kaori-seasons commented on code in PR #6773: URL: https://github.com/apache/paimon/pull/6773#discussion_r2601284902
########## paimon-lucene/src/main/java/org/apache/paimon/lucene/index/VectorGlobalIndexWriter.java: ########## @@ -0,0 +1,197 @@ +/* + * 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.paimon.lucene.index; + +import org.apache.paimon.globalindex.GlobalIndexWriter; +import org.apache.paimon.globalindex.io.GlobalIndexFileWriter; +import org.apache.paimon.options.Options; +import org.apache.paimon.types.ArrayType; +import org.apache.paimon.types.DataType; +import org.apache.paimon.utils.Range; + +import org.apache.lucene.codecs.KnnVectorsFormat; +import org.apache.lucene.codecs.lucene912.Lucene912Codec; +import org.apache.lucene.codecs.lucene99.Lucene99HnswVectorsFormat; +import org.apache.lucene.document.Document; +import org.apache.lucene.index.IndexWriter; +import org.apache.lucene.index.IndexWriterConfig; +import org.apache.lucene.index.VectorSimilarityFunction; +import org.apache.lucene.store.Directory; +import org.apache.lucene.store.IOContext; + +import java.io.BufferedOutputStream; +import java.io.IOException; +import java.io.OutputStream; +import java.nio.ByteBuffer; +import java.nio.charset.StandardCharsets; +import java.util.ArrayList; +import java.util.List; + +import static org.apache.paimon.utils.Preconditions.checkArgument; + +/** + * Vector global index writer using Apache Lucene 9.x. + * + * <p>This implementation uses Lucene's native KnnFloatVectorField with HNSW algorithm for efficient + * approximate nearest neighbor search. + */ +public class VectorGlobalIndexWriter implements GlobalIndexWriter { + + private final GlobalIndexFileWriter fileWriter; + private final VectorIndexOptions vectorOptions; + private final VectorSimilarityFunction similarityFunction; + private final int sizePerIndex; + + private final List<VectorIndex> vectorIndices; + private final List<ResultEntry> results; + + public VectorGlobalIndexWriter( + GlobalIndexFileWriter fileWriter, DataType fieldType, Options options) { + checkArgument( + fieldType instanceof ArrayType, + "Vector field type must be ARRAY, but was: " + fieldType); + this.fileWriter = fileWriter; + this.vectorIndices = new ArrayList<>(); + this.results = new ArrayList<>(); + this.vectorOptions = new VectorIndexOptions(options); + this.similarityFunction = vectorOptions.metric().vectorSimilarityFunction(); + this.sizePerIndex = vectorOptions.sizePerIndex(); + } + + @Override + public void write(Object key) { + VectorIndex index; + if (key instanceof FloatVectorIndex) { + index = (FloatVectorIndex) key; + } else if (key instanceof ByteVectorIndex) { + index = (ByteVectorIndex) key; + } else { + throw new IllegalArgumentException( + "Unsupported index type: " + key.getClass().getName()); + } + index.checkDimension(vectorOptions.dimension()); + vectorIndices.add(index); + if (vectorIndices.size() >= sizePerIndex) { + try { + flush(); + } catch (IOException e) { + throw new RuntimeException(e); + } + } + } + + @Override + public List<ResultEntry> finish() { + try { + if (!vectorIndices.isEmpty()) { + flush(); + } + + return results; + } catch (IOException e) { + throw new RuntimeException("Failed to write vector global index", e); + } + } + + private void flush() throws IOException { + String fileName = fileWriter.newFileName(VectorGlobalIndexerFactory.IDENTIFIER); + try (OutputStream out = new BufferedOutputStream(fileWriter.newOutputStream(fileName))) { + buildIndex( + vectorIndices, + this.vectorOptions.m(), + this.vectorOptions.efConstruction(), + this.vectorOptions.writeBufferSize(), + out); + } + long minRowIdInBatch = vectorIndices.get(0).rowId(); + long maxRowIdInBatch = vectorIndices.get(vectorIndices.size() - 1).rowId(); + results.add(ResultEntry.of(fileName, null, new Range(minRowIdInBatch, maxRowIdInBatch))); + vectorIndices.clear(); + } + + private void buildIndex( + List<VectorIndex> batchVectors, + int m, + int efConstruction, + int writeBufferSize, + OutputStream out) + throws IOException { + + IndexWriterConfig config = getIndexWriterConfig(m, efConstruction, writeBufferSize); + try (IndexMMapDirectory indexMMapDirectory = new IndexMMapDirectory()) { + try (IndexWriter writer = new IndexWriter(indexMMapDirectory.directory(), config)) { + for (VectorIndex vectorIndex : batchVectors) { + Document doc = new Document(); + doc.add(vectorIndex.indexableField(similarityFunction)); + doc.add(vectorIndex.rowIdStoredField()); + writer.addDocument(doc); + } + writer.commit(); + } + + serializeDirectory(indexMMapDirectory.directory(), out); + } catch (Exception e) { + throw new RuntimeException(e); + } + } + + private static IndexWriterConfig getIndexWriterConfig( + int m, int efConstruction, int writeBufferSize) { + IndexWriterConfig config = new IndexWriterConfig(); + config.setRAMBufferSizeMB(writeBufferSize); + config.setCodec( + new Lucene912Codec(Lucene912Codec.Mode.BEST_SPEED) { + @Override + public KnnVectorsFormat getKnnVectorsFormatForField(String field) { + return new Lucene99HnswVectorsFormat(m, efConstruction); + } + }); + return config; + } + + private void serializeDirectory(Directory directory, OutputStream out) throws IOException { + String[] files = directory.listAll(); + out.write(intToBytes(files.length)); + + for (String fileName : files) { + byte[] nameBytes = fileName.getBytes(StandardCharsets.UTF_8); + out.write(intToBytes(nameBytes.length)); + out.write(nameBytes); + long fileLength = directory.fileLength(fileName); + out.write(ByteBuffer.allocate(8).putLong(fileLength).array()); + + try (org.apache.lucene.store.IndexInput input = Review Comment: Current situation: `VectorGlobalIndexWriter#getIndexWriterConfig` uses the parameterless constructor `new IndexWriterConfig()`, and calls `directory.createOutput(fileName, null)` (passing null) during serialization/deserialization. Recommendation: Use `IOContext.DEFAULT` in `createOutput`. Also use `IOContext.DEFAULT` when reading/writing `IndexInput`/`IndexOutput`. Examples: ``` try (IndexInput input = directory.openInput(fileName, IOContext.DEFAULT)) { byte[] buffer = new byte[32 * 1024]; long remaining = fileLength; while (remaining > 0) { int toRead = (int) Math.min(buffer.length, remaining); input.readBytes(buffer, 0, toRead); out.write(buffer, 0, toRead); remaining -= toRead; } } ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
