|
| 1 | +/* |
| 2 | + * Licensed to the Apache Software Foundation (ASF) under one |
| 3 | + * or more contributor license agreements. See the NOTICE file |
| 4 | + * distributed with this work for additional information |
| 5 | + * regarding copyright ownership. The ASF licenses this file |
| 6 | + * to you under the Apache License, Version 2.0 (the |
| 7 | + * "License"); you may not use this file except in compliance |
| 8 | + * with the License. You may obtain a copy of the License at |
| 9 | + * |
| 10 | + * http://www.apache.org/licenses/LICENSE-2.0 |
| 11 | + * |
| 12 | + * Unless required by applicable law or agreed to in writing, |
| 13 | + * software distributed under the License is distributed on an |
| 14 | + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY |
| 15 | + * KIND, either express or implied. See the License for the |
| 16 | + * specific language governing permissions and limitations |
| 17 | + * under the License. |
| 18 | + */ |
| 19 | + |
| 20 | +package org.apache.hudi.table.action.commit; |
| 21 | + |
| 22 | +import org.apache.hudi.common.engine.HoodieEngineContext; |
| 23 | +import org.apache.hudi.common.model.HoodieKey; |
| 24 | +import org.apache.hudi.common.model.HoodieRecordLocation; |
| 25 | +import org.apache.hudi.common.model.WriteOperationType; |
| 26 | +import org.apache.hudi.common.util.Option; |
| 27 | +import org.apache.hudi.common.util.ValidationUtils; |
| 28 | +import org.apache.hudi.common.util.collection.Pair; |
| 29 | +import org.apache.hudi.config.HoodieWriteConfig; |
| 30 | +import org.apache.hudi.exception.HoodieException; |
| 31 | +import org.apache.hudi.index.bucket.BucketIdentifier; |
| 32 | +import org.apache.hudi.index.bucket.HoodieBucketIndex; |
| 33 | +import org.apache.hudi.table.HoodieTable; |
| 34 | +import org.apache.hudi.table.WorkloadProfile; |
| 35 | +import org.apache.hudi.table.WorkloadStat; |
| 36 | + |
| 37 | +import java.util.ArrayList; |
| 38 | +import java.util.Collections; |
| 39 | +import java.util.HashMap; |
| 40 | +import java.util.HashSet; |
| 41 | +import java.util.List; |
| 42 | +import java.util.Map; |
| 43 | +import java.util.Set; |
| 44 | + |
| 45 | +import static org.apache.hudi.common.model.WriteOperationType.INSERT_OVERWRITE; |
| 46 | +import static org.apache.hudi.common.model.WriteOperationType.INSERT_OVERWRITE_TABLE; |
| 47 | + |
| 48 | +/** |
| 49 | + * Packs incoming records to be inserted into buckets (1 bucket = 1 partition) for Java engine. |
| 50 | + * Equivalent to SparkBucketIndexPartitioner. |
| 51 | + * TODO: Reduce duplicate code between Spark and Java version. |
| 52 | + */ |
| 53 | +public class JavaBucketIndexPartitioner implements JavaPartitioner { |
| 54 | + |
| 55 | + private final int numBuckets; |
| 56 | + private final String indexKeyField; |
| 57 | + private final int totalPartitionPaths; |
| 58 | + private final List<String> partitionPaths; |
| 59 | + /** |
| 60 | + * Helps get the partition id, partition id is partition offset + bucket id. |
| 61 | + * The partition offset is a multiple of the bucket num. |
| 62 | + */ |
| 63 | + private final Map<String, Integer> partitionPathOffset; |
| 64 | + private final boolean isOverwrite; |
| 65 | + |
| 66 | + /** |
| 67 | + * Partition path and file groups in it pair. Decide the file group an incoming update should go to. |
| 68 | + */ |
| 69 | + private final Map<String, Set<String>> updatePartitionPathFileIds; |
| 70 | + |
| 71 | + /** |
| 72 | + * Bucket number to bucket info mapping. |
| 73 | + */ |
| 74 | + private final Map<Integer, BucketInfo> bucketInfoMap; |
| 75 | + |
| 76 | + private final boolean isNonBlockingConcurrencyControl; |
| 77 | + |
| 78 | + public JavaBucketIndexPartitioner(WorkloadProfile profile, |
| 79 | + HoodieEngineContext context, |
| 80 | + HoodieTable table, |
| 81 | + HoodieWriteConfig config) { |
| 82 | + if (!(table.getIndex() instanceof HoodieBucketIndex)) { |
| 83 | + throw new HoodieException( |
| 84 | + "Bucket index partitioner should only be used by BucketIndex other than " |
| 85 | + + table.getIndex().getClass().getSimpleName()); |
| 86 | + } |
| 87 | + this.numBuckets = ((HoodieBucketIndex) table.getIndex()).getNumBuckets(); |
| 88 | + this.indexKeyField = config.getBucketIndexHashField(); |
| 89 | + this.totalPartitionPaths = profile.getPartitionPaths().size(); |
| 90 | + this.partitionPaths = new ArrayList<>(profile.getPartitionPaths()); |
| 91 | + this.partitionPathOffset = new HashMap<>(); |
| 92 | + int i = 0; |
| 93 | + for (Object partitionPath : profile.getPartitionPaths()) { |
| 94 | + partitionPathOffset.put(partitionPath.toString(), i); |
| 95 | + i += numBuckets; |
| 96 | + } |
| 97 | + this.updatePartitionPathFileIds = new HashMap<>(); |
| 98 | + this.bucketInfoMap = new HashMap<>(); |
| 99 | + assignUpdates(profile); |
| 100 | + WriteOperationType operationType = profile.getOperationType(); |
| 101 | + this.isOverwrite = INSERT_OVERWRITE.equals(operationType) || INSERT_OVERWRITE_TABLE.equals(operationType); |
| 102 | + this.isNonBlockingConcurrencyControl = config.isNonBlockingConcurrencyControl(); |
| 103 | + |
| 104 | + if (isOverwrite) { |
| 105 | + ValidationUtils.checkArgument(!isNonBlockingConcurrencyControl, |
| 106 | + "Insert overwrite is not supported with non-blocking concurrency control"); |
| 107 | + } |
| 108 | + } |
| 109 | + |
| 110 | + private void assignUpdates(WorkloadProfile profile) { |
| 111 | + // Each update location gets tracked |
| 112 | + Set<Map.Entry<String, WorkloadStat>> partitionStatEntries = profile.getInputPartitionPathStatMap() |
| 113 | + .entrySet(); |
| 114 | + for (Map.Entry<String, WorkloadStat> partitionStat : partitionStatEntries) { |
| 115 | + if (!updatePartitionPathFileIds.containsKey(partitionStat.getKey())) { |
| 116 | + updatePartitionPathFileIds.put(partitionStat.getKey(), new HashSet<>()); |
| 117 | + } |
| 118 | + for (Map.Entry<String, Pair<String, Long>> updateLocEntry : |
| 119 | + partitionStat.getValue().getUpdateLocationToCount().entrySet()) { |
| 120 | + updatePartitionPathFileIds.get(partitionStat.getKey()).add(updateLocEntry.getKey()); |
| 121 | + } |
| 122 | + } |
| 123 | + } |
| 124 | + |
| 125 | + @Override |
| 126 | + public int getNumPartitions() { |
| 127 | + return totalPartitionPaths * numBuckets; |
| 128 | + } |
| 129 | + |
| 130 | + @Override |
| 131 | + public int getPartition(Object key) { |
| 132 | + Pair<HoodieKey, Option<HoodieRecordLocation>> keyLocation = (Pair<HoodieKey, Option<HoodieRecordLocation>>) key; |
| 133 | + String partitionPath = keyLocation.getLeft().getPartitionPath(); |
| 134 | + Option<HoodieRecordLocation> location = keyLocation.getRight(); |
| 135 | + int bucketId = location.isPresent() |
| 136 | + ? BucketIdentifier.bucketIdFromFileId(location.get().getFileId()) |
| 137 | + : BucketIdentifier.getBucketId(keyLocation.getLeft().getRecordKey(), indexKeyField, numBuckets); |
| 138 | + return partitionPathOffset.get(partitionPath) + bucketId; |
| 139 | + } |
| 140 | + |
| 141 | + public BucketInfo getBucketInfo(int bucketNumber) { |
| 142 | + return bucketInfoMap.computeIfAbsent(bucketNumber, k -> { |
| 143 | + int bucketId = bucketNumber % numBuckets; |
| 144 | + String partitionPath = partitionPaths.get(bucketNumber / numBuckets); |
| 145 | + return getBucketInfo(bucketId, partitionPath); |
| 146 | + }); |
| 147 | + } |
| 148 | + |
| 149 | + protected BucketInfo getBucketInfo(int bucketId, String partitionPath) { |
| 150 | + String bucketIdStr = BucketIdentifier.bucketIdStr(bucketId); |
| 151 | + // Insert overwrite always generates new bucket file id |
| 152 | + if (isOverwrite) { |
| 153 | + return new BucketInfo(BucketType.INSERT, BucketIdentifier.newBucketFileIdPrefix(bucketIdStr), partitionPath); |
| 154 | + } |
| 155 | + Option<String> fileIdOption = Option.fromJavaOptional(updatePartitionPathFileIds |
| 156 | + .getOrDefault(partitionPath, Collections.emptySet()).stream() |
| 157 | + .filter(e -> e.startsWith(bucketIdStr)) |
| 158 | + .findFirst()); |
| 159 | + if (fileIdOption.isPresent()) { |
| 160 | + return new BucketInfo(BucketType.UPDATE, fileIdOption.get(), partitionPath); |
| 161 | + } else { |
| 162 | + // Always write into log file instead of base file if using NB-CC |
| 163 | + if (isNonBlockingConcurrencyControl) { |
| 164 | + return new BucketInfo(BucketType.UPDATE, BucketIdentifier.newBucketFileIdForNBCC(bucketIdStr), partitionPath); |
| 165 | + } |
| 166 | + return new BucketInfo(BucketType.INSERT, BucketIdentifier.newBucketFileIdPrefix(bucketIdStr), partitionPath); |
| 167 | + } |
| 168 | + } |
| 169 | + |
| 170 | + @Override |
| 171 | + public List<String> getSmallFileIds() { |
| 172 | + return Collections.emptyList(); |
| 173 | + } |
| 174 | +} |
0 commit comments