org.apache.hudi.table.action.commit.FlinkWriteHelper Maven / Gradle / Ivy

Go to download
/*
 * 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.table.action.commit;

import org.apache.hudi.client.WriteStatus;
import org.apache.hudi.common.config.TypedProperties;
import org.apache.hudi.common.data.HoodieListData;
import org.apache.hudi.common.engine.HoodieEngineContext;
import org.apache.hudi.common.model.HoodieKey;
import org.apache.hudi.common.model.HoodieOperation;
import org.apache.hudi.common.model.HoodieRecord;
import org.apache.hudi.common.model.HoodieRecordMerger;
import org.apache.hudi.common.model.WriteOperationType;
import org.apache.hudi.exception.HoodieException;
import org.apache.hudi.exception.HoodieUpsertException;
import org.apache.hudi.index.HoodieIndex;
import org.apache.hudi.table.HoodieTable;
import org.apache.hudi.table.action.HoodieWriteMetadata;

import org.apache.avro.Schema;

import java.io.IOException;
import java.time.Duration;
import java.time.Instant;
import java.util.List;
import java.util.Map;
import java.util.Objects;
import java.util.stream.Collectors;

/**
 * Overrides the {@link #write} method to not look up index and partition the records, because
 * with {@code org.apache.hudi.operator.partitioner.BucketAssigner}, each hoodie record
 * is tagged with a bucket ID (partition path + fileID) in streaming way. The FlinkWriteHelper only hands over
 * the records to the action executor {@link BaseCommitActionExecutor} to execute.
 *
 * Computing the records batch locations all at a time is a pressure to the engine,
 * we should avoid that in streaming system.
 */
public class FlinkWriteHelper extends BaseWriteHelper>,
    List, List, R> {

  private FlinkWriteHelper() {
    super(ignored -> -1);
  }

  private static class WriteHelperHolder {
    private static final FlinkWriteHelper FLINK_WRITE_HELPER = new FlinkWriteHelper();
  }

  public static FlinkWriteHelper newInstance() {
    return WriteHelperHolder.FLINK_WRITE_HELPER;
  }

  @Override
  public HoodieWriteMetadata> write(String instantTime, List> inputRecords, HoodieEngineContext context,
                                                      HoodieTable>, List, List> table, boolean shouldCombine, int configuredShuffleParallelism,
                                                      BaseCommitActionExecutor>, List, List, R> executor, WriteOperationType operationType) {
    try {
      Instant lookupBegin = Instant.now();
      Duration indexLookupDuration = Duration.between(lookupBegin, Instant.now());

      HoodieWriteMetadata> result = executor.execute(inputRecords);
      result.setIndexLookupDuration(indexLookupDuration);
      return result;
    } catch (Throwable e) {
      if (e instanceof HoodieUpsertException) {
        throw (HoodieUpsertException) e;
      }
      throw new HoodieUpsertException("Failed to upsert for commit time " + instantTime, e);
    }
  }

  @Override
  protected List> tag(List> dedupedRecords, HoodieEngineContext context, HoodieTable>, List, List> table) {
    return table.getIndex().tagLocation(HoodieListData.eager(dedupedRecords), context, table).collectAsList();
  }

  @Override
  public List> deduplicateRecords(
      List> records, HoodieIndex index, int parallelism, String schemaStr, TypedProperties props, HoodieRecordMerger merger) {
    // If index used is global, then records are expected to differ in their partitionPath
    Map>> keyedRecords = records.stream()
        .collect(Collectors.groupingBy(record -> record.getKey().getRecordKey()));

    // caution that the avro schema is not serializable
    final Schema schema = new Schema.Parser().parse(schemaStr);
    return keyedRecords.values().stream().map(x -> x.stream().reduce((rec1, rec2) -> {
      HoodieRecord reducedRecord;
      try {
        // Precombine do not need schema and do not return null
        reducedRecord =  merger.merge(rec1, schema, rec2, schema, props).get().getLeft();
      } catch (IOException e) {
        throw new HoodieException(String.format("Error to merge two records, %s, %s", rec1, rec2), e);
      }
      // we cannot allow the user to change the key or partitionPath, since that will affect
      // everything
      // so pick it from one of the records.
      boolean choosePrev = rec1.getData() == reducedRecord.getData();
      HoodieKey reducedKey = choosePrev ? rec1.getKey() : rec2.getKey();
      HoodieOperation operation = choosePrev ? rec1.getOperation() : rec2.getOperation();
      HoodieRecord hoodieRecord = reducedRecord.newInstance(reducedKey, operation);
      // reuse the location from the first record.
      hoodieRecord.setCurrentLocation(rec1.getCurrentLocation());
      return hoodieRecord;
    }).orElse(null)).filter(Objects::nonNull).collect(Collectors.toList());
  }
}