cshuo commented on code in PR #18083: URL: https://github.com/apache/hudi/pull/18083#discussion_r2972558964
########## hudi-flink-datasource/hudi-flink/src/main/java/org/apache/hudi/sink/append/AppendWriteFunctionWithContinuousSort.java: ########## @@ -0,0 +1,330 @@ +/* + * 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.hudi.sink.append; + +import org.apache.hudi.configuration.FlinkOptions; +import org.apache.hudi.exception.HoodieException; +import org.apache.hudi.exception.HoodieIOException; +import org.apache.hudi.sink.StreamWriteOperatorCoordinator; +import org.apache.hudi.sink.bulk.sort.SortOperatorGen; + +import org.apache.flink.configuration.Configuration; +import org.apache.flink.core.memory.MemorySegment; +import org.apache.flink.core.memory.MemorySegmentFactory; +import org.apache.flink.table.data.RowData; +import org.apache.flink.table.planner.codegen.sort.SortCodeGenerator; +import org.apache.flink.table.runtime.typeutils.RowDataSerializer; +import org.apache.flink.table.runtime.generated.GeneratedNormalizedKeyComputer; +import org.apache.flink.table.runtime.generated.GeneratedRecordComparator; +import org.apache.flink.table.runtime.generated.NormalizedKeyComputer; +import org.apache.flink.table.runtime.generated.RecordComparator; +import org.apache.flink.table.types.logical.RowType; +import org.apache.flink.util.Collector; + +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.io.IOException; +import java.util.Arrays; +import java.util.Iterator; +import java.util.List; +import java.util.Map; +import java.util.TreeMap; +import java.util.stream.Collectors; + +/** + * Sink function to write data with continuous sorting for improved compression. + * + * <p>Unlike {@link AppendWriteFunctionWithBIMBufferSort} which uses batch sorting, + * this function maintains sorted order continuously using a TreeMap, providing: + * <ul> + * <li>Non-blocking inserts (O(log n) vs O(1) + periodic O(n log n))</li> + * <li>Incremental draining without re-sorting</li> + * <li>Predictable latency (no sort spikes)</li> + * </ul> + * + * <p>Strategy: + * <ol> + * <li>Records are inserted in sorted order (TreeMap)</li> + * <li>When buffer reaches max capacity, oldest record(s) are drained synchronously</li> + * <li>Drain size is configurable to balance latency vs. throughput vs compression ratio</li> + * </ol> + * + * @param <T> Type of the input record + * @see StreamWriteOperatorCoordinator + */ +public class AppendWriteFunctionWithContinuousSort<T> extends AppendWriteFunction<T> { + + private static final Logger LOG = LoggerFactory.getLogger(AppendWriteFunctionWithContinuousSort.class); + + private final long maxCapacity; + private final int drainSize; + + private transient TreeMap<SortKey, RowData> sortedRecords; + private transient long insertionSequence; + + // Sort key computation + private transient NormalizedKeyComputer normalizedKeyComputer; + private transient RecordComparator recordComparator; + private transient MemorySegment reusableKeySegment; + private transient int normalizedKeySize; + private transient boolean objectReuseEnabled; + private transient RowDataSerializer rowDataSerializer; + + // Metrics + private transient long totalDrainOperations; + private transient long totalDrainedRecords; + private transient long totalInserted; + + public AppendWriteFunctionWithContinuousSort(Configuration config, RowType rowType) { + super(config, rowType); + + // Configuration + this.maxCapacity = config.get(FlinkOptions.WRITE_BUFFER_SIZE); + this.drainSize = config.get(FlinkOptions.WRITE_BUFFER_SORT_CONTINUOUS_DRAIN_SIZE); + + LOG.info("AppendWriteFunctionWithContinuousSort created: maxCapacity={}, drainSize={}", + maxCapacity, drainSize); + } + + @Override + public void open(Configuration parameters) throws Exception { + // Validate configuration before calling super.open() which requires Flink runtime context + if (maxCapacity <= 0) { + throw new IllegalArgumentException( + String.format("Buffer capacity must be positive, got: %d", maxCapacity)); + } + + if (drainSize <= 0) { + throw new IllegalArgumentException( + String.format("Drain size must be positive, got: %d", drainSize)); + } + + // Resolve sort keys, falling back to record key if not specified + List<String> sortKeyList = AppendWriteFunctions.resolveSortKeys(config); + + super.open(parameters); + + LOG.info("Initializing continuous sort with keys: {}", sortKeyList); + + // Create sort code generator for normalized key computation and record comparison + SortOperatorGen sortOperatorGen = new SortOperatorGen(rowType, sortKeyList.toArray(new String[0])); + SortCodeGenerator codeGenerator = sortOperatorGen.createSortCodeGenerator(); + GeneratedNormalizedKeyComputer generatedKeyComputer = codeGenerator.generateNormalizedKeyComputer("ContinuousSortKeyComputer"); + GeneratedRecordComparator generatedComparator = codeGenerator.generateRecordComparator("ContinuousSortComparator"); + + // Instantiate code-generated components + ClassLoader classLoader = Thread.currentThread().getContextClassLoader(); + this.normalizedKeyComputer = generatedKeyComputer.newInstance(classLoader); + this.recordComparator = generatedComparator.newInstance(classLoader); + this.normalizedKeySize = normalizedKeyComputer.getNumKeyBytes(); + + // Initialize TreeMap with comparator that uses normalized keys for fast comparison + // and falls back to RecordComparator for full comparison when normalized keys are equal + this.sortedRecords = new TreeMap<>((k1, k2) -> { + MemorySegment seg1 = MemorySegmentFactory.wrap(k1.keyBytes); + MemorySegment seg2 = MemorySegmentFactory.wrap(k2.keyBytes); + + int cmp = normalizedKeyComputer.compareKey(seg1, 0, seg2, 0); + if (cmp != 0) { + return cmp; + } + // Normalized keys are equal - use full record comparison for correct ordering + cmp = recordComparator.compare(k1.record, k2.record); + if (cmp != 0) { + return cmp; + } + // Records are equal by sort keys - use insertion order for stability + return Long.compare(k1.insertionOrder, k2.insertionOrder); + }); + this.insertionSequence = 0L; + + // Allocate reusable on-heap buffer for computing keys + byte[] reusableKeyBuffer = new byte[normalizedKeySize]; + this.reusableKeySegment = MemorySegmentFactory.wrap(reusableKeyBuffer); + + // Detect object reuse mode and create serializer for copying if needed + this.objectReuseEnabled = getRuntimeContext().isObjectReuseEnabled(); + if (this.objectReuseEnabled) { + this.rowDataSerializer = new RowDataSerializer(rowType); + } + + // Initialize metrics + this.totalDrainOperations = 0; + this.totalDrainedRecords = 0; + this.totalInserted = 0; + + LOG.info("AppendWriteFunctionWithContinuousSort initialized successfully"); + } + + @Override + public void processElement(T value, Context ctx, Collector<RowData> out) throws Exception { + RowData data = (RowData) value; + + // Check if buffer has reached max capacity + if (sortedRecords.size() >= maxCapacity) { + drainRecords(drainSize); + + // Verify there's space after draining + if (sortedRecords.size() >= maxCapacity) { + throw new HoodieException( + String.format("Buffer cannot accept record after draining. " + + "Buffer size: %d, maxCapacity: %d, drainSize: %d", + sortedRecords.size(), maxCapacity, drainSize)); + } + } + + // Copy RowData when object reuse is enabled to prevent mutation after insertion + if (objectReuseEnabled) { + data = rowDataSerializer.copy(data); + } + + // Write to buffer (maintains sorted order) + // Compute normalized key into reusable segment + normalizedKeyComputer.putKey(data, reusableKeySegment, 0); + + // Create sort key (copies the normalized key from reusable segment) + SortKey key = new SortKey(reusableKeySegment, normalizedKeySize, data, insertionSequence++); + + // Store the RowData + sortedRecords.put(key, data); + + totalInserted++; + } + + /** + * Drain oldest records from buffer and write to storage. + */ + private void drainRecords(int count) throws IOException { + if (sortedRecords.isEmpty()) { + return; + } + + // Initialize writer if needed + if (this.writerHelper == null) { + initWriterHelper(); + } + + // Drain records from TreeMap + int actualCount = Math.min(count, sortedRecords.size()); + int drained = 0; + + Iterator<Map.Entry<SortKey, RowData>> iterator = sortedRecords.entrySet().iterator(); + while (iterator.hasNext() && drained < actualCount) { + Map.Entry<SortKey, RowData> entry = iterator.next(); + RowData record = entry.getValue(); + + // Write record + writerHelper.write(record); + + // Remove from TreeMap - memory immediately reclaimed + iterator.remove(); + drained++; + } + + totalDrainOperations++; + totalDrainedRecords += drained; + } + + @Override + public void snapshotState() { + try { + // Drain all remaining records and reset for next checkpoint interval + if (!sortedRecords.isEmpty()) { + LOG.info("Snapshot: draining {} remaining records", sortedRecords.size()); + drainRecords(sortedRecords.size()); + sortedRecords.clear(); + insertionSequence = 0L; + } + + LOG.info("Snapshot complete: total drained={}, operations={}", + totalDrainedRecords, totalDrainOperations); + + } catch (IOException e) { + throw new HoodieIOException("Failed to drain buffer during snapshot", e); + } finally { + super.snapshotState(); Review Comment: Let's move `super.snapshotState();` out of the `finally` clause like other append write functions, since write meta event should not be sent to coordinator when there is ingestion exception. -- 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]
