Use read advice consistently in the knn vector formats

lucene/backward-codecs/src/test/org/apache/lucene/backward_codecs/lucene99/Lucene99RWHnswScalarQuantizationVectorsFormat.java

@@ -30,6 +30,7 @@
 import org.apache.lucene.codecs.lucene99.Lucene99ScalarQuantizedVectorsFormat;
 import org.apache.lucene.codecs.lucene99.Lucene99ScalarQuantizedVectorsWriter;
 import org.apache.lucene.index.SegmentWriteState;
+import org.apache.lucene.store.ReadAdvice;
 
 class Lucene99RWHnswScalarQuantizationVectorsFormat
     extends Lucene99HnswScalarQuantizedVectorsFormat {
@@ -54,15 +55,16 @@ public KnnVectorsWriter fieldsWriter(SegmentWriteState state) throws IOException
 
   static class Lucene99RWScalarQuantizedFormat extends Lucene99ScalarQuantizedVectorsFormat {
     private static final FlatVectorsFormat rawVectorFormat =
-        new Lucene99FlatVectorsFormat(new DefaultFlatVectorScorer());
+        new Lucene99FlatVectorsFormat(new DefaultFlatVectorScorer(), ReadAdvice.RANDOM);
 
     @Override
     public FlatVectorsWriter fieldsWriter(SegmentWriteState state) throws IOException {
       return new Lucene99ScalarQuantizedVectorsWriter(
           state,
           null,
           rawVectorFormat.fieldsWriter(state),
-          new ScalarQuantizedVectorScorer(new DefaultFlatVectorScorer()));
+          new ScalarQuantizedVectorScorer(new DefaultFlatVectorScorer()),
+          ReadAdvice.RANDOM);
     }
   }
 }

lucene/codecs/src/java/org/apache/lucene/codecs/bitvectors/HnswBitVectorsFormat.java

@@ -41,6 +41,7 @@
 import org.apache.lucene.index.Sorter;
 import org.apache.lucene.index.VectorEncoding;
 import org.apache.lucene.search.TaskExecutor;
+import org.apache.lucene.store.ReadAdvice;
 import org.apache.lucene.util.hnsw.HnswGraph;
 
 /**
@@ -128,7 +129,8 @@ public HnswBitVectorsFormat(
     } else {
       this.mergeExec = null;
     }
-    this.flatVectorsFormat = new Lucene99FlatVectorsFormat(new FlatBitVectorsScorer());
+    this.flatVectorsFormat =
+        new Lucene99FlatVectorsFormat(new FlatBitVectorsScorer(), ReadAdvice.RANDOM);
   }
 
   @Override

lucene/core/src/java/org/apache/lucene/codecs/KnnVectorsReader.java

@@ -123,11 +123,4 @@ public abstract void search(
   public KnnVectorsReader getMergeInstance() {
     return this;
   }
-
-  /**
-   * Optional: reset or close merge resources used in the reader
-   *
-   * <p>The default implementation is empty
-   */
-  public void finishMerge() throws IOException {}
 }

lucene/core/src/java/org/apache/lucene/codecs/KnnVectorsWriter.java

@@ -109,18 +109,9 @@ public final void merge(MergeState mergeState) throws IOException {
         }
       }
     }
-    finishMerge(mergeState);
     finish();
   }
 
-  private void finishMerge(MergeState mergeState) throws IOException {
-    for (KnnVectorsReader reader : mergeState.knnVectorsReaders) {
-      if (reader != null) {
-        reader.finishMerge();
-      }
-    }
-  }
-
   /** Tracks state of one sub-reader that we are merging */
   private static class FloatVectorValuesSub extends DocIDMerger.Sub {
 

lucene/core/src/java/org/apache/lucene/codecs/hnsw/FlatVectorsReader.java

@@ -88,15 +88,4 @@ public abstract RandomVectorScorer getRandomVectorScorer(String field, float[] t
    */
   public abstract RandomVectorScorer getRandomVectorScorer(String field, byte[] target)
       throws IOException;
-
-  /**
-   * Returns an instance optimized for merging. This instance may only be consumed in the thread
-   * that called {@link #getMergeInstance()}.
-   *
-   * <p>The default implementation returns {@code this}
-   */
-  @Override
-  public FlatVectorsReader getMergeInstance() {
-    return this;
-  }
 }

lucene/core/src/java/org/apache/lucene/codecs/lucene102/Lucene102BinaryQuantizedVectorsFormat.java

@@ -24,6 +24,7 @@
 import org.apache.lucene.codecs.lucene99.Lucene99FlatVectorsFormat;
 import org.apache.lucene.index.SegmentReadState;
 import org.apache.lucene.index.SegmentWriteState;
+import org.apache.lucene.store.ReadAdvice;
 
 /**
  * The binary quantization format used here is a per-vector optimized scalar quantization. These
@@ -103,15 +104,27 @@ public class Lucene102BinaryQuantizedVectorsFormat extends FlatVectorsFormat {
   static final String VECTOR_DATA_EXTENSION = "veb";
   static final int DIRECT_MONOTONIC_BLOCK_SHIFT = 16;
 
+  /**
+   * Specifies the format used for storing, reading, and merging raw vectors on disk.
+   * Since these vectors are rarely accessed, we optimize for merge operations by using
+   * {@link ReadAdvice#SEQUENTIAL} when opening the underlying file.
+   */
   private static final FlatVectorsFormat rawVectorFormat =
-      new Lucene99FlatVectorsFormat(FlatVectorScorerUtil.getLucene99FlatVectorsScorer());
+      new Lucene99FlatVectorsFormat(FlatVectorScorerUtil.getLucene99FlatVectorsScorer(), ReadAdvice.SEQUENTIAL);
 
   private static final Lucene102BinaryFlatVectorsScorer scorer =
       new Lucene102BinaryFlatVectorsScorer(FlatVectorScorerUtil.getLucene99FlatVectorsScorer());
 
+  private final ReadAdvice readAdvice;
+
   /** Creates a new instance with the default number of vectors per cluster. */
   public Lucene102BinaryQuantizedVectorsFormat() {
+    this(ReadAdvice.SEQUENTIAL);
+  }
+
+  public Lucene102BinaryQuantizedVectorsFormat(ReadAdvice readAdvice) {
     super(NAME);
+    this.readAdvice = readAdvice;
   }
 
   @Override
@@ -123,7 +136,7 @@ public FlatVectorsWriter fieldsWriter(SegmentWriteState state) throws IOExceptio
   @Override
   public FlatVectorsReader fieldsReader(SegmentReadState state) throws IOException {
     return new Lucene102BinaryQuantizedVectorsReader(
-        state, rawVectorFormat.fieldsReader(state), scorer);
+        state, rawVectorFormat.fieldsReader(state), scorer, readAdvice);
   }
 
   @Override

lucene/core/src/java/org/apache/lucene/codecs/lucene102/Lucene102BinaryQuantizedVectorsReader.java

@@ -62,7 +62,8 @@ class Lucene102BinaryQuantizedVectorsReader extends FlatVectorsReader {
   Lucene102BinaryQuantizedVectorsReader(
       SegmentReadState state,
       FlatVectorsReader rawVectorsReader,
-      Lucene102BinaryFlatVectorsScorer vectorsScorer)
+      Lucene102BinaryFlatVectorsScorer vectorsScorer,
+      ReadAdvice readAdvice)
       throws IOException {
     super(vectorsScorer);
     this.vectorScorer = vectorsScorer;
@@ -97,9 +98,7 @@ class Lucene102BinaryQuantizedVectorsReader extends FlatVectorsReader {
               versionMeta,
               Lucene102BinaryQuantizedVectorsFormat.VECTOR_DATA_EXTENSION,
               Lucene102BinaryQuantizedVectorsFormat.VECTOR_DATA_CODEC_NAME,
-              // Quantized vectors are accessed randomly from their node ID stored in the HNSW
-              // graph.
-              state.context.withReadAdvice(ReadAdvice.RANDOM));
+              state.context.withReadAdvice(readAdvice));
       success = true;
     } finally {
       if (success == false) {

lucene/core/src/java/org/apache/lucene/codecs/lucene102/Lucene102HnswBinaryQuantizedVectorsFormat.java

@@ -34,6 +34,7 @@
 import org.apache.lucene.index.SegmentReadState;
 import org.apache.lucene.index.SegmentWriteState;
 import org.apache.lucene.search.TaskExecutor;
+import org.apache.lucene.store.ReadAdvice;
 import org.apache.lucene.util.hnsw.HnswGraph;
 
 /**
@@ -57,9 +58,13 @@ public class Lucene102HnswBinaryQuantizedVectorsFormat extends KnnVectorsFormat
    */
   private final int beamWidth;
 
-  /** The format for storing, reading, merging vectors on disk */
+  /**
+   * Specifies the format used for storing, reading, and merging vectors on disk.
+   * Since these vectors are primarily accessed randomly via the HSNW graph,
+   * we use {@link ReadAdvice#RANDOM} when opening the underlying file.
+   */
   private static final FlatVectorsFormat flatVectorsFormat =
-      new Lucene102BinaryQuantizedVectorsFormat();
+      new Lucene102BinaryQuantizedVectorsFormat(ReadAdvice.RANDOM);
 
   private final int numMergeWorkers;
   private final TaskExecutor mergeExec;

lucene/core/src/java/org/apache/lucene/codecs/lucene99/Lucene99FlatVectorsFormat.java

@@ -27,6 +27,7 @@
 import org.apache.lucene.index.SegmentWriteState;
 import org.apache.lucene.search.DocIdSetIterator;
 import org.apache.lucene.store.IndexOutput;
+import org.apache.lucene.store.ReadAdvice;
 
 /**
  * Lucene 9.9 flat vector format, which encodes numeric vector values
@@ -78,21 +79,23 @@ public final class Lucene99FlatVectorsFormat extends FlatVectorsFormat {
 
   static final int DIRECT_MONOTONIC_BLOCK_SHIFT = 16;
   private final FlatVectorsScorer vectorsScorer;
+  private final ReadAdvice readAdvice;
 
   /** Constructs a format */
-  public Lucene99FlatVectorsFormat(FlatVectorsScorer vectorsScorer) {
+  public Lucene99FlatVectorsFormat(FlatVectorsScorer vectorsScorer, ReadAdvice readAdvice) {
     super(NAME);
     this.vectorsScorer = vectorsScorer;
+    this.readAdvice = readAdvice;
   }
 
   @Override
   public FlatVectorsWriter fieldsWriter(SegmentWriteState state) throws IOException {
-    return new Lucene99FlatVectorsWriter(state, vectorsScorer);
+    return new Lucene99FlatVectorsWriter(state, vectorsScorer, readAdvice);
   }
 
   @Override
   public FlatVectorsReader fieldsReader(SegmentReadState state) throws IOException {
-    return new Lucene99FlatVectorsReader(state, vectorsScorer);
+    return new Lucene99FlatVectorsReader(state, vectorsScorer, readAdvice);
   }
 
   @Override

lucene/core/src/java/org/apache/lucene/codecs/lucene99/Lucene99FlatVectorsReader.java

@@ -21,7 +21,6 @@
 import static org.apache.lucene.codecs.lucene99.Lucene99HnswVectorsReader.readVectorEncoding;
 
 import java.io.IOException;
-import java.io.UncheckedIOException;
 import org.apache.lucene.codecs.CodecUtil;
 import org.apache.lucene.codecs.hnsw.FlatVectorsReader;
 import org.apache.lucene.codecs.hnsw.FlatVectorsScorer;
@@ -60,8 +59,8 @@ public final class Lucene99FlatVectorsReader extends FlatVectorsReader {
   private final IndexInput vectorData;
   private final FieldInfos fieldInfos;
 
-  public Lucene99FlatVectorsReader(SegmentReadState state, FlatVectorsScorer scorer)
-      throws IOException {
+  public Lucene99FlatVectorsReader(
+      SegmentReadState state, FlatVectorsScorer scorer, ReadAdvice readAdvice) throws IOException {
     super(scorer);
     int versionMeta = readMetadata(state);
     this.fieldInfos = state.fieldInfos;
@@ -73,9 +72,7 @@ public Lucene99FlatVectorsReader(SegmentReadState state, FlatVectorsScorer score
               versionMeta,
               Lucene99FlatVectorsFormat.VECTOR_DATA_EXTENSION,
               Lucene99FlatVectorsFormat.VECTOR_DATA_CODEC_NAME,
-              // Flat formats are used to randomly access vectors from their node ID that is stored
-              // in the HNSW graph.
-              state.context.withReadAdvice(ReadAdvice.RANDOM));
+              state.context.withReadAdvice(readAdvice));
       success = true;
     } finally {
       if (success == false) {
@@ -171,17 +168,6 @@ public void checkIntegrity() throws IOException {
     CodecUtil.checksumEntireFile(vectorData);
   }
 
-  @Override
-  public FlatVectorsReader getMergeInstance() {
-    try {
-      // Update the read advice since vectors are guaranteed to be accessed sequentially for merge
-      this.vectorData.updateReadAdvice(ReadAdvice.SEQUENTIAL);
-      return this;
-    } catch (IOException exception) {
-      throw new UncheckedIOException(exception);
-    }
-  }
-
   private FieldEntry getFieldEntry(String field, VectorEncoding expectedEncoding) {
     final FieldInfo info = fieldInfos.fieldInfo(field);
     final FieldEntry fieldEntry;
@@ -262,13 +248,6 @@ public RandomVectorScorer getRandomVectorScorer(String field, byte[] target) thr
         target);
   }
 
-  @Override
-  public void finishMerge() throws IOException {
-    // This makes sure that the access pattern hint is reverted back since HNSW implementation
-    // needs it
-    this.vectorData.updateReadAdvice(ReadAdvice.RANDOM);
-  }
-
   @Override
   public void close() throws IOException {
     IOUtils.close(vectorData);

lucene/core/src/java/org/apache/lucene/codecs/lucene99/Lucene99FlatVectorsWriter.java

@@ -66,15 +66,18 @@ public final class Lucene99FlatVectorsWriter extends FlatVectorsWriter {
       RamUsageEstimator.shallowSizeOfInstance(Lucene99FlatVectorsWriter.class);
 
   private final SegmentWriteState segmentWriteState;
+  private final ReadAdvice readAdvice;
   private final IndexOutput meta, vectorData;
 
   private final List<FieldWriter<?>> fields = new ArrayList<>();
+
   private boolean finished;
 
-  public Lucene99FlatVectorsWriter(SegmentWriteState state, FlatVectorsScorer scorer)
-      throws IOException {
+  public Lucene99FlatVectorsWriter(
+      SegmentWriteState state, FlatVectorsScorer scorer, ReadAdvice readAdvice) throws IOException {
     super(scorer);
     segmentWriteState = state;
+    this.readAdvice = readAdvice;
     String metaFileName =
         IndexFileNames.segmentFileName(
             state.segmentInfo.name, state.segmentSuffix, Lucene99FlatVectorsFormat.META_EXTENSION);
@@ -282,7 +285,7 @@ public CloseableRandomVectorScorerSupplier mergeOneFieldToIndex(
       // to perform random reads.
       vectorDataInput =
           segmentWriteState.directory.openInput(
-              tempVectorData.getName(), IOContext.DEFAULT.withReadAdvice(ReadAdvice.RANDOM));
+              tempVectorData.getName(), IOContext.DEFAULT.withReadAdvice(readAdvice));
       // copy the temporary file vectors to the actual data file
       vectorData.copyBytes(vectorDataInput, vectorDataInput.length() - CodecUtil.footerLength());
       CodecUtil.retrieveChecksum(vectorDataInput);

lucene/core/src/java/org/apache/lucene/codecs/lucene99/Lucene99HnswScalarQuantizedVectorsFormat.java

@@ -32,6 +32,7 @@
 import org.apache.lucene.index.SegmentReadState;
 import org.apache.lucene.index.SegmentWriteState;
 import org.apache.lucene.search.TaskExecutor;
+import org.apache.lucene.store.ReadAdvice;
 import org.apache.lucene.util.hnsw.HnswGraph;
 
 /**
@@ -134,8 +135,15 @@ public Lucene99HnswScalarQuantizedVectorsFormat(
     } else {
       this.mergeExec = null;
     }
+
+    /*
+     * Defines the format used for storing, reading, and merging vectors on disk. Flat formats
+     * enable random access to vectors based on their node ID, as recorded in the HNSW graph. To
+     * ensure consistent access, the {@link ReadAdvice#RANDOM} read advice is used.
+     */
     this.flatVectorsFormat =
-        new Lucene99ScalarQuantizedVectorsFormat(confidenceInterval, bits, compress);
+        new Lucene99ScalarQuantizedVectorsFormat(
+            confidenceInterval, bits, compress, ReadAdvice.RANDOM);
   }
 
   @Override

lucene/core/src/java/org/apache/lucene/codecs/lucene99/Lucene99HnswVectorsFormat.java

@@ -29,6 +29,7 @@
 import org.apache.lucene.index.SegmentReadState;
 import org.apache.lucene.index.SegmentWriteState;
 import org.apache.lucene.search.TaskExecutor;
+import org.apache.lucene.store.ReadAdvice;
 import org.apache.lucene.util.hnsw.HnswGraph;
 import org.apache.lucene.util.hnsw.HnswGraphBuilder;
 
@@ -130,9 +131,14 @@ public final class Lucene99HnswVectorsFormat extends KnnVectorsFormat {
    */
   private final int beamWidth;
 
-  /** The format for storing, reading, and merging vectors on disk. */
+  /**
+   * Specifies the format used for storing, reading, and merging vectors on disk.
+   * Since these vectors are primarily accessed randomly via the HSNW graph,
+   * we use {@link ReadAdvice#RANDOM} when opening the underlying file.
+   */
   private static final FlatVectorsFormat flatVectorsFormat =
-      new Lucene99FlatVectorsFormat(FlatVectorScorerUtil.getLucene99FlatVectorsScorer());
+      new Lucene99FlatVectorsFormat(
+          FlatVectorScorerUtil.getLucene99FlatVectorsScorer(), ReadAdvice.RANDOM);
 
   private final int numMergeWorkers;
   private final TaskExecutor mergeExec;

lucene/core/src/java/org/apache/lucene/codecs/lucene99/Lucene99HnswVectorsReader.java

@@ -70,12 +70,11 @@ public final class Lucene99HnswVectorsReader extends KnnVectorsReader
 
   private final FlatVectorsReader flatVectorsReader;
   private final FieldInfos fieldInfos;
-  private final IntObjectHashMap<FieldEntry> fields;
+  private final IntObjectHashMap<FieldEntry> fields = new IntObjectHashMap<>();
   private final IndexInput vectorIndex;
 
   public Lucene99HnswVectorsReader(SegmentReadState state, FlatVectorsReader flatVectorsReader)
       throws IOException {
-    this.fields = new IntObjectHashMap<>();
     this.flatVectorsReader = flatVectorsReader;
     boolean success = false;
     this.fieldInfos = state.fieldInfos;
@@ -115,24 +114,6 @@ public Lucene99HnswVectorsReader(SegmentReadState state, FlatVectorsReader flatV
     }
   }
 
-  private Lucene99HnswVectorsReader(
-      Lucene99HnswVectorsReader reader, FlatVectorsReader flatVectorsReader) {
-    this.flatVectorsReader = flatVectorsReader;
-    this.fieldInfos = reader.fieldInfos;
-    this.fields = reader.fields;
-    this.vectorIndex = reader.vectorIndex;
-  }
-
-  @Override
-  public KnnVectorsReader getMergeInstance() {
-    return new Lucene99HnswVectorsReader(this, this.flatVectorsReader.getMergeInstance());
-  }
-
-  @Override
-  public void finishMerge() throws IOException {
-    flatVectorsReader.finishMerge();
-  }
-
   private static IndexInput openDataInput(
       SegmentReadState state,
       int versionMeta,