Apache Solr open-source search software
LUCENE-9582: rename VectorValues.ScoreFunction to SearchStrategy (#2018)
Michael Sokolov840a353bc706
diff --git a/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextFieldInfosFormat.java b/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextFieldInfosFormat.javaindex e68a7e50254..2cdd00a2380 100644--- a/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextFieldInfosFormat.java+++ b/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextFieldInfosFormat.java@@ -69,7 +69,7 @@ public class SimpleTextFieldInfosFormat extends FieldInfosFormat {static final BytesRef INDEX_DIM_COUNT = new BytesRef(" index dimensional count ");static final BytesRef DIM_NUM_BYTES = new BytesRef(" dimensional num bytes ");static final BytesRef VECTOR_NUM_DIMS = new BytesRef(" vector number of dimensions ");- static final BytesRef VECTOR_SCORE_FUNC = new BytesRef(" vector score function ");+ static final BytesRef VECTOR_SEARCH_STRATEGY = new BytesRef(" vector search strategy ");static final BytesRef SOFT_DELETES = new BytesRef(" soft-deletes ");@Override@@ -154,9 +154,9 @@ public class SimpleTextFieldInfosFormat extends FieldInfosFormat {int vectorNumDimensions = Integer.parseInt(readString(VECTOR_NUM_DIMS.length, scratch));SimpleTextUtil.readLine(input, scratch);- assert StringHelper.startsWith(scratch.get(), VECTOR_SCORE_FUNC);- String scoreFunction = readString(VECTOR_SCORE_FUNC.length, scratch);- VectorValues.ScoreFunction vectorDistFunc = distanceFunction(scoreFunction);+ assert StringHelper.startsWith(scratch.get(), VECTOR_SEARCH_STRATEGY);+ String scoreFunction = readString(VECTOR_SEARCH_STRATEGY.length, scratch);+ VectorValues.SearchStrategy vectorDistFunc = distanceFunction(scoreFunction);SimpleTextUtil.readLine(input, scratch);assert StringHelper.startsWith(scratch.get(), SOFT_DELETES);@@ -186,8 +186,8 @@ public class SimpleTextFieldInfosFormat extends FieldInfosFormat {return DocValuesType.valueOf(dvType);}- public VectorValues.ScoreFunction distanceFunction(String scoreFunction) {- return VectorValues.ScoreFunction.valueOf(scoreFunction);+ public VectorValues.SearchStrategy distanceFunction(String scoreFunction) {+ return VectorValues.SearchStrategy.valueOf(scoreFunction);}private String readString(int offset, BytesRefBuilder scratch) {@@ -274,8 +274,8 @@ public class SimpleTextFieldInfosFormat extends FieldInfosFormat {SimpleTextUtil.write(out, Integer.toString(fi.getVectorDimension()), scratch);SimpleTextUtil.writeNewline(out);- SimpleTextUtil.write(out, VECTOR_SCORE_FUNC);- SimpleTextUtil.write(out, fi.getVectorScoreFunction().name(), scratch);+ SimpleTextUtil.write(out, VECTOR_SEARCH_STRATEGY);+ SimpleTextUtil.write(out, fi.getVectorSearchStrategy().name(), scratch);SimpleTextUtil.writeNewline(out);SimpleTextUtil.write(out, SOFT_DELETES);diff --git a/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextVectorReader.java b/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextVectorReader.javaindex 11494c97ed9..4c196253c2a 100644--- a/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextVectorReader.java+++ b/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextVectorReader.java@@ -63,7 +63,7 @@ public class SimpleTextVectorReader extends VectorReader {while (fieldNumber != -1) {String fieldName = readString(in, FIELD_NAME);String scoreFunctionName = readString(in, SCORE_FUNCTION);- VectorValues.ScoreFunction scoreFunction = VectorValues.ScoreFunction.valueOf(scoreFunctionName);+ VectorValues.SearchStrategy searchStrategy = VectorValues.SearchStrategy.valueOf(scoreFunctionName);long vectorDataOffset = readLong(in, VECTOR_DATA_OFFSET);long vectorDataLength = readLong(in, VECTOR_DATA_LENGTH);int dimension = readInt(in, VECTOR_DIMENSION);@@ -73,7 +73,7 @@ public class SimpleTextVectorReader extends VectorReader {docIds[i] = readInt(in, EMPTY);}assert fieldEntries.containsKey(fieldName) == false;- fieldEntries.put(fieldName, new FieldEntry(dimension, scoreFunction, vectorDataOffset, vectorDataLength, docIds));+ fieldEntries.put(fieldName, new FieldEntry(dimension, searchStrategy, vectorDataOffset, vectorDataLength, docIds));fieldNumber = readInt(in, FIELD_NUMBER);}SimpleTextUtil.checkFooter(in);@@ -138,16 +138,16 @@ public class SimpleTextVectorReader extends VectorReader {private static class FieldEntry {final int dimension;- final VectorValues.ScoreFunction scoreFunction;+ final VectorValues.SearchStrategy searchStrategy;final long vectorDataOffset;final long vectorDataLength;final int[] ordToDoc;- FieldEntry(int dimension, VectorValues.ScoreFunction scoreFunction,+ FieldEntry(int dimension, VectorValues.SearchStrategy searchStrategy,long vectorDataOffset, long vectorDataLength, int[] ordToDoc) {this.dimension = dimension;- this.scoreFunction = scoreFunction;+ this.searchStrategy = searchStrategy;this.vectorDataOffset = vectorDataOffset;this.vectorDataLength = vectorDataLength;this.ordToDoc = ordToDoc;@@ -189,8 +189,8 @@ public class SimpleTextVectorReader extends VectorReader {}@Override- public ScoreFunction scoreFunction() {- return entry.scoreFunction;+ public SearchStrategy searchStrategy() {+ return entry.searchStrategy;}@Overridediff --git a/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextVectorWriter.java b/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextVectorWriter.javaindex 3f076cf4a0d..36809befdad 100644--- a/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextVectorWriter.java+++ b/lucene/codecs/src/java/org/apache/lucene/codecs/simpletext/SimpleTextVectorWriter.java@@ -84,7 +84,7 @@ public class SimpleTextVectorWriter extends VectorWriter {private void writeMeta(FieldInfo field, long vectorDataOffset, long vectorDataLength, List<Integer> docIds) throws IOException {writeField(meta, FIELD_NUMBER, field.number);writeField(meta, FIELD_NAME, field.name);- writeField(meta, SCORE_FUNCTION, field.getVectorScoreFunction().name());+ writeField(meta, SCORE_FUNCTION, field.getVectorSearchStrategy().name());writeField(meta, VECTOR_DATA_OFFSET, vectorDataOffset);writeField(meta, VECTOR_DATA_LENGTH, vectorDataLength);writeField(meta, VECTOR_DIMENSION, field.getVectorDimension());diff --git a/lucene/codecs/src/test/org/apache/lucene/codecs/uniformsplit/TestBlockWriter.java b/lucene/codecs/src/test/org/apache/lucene/codecs/uniformsplit/TestBlockWriter.javaindex f2117785522..18540e77955 100644--- a/lucene/codecs/src/test/org/apache/lucene/codecs/uniformsplit/TestBlockWriter.java+++ b/lucene/codecs/src/test/org/apache/lucene/codecs/uniformsplit/TestBlockWriter.java@@ -122,7 +122,7 @@ public class TestBlockWriter extends LuceneTestCase {0,0,0,- VectorValues.ScoreFunction.NONE,+ VectorValues.SearchStrategy.NONE,true);}diff --git a/lucene/codecs/src/test/org/apache/lucene/codecs/uniformsplit/sharedterms/TestSTBlockReader.java b/lucene/codecs/src/test/org/apache/lucene/codecs/uniformsplit/sharedterms/TestSTBlockReader.javaindex c08bb550693..31e9ab4a9da 100644--- a/lucene/codecs/src/test/org/apache/lucene/codecs/uniformsplit/sharedterms/TestSTBlockReader.java+++ b/lucene/codecs/src/test/org/apache/lucene/codecs/uniformsplit/sharedterms/TestSTBlockReader.java@@ -205,7 +205,7 @@ public class TestSTBlockReader extends LuceneTestCase {0,0,0,- VectorValues.ScoreFunction.NONE,+ VectorValues.SearchStrategy.NONE,false);}diff --git a/lucene/core/src/java/org/apache/lucene/codecs/VectorWriter.java b/lucene/core/src/java/org/apache/lucene/codecs/VectorWriter.javaindex 5dda312466b..a5f9e7f8e18 100644--- a/lucene/core/src/java/org/apache/lucene/codecs/VectorWriter.java+++ b/lucene/core/src/java/org/apache/lucene/codecs/VectorWriter.java@@ -69,23 +69,23 @@ public abstract class VectorWriter implements Closeable {}List<VectorValuesSub> subs = new ArrayList<>();int dimension = -1;- VectorValues.ScoreFunction scoreFunction = null;+ VectorValues.SearchStrategy searchStrategy = null;int nonEmptySegmentIndex = 0;for (int i = 0; i < mergeState.vectorReaders.length; i++) {VectorReader vectorReader = mergeState.vectorReaders[i];if (vectorReader != null) {if (mergeFieldInfo != null && mergeFieldInfo.hasVectorValues()) {int segmentDimension = mergeFieldInfo.getVectorDimension();- VectorValues.ScoreFunction segmentScoreFunction = mergeFieldInfo.getVectorScoreFunction();+ VectorValues.SearchStrategy segmentSearchStrategy = mergeFieldInfo.getVectorSearchStrategy();if (dimension == -1) {dimension = segmentDimension;- scoreFunction = mergeFieldInfo.getVectorScoreFunction();+ searchStrategy = mergeFieldInfo.getVectorSearchStrategy();} else if (dimension != segmentDimension) {throw new IllegalStateException("Varying dimensions for vector-valued field " + mergeFieldInfo.name+ ": " + dimension + "!=" + segmentDimension);- } else if (scoreFunction != segmentScoreFunction) {- throw new IllegalStateException("Varying score functions for vector-valued field " + mergeFieldInfo.name- + ": " + scoreFunction + "!=" + segmentScoreFunction);+ } else if (searchStrategy != segmentSearchStrategy) {+ throw new IllegalStateException("Varying search strategys for vector-valued field " + mergeFieldInfo.name+ + ": " + searchStrategy + "!=" + segmentSearchStrategy);}VectorValues values = vectorReader.getVectorValues(mergeFieldInfo.name);if (values != null) {@@ -223,8 +223,8 @@ public abstract class VectorWriter implements Closeable {}@Override- public VectorValues.ScoreFunction scoreFunction() {- return subs.get(0).values.scoreFunction();+ public SearchStrategy searchStrategy() {+ return subs.get(0).values.searchStrategy();}class MergerRandomAccess implements VectorValues.RandomAccess {@@ -249,8 +249,8 @@ public abstract class VectorWriter implements Closeable {}@Override- public ScoreFunction scoreFunction() {- return VectorValuesMerger.this.scoreFunction();+ public SearchStrategy searchStrategy() {+ return VectorValuesMerger.this.searchStrategy();}@Overridediff --git a/lucene/core/src/java/org/apache/lucene/codecs/lucene50/Lucene50FieldInfosFormat.java b/lucene/core/src/java/org/apache/lucene/codecs/lucene50/Lucene50FieldInfosFormat.javaindex 0f0fe8c642f..b632dada64f 100644--- a/lucene/core/src/java/org/apache/lucene/codecs/lucene50/Lucene50FieldInfosFormat.java+++ b/lucene/core/src/java/org/apache/lucene/codecs/lucene50/Lucene50FieldInfosFormat.java@@ -150,7 +150,7 @@ public final class Lucene50FieldInfosFormat extends FieldInfosFormat {try {infos[i] = new FieldInfo(name, fieldNumber, storeTermVector, omitNorms, storePayloads,indexOptions, docValuesType, dvGen, attributes, 0, 0, 0,- 0, VectorValues.ScoreFunction.NONE, false);+ 0, VectorValues.SearchStrategy.NONE, false);} catch (IllegalStateException e) {throw new CorruptIndexException("invalid fieldinfo for field: " + name + ", fieldNumber=" + fieldNumber, input, e);}diff --git a/lucene/core/src/java/org/apache/lucene/codecs/lucene60/Lucene60FieldInfosFormat.java b/lucene/core/src/java/org/apache/lucene/codecs/lucene60/Lucene60FieldInfosFormat.javaindex 4f6f42a2f99..972e3c6ff08 100644--- a/lucene/core/src/java/org/apache/lucene/codecs/lucene60/Lucene60FieldInfosFormat.java+++ b/lucene/core/src/java/org/apache/lucene/codecs/lucene60/Lucene60FieldInfosFormat.java@@ -166,7 +166,7 @@ public final class Lucene60FieldInfosFormat extends FieldInfosFormat {infos[i] = new FieldInfo(name, fieldNumber, storeTermVector, omitNorms, storePayloads,indexOptions, docValuesType, dvGen, attributes,pointDataDimensionCount, pointIndexDimensionCount, pointNumBytes,- 0, VectorValues.ScoreFunction.NONE, isSoftDeletesField);+ 0, VectorValues.SearchStrategy.NONE, isSoftDeletesField);} catch (IllegalStateException e) {throw new CorruptIndexException("invalid fieldinfo for field: " + name + ", fieldNumber=" + fieldNumber, input, e);}diff --git a/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90FieldInfosFormat.java b/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90FieldInfosFormat.javaindex cab87e9102e..b7b7700fbc4 100644--- a/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90FieldInfosFormat.java+++ b/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90FieldInfosFormat.java@@ -32,6 +32,7 @@ import org.apache.lucene.index.IndexFileNames;import org.apache.lucene.index.IndexOptions;import org.apache.lucene.index.SegmentInfo;import org.apache.lucene.index.VectorValues;+import org.apache.lucene.index.VectorValues.SearchStrategy;import org.apache.lucene.store.ChecksumIndexInput;import org.apache.lucene.store.DataOutput;import org.apache.lucene.store.Directory;@@ -102,8 +103,8 @@ import org.apache.lucene.store.IndexOutput;* <li>VectorDistFunction: a byte containing distance function used for similarity calculation.* <ul>* <li>0: no distance function is defined for this field.</li>- * <li>1: EUCLIDEAN distance. ({@link org.apache.lucene.index.VectorValues.ScoreFunction#EUCLIDEAN})</li>- * <li>2: DOT_PRODUCT score. ({@link org.apache.lucene.index.VectorValues.ScoreFunction#DOT_PRODUCT})</li>+ * <li>1: EUCLIDEAN_HNSW distance. ({@link SearchStrategy#EUCLIDEAN_HNSW})</li>+ * <li>2: DOT_PRODUCT_HNSW score. ({@link SearchStrategy#DOT_PRODUCT_HNSW})</li>* </ul>* </li>* </ul>@@ -170,7 +171,7 @@ public final class Lucene90FieldInfosFormat extends FieldInfosFormat {pointNumBytes = 0;}final int vectorDimension = input.readVInt();- final VectorValues.ScoreFunction vectorDistFunc = getDistFunc(input, input.readByte());+ final VectorValues.SearchStrategy vectorDistFunc = getDistFunc(input, input.readByte());try {infos[i] = new FieldInfo(name, fieldNumber, storeTermVector, omitNorms, storePayloads,@@ -235,11 +236,11 @@ public final class Lucene90FieldInfosFormat extends FieldInfosFormat {}}- private static VectorValues.ScoreFunction getDistFunc(IndexInput input, byte b) throws IOException {- if (b < 0 || b >= VectorValues.ScoreFunction.values().length) {+ private static VectorValues.SearchStrategy getDistFunc(IndexInput input, byte b) throws IOException {+ if (b < 0 || b >= VectorValues.SearchStrategy.values().length) {throw new CorruptIndexException("invalid distance function: " + b, input);}- return VectorValues.ScoreFunction.values()[b];+ return VectorValues.SearchStrategy.values()[b];}static {@@ -315,7 +316,7 @@ public final class Lucene90FieldInfosFormat extends FieldInfosFormat {output.writeVInt(fi.getPointNumBytes());}output.writeVInt(fi.getVectorDimension());- output.writeByte((byte) fi.getVectorScoreFunction().ordinal());+ output.writeByte((byte) fi.getVectorSearchStrategy().ordinal());}CodecUtil.writeFooter(output);}diff --git a/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90VectorReader.java b/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90VectorReader.javaindex 4a32a361fb7..9d313d0c432 100644--- a/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90VectorReader.java+++ b/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90VectorReader.java@@ -102,11 +102,11 @@ public final class Lucene90VectorReader extends VectorReader {if (info == null) {throw new CorruptIndexException("Invalid field number: " + fieldNumber, meta);}- int scoreFunctionId = meta.readInt();- if (scoreFunctionId < 0 || scoreFunctionId >= VectorValues.ScoreFunction.values().length) {- throw new CorruptIndexException("Invalid score function id: " + scoreFunctionId, meta);+ int searchStrategyId = meta.readInt();+ if (searchStrategyId < 0 || searchStrategyId >= VectorValues.SearchStrategy.values().length) {+ throw new CorruptIndexException("Invalid search strategy id: " + searchStrategyId, meta);}- VectorValues.ScoreFunction scoreFunction = VectorValues.ScoreFunction.values()[scoreFunctionId];+ VectorValues.SearchStrategy searchStrategy = VectorValues.SearchStrategy.values()[searchStrategyId];long vectorDataOffset = meta.readVLong();long vectorDataLength = meta.readVLong();int dimension = meta.readInt();@@ -116,7 +116,7 @@ public final class Lucene90VectorReader extends VectorReader {int doc = meta.readVInt();ordToDoc[i] = doc;}- FieldEntry fieldEntry = new FieldEntry(dimension, scoreFunction, maxDoc, vectorDataOffset, vectorDataLength,+ FieldEntry fieldEntry = new FieldEntry(dimension, searchStrategy, maxDoc, vectorDataOffset, vectorDataLength,ordToDoc);fields.put(info.name, fieldEntry);}@@ -173,17 +173,17 @@ public final class Lucene90VectorReader extends VectorReader {private static class FieldEntry {final int dimension;- final VectorValues.ScoreFunction scoreFunction;+ final VectorValues.SearchStrategy searchStrategy;final int maxDoc;final long vectorDataOffset;final long vectorDataLength;final int[] ordToDoc;- FieldEntry(int dimension, VectorValues.ScoreFunction scoreFunction, int maxDoc,+ FieldEntry(int dimension, VectorValues.SearchStrategy searchStrategy, int maxDoc,long vectorDataOffset, long vectorDataLength, int[] ordToDoc) {this.dimension = dimension;- this.scoreFunction = scoreFunction;+ this.searchStrategy = searchStrategy;this.maxDoc = maxDoc;this.vectorDataOffset = vectorDataOffset;this.vectorDataLength = vectorDataLength;@@ -231,8 +231,8 @@ public final class Lucene90VectorReader extends VectorReader {}@Override- public ScoreFunction scoreFunction() {- return fieldEntry.scoreFunction;+ public SearchStrategy searchStrategy() {+ return fieldEntry.searchStrategy;}@Override@@ -312,8 +312,8 @@ public final class Lucene90VectorReader extends VectorReader {}@Override- public VectorValues.ScoreFunction scoreFunction() {- return fieldEntry.scoreFunction;+ public SearchStrategy searchStrategy() {+ return fieldEntry.searchStrategy;}@Overridediff --git a/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90VectorWriter.java b/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90VectorWriter.javaindex cdafb665251..e64e061f464 100644--- a/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90VectorWriter.java+++ b/lucene/core/src/java/org/apache/lucene/codecs/lucene90/Lucene90VectorWriter.java@@ -92,7 +92,7 @@ public final class Lucene90VectorWriter extends VectorWriter {private void writeMeta(FieldInfo field, long vectorDataOffset, long vectorDataLength, List<Integer> docIds) throws IOException {meta.writeInt(field.number);- meta.writeInt(field.getVectorScoreFunction().ordinal());+ meta.writeInt(field.getVectorSearchStrategy().ordinal());meta.writeVLong(vectorDataOffset);meta.writeVLong(vectorDataLength);meta.writeInt(field.getVectorDimension());diff --git a/lucene/core/src/java/org/apache/lucene/document/FieldType.java b/lucene/core/src/java/org/apache/lucene/document/FieldType.javaindex 21d9edeb77a..b89be8baf57 100644--- a/lucene/core/src/java/org/apache/lucene/document/FieldType.java+++ b/lucene/core/src/java/org/apache/lucene/document/FieldType.java@@ -46,7 +46,7 @@ public class FieldType implements IndexableFieldType {private int indexDimensionCount;private int dimensionNumBytes;private int vectorDimension;- private VectorValues.ScoreFunction vectorScoreFunction = VectorValues.ScoreFunction.NONE;+ private VectorValues.SearchStrategy vectorSearchStrategy = VectorValues.SearchStrategy.NONE;private Map<String, String> attributes;/**@@ -66,7 +66,7 @@ public class FieldType implements IndexableFieldType {this.indexDimensionCount = ref.pointIndexDimensionCount();this.dimensionNumBytes = ref.pointNumBytes();this.vectorDimension = ref.vectorDimension();- this.vectorScoreFunction = ref.vectorScoreFunction();+ this.vectorSearchStrategy = ref.vectorSearchStrategy();if (ref.getAttributes() != null) {this.attributes = new HashMap<>(ref.getAttributes());}@@ -357,7 +357,7 @@ public class FieldType implements IndexableFieldType {return dimensionNumBytes;}- void setVectorDimensionsAndScoreFunction(int numDimensions, VectorValues.ScoreFunction distFunc) {+ void setVectorDimensionsAndSearchStrategy(int numDimensions, VectorValues.SearchStrategy distFunc) {checkIfFrozen();if (numDimensions <= 0) {throw new IllegalArgumentException("vector numDimensions must be > 0; got " + numDimensions);@@ -366,7 +366,7 @@ public class FieldType implements IndexableFieldType {throw new IllegalArgumentException("vector numDimensions must be <= VectorValues.MAX_DIMENSIONS (=" + VectorValues.MAX_DIMENSIONS + "); got " + numDimensions);}this.vectorDimension = numDimensions;- this.vectorScoreFunction = distFunc;+ this.vectorSearchStrategy = distFunc;}@Override@@ -375,8 +375,8 @@ public class FieldType implements IndexableFieldType {}@Override- public VectorValues.ScoreFunction vectorScoreFunction() {- return vectorScoreFunction;+ public VectorValues.SearchStrategy vectorSearchStrategy() {+ return vectorSearchStrategy;}/**diff --git a/lucene/core/src/java/org/apache/lucene/document/VectorField.java b/lucene/core/src/java/org/apache/lucene/document/VectorField.javaindex 24d9bd8ddaf..eff6e120000 100644--- a/lucene/core/src/java/org/apache/lucene/document/VectorField.java+++ b/lucene/core/src/java/org/apache/lucene/document/VectorField.java@@ -23,15 +23,15 @@ import org.apache.lucene.index.VectorValues;* Vectors are dense - that is, every dimension of a vector contains an explicit value, stored* packed into an array (of type float[]) whose length is the vector dimension. Values can be* retrieved using {@link VectorValues}, which is a forward-only docID-based iterator and also- * offers random-access by dense ordinal (not docId). VectorValues.ScoreFunctions may be+ * offers random-access by dense ordinal (not docId). VectorValues.SearchStrategys may be* used to compare vectors at query time (for example as part of result ranking). A VectorField may- * be associated with a score function that defines the metric used for nearest-neighbor search+ * be associated with a search strategy that defines the metric used for nearest-neighbor search* among vectors of that field, but at the moment this association is purely nominal: it is intended* for future use by the to-be-implemented nearest neighbors search.*/public class VectorField extends Field {- private static FieldType getType(float[] v, VectorValues.ScoreFunction scoreFunction) {+ private static FieldType getType(float[] v, VectorValues.SearchStrategy searchStrategy) {if (v == null) {throw new IllegalArgumentException("vector value must not be null");}@@ -42,38 +42,38 @@ public class VectorField extends Field {if (dimension > VectorValues.MAX_DIMENSIONS) {throw new IllegalArgumentException("cannot index vectors with dimension greater than " + VectorValues.MAX_DIMENSIONS);}- if (scoreFunction == null) {- throw new IllegalArgumentException("score function must not be null");+ if (searchStrategy == null) {+ throw new IllegalArgumentException("search strategy must not be null");}FieldType type = new FieldType();- type.setVectorDimensionsAndScoreFunction(dimension, scoreFunction);+ type.setVectorDimensionsAndSearchStrategy(dimension, searchStrategy);type.freeze();return type;}/** Creates a numeric vector field. Fields are single-valued: each document has either one value- * or no value. Vectors of a single field share the same dimension and score function.+ * or no value. Vectors of a single field share the same dimension and search strategy.** @param name field name* @param vector value- * @param scoreFunction a function defining vector proximity.+ * @param searchStrategy a function defining vector proximity.* @throws IllegalArgumentException if any parameter is null, or the vector is empty or has dimension > 1024.*/- public VectorField(String name, float[] vector, VectorValues.ScoreFunction scoreFunction) {- super(name, getType(vector, scoreFunction));+ public VectorField(String name, float[] vector, VectorValues.SearchStrategy searchStrategy) {+ super(name, getType(vector, searchStrategy));fieldsData = vector;}- /** Creates a numeric vector field with the default EUCLIDEAN (L2) score function. Fields are+ /** Creates a numeric vector field with the default EUCLIDEAN_HNSW (L2) search strategy. Fields are* single-valued: each document has either one value or no value. Vectors of a single field share- * the same dimension and score function.+ * the same dimension and search strategy.** @param name field name* @param vector value* @throws IllegalArgumentException if any parameter is null, or the vector is empty or has dimension > 1024.*/public VectorField(String name, float[] vector) {- this(name, vector, VectorValues.ScoreFunction.EUCLIDEAN);+ this(name, vector, VectorValues.SearchStrategy.EUCLIDEAN_HNSW);}/**diff --git a/lucene/core/src/java/org/apache/lucene/index/FieldInfo.java b/lucene/core/src/java/org/apache/lucene/index/FieldInfo.javaindex 4a3463c58fc..cedc7a72ffb 100644--- a/lucene/core/src/java/org/apache/lucene/index/FieldInfo.java+++ b/lucene/core/src/java/org/apache/lucene/index/FieldInfo.java@@ -55,7 +55,7 @@ public final class FieldInfo {private int pointNumBytes;private int vectorDimension; // if it is a positive value, it means this field indexes vectors- private VectorValues.ScoreFunction vectorScoreFunction = VectorValues.ScoreFunction.NONE;+ private VectorValues.SearchStrategy vectorSearchStrategy = VectorValues.SearchStrategy.NONE;// whether this field is used as the soft-deletes fieldprivate final boolean softDeletesField;@@ -68,7 +68,7 @@ public final class FieldInfo {public FieldInfo(String name, int number, boolean storeTermVector, boolean omitNorms, boolean storePayloads,IndexOptions indexOptions, DocValuesType docValues, long dvGen, Map<String,String> attributes,int pointDimensionCount, int pointIndexDimensionCount, int pointNumBytes,- int vectorDimension, VectorValues.ScoreFunction vectorScoreFunction, boolean softDeletesField) {+ int vectorDimension, VectorValues.SearchStrategy vectorSearchStrategy, boolean softDeletesField) {this.name = Objects.requireNonNull(name);this.number = number;this.docValuesType = Objects.requireNonNull(docValues, "DocValuesType must not be null (field: \"" + name + "\")");@@ -88,7 +88,7 @@ public final class FieldInfo {this.pointIndexDimensionCount = pointIndexDimensionCount;this.pointNumBytes = pointNumBytes;this.vectorDimension = vectorDimension;- this.vectorScoreFunction = vectorScoreFunction;+ this.vectorSearchStrategy = vectorSearchStrategy;this.softDeletesField = softDeletesField;this.checkConsistency();}@@ -147,8 +147,8 @@ public final class FieldInfo {throw new IllegalStateException("vectorDimension must be >=0; got " + vectorDimension);}- if (vectorDimension == 0 && vectorScoreFunction != VectorValues.ScoreFunction.NONE) {- throw new IllegalStateException("vector score function must be NONE when dimension = 0; got " + vectorScoreFunction);+ if (vectorDimension == 0 && vectorSearchStrategy != VectorValues.SearchStrategy.NONE) {+ throw new IllegalStateException("vector search strategy must be NONE when dimension = 0; got " + vectorSearchStrategy);}return true;@@ -247,25 +247,25 @@ public final class FieldInfo {}/** Record that this field is indexed with vectors, with the specified num of dimensions and distance function */- public void setVectorDimensionAndScoreFunction(int dimension, VectorValues.ScoreFunction scoreFunction) {+ public void setVectorDimensionAndSearchStrategy(int dimension, VectorValues.SearchStrategy searchStrategy) {if (dimension < 0) {throw new IllegalArgumentException("vector dimension must be >= 0; got " + dimension);}if (dimension > VectorValues.MAX_DIMENSIONS) {throw new IllegalArgumentException("vector dimension must be <= VectorValues.MAX_DIMENSIONS (=" + VectorValues.MAX_DIMENSIONS + "); got " + dimension);}- if (dimension == 0 && scoreFunction != VectorValues.ScoreFunction.NONE) {- throw new IllegalArgumentException("vector score function must be NONE when the vector dimension = 0; got " + scoreFunction);+ if (dimension == 0 && searchStrategy != VectorValues.SearchStrategy.NONE) {+ throw new IllegalArgumentException("vector search strategy must be NONE when the vector dimension = 0; got " + searchStrategy);}if (vectorDimension != 0 && vectorDimension != dimension) {throw new IllegalArgumentException("cannot change vector dimension from " + vectorDimension + " to " + dimension + " for field=\"" + name + "\"");}- if (vectorScoreFunction != VectorValues.ScoreFunction.NONE && vectorScoreFunction != scoreFunction) {- throw new IllegalArgumentException("cannot change vector score function from " + vectorScoreFunction + " to " + scoreFunction + " for field=\"" + name + "\"");+ if (vectorSearchStrategy != VectorValues.SearchStrategy.NONE && vectorSearchStrategy != searchStrategy) {+ throw new IllegalArgumentException("cannot change vector search strategy from " + vectorSearchStrategy + " to " + searchStrategy + " for field=\"" + name + "\"");}this.vectorDimension = dimension;- this.vectorScoreFunction = scoreFunction;+ this.vectorSearchStrategy = searchStrategy;assert checkConsistency();}@@ -275,9 +275,9 @@ public final class FieldInfo {return vectorDimension;}- /** Returns {@link org.apache.lucene.index.VectorValues.ScoreFunction} for the field */- public VectorValues.ScoreFunction getVectorScoreFunction() {- return vectorScoreFunction;+ /** Returns {@link VectorValues.SearchStrategy} for the field */+ public VectorValues.SearchStrategy getVectorSearchStrategy() {+ return vectorSearchStrategy;}/** Record that this field is indexed with docvalues, with the specified type */diff --git a/lucene/core/src/java/org/apache/lucene/index/FieldInfos.java b/lucene/core/src/java/org/apache/lucene/index/FieldInfos.javaindex 8d8ff15ee5a..95755ff2d87 100644--- a/lucene/core/src/java/org/apache/lucene/index/FieldInfos.java+++ b/lucene/core/src/java/org/apache/lucene/index/FieldInfos.java@@ -273,11 +273,11 @@ public class FieldInfos implements Iterable<FieldInfo> {static final class FieldVectorProperties {final int numDimensions;- final VectorValues.ScoreFunction scoreFunction;+ final VectorValues.SearchStrategy searchStrategy;- FieldVectorProperties(int numDimensions, VectorValues.ScoreFunction scoreFunction) {+ FieldVectorProperties(int numDimensions, VectorValues.SearchStrategy searchStrategy) {this.numDimensions = numDimensions;- this.scoreFunction = scoreFunction;+ this.searchStrategy = searchStrategy;}}@@ -319,7 +319,7 @@ public class FieldInfos implements Iterable<FieldInfo> {* number assigned if possible otherwise the first unassigned field number* is used as the field number.*/- synchronized int addOrGet(String fieldName, int preferredFieldNumber, IndexOptions indexOptions, DocValuesType dvType, int dimensionCount, int indexDimensionCount, int dimensionNumBytes, int vectorDimension, VectorValues.ScoreFunction scoreFunction, boolean isSoftDeletesField) {+ synchronized int addOrGet(String fieldName, int preferredFieldNumber, IndexOptions indexOptions, DocValuesType dvType, int dimensionCount, int indexDimensionCount, int dimensionNumBytes, int vectorDimension, VectorValues.SearchStrategy searchStrategy, boolean isSoftDeletesField) {if (indexOptions != IndexOptions.NONE) {IndexOptions currentOpts = this.indexOptions.get(fieldName);if (currentOpts == null) {@@ -358,11 +358,11 @@ public class FieldInfos implements Iterable<FieldInfo> {if (props.numDimensions != vectorDimension) {throw new IllegalArgumentException("cannot change vector dimension from " + props.numDimensions + " to " + vectorDimension + " for field=\"" + fieldName + "\"");}- if (props.scoreFunction != scoreFunction) {- throw new IllegalArgumentException("cannot change vector score function from " + props.scoreFunction + " to " + scoreFunction + " for field=\"" + fieldName + "\"");+ if (props.searchStrategy != searchStrategy) {+ throw new IllegalArgumentException("cannot change vector search strategy from " + props.searchStrategy + " to " + searchStrategy + " for field=\"" + fieldName + "\"");}} else {- vectorProps.put(fieldName, new FieldVectorProperties(vectorDimension, scoreFunction));+ vectorProps.put(fieldName, new FieldVectorProperties(vectorDimension, searchStrategy));}}Integer fieldNumber = nameToNumber.get(fieldName);@@ -443,7 +443,7 @@ public class FieldInfos implements Iterable<FieldInfo> {}}- synchronized void verifyConsistentVectorProperties(Integer number, String name, int numDimensions, VectorValues.ScoreFunction scoreFunction) {+ synchronized void verifyConsistentVectorProperties(Integer number, String name, int numDimensions, VectorValues.SearchStrategy searchStrategy) {if (name.equals(numberToName.get(number)) == false) {throw new IllegalArgumentException("field number " + number + " is already mapped to field name \"" + numberToName.get(number) + "\", not \"" + name + "\"");}@@ -455,8 +455,8 @@ public class FieldInfos implements Iterable<FieldInfo> {if (props.numDimensions != numDimensions) {throw new IllegalArgumentException("cannot change vector dimension from " + props.numDimensions + " to " + numDimensions + " for field=\"" + name + "\"");}- if (props.scoreFunction != scoreFunction) {- throw new IllegalArgumentException("cannot change vector score function from " + props.scoreFunction + " to " + scoreFunction + " for field=\"" + name + "\"");+ if (props.searchStrategy != searchStrategy) {+ throw new IllegalArgumentException("cannot change vector search strategy from " + props.searchStrategy + " to " + searchStrategy + " for field=\"" + name + "\"");}}}@@ -510,15 +510,15 @@ public class FieldInfos implements Iterable<FieldInfo> {dimensions.put(name, new FieldDimensions(dimensionCount, indexDimensionCount, dimensionNumBytes));}- synchronized void setVectorDimensionsAndScoreFunction(int number, String name, int numDimensions, VectorValues.ScoreFunction scoreFunction) {+ synchronized void setVectorDimensionsAndSearchStrategy(int number, String name, int numDimensions, VectorValues.SearchStrategy searchStrategy) {if (numDimensions <= 0) {throw new IllegalArgumentException("vector numDimensions must be > 0; got " + numDimensions);}if (numDimensions > VectorValues.MAX_DIMENSIONS) {throw new IllegalArgumentException("vector numDimensions must be <= VectorValues.MAX_DIMENSIONS (=" + VectorValues.MAX_DIMENSIONS + "); got " + numDimensions);}- verifyConsistentVectorProperties(number, name, numDimensions, scoreFunction);- vectorProps.put(name, new FieldVectorProperties(numDimensions, scoreFunction));+ verifyConsistentVectorProperties(number, name, numDimensions, searchStrategy);+ vectorProps.put(name, new FieldVectorProperties(numDimensions, searchStrategy));}}@@ -553,8 +553,8 @@ public class FieldInfos implements Iterable<FieldInfo> {// before then we'll get the same name and number,// else we'll allocate a new one:final boolean isSoftDeletesField = name.equals(globalFieldNumbers.softDeletesFieldName);- final int fieldNumber = globalFieldNumbers.addOrGet(name, -1, IndexOptions.NONE, DocValuesType.NONE, 0, 0, 0, 0, VectorValues.ScoreFunction.NONE, isSoftDeletesField);- fi = new FieldInfo(name, fieldNumber, false, false, false, IndexOptions.NONE, DocValuesType.NONE, -1, new HashMap<>(), 0, 0, 0, 0, VectorValues.ScoreFunction.NONE, isSoftDeletesField);+ final int fieldNumber = globalFieldNumbers.addOrGet(name, -1, IndexOptions.NONE, DocValuesType.NONE, 0, 0, 0, 0, VectorValues.SearchStrategy.NONE, isSoftDeletesField);+ fi = new FieldInfo(name, fieldNumber, false, false, false, IndexOptions.NONE, DocValuesType.NONE, -1, new HashMap<>(), 0, 0, 0, 0, VectorValues.SearchStrategy.NONE, isSoftDeletesField);assert !byName.containsKey(fi.name);globalFieldNumbers.verifyConsistent(Integer.valueOf(fi.number), fi.name, DocValuesType.NONE);byName.put(fi.name, fi);@@ -569,7 +569,7 @@ public class FieldInfos implements Iterable<FieldInfo> {DocValuesType docValues, long dvGen,Map<String, String> attributes,int dataDimensionCount, int indexDimensionCount, int dimensionNumBytes,- int vectorDimension, VectorValues.ScoreFunction vectorScoreFunction,+ int vectorDimension, VectorValues.SearchStrategy vectorSearchStrategy,boolean isSoftDeletesField) {assert assertNotFinished();if (docValues == null) {@@ -587,8 +587,8 @@ public class FieldInfos implements Iterable<FieldInfo> {// number for this field. If the field was seen// before then we'll get the same name and number,// else we'll allocate a new one:- final int fieldNumber = globalFieldNumbers.addOrGet(name, preferredFieldNumber, indexOptions, docValues, dataDimensionCount, indexDimensionCount, dimensionNumBytes, vectorDimension, vectorScoreFunction, isSoftDeletesField);- fi = new FieldInfo(name, fieldNumber, storeTermVector, omitNorms, storePayloads, indexOptions, docValues, dvGen, attributes, dataDimensionCount, indexDimensionCount, dimensionNumBytes, vectorDimension, vectorScoreFunction, isSoftDeletesField);+ final int fieldNumber = globalFieldNumbers.addOrGet(name, preferredFieldNumber, indexOptions, docValues, dataDimensionCount, indexDimensionCount, dimensionNumBytes, vectorDimension, vectorSearchStrategy, isSoftDeletesField);+ fi = new FieldInfo(name, fieldNumber, storeTermVector, omitNorms, storePayloads, indexOptions, docValues, dvGen, attributes, dataDimensionCount, indexDimensionCount, dimensionNumBytes, vectorDimension, vectorSearchStrategy, isSoftDeletesField);assert !byName.containsKey(fi.name);globalFieldNumbers.verifyConsistent(Integer.valueOf(fi.number), fi.name, fi.getDocValuesType());byName.put(fi.name, fi);@@ -623,7 +623,7 @@ public class FieldInfos implements Iterable<FieldInfo> {fi.getIndexOptions(), fi.getDocValuesType(), dvGen,fi.attributes(),fi.getPointDimensionCount(), fi.getPointIndexDimensionCount(), fi.getPointNumBytes(),- fi.getVectorDimension(), fi.getVectorScoreFunction(),+ fi.getVectorDimension(), fi.getVectorSearchStrategy(),fi.isSoftDeletesField());}diff --git a/lucene/core/src/java/org/apache/lucene/index/IndexWriter.java b/lucene/core/src/java/org/apache/lucene/index/IndexWriter.javaindex 9c557184629..73a18a49465 100644--- a/lucene/core/src/java/org/apache/lucene/index/IndexWriter.java+++ b/lucene/core/src/java/org/apache/lucene/index/IndexWriter.java@@ -1185,7 +1185,7 @@ public class IndexWriter implements Closeable, TwoPhaseCommit, Accountable,FieldInfos fis = readFieldInfos(info);for(FieldInfo fi : fis) {map.addOrGet(fi.name, fi.number, fi.getIndexOptions(), fi.getDocValuesType(), fi.getPointDimensionCount(), fi.getPointIndexDimensionCount(), fi.getPointNumBytes(),- fi.getVectorDimension(), fi.getVectorScoreFunction(), fi.isSoftDeletesField());+ fi.getVectorDimension(), fi.getVectorSearchStrategy(), fi.isSoftDeletesField());}}@@ -1922,7 +1922,7 @@ public class IndexWriter implements Closeable, TwoPhaseCommit, Accountable,if (globalFieldNumberMap.contains(f.name(), dvType) == false) {// if this field doesn't exists we try to add it. if it exists and the DV type doesn't match we// get a consistent error message as if you try to do that during an indexing operation.- globalFieldNumberMap.addOrGet(f.name(), -1, IndexOptions.NONE, dvType, 0, 0, 0, 0, VectorValues.ScoreFunction.NONE, f.name().equals(config.softDeletesField));+ globalFieldNumberMap.addOrGet(f.name(), -1, IndexOptions.NONE, dvType, 0, 0, 0, 0, VectorValues.SearchStrategy.NONE, f.name().equals(config.softDeletesField));assert globalFieldNumberMap.contains(f.name(), dvType);}if (config.getIndexSortFields().contains(f.name())) {@@ -2969,7 +2969,7 @@ public class IndexWriter implements Closeable, TwoPhaseCommit, Accountable,// This will throw exceptions if any of the incoming fields have an illegal schema change:globalFieldNumberMap.addOrGet(fi.name, fi.number, fi.getIndexOptions(), fi.getDocValuesType(),fi.getPointDimensionCount(), fi.getPointIndexDimensionCount(), fi.getPointNumBytes(),- fi.getVectorDimension(), fi.getVectorScoreFunction(), fi.isSoftDeletesField());+ fi.getVectorDimension(), fi.getVectorSearchStrategy(), fi.isSoftDeletesField());}infos.add(copySegmentAsIs(info, newSegName, context));}diff --git a/lucene/core/src/java/org/apache/lucene/index/IndexableFieldType.java b/lucene/core/src/java/org/apache/lucene/index/IndexableFieldType.javaindex 9f85d04c7aa..ce19472caca 100644--- a/lucene/core/src/java/org/apache/lucene/index/IndexableFieldType.java+++ b/lucene/core/src/java/org/apache/lucene/index/IndexableFieldType.java@@ -120,9 +120,9 @@ public interface IndexableFieldType {public int vectorDimension();/**- * The {@link org.apache.lucene.index.VectorValues.ScoreFunction} of the field's vector value+ * The {@link VectorValues.SearchStrategy} of the field's vector value*/- public VectorValues.ScoreFunction vectorScoreFunction();+ public VectorValues.SearchStrategy vectorSearchStrategy();/*** Attributes for the field type.diff --git a/lucene/core/src/java/org/apache/lucene/index/IndexingChain.java b/lucene/core/src/java/org/apache/lucene/index/IndexingChain.javaindex 071e6ceeda9..c20a453d869 100644--- a/lucene/core/src/java/org/apache/lucene/index/IndexingChain.java+++ b/lucene/core/src/java/org/apache/lucene/index/IndexingChain.java@@ -783,14 +783,14 @@ final class IndexingChain implements Accountable {/** Called from processDocument to index one field's vector value */private void indexVector(int docID, PerField fp, IndexableField field) {int dimension = field.fieldType().vectorDimension();- VectorValues.ScoreFunction scoreFunction = field.fieldType().vectorScoreFunction();+ VectorValues.SearchStrategy searchStrategy = field.fieldType().vectorSearchStrategy();// Record dimensions and distance function for this field; this setter will throw IllegalArgExc if// the dimensions or distance function were already set to something different:if (fp.fieldInfo.getVectorDimension() == 0) {- fieldInfos.globalFieldNumbers.setVectorDimensionsAndScoreFunction(fp.fieldInfo.number, fp.fieldInfo.name, dimension, scoreFunction);+ fieldInfos.globalFieldNumbers.setVectorDimensionsAndSearchStrategy(fp.fieldInfo.number, fp.fieldInfo.name, dimension, searchStrategy);}- fp.fieldInfo.setVectorDimensionAndScoreFunction(dimension, scoreFunction);+ fp.fieldInfo.setVectorDimensionAndSearchStrategy(dimension, searchStrategy);if (fp.vectorValuesWriter == null) {fp.vectorValuesWriter = new VectorValuesWriter(fp.fieldInfo, bytesUsed);diff --git a/lucene/core/src/java/org/apache/lucene/index/ReadersAndUpdates.java b/lucene/core/src/java/org/apache/lucene/index/ReadersAndUpdates.javaindex 10127fb2b17..07b1f237040 100644--- a/lucene/core/src/java/org/apache/lucene/index/ReadersAndUpdates.java+++ b/lucene/core/src/java/org/apache/lucene/index/ReadersAndUpdates.java@@ -658,7 +658,7 @@ final class ReadersAndUpdates {return new FieldInfo(fi.name, fieldNumber, fi.hasVectors(), fi.omitsNorms(), fi.hasPayloads(),fi.getIndexOptions(), fi.getDocValuesType(), fi.getDocValuesGen(), new HashMap<>(fi.attributes()),fi.getPointDimensionCount(), fi.getPointIndexDimensionCount(), fi.getPointNumBytes(),- fi.getVectorDimension(), fi.getVectorScoreFunction(), fi.isSoftDeletesField());+ fi.getVectorDimension(), fi.getVectorSearchStrategy(), fi.isSoftDeletesField());}private SegmentReader createNewReaderWithLatestLiveDocs(SegmentReader reader) throws IOException {diff --git a/lucene/core/src/java/org/apache/lucene/index/VectorValues.java b/lucene/core/src/java/org/apache/lucene/index/VectorValues.javaindex 0404ef05b56..c6534f0c47d 100644--- a/lucene/core/src/java/org/apache/lucene/index/VectorValues.java+++ b/lucene/core/src/java/org/apache/lucene/index/VectorValues.java@@ -51,9 +51,9 @@ public abstract class VectorValues extends DocIdSetIterator {public abstract int size();/**- * Return the score function used to compare these vectors+ * Return the search strategy used to compare these vectors*/- public abstract ScoreFunction scoreFunction();+ public abstract SearchStrategy searchStrategy();/*** Return the vector value for the current document ID.@@ -99,9 +99,9 @@ public abstract class VectorValues extends DocIdSetIterator {int dimension();/**- * Return the score function used to compare these vectors+ * Return the search strategy used to compare these vectors*/- ScoreFunction scoreFunction();+ SearchStrategy searchStrategy();/*** Return the vector value indexed at the given ordinal. The provided floating point array may@@ -120,106 +120,31 @@ public abstract class VectorValues extends DocIdSetIterator {/*** Return the k nearest neighbor documents as determined by comparison of their vector values- * for this field, to the given vector, by the field's score function. If the score function is+ * for this field, to the given vector, by the field's search strategy. If the search strategy is* reversed, lower values indicate nearer vectors, otherwise higher scores indicate nearer* vectors. Unlike relevance scores, vector scores may be negative.* @param target the vector-valued query* @param k the number of docs to return* @param fanout control the accuracy/speed tradeoff - larger values give better recall at higher cost- * @return the k nearest neighbor documents, along with their (scoreFunction-specific) scores.+ * @return the k nearest neighbor documents, along with their (searchStrategy-specific) scores.*/TopDocs search(float[] target, int k, int fanout) throws IOException;}/**- * Score function. This is used during indexing and searching of the vectors to determine the nearest neighbors.- * Score values may be negative. By default high scores indicate nearer documents, unless the function is reversed.+ * Search strategy. This is a label describing the method used during indexing and searching of the vectors in order to+ * determine the nearest neighbors.*/- public enum ScoreFunction {- /** No distance function is used. Note: {@link VectorValues.RandomAccess#search(float[], int, int)}- * is not supported for fields specifying this score function. */+ public enum SearchStrategy {+ /** No search strategy is provided. Note: {@link VectorValues.RandomAccess#search(float[], int, int)}+ * is not supported for fields specifying this strategy. */NONE,- /** Euclidean distance */- EUCLIDEAN(true) {- @Override- public float score(float[] v1, float[] v2) {- assert v1.length == v2.length;- float squareSum = 0.0f;- int dim = v1.length;- for (int i = 0; i < dim; i++) {- float diff = v1[i] - v2[i];- squareSum += diff * diff;- }- return squareSum;- }- },-- /** dot product - note, may be negative; larger values are better */- DOT_PRODUCT() {- @Override- public float score(float[] a, float[] b) {- float res = 0f;- /*- * If length of vector is larger than 8, we use unrolled dot product to accelerate the- * calculation.- */- int i;- for (i = 0; i < a.length % 8; i++) {- res += b[i] * a[i];- }- if (a.length < 8) {- return res;- }- float s0 = 0f;- float s1 = 0f;- float s2 = 0f;- float s3 = 0f;- float s4 = 0f;- float s5 = 0f;- float s6 = 0f;- float s7 = 0f;- for (; i + 7 < a.length; i += 8) {- s0 += b[i] * a[i];- s1 += b[i + 1] * a[i + 1];- s2 += b[i + 2] * a[i + 2];- s3 += b[i + 3] * a[i + 3];- s4 += b[i + 4] * a[i + 4];- s5 += b[i + 5] * a[i + 5];- s6 += b[i + 6] * a[i + 6];- s7 += b[i + 7] * a[i + 7];- }- res += s0 + s1 + s2 + s3 + s4 + s5 + s6 + s7;- return res;- }- };-- /** If reversed, smaller values are better */- final public boolean reversed;-- ScoreFunction(boolean reversed) {- this.reversed = reversed;- }-- ScoreFunction() {- this(false);- }+ /** HNSW graph built using Euclidean distance */+ EUCLIDEAN_HNSW,- /**- * Calculates the score between the specified two vectors.- */- public float score(float[] v1, float[] v2) {- throw new UnsupportedOperationException();- }-- }-- /**- * Calculates a similarity score between the two vectors with specified function.- */- public static float compare(float[] v1, float[] v2, ScoreFunction scoreFunction) {- assert v1.length == v2.length : "attempt to compare vectors of lengths: " + v1.length + " " + v2.length;- return scoreFunction.score(v1, v2);+ /** HNSW graph buit using dot product */+ DOT_PRODUCT_HNSW}/**@@ -239,8 +164,8 @@ public abstract class VectorValues extends DocIdSetIterator {}@Override- public ScoreFunction scoreFunction() {- return ScoreFunction.NONE;+ public SearchStrategy searchStrategy() {+ return SearchStrategy.NONE;}@Overridediff --git a/lucene/core/src/java/org/apache/lucene/index/VectorValuesWriter.java b/lucene/core/src/java/org/apache/lucene/index/VectorValuesWriter.javaindex 564b7b61a22..c4e97018b04 100644--- a/lucene/core/src/java/org/apache/lucene/index/VectorValuesWriter.java+++ b/lucene/core/src/java/org/apache/lucene/index/VectorValuesWriter.java@@ -90,7 +90,7 @@ class VectorValuesWriter {* @throws IOException if there is an error writing the field and its values*/public void flush(Sorter.DocMap sortMap, VectorWriter vectorWriter) throws IOException {- VectorValues vectorValues = new BufferedVectorValues(docsWithField, vectors, fieldInfo.getVectorDimension(), fieldInfo.getVectorScoreFunction());+ VectorValues vectorValues = new BufferedVectorValues(docsWithField, vectors, fieldInfo.getVectorDimension(), fieldInfo.getVectorSearchStrategy());if (sortMap != null) {vectorWriter.writeField(fieldInfo, new SortingVectorValues(vectorValues, sortMap));} else {@@ -167,8 +167,8 @@ class VectorValuesWriter {}@Override- public ScoreFunction scoreFunction() {- return delegate.scoreFunction();+ public SearchStrategy searchStrategy() {+ return delegate.searchStrategy();}@Override@@ -197,8 +197,8 @@ class VectorValuesWriter {}@Override- public ScoreFunction scoreFunction() {- return delegate.scoreFunction();+ public SearchStrategy searchStrategy() {+ return delegate.searchStrategy();}@Override@@ -225,7 +225,7 @@ class VectorValuesWriter {// These are always the vectors of a VectorValuesWriter, which are copied when added to itfinal List<float[]> vectors;- final VectorValues.ScoreFunction scoreFunction;+ final SearchStrategy searchStrategy;final int dimension;final ByteBuffer buffer;@@ -236,11 +236,11 @@ class VectorValuesWriter {DocIdSetIterator docsWithFieldIter;int ord = -1;- BufferedVectorValues(DocsWithFieldSet docsWithField, List<float[]> vectors, int dimension, VectorValues.ScoreFunction scoreFunction) {+ BufferedVectorValues(DocsWithFieldSet docsWithField, List<float[]> vectors, int dimension, SearchStrategy searchStrategy) {this.docsWithField = docsWithField;this.vectors = vectors;this.dimension = dimension;- this.scoreFunction = scoreFunction;+ this.searchStrategy = searchStrategy;buffer = ByteBuffer.allocate(dimension * Float.BYTES);binaryValue = new BytesRef(buffer.array());raBuffer = ByteBuffer.allocate(dimension * Float.BYTES);@@ -264,8 +264,8 @@ class VectorValuesWriter {}@Override- public VectorValues.ScoreFunction scoreFunction() {- return scoreFunction;+ public SearchStrategy searchStrategy() {+ return searchStrategy;}@Overridediff --git a/lucene/core/src/java/org/apache/lucene/util/VectorUtil.java b/lucene/core/src/java/org/apache/lucene/util/VectorUtil.javanew file mode 100644index 00000000000..e2f8fea38b4--- /dev/null+++ b/lucene/core/src/java/org/apache/lucene/util/VectorUtil.java@@ -0,0 +1,74 @@+/*+ * Licensed to the Apache Software Foundation (ASF) under one or more+ * contributor license agreements. See the NOTICE file distributed with+ * this work for additional information regarding copyright ownership.+ * The ASF licenses this file to You under the Apache License, Version 2.0+ * (the "License"); you may not use this file except in compliance with+ * the License. You may obtain a copy of the License at+ *+ * http://www.apache.org/licenses/LICENSE-2.0+ *+ * Unless required by applicable law or agreed to in writing, software+ * distributed under the License is distributed on an "AS IS" BASIS,+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.+ * See the License for the specific language governing permissions and+ * limitations under the License.+ */++package org.apache.lucene.util;++/**+ * Utilities for computations with numeric arrays+ */+public final class VectorUtil {++ private VectorUtil() {+ }++ public static float dotProduct(float[] a, float[] b) {+ float res = 0f;+ /*+ * If length of vector is larger than 8, we use unrolled dot product to accelerate the+ * calculation.+ */+ int i;+ for (i = 0; i < a.length % 8; i++) {+ res += b[i] * a[i];+ }+ if (a.length < 8) {+ return res;+ }+ float s0 = 0f;+ float s1 = 0f;+ float s2 = 0f;+ float s3 = 0f;+ float s4 = 0f;+ float s5 = 0f;+ float s6 = 0f;+ float s7 = 0f;+ for (; i + 7 < a.length; i += 8) {+ s0 += b[i] * a[i];+ s1 += b[i + 1] * a[i + 1];+ s2 += b[i + 2] * a[i + 2];+ s3 += b[i + 3] * a[i + 3];+ s4 += b[i + 4] * a[i + 4];+ s5 += b[i + 5] * a[i + 5];+ s6 += b[i + 6] * a[i + 6];+ s7 += b[i + 7] * a[i + 7];+ }+ res += s0 + s1 + s2 + s3 + s4 + s5 + s6 + s7;+ return res;+ }++ public static float squareDistance(float[] v1, float[] v2) {+ assert v1.length == v2.length;+ float squareSum = 0.0f;+ int dim = v1.length;+ for (int i = 0; i < dim; i++) {+ float diff = v1[i] - v2[i];+ squareSum += diff * diff;+ }+ return squareSum;+ }++}diff --git a/lucene/core/src/test/org/apache/lucene/index/TestIndexableField.java b/lucene/core/src/test/org/apache/lucene/index/TestIndexableField.javaindex 81e2387d038..1cd3ebf9ffe 100644--- a/lucene/core/src/test/org/apache/lucene/index/TestIndexableField.java+++ b/lucene/core/src/test/org/apache/lucene/index/TestIndexableField.java@@ -112,8 +112,8 @@ public class TestIndexableField extends LuceneTestCase {}@Override- public VectorValues.ScoreFunction vectorScoreFunction() {- return VectorValues.ScoreFunction.NONE;+ public VectorValues.SearchStrategy vectorSearchStrategy() {+ return VectorValues.SearchStrategy.NONE;}@Overridediff --git a/lucene/core/src/test/org/apache/lucene/index/TestPendingSoftDeletes.java b/lucene/core/src/test/org/apache/lucene/index/TestPendingSoftDeletes.javaindex 226f199c9de..4ee913bc111 100644--- a/lucene/core/src/test/org/apache/lucene/index/TestPendingSoftDeletes.java+++ b/lucene/core/src/test/org/apache/lucene/index/TestPendingSoftDeletes.java@@ -37,7 +37,7 @@ import org.apache.lucene.util.IOUtils;import org.apache.lucene.util.StringHelper;import org.apache.lucene.util.Version;-import static org.apache.lucene.index.VectorValues.ScoreFunction.NONE;+import static org.apache.lucene.index.VectorValues.SearchStrategy.NONE;public class TestPendingSoftDeletes extends TestPendingDeletes {diff --git a/lucene/core/src/test/org/apache/lucene/index/TestVectorValues.java b/lucene/core/src/test/org/apache/lucene/index/TestVectorValues.javaindex 153d063fff7..fe280c0afa3 100644--- a/lucene/core/src/test/org/apache/lucene/index/TestVectorValues.java+++ b/lucene/core/src/test/org/apache/lucene/index/TestVectorValues.java@@ -27,7 +27,7 @@ import org.apache.lucene.document.Field.Store;import org.apache.lucene.document.NumericDocValuesField;import org.apache.lucene.document.StringField;import org.apache.lucene.document.VectorField;-import org.apache.lucene.index.VectorValues.ScoreFunction;+import org.apache.lucene.index.VectorValues.SearchStrategy;import org.apache.lucene.search.Sort;import org.apache.lucene.search.SortField;import org.apache.lucene.store.Directory;@@ -57,7 +57,7 @@ public class TestVectorValues extends LuceneTestCase {}try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);}}@@ -67,7 +67,7 @@ public class TestVectorValues extends LuceneTestCase {float[] v = new float[1];VectorField field = new VectorField("f", v);assertEquals(1, field.fieldType().vectorDimension());- assertEquals(ScoreFunction.EUCLIDEAN, field.fieldType().vectorScoreFunction());+ assertEquals(VectorValues.SearchStrategy.EUCLIDEAN_HNSW, field.fieldType().vectorSearchStrategy());assertSame(v, field.vectorValue());}@@ -94,7 +94,7 @@ public class TestVectorValues extends LuceneTestCase {try (Directory dir = newDirectory();IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);if (random().nextBoolean()) {// sometimes test with two segments@@ -102,18 +102,18 @@ public class TestVectorValues extends LuceneTestCase {}Document doc2 = new Document();- doc2.add(new VectorField("f", new float[3], ScoreFunction.DOT_PRODUCT));+ doc2.add(new VectorField("f", new float[3], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));IllegalArgumentException expected = expectThrows(IllegalArgumentException.class,() -> w.addDocument(doc2));assertEquals("cannot change vector dimension from 4 to 3 for field=\"f\"", expected.getMessage());}}- public void testIllegalScoreFunctionChange() throws Exception {+ public void testIllegalSearchStrategyChange() throws Exception {try (Directory dir = newDirectory();IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);if (random().nextBoolean()) {// sometimes test with two segments@@ -121,10 +121,10 @@ public class TestVectorValues extends LuceneTestCase {}Document doc2 = new Document();- doc2.add(new VectorField("f", new float[4], ScoreFunction.EUCLIDEAN));+ doc2.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.EUCLIDEAN_HNSW));IllegalArgumentException expected = expectThrows(IllegalArgumentException.class,() -> w.addDocument(doc2));- assertEquals("cannot change vector score function from DOT_PRODUCT to EUCLIDEAN for field=\"f\"", expected.getMessage());+ assertEquals("cannot change vector search strategy from DOT_PRODUCT_HNSW to EUCLIDEAN_HNSW for field=\"f\"", expected.getMessage());}}@@ -132,13 +132,13 @@ public class TestVectorValues extends LuceneTestCase {try (Directory dir = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);}try (IndexWriter w2 = new IndexWriter(dir, createIndexWriterConfig())) {Document doc2 = new Document();- doc2.add(new VectorField("f", new float[1], ScoreFunction.DOT_PRODUCT));+ doc2.add(new VectorField("f", new float[1], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));IllegalArgumentException expected = expectThrows(IllegalArgumentException.class,() -> w2.addDocument(doc2));assertEquals("cannot change vector dimension from 4 to 1 for field=\"f\"", expected.getMessage());@@ -146,20 +146,20 @@ public class TestVectorValues extends LuceneTestCase {}}- public void testIllegalScoreFunctionChangeTwoWriters() throws Exception {+ public void testIllegalSearchStrategyChangeTwoWriters() throws Exception {try (Directory dir = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);}try (IndexWriter w2 = new IndexWriter(dir, createIndexWriterConfig())) {Document doc2 = new Document();- doc2.add(new VectorField("f", new float[4], ScoreFunction.EUCLIDEAN));+ doc2.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.EUCLIDEAN_HNSW));IllegalArgumentException expected = expectThrows(IllegalArgumentException.class,() -> w2.addDocument(doc2));- assertEquals("cannot change vector score function from DOT_PRODUCT to EUCLIDEAN for field=\"f\"", expected.getMessage());+ assertEquals("cannot change vector search strategy from DOT_PRODUCT_HNSW to EUCLIDEAN_HNSW for field=\"f\"", expected.getMessage());}}}@@ -167,7 +167,7 @@ public class TestVectorValues extends LuceneTestCase {public void testAddIndexesDirectory0() throws Exception {String fieldName = "field";Document doc = new Document();- doc.add(new VectorField(fieldName, new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField(fieldName, new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));try (Directory dir = newDirectory();Directory dir2 = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {@@ -194,7 +194,7 @@ public class TestVectorValues extends LuceneTestCase {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {w.addDocument(doc);}- doc.add(new VectorField(fieldName, new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField(fieldName, new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));try (IndexWriter w2 = new IndexWriter(dir2, createIndexWriterConfig())) {w2.addDocument(doc);w2.addIndexes(new Directory[]{dir});@@ -213,7 +213,7 @@ public class TestVectorValues extends LuceneTestCase {String fieldName = "field";float[] vector = new float[1];Document doc = new Document();- doc.add(new VectorField(fieldName, vector, ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField(fieldName, vector, VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));try (Directory dir = newDirectory();Directory dir2 = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {@@ -244,12 +244,12 @@ public class TestVectorValues extends LuceneTestCase {Directory dir2 = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);}try (IndexWriter w2 = new IndexWriter(dir2, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[5], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[5], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w2.addDocument(doc);IllegalArgumentException expected = expectThrows(IllegalArgumentException.class,() -> w2.addIndexes(new Directory[]{dir}));@@ -258,21 +258,21 @@ public class TestVectorValues extends LuceneTestCase {}}- public void testIllegalScoreFunctionChangeViaAddIndexesDirectory() throws Exception {+ public void testIllegalSearchStrategyChangeViaAddIndexesDirectory() throws Exception {try (Directory dir = newDirectory();Directory dir2 = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);}try (IndexWriter w2 = new IndexWriter(dir2, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.EUCLIDEAN));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.EUCLIDEAN_HNSW));w2.addDocument(doc);IllegalArgumentException expected = expectThrows(IllegalArgumentException.class,() -> w2.addIndexes(dir));- assertEquals("cannot change vector score function from EUCLIDEAN to DOT_PRODUCT for field=\"f\"", expected.getMessage());+ assertEquals("cannot change vector search strategy from EUCLIDEAN_HNSW to DOT_PRODUCT_HNSW for field=\"f\"", expected.getMessage());}}}@@ -282,12 +282,12 @@ public class TestVectorValues extends LuceneTestCase {Directory dir2 = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);}try (IndexWriter w2 = new IndexWriter(dir2, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[5], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[5], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w2.addDocument(doc);try (DirectoryReader r = DirectoryReader.open(dir)) {IllegalArgumentException expected = expectThrows(IllegalArgumentException.class,@@ -298,22 +298,22 @@ public class TestVectorValues extends LuceneTestCase {}}- public void testIllegalScoreFunctionChangeViaAddIndexesCodecReader() throws Exception {+ public void testIllegalSearchStrategyChangeViaAddIndexesCodecReader() throws Exception {try (Directory dir = newDirectory();Directory dir2 = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);}try (IndexWriter w2 = new IndexWriter(dir2, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.EUCLIDEAN));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.EUCLIDEAN_HNSW));w2.addDocument(doc);try (DirectoryReader r = DirectoryReader.open(dir)) {IllegalArgumentException expected = expectThrows(IllegalArgumentException.class,() -> w2.addIndexes(new CodecReader[]{(CodecReader) getOnlyLeafReader(r)}));- assertEquals("cannot change vector score function from EUCLIDEAN to DOT_PRODUCT for field=\"f\"", expected.getMessage());+ assertEquals("cannot change vector search strategy from EUCLIDEAN_HNSW to DOT_PRODUCT_HNSW for field=\"f\"", expected.getMessage());}}}@@ -324,12 +324,12 @@ public class TestVectorValues extends LuceneTestCase {Directory dir2 = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);}try (IndexWriter w2 = new IndexWriter(dir2, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[5], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[5], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w2.addDocument(doc);try (DirectoryReader r = DirectoryReader.open(dir)) {IllegalArgumentException expected = expectThrows(IllegalArgumentException.class,@@ -340,22 +340,22 @@ public class TestVectorValues extends LuceneTestCase {}}- public void testIllegalScoreFunctionChangeViaAddIndexesSlowCodecReader() throws Exception {+ public void testIllegalSearchStrategyChangeViaAddIndexesSlowCodecReader() throws Exception {try (Directory dir = newDirectory();Directory dir2 = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);}try (IndexWriter w2 = new IndexWriter(dir2, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.EUCLIDEAN));+ doc.add(new VectorField("f", new float[4], SearchStrategy.EUCLIDEAN_HNSW));w2.addDocument(doc);try (DirectoryReader r = DirectoryReader.open(dir)) {IllegalArgumentException expected = expectThrows(IllegalArgumentException.class,() -> TestUtil.addIndexesSlowly(w2, r));- assertEquals("cannot change vector score function from EUCLIDEAN to DOT_PRODUCT for field=\"f\"", expected.getMessage());+ assertEquals("cannot change vector search strategy from EUCLIDEAN_HNSW to DOT_PRODUCT_HNSW for field=\"f\"", expected.getMessage());}}}@@ -365,8 +365,8 @@ public class TestVectorValues extends LuceneTestCase {try (Directory dir = newDirectory();IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));IllegalArgumentException expected = expectThrows(IllegalArgumentException.class,() -> w.addDocument(doc));assertEquals("VectorValuesField \"f\" appears more than once in this document (only one value is allowed per field)",@@ -379,10 +379,10 @@ public class TestVectorValues extends LuceneTestCase {IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();expectThrows(IllegalArgumentException.class,- () -> doc.add(new VectorField("f", new float[VectorValues.MAX_DIMENSIONS + 1], ScoreFunction.DOT_PRODUCT)));+ () -> doc.add(new VectorField("f", new float[VectorValues.MAX_DIMENSIONS + 1], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW)));Document doc2 = new Document();- doc2.add(new VectorField("f", new float[1], ScoreFunction.EUCLIDEAN));+ doc2.add(new VectorField("f", new float[1], VectorValues.SearchStrategy.EUCLIDEAN_HNSW));w.addDocument(doc2);}}@@ -392,11 +392,11 @@ public class TestVectorValues extends LuceneTestCase {IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();Exception e = expectThrows(IllegalArgumentException.class,- () -> doc.add(new VectorField("f", new float[0], ScoreFunction.NONE)));+ () -> doc.add(new VectorField("f", new float[0], SearchStrategy.NONE)));assertEquals("cannot index an empty vector", e.getMessage());Document doc2 = new Document();- doc2.add(new VectorField("f", new float[1], ScoreFunction.NONE));+ doc2.add(new VectorField("f", new float[1], VectorValues.SearchStrategy.NONE));w.addDocument(doc2);}}@@ -406,14 +406,14 @@ public class TestVectorValues extends LuceneTestCase {try (Directory dir = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);}IndexWriterConfig iwc = newIndexWriterConfig();iwc.setCodec(Codec.forName("SimpleText"));try (IndexWriter w = new IndexWriter(dir, iwc)) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);w.forceMerge(1);}@@ -427,12 +427,12 @@ public class TestVectorValues extends LuceneTestCase {try (Directory dir = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, iwc)) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);}try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("f", new float[4], ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("f", new float[4], VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);w.forceMerge(1);}@@ -440,7 +440,7 @@ public class TestVectorValues extends LuceneTestCase {}public void testInvalidVectorFieldUsage() {- VectorField field = new VectorField("field", new float[2], ScoreFunction.NONE);+ VectorField field = new VectorField("field", new float[2], VectorValues.SearchStrategy.NONE);expectThrows(IllegalArgumentException.class, () -> field.setIntValue(14));@@ -454,7 +454,7 @@ public class TestVectorValues extends LuceneTestCase {IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();doc.add(new StringField("id", "0", Store.NO));- doc.add(new VectorField("v", new float[]{2, 3, 5}, ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("v", new float[]{2, 3, 5}, VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);w.addDocument(new Document());w.commit();@@ -475,12 +475,12 @@ public class TestVectorValues extends LuceneTestCase {IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();doc.add(new StringField("id", "0", Store.NO));- doc.add(new VectorField("v0", new float[]{2, 3, 5}, ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("v0", new float[]{2, 3, 5}, VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);w.commit();doc = new Document();- doc.add(new VectorField("v1", new float[]{2, 3, 5}, ScoreFunction.DOT_PRODUCT));+ doc.add(new VectorField("v1", new float[]{2, 3, 5}, VectorValues.SearchStrategy.DOT_PRODUCT_HNSW));w.addDocument(doc);w.forceMerge(1);}@@ -492,10 +492,10 @@ public class TestVectorValues extends LuceneTestCase {int[] fieldDocCounts = new int[numFields];float[] fieldTotals= new float[numFields];int[] fieldDims = new int[numFields];- ScoreFunction[] fieldScoreFunctions = new ScoreFunction[numFields];+ VectorValues.SearchStrategy[] fieldSearchStrategies = new VectorValues.SearchStrategy[numFields];for (int i = 0; i < numFields; i++) {fieldDims[i] = random().nextInt(20) + 1;- fieldScoreFunctions[i] = ScoreFunction.values()[random().nextInt(ScoreFunction.values().length)];+ fieldSearchStrategies[i] = VectorValues.SearchStrategy.values()[random().nextInt(VectorValues.SearchStrategy.values().length)];}try (Directory dir = newDirectory();RandomIndexWriter w = new RandomIndexWriter(random(), dir, createIndexWriterConfig())) {@@ -505,7 +505,7 @@ public class TestVectorValues extends LuceneTestCase {String fieldName = "int" + field;if (random().nextInt(100) == 17) {float[] v = randomVector(fieldDims[field]);- doc.add(new VectorField(fieldName, v, fieldScoreFunctions[field]));+ doc.add(new VectorField(fieldName, v, fieldSearchStrategies[field]));fieldDocCounts[field]++;fieldTotals[field] += v[0];}@@ -542,15 +542,15 @@ public class TestVectorValues extends LuceneTestCase {try (Directory dir = newDirectory();IndexWriter iw = new IndexWriter(dir, createIndexWriterConfig())) {Document doc1 = new Document();- doc1.add(new VectorField(fieldName, v, VectorValues.ScoreFunction.EUCLIDEAN));+ doc1.add(new VectorField(fieldName, v, VectorValues.SearchStrategy.EUCLIDEAN_HNSW));v[0] = 1;Document doc2 = new Document();- doc2.add(new VectorField(fieldName, v, VectorValues.ScoreFunction.EUCLIDEAN));+ doc2.add(new VectorField(fieldName, v, VectorValues.SearchStrategy.EUCLIDEAN_HNSW));iw.addDocument(doc1);iw.addDocument(doc2);v[0] = 2;Document doc3 = new Document();- doc3.add(new VectorField(fieldName, v, VectorValues.ScoreFunction.EUCLIDEAN));+ doc3.add(new VectorField(fieldName, v, VectorValues.SearchStrategy.EUCLIDEAN_HNSW));iw.addDocument(doc3);try (IndexReader reader = iw.getReader()) {LeafReader r = reader.leaves().get(0).reader();@@ -693,10 +693,10 @@ public class TestVectorValues extends LuceneTestCase {try (Directory dir = newDirectory()) {try (IndexWriter w = new IndexWriter(dir, createIndexWriterConfig())) {Document doc = new Document();- doc.add(new VectorField("v1", randomVector(3), ScoreFunction.NONE));+ doc.add(new VectorField("v1", randomVector(3), VectorValues.SearchStrategy.NONE));w.addDocument(doc);- doc.add(new VectorField("v2", randomVector(3), ScoreFunction.NONE));+ doc.add(new VectorField("v2", randomVector(3), VectorValues.SearchStrategy.NONE));w.addDocument(doc);}@@ -714,12 +714,12 @@ public class TestVectorValues extends LuceneTestCase {}}- public void testScoreFunctionIdentifiers() throws Exception {- // make sure we don't accidentally mess up score function identifiers by re-ordering their enumerators- assertEquals(0, ScoreFunction.NONE.ordinal());- assertEquals(1, ScoreFunction.EUCLIDEAN.ordinal());- assertEquals(2, ScoreFunction.DOT_PRODUCT.ordinal());- assertEquals(3, ScoreFunction.values().length);+ public void testSearchStrategyIdentifiers() throws Exception {+ // make sure we don't accidentally mess up search strategy identifiers by re-ordering their enumerators+ assertEquals(0, VectorValues.SearchStrategy.NONE.ordinal());+ assertEquals(1, VectorValues.SearchStrategy.EUCLIDEAN_HNSW.ordinal());+ assertEquals(2, VectorValues.SearchStrategy.DOT_PRODUCT_HNSW.ordinal());+ assertEquals(3, VectorValues.SearchStrategy.values().length);}}diff --git a/lucene/core/src/test/org/apache/lucene/util/TestVectorUtil.java b/lucene/core/src/test/org/apache/lucene/util/TestVectorUtil.javanew file mode 100644index 00000000000..14666766a40--- /dev/null+++ b/lucene/core/src/test/org/apache/lucene/util/TestVectorUtil.java@@ -0,0 +1,83 @@+/*+ * Licensed to the Apache Software Foundation (ASF) under one or more+ * contributor license agreements. See the NOTICE file distributed with+ * this work for additional information regarding copyright ownership.+ * The ASF licenses this file to You under the Apache License, Version 2.0+ * (the "License"); you may not use this file except in compliance with+ * the License. You may obtain a copy of the License at+ *+ * http://www.apache.org/licenses/LICENSE-2.0+ *+ * Unless required by applicable law or agreed to in writing, software+ * distributed under the License is distributed on an "AS IS" BASIS,+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.+ * See the License for the specific language governing permissions and+ * limitations under the License.+ */+package org.apache.lucene.util;++public class TestVectorUtil extends LuceneTestCase {++ public void testBasicDotProduct() {+ assertEquals(5, VectorUtil.dotProduct(new float[]{1, 2, 3}, new float[]{-10, 0, 5}), 0);+ }++ public void testSelfDotProduct() {+ // the dot product of a vector with itself is equal to the sum of the squares of its components+ float[] v = randomVector();+ assertEquals(l2(v), VectorUtil.dotProduct(v, v), 1e-5);+ }++ public void testOrthogonalDotProduct() {+ // the dot product of two perpendicular vectors is 0+ float[] v = new float[2];+ v[0] = random().nextInt(100);+ v[1] = random().nextInt(100);+ float[] u = new float[2];+ u[0] = v[1];+ u[1] = -v[0];+ assertEquals(0, VectorUtil.dotProduct(u, v), 1e-5);+ }++ public void testSelfSquareDistance() {+ // the l2 distance of a vector with itself is zero+ float[] v = randomVector();+ assertEquals(0, VectorUtil.squareDistance(v, v), 1e-5);+ }++ public void testBasicSquareDistance() {+ assertEquals(12, VectorUtil.squareDistance(new float[]{1, 2, 3}, new float[]{-1, 0, 5}), 0);+ }++ public void testRandomSquareDistance() {+ // the square distance of a vector with its inverse is equal to four times the sum of squares of its components+ float[] v = randomVector();+ float[] u = negative(v);+ assertEquals(4 * l2(v), VectorUtil.squareDistance(u, v), 1e-5);+ }++ private float l2(float[] v) {+ float l2 = 0;+ for (float x : v) {+ l2 += x * x;+ }+ return l2;+ }++ private float[] negative(float[] v) {+ float[] u = new float[v.length];+ for (int i = 0; i < v.length; i++) {+ u[i] = - v[i];+ }+ return u;+ }++ private float[] randomVector() {+ float[] v = new float[random().nextInt(100) + 1];+ for (int i = 0; i < v.length; i++) {+ v[i] = random().nextFloat();+ }+ return v;+ }++}diff --git a/lucene/highlighter/src/java/org/apache/lucene/search/highlight/TermVectorLeafReader.java b/lucene/highlighter/src/java/org/apache/lucene/search/highlight/TermVectorLeafReader.javaindex f5f8c5c8a60..0b8cb906f77 100644--- a/lucene/highlighter/src/java/org/apache/lucene/search/highlight/TermVectorLeafReader.java+++ b/lucene/highlighter/src/java/org/apache/lucene/search/highlight/TermVectorLeafReader.java@@ -82,7 +82,7 @@ public class TermVectorLeafReader extends LeafReader {}FieldInfo fieldInfo = new FieldInfo(field, 0,true, true, terms.hasPayloads(),- indexOptions, DocValuesType.NONE, -1, Collections.emptyMap(), 0, 0, 0, 0, VectorValues.ScoreFunction.NONE, false);+ indexOptions, DocValuesType.NONE, -1, Collections.emptyMap(), 0, 0, 0, 0, VectorValues.SearchStrategy.NONE, false);fieldInfos = new FieldInfos(new FieldInfo[]{fieldInfo});}diff --git a/lucene/memory/src/java/org/apache/lucene/index/memory/MemoryIndex.java b/lucene/memory/src/java/org/apache/lucene/index/memory/MemoryIndex.javaindex ed00f85f7c3..3bdbee655ec 100644--- a/lucene/memory/src/java/org/apache/lucene/index/memory/MemoryIndex.java+++ b/lucene/memory/src/java/org/apache/lucene/index/memory/MemoryIndex.java@@ -502,7 +502,7 @@ public class MemoryIndex {return new FieldInfo(fieldName, ord, fieldType.storeTermVectors(), fieldType.omitNorms(), storePayloads,indexOptions, fieldType.docValuesType(), -1, Collections.emptyMap(),fieldType.pointDimensionCount(), fieldType.pointIndexDimensionCount(), fieldType.pointNumBytes(),- fieldType.vectorDimension(), fieldType.vectorScoreFunction(), false);+ fieldType.vectorDimension(), fieldType.vectorSearchStrategy(), false);}private void storePointValues(Info info, BytesRef pointValue) {@@ -522,7 +522,7 @@ public class MemoryIndex {info.fieldInfo.name, info.fieldInfo.number, info.fieldInfo.hasVectors(), info.fieldInfo.hasPayloads(),info.fieldInfo.hasPayloads(), info.fieldInfo.getIndexOptions(), docValuesType, -1, info.fieldInfo.attributes(),info.fieldInfo.getPointDimensionCount(), info.fieldInfo.getPointIndexDimensionCount(), info.fieldInfo.getPointNumBytes(),- info.fieldInfo.getVectorDimension(), info.fieldInfo.getVectorScoreFunction(),+ info.fieldInfo.getVectorDimension(), info.fieldInfo.getVectorSearchStrategy(),info.fieldInfo.isSoftDeletesField());} else if (existingDocValuesType != docValuesType) {diff --git a/lucene/test-framework/src/java/org/apache/lucene/index/BaseIndexFileFormatTestCase.java b/lucene/test-framework/src/java/org/apache/lucene/index/BaseIndexFileFormatTestCase.javaindex 6d6b3115808..5a5e4fb4fb1 100644--- a/lucene/test-framework/src/java/org/apache/lucene/index/BaseIndexFileFormatTestCase.java+++ b/lucene/test-framework/src/java/org/apache/lucene/index/BaseIndexFileFormatTestCase.java@@ -355,7 +355,7 @@ abstract class BaseIndexFileFormatTestCase extends LuceneTestCase {FieldInfo field = new FieldInfo(proto.name, proto.number, proto.hasVectors(), proto.omitsNorms(), proto.hasPayloads(),proto.getIndexOptions(), proto.getDocValuesType(), proto.getDocValuesGen(), new HashMap<>(),proto.getPointDimensionCount(), proto.getPointIndexDimensionCount(), proto.getPointNumBytes(),- proto.getVectorDimension(), proto.getVectorScoreFunction(), proto.isSoftDeletesField());+ proto.getVectorDimension(), proto.getVectorSearchStrategy(), proto.isSoftDeletesField());FieldInfos fieldInfos = new FieldInfos(new FieldInfo[] { field } );diff --git a/lucene/test-framework/src/java/org/apache/lucene/index/MismatchedLeafReader.java b/lucene/test-framework/src/java/org/apache/lucene/index/MismatchedLeafReader.javaindex ed8a1107c13..3fd866b6961 100644--- a/lucene/test-framework/src/java/org/apache/lucene/index/MismatchedLeafReader.java+++ b/lucene/test-framework/src/java/org/apache/lucene/index/MismatchedLeafReader.java@@ -81,7 +81,7 @@ public class MismatchedLeafReader extends FilterLeafReader {oldInfo.getPointIndexDimensionCount(), // index dimension countoldInfo.getPointNumBytes(), // dimension numBytesoldInfo.getVectorDimension(), // number of dimensions of the field's vector- oldInfo.getVectorScoreFunction(), // distance function for calculating similarity of the field's vector+ oldInfo.getVectorSearchStrategy(), // distance function for calculating similarity of the field's vectoroldInfo.isSoftDeletesField()); // used as soft-deletes fieldshuffled.set(i, newInfo);}diff --git a/lucene/test-framework/src/java/org/apache/lucene/index/RandomPostingsTester.java b/lucene/test-framework/src/java/org/apache/lucene/index/RandomPostingsTester.javaindex 366a0d9206f..b420e7f251d 100644--- a/lucene/test-framework/src/java/org/apache/lucene/index/RandomPostingsTester.java+++ b/lucene/test-framework/src/java/org/apache/lucene/index/RandomPostingsTester.java@@ -130,7 +130,7 @@ public class RandomPostingsTester {fieldInfoArray[fieldUpto] = new FieldInfo(field, fieldUpto, false, false, true,IndexOptions.DOCS_AND_FREQS_AND_POSITIONS_AND_OFFSETS,DocValuesType.NONE, -1, new HashMap<>(),- 0, 0, 0, 0, VectorValues.ScoreFunction.NONE, false);+ 0, 0, 0, 0, VectorValues.SearchStrategy.NONE, false);fieldUpto++;SortedMap<BytesRef,SeedAndOrd> postings = new TreeMap<>();@@ -651,7 +651,7 @@ public class RandomPostingsTester {DocValuesType.NONE,-1,new HashMap<>(),- 0, 0, 0, 0, VectorValues.ScoreFunction.NONE, false);+ 0, 0, 0, 0, VectorValues.SearchStrategy.NONE, false);}FieldInfos newFieldInfos = new FieldInfos(newFieldInfoArray);diff --git a/solr/core/src/java/org/apache/solr/schema/SchemaField.java b/solr/core/src/java/org/apache/solr/schema/SchemaField.javaindex cea4536ce4f..8dd8c869872 100644--- a/solr/core/src/java/org/apache/solr/schema/SchemaField.java+++ b/solr/core/src/java/org/apache/solr/schema/SchemaField.java@@ -454,8 +454,8 @@ public final class SchemaField extends FieldProperties implements IndexableField}@Override- public VectorValues.ScoreFunction vectorScoreFunction() {- return VectorValues.ScoreFunction.NONE;+ public VectorValues.SearchStrategy vectorSearchStrategy() {+ return VectorValues.SearchStrategy.NONE;}@Overridediff --git a/solr/core/src/java/org/apache/solr/search/CollapsingQParserPlugin.java b/solr/core/src/java/org/apache/solr/search/CollapsingQParserPlugin.javaindex a82a64174ea..db9a56def33 100644--- a/solr/core/src/java/org/apache/solr/search/CollapsingQParserPlugin.java+++ b/solr/core/src/java/org/apache/solr/search/CollapsingQParserPlugin.java@@ -494,7 +494,7 @@ public class CollapsingQParserPlugin extends QParserPlugin {fieldInfo.getPointIndexDimensionCount(),fieldInfo.getPointNumBytes(),fieldInfo.getVectorDimension(),- fieldInfo.getVectorScoreFunction(),+ fieldInfo.getVectorSearchStrategy(),fieldInfo.isSoftDeletesField());newInfos.add(f);} else {diff --git a/solr/core/src/java/org/apache/solr/search/Insanity.java b/solr/core/src/java/org/apache/solr/search/Insanity.javaindex 7eae21750a5..273a2235633 100644--- a/solr/core/src/java/org/apache/solr/search/Insanity.java+++ b/solr/core/src/java/org/apache/solr/search/Insanity.java@@ -68,7 +68,7 @@ public class Insanity {filteredInfos.add(new FieldInfo(fi.name, fi.number, fi.hasVectors(), fi.omitsNorms(),fi.hasPayloads(), fi.getIndexOptions(), DocValuesType.NONE, -1, Collections.emptyMap(),fi.getPointDimensionCount(), fi.getPointIndexDimensionCount(), fi.getPointNumBytes(),- fi.getVectorDimension(), fi.getVectorScoreFunction(), fi.isSoftDeletesField()));+ fi.getVectorDimension(), fi.getVectorSearchStrategy(), fi.isSoftDeletesField()));} else {filteredInfos.add(fi);}diff --git a/solr/core/src/java/org/apache/solr/uninverting/UninvertingReader.java b/solr/core/src/java/org/apache/solr/uninverting/UninvertingReader.javaindex ed62cc4e3fc..8ed09fe2329 100644--- a/solr/core/src/java/org/apache/solr/uninverting/UninvertingReader.java+++ b/solr/core/src/java/org/apache/solr/uninverting/UninvertingReader.java@@ -285,7 +285,7 @@ public class UninvertingReader extends FilterLeafReader {newFieldInfos.add(new FieldInfo(fi.name, fi.number, fi.hasVectors(), fi.omitsNorms(),fi.hasPayloads(), fi.getIndexOptions(), type, fi.getDocValuesGen(), fi.attributes(),fi.getPointDimensionCount(), fi.getPointIndexDimensionCount(), fi.getPointNumBytes(),- fi.getVectorDimension(), fi.getVectorScoreFunction(), fi.isSoftDeletesField()));+ fi.getVectorDimension(), fi.getVectorSearchStrategy(), fi.isSoftDeletesField()));} else {newFieldInfos.add(fi);}