(min);
     while (top.next < max) {
-
+      top = scoreWindow(top, collector, singleClauseCollector, acceptDocs, min, max);
-      final int windowBase = top.next & ~MASK; // find the window that the next match belongs to
-      final int windowMin = Math.max(min, windowBase);
-      final int windowMax = Math.min(max, windowBase + SIZE);
-
-      // general case
-      scoreWindow(collector, acceptDocs, windowBase, windowMin, windowMax);
-      top = head.top();
     }
 
     return top.next;

lucene/core/src/java/org/apache/lucene/search/BooleanTopLevelScorers.java

@@ -22,6 +22,8 @@ import java.util.Arrays;
 import java.util.Collection;
 import java.util.Collections;
 
+import org.apache.lucene.util.Bits;
+
 /** Internal document-at-a-time scorers used to deal with stupid coord() computation */
 class BooleanTopLevelScorers {
   
@@ -49,6 +51,38 @@ class BooleanTopLevelScorers {
     }
   }
 
+  /**
+   * Used when there is more than one scorer in a query, but a segment
+   * only had one non-null scorer.
+   */
+  static class BoostedBulkScorer extends BulkScorer {
+
+    final BulkScorer in;
+    final float boost;
+
+    BoostedBulkScorer(BulkScorer scorer, float boost) {
+      this.in = scorer;
+      this.boost = boost;
+    }
+
+    @Override
+    public int score(LeafCollector collector, Bits acceptDocs, int min, int max) throws IOException {
+      final LeafCollector wrapped = new FilterLeafCollector(collector) {
+        @Override
+        public void setScorer(Scorer scorer) throws IOException {
+          super.setScorer(new BoostedScorer(scorer, boost));
+        }
+      };
+      return in.score(wrapped, acceptDocs, min, max);
+    }
+
+    @Override
+    public long cost() {
+      return in.cost();
+    }
+
+  }
+
   /** 
    * Used when there are both mandatory and optional clauses, but minShouldMatch
    * dictates that some of the optional clauses must match. The query is a conjunction,

lucene/core/src/java/org/apache/lucene/search/BooleanWeight.java

@@ -190,7 +190,7 @@ final class BooleanWeight extends Weight {
   /** Try to build a boolean scorer for this weight. Returns null if {@link BooleanScorer}
    *  cannot be used. */
   // pkg-private for forcing use of BooleanScorer in tests
-  BooleanScorer booleanScorer(LeafReaderContext context) throws IOException {
+  BulkScorer booleanScorer(LeafReaderContext context) throws IOException {
     List<BulkScorer> optional = new ArrayList<BulkScorer>();
     Iterator<BooleanClause> cIter = query.iterator();
     for (Weight w  : weights) {
@@ -222,12 +222,21 @@ final class BooleanWeight extends Weight {
       return null;
     }
 
+    if (optional.size() == 1) {
+      BulkScorer opt = optional.get(0);
+      if (!disableCoord && maxCoord > 1) {
+        return new BooleanTopLevelScorers.BoostedBulkScorer(opt, coord(1, maxCoord));
+      } else {
+        return opt;
+      }
+    }
+
     return new BooleanScorer(this, disableCoord, maxCoord, optional, Math.max(1, query.getMinimumNumberShouldMatch()), needsScores);
   }
 
   @Override
   public BulkScorer bulkScorer(LeafReaderContext context) throws IOException {
-    final BooleanScorer bulkScorer = booleanScorer(context);
+    final BulkScorer bulkScorer = booleanScorer(context);
     if (bulkScorer != null) { // BooleanScorer is applicable
       // TODO: what is the right heuristic here?
       final long costThreshold;

lucene/core/src/test/org/apache/lucene/search/TestBooleanScorer.java

@@ -18,17 +18,24 @@ package org.apache.lucene.search;
  */
 
 import java.io.IOException;
+import java.util.Arrays;
 import java.util.Set;
 
 import org.apache.lucene.document.Document;
 import org.apache.lucene.document.Field;
+import org.apache.lucene.document.StringField;
+import org.apache.lucene.document.Field.Store;
 import org.apache.lucene.index.IndexReader;
 import org.apache.lucene.index.LeafReaderContext;
 import org.apache.lucene.index.RandomIndexWriter;
 import org.apache.lucene.index.Term;
+import org.apache.lucene.search.BooleanClause.Occur;
+import org.apache.lucene.search.Weight.DefaultBulkScorer;
+import org.apache.lucene.search.similarities.ClassicSimilarity;
 import org.apache.lucene.store.Directory;
 import org.apache.lucene.util.Bits;
 import org.apache.lucene.util.LuceneTestCase;
+import org.apache.lucene.util.TestUtil;
 
 public class TestBooleanScorer extends LuceneTestCase {
   private static final String FIELD = "category";
@@ -141,4 +148,93 @@ public class TestBooleanScorer extends LuceneTestCase {
     r.close();
     dir.close();
   }
+
+  public void testOptimizeTopLevelClauseOrNull() throws IOException {
+    // When there is a single non-null scorer, this scorer should be used
+    // directly
+    Directory dir = newDirectory();
+    RandomIndexWriter w = new RandomIndexWriter(random(), dir);
+    Document doc = new Document();
+    doc.add(new StringField("foo", "bar", Store.NO));
+    w.addDocument(doc);
+    IndexReader reader = w.getReader();
+    IndexSearcher searcher = new IndexSearcher(reader);
+    searcher.setQueryCache(null); // so that weights are not wrapped
+    final LeafReaderContext ctx = reader.leaves().get(0);
+    Query query = new BooleanQuery.Builder()
+      .add(new TermQuery(new Term("foo", "bar")), Occur.SHOULD) // existing term
+      .add(new TermQuery(new Term("foo", "baz")), Occur.SHOULD) // missing term
+      .build();
+
+    // no scores -> term scorer
+    Weight weight = searcher.createNormalizedWeight(query, false);
+    BulkScorer scorer = ((BooleanWeight) weight).booleanScorer(ctx);
+    assertTrue(scorer instanceof DefaultBulkScorer); // term scorer
+
+    // disabled coords -> term scorer
+    query = new BooleanQuery.Builder()
+      .add(new TermQuery(new Term("foo", "bar")), Occur.SHOULD) // existing term
+      .add(new TermQuery(new Term("foo", "baz")), Occur.SHOULD) // missing term
+      .setDisableCoord(true)
+      .build();
+    weight = searcher.createNormalizedWeight(query, true);
+    scorer = ((BooleanWeight) weight).booleanScorer(ctx);
+    assertTrue(scorer instanceof DefaultBulkScorer); // term scorer
+
+    // enabled coords -> BoostedBulkScorer
+    searcher.setSimilarity(new ClassicSimilarity());
+    query = new BooleanQuery.Builder()
+      .add(new TermQuery(new Term("foo", "bar")), Occur.SHOULD) // existing term
+      .add(new TermQuery(new Term("foo", "baz")), Occur.SHOULD) // missing term
+      .build();
+    weight = searcher.createNormalizedWeight(query, true);
+    scorer = ((BooleanWeight) weight).booleanScorer(ctx);
+    assertTrue(scorer instanceof BooleanTopLevelScorers.BoostedBulkScorer);
+
+    w.close();
+    reader.close();
+    dir.close();
+  }
+
+  public void testSparseClauseOptimization() throws IOException {
+    // When some windows have only one scorer that can match, the scorer will
+    // directly call the collector in this window
+    Directory dir = newDirectory();
+    RandomIndexWriter w = new RandomIndexWriter(random(), dir);
+    Document emptyDoc = new Document();
+    final int numDocs = atLeast(10);
+    for (int d = 0; d < numDocs; ++d) {
+      for (int i = random().nextInt(5000); i >= 0; --i) {
+        w.addDocument(emptyDoc);
+      }
+      Document doc = new Document();
+      for (String value : Arrays.asList("foo", "bar", "baz")) {
+        if (random().nextBoolean()) {
+          doc.add(new StringField("field", value, Store.NO));
+        }
+      }
+    }
+    for (int i = TestUtil.nextInt(random(), 3000, 5000); i >= 0; --i) {
+      w.addDocument(emptyDoc);
+    }
+    if (random().nextBoolean()) {
+      w.forceMerge(1);
+    }
+    IndexReader reader = w.getReader();
+    IndexSearcher searcher = newSearcher(reader);
+
+    Query query = new BooleanQuery.Builder()
+      .add(new BoostQuery(new TermQuery(new Term("field", "foo")), 3), Occur.SHOULD)
+      .add(new BoostQuery(new TermQuery(new Term("field", "bar")), 3), Occur.SHOULD)
+      .add(new BoostQuery(new TermQuery(new Term("field", "baz")), 3), Occur.SHOULD)
+      .setDisableCoord(random().nextBoolean())
+      .build();
+
+    // duel BS1 vs. BS2
+    QueryUtils.check(random(), query, searcher);
+
+    reader.close();
+    w.close();
+    dir.close();
+  }
 }