Examples with HoodieData org.apache.hudi.common.data.HoodieData used on opensource projects

Example 11 with HoodieData

use of org.apache.hudi.common.data.HoodieData in project hudi by apache.

the class HoodieTableMetadataUtil method convertFilesToColumnStatsRecords.

/**
 * Convert added and deleted action metadata to column stats index records.
 */
public static HoodieData<HoodieRecord> convertFilesToColumnStatsRecords(HoodieEngineContext engineContext, Map<String, List<String>> partitionToDeletedFiles, Map<String, Map<String, Long>> partitionToAppendedFiles, MetadataRecordsGenerationParams recordsGenerationParams) {
    HoodieData<HoodieRecord> allRecordsRDD = engineContext.emptyHoodieData();
    final List<String> columnsToIndex = getColumnsToIndex(recordsGenerationParams.getDataMetaClient(), recordsGenerationParams.isAllColumnStatsIndexEnabled());
    final List<Pair<String, List<String>>> partitionToDeletedFilesList = partitionToDeletedFiles.entrySet().stream().map(e -> Pair.of(e.getKey(), e.getValue())).collect(Collectors.toList());
    int parallelism = Math.max(Math.min(partitionToDeletedFilesList.size(), recordsGenerationParams.getColumnStatsIndexParallelism()), 1);
    final HoodieData<Pair<String, List<String>>> partitionToDeletedFilesRDD = engineContext.parallelize(partitionToDeletedFilesList, parallelism);
    HoodieData<HoodieRecord> deletedFilesRecordsRDD = partitionToDeletedFilesRDD.flatMap(partitionToDeletedFilesPair -> {
        final String partitionName = partitionToDeletedFilesPair.getLeft();
        final String partition = getPartition(partitionName);
        final List<String> deletedFileList = partitionToDeletedFilesPair.getRight();
        return deletedFileList.stream().flatMap(deletedFile -> {
            final String filePathWithPartition = partitionName + "/" + deletedFile;
            return getColumnStats(partition, filePathWithPartition, recordsGenerationParams.getDataMetaClient(), columnsToIndex, true);
        }).iterator();
    });
    allRecordsRDD = allRecordsRDD.union(deletedFilesRecordsRDD);
    final List<Pair<String, Map<String, Long>>> partitionToAppendedFilesList = partitionToAppendedFiles.entrySet().stream().map(entry -> Pair.of(entry.getKey(), entry.getValue())).collect(Collectors.toList());
    parallelism = Math.max(Math.min(partitionToAppendedFilesList.size(), recordsGenerationParams.getColumnStatsIndexParallelism()), 1);
    final HoodieData<Pair<String, Map<String, Long>>> partitionToAppendedFilesRDD = engineContext.parallelize(partitionToAppendedFilesList, parallelism);
    HoodieData<HoodieRecord> appendedFilesRecordsRDD = partitionToAppendedFilesRDD.flatMap(partitionToAppendedFilesPair -> {
        final String partitionName = partitionToAppendedFilesPair.getLeft();
        final String partition = getPartition(partitionName);
        final Map<String, Long> appendedFileMap = partitionToAppendedFilesPair.getRight();
        return appendedFileMap.entrySet().stream().flatMap(appendedFileNameLengthEntry -> {
            if (!FSUtils.isBaseFile(new Path(appendedFileNameLengthEntry.getKey())) || !appendedFileNameLengthEntry.getKey().endsWith(HoodieFileFormat.PARQUET.getFileExtension())) {
                return Stream.empty();
            }
            final String filePathWithPartition = partitionName + "/" + appendedFileNameLengthEntry.getKey();
            return getColumnStats(partition, filePathWithPartition, recordsGenerationParams.getDataMetaClient(), columnsToIndex, false);
        }).iterator();
    });
    allRecordsRDD = allRecordsRDD.union(appendedFilesRecordsRDD);
    return allRecordsRDD;
}

Example 12 with HoodieData

use of org.apache.hudi.common.data.HoodieData in project hudi by apache.

the class HoodieTableMetadataUtil method convertMetadataToBloomFilterRecords.

/**
 * Convert commit action metadata to bloom filter records.
 *
 * @param context                 - Engine context to use
 * @param commitMetadata          - Commit action metadata
 * @param instantTime             - Action instant time
 * @param recordsGenerationParams - Parameters for bloom filter record generation
 * @return HoodieData of metadata table records
 */
public static HoodieData<HoodieRecord> convertMetadataToBloomFilterRecords(HoodieEngineContext context, HoodieCommitMetadata commitMetadata, String instantTime, MetadataRecordsGenerationParams recordsGenerationParams) {
    final List<HoodieWriteStat> allWriteStats = commitMetadata.getPartitionToWriteStats().values().stream().flatMap(entry -> entry.stream()).collect(Collectors.toList());
    if (allWriteStats.isEmpty()) {
        return context.emptyHoodieData();
    }
    final int parallelism = Math.max(Math.min(allWriteStats.size(), recordsGenerationParams.getBloomIndexParallelism()), 1);
    HoodieData<HoodieWriteStat> allWriteStatsRDD = context.parallelize(allWriteStats, parallelism);
    return allWriteStatsRDD.flatMap(hoodieWriteStat -> {
        final String partition = hoodieWriteStat.getPartitionPath();
        // For bloom filter index, delta writes do not change the base file bloom filter entries
        if (hoodieWriteStat instanceof HoodieDeltaWriteStat) {
            return Collections.emptyListIterator();
        }
        String pathWithPartition = hoodieWriteStat.getPath();
        if (pathWithPartition == null) {
            // Empty partition
            LOG.error("Failed to find path in write stat to update metadata table " + hoodieWriteStat);
            return Collections.emptyListIterator();
        }
        int offset = partition.equals(NON_PARTITIONED_NAME) ? (pathWithPartition.startsWith("/") ? 1 : 0) : partition.length() + 1;
        final String fileName = pathWithPartition.substring(offset);
        if (!FSUtils.isBaseFile(new Path(fileName))) {
            return Collections.emptyListIterator();
        }
        final Path writeFilePath = new Path(recordsGenerationParams.getDataMetaClient().getBasePath(), pathWithPartition);
        try (HoodieFileReader<IndexedRecord> fileReader = HoodieFileReaderFactory.getFileReader(recordsGenerationParams.getDataMetaClient().getHadoopConf(), writeFilePath)) {
            try {
                final BloomFilter fileBloomFilter = fileReader.readBloomFilter();
                if (fileBloomFilter == null) {
                    LOG.error("Failed to read bloom filter for " + writeFilePath);
                    return Collections.emptyListIterator();
                }
                ByteBuffer bloomByteBuffer = ByteBuffer.wrap(fileBloomFilter.serializeToString().getBytes());
                HoodieRecord record = HoodieMetadataPayload.createBloomFilterMetadataRecord(partition, fileName, instantTime, recordsGenerationParams.getBloomFilterType(), bloomByteBuffer, false);
                return Collections.singletonList(record).iterator();
            } catch (Exception e) {
                LOG.error("Failed to read bloom filter for " + writeFilePath);
                return Collections.emptyListIterator();
            } finally {
                fileReader.close();
            }
        } catch (IOException e) {
            LOG.error("Failed to get bloom filter for file: " + writeFilePath + ", write stat: " + hoodieWriteStat);
        }
        return Collections.emptyListIterator();
    });
}

Example 13 with HoodieData

use of org.apache.hudi.common.data.HoodieData in project hudi by apache.

the class HoodieDeleteHelper method execute.

@Override
public HoodieWriteMetadata<HoodieData<WriteStatus>> execute(String instantTime, HoodieData<HoodieKey> keys, HoodieEngineContext context, HoodieWriteConfig config, HoodieTable<T, HoodieData<HoodieRecord<T>>, HoodieData<HoodieKey>, HoodieData<WriteStatus>> table, BaseCommitActionExecutor<T, HoodieData<HoodieRecord<T>>, HoodieData<HoodieKey>, HoodieData<WriteStatus>, R> deleteExecutor) {
    try {
        HoodieData<HoodieKey> dedupedKeys = keys;
        final int parallelism = config.getDeleteShuffleParallelism();
        if (config.shouldCombineBeforeDelete()) {
            // De-dupe/merge if needed
            dedupedKeys = deduplicateKeys(keys, table, parallelism);
        } else if (!keys.isEmpty()) {
            dedupedKeys = keys.repartition(parallelism);
        }
        HoodieData<HoodieRecord<T>> dedupedRecords = dedupedKeys.map(key -> new HoodieAvroRecord(key, new EmptyHoodieRecordPayload()));
        Instant beginTag = Instant.now();
        // perform index loop up to get existing location of records
        HoodieData<HoodieRecord<T>> taggedRecords = table.getIndex().tagLocation(dedupedRecords, context, table);
        Duration tagLocationDuration = Duration.between(beginTag, Instant.now());
        // filter out non existent keys/records
        HoodieData<HoodieRecord<T>> taggedValidRecords = taggedRecords.filter(HoodieRecord::isCurrentLocationKnown);
        HoodieWriteMetadata<HoodieData<WriteStatus>> result;
        if (!taggedValidRecords.isEmpty()) {
            result = deleteExecutor.execute(taggedValidRecords);
            result.setIndexLookupDuration(tagLocationDuration);
        } else {
            // if entire set of keys are non existent
            deleteExecutor.saveWorkloadProfileMetadataToInflight(new WorkloadProfile(Pair.of(new HashMap<>(), new WorkloadStat())), instantTime);
            result = new HoodieWriteMetadata<>();
            result.setWriteStatuses(context.emptyHoodieData());
            deleteExecutor.commitOnAutoCommit(result);
        }
        return result;
    } catch (Throwable e) {
        if (e instanceof HoodieUpsertException) {
            throw (HoodieUpsertException) e;
        }
        throw new HoodieUpsertException("Failed to delete for commit time " + instantTime, e);
    }
}

Example 14 with HoodieData

use of org.apache.hudi.common.data.HoodieData in project hudi by apache.

the class HoodieMergeHelper method runMerge.

@Override
public void runMerge(HoodieTable<T, HoodieData<HoodieRecord<T>>, HoodieData<HoodieKey>, HoodieData<WriteStatus>> table, HoodieMergeHandle<T, HoodieData<HoodieRecord<T>>, HoodieData<HoodieKey>, HoodieData<WriteStatus>> mergeHandle) throws IOException {
    final boolean externalSchemaTransformation = table.getConfig().shouldUseExternalSchemaTransformation();
    Configuration cfgForHoodieFile = new Configuration(table.getHadoopConf());
    HoodieBaseFile baseFile = mergeHandle.baseFileForMerge();
    final GenericDatumWriter<GenericRecord> gWriter;
    final GenericDatumReader<GenericRecord> gReader;
    Schema readSchema;
    if (externalSchemaTransformation || baseFile.getBootstrapBaseFile().isPresent()) {
        readSchema = HoodieFileReaderFactory.getFileReader(table.getHadoopConf(), mergeHandle.getOldFilePath()).getSchema();
        gWriter = new GenericDatumWriter<>(readSchema);
        gReader = new GenericDatumReader<>(readSchema, mergeHandle.getWriterSchemaWithMetaFields());
    } else {
        gReader = null;
        gWriter = null;
        readSchema = mergeHandle.getWriterSchemaWithMetaFields();
    }
    BoundedInMemoryExecutor<GenericRecord, GenericRecord, Void> wrapper = null;
    HoodieFileReader<GenericRecord> reader = HoodieFileReaderFactory.getFileReader(cfgForHoodieFile, mergeHandle.getOldFilePath());
    try {
        final Iterator<GenericRecord> readerIterator;
        if (baseFile.getBootstrapBaseFile().isPresent()) {
            readerIterator = getMergingIterator(table, mergeHandle, baseFile, reader, readSchema, externalSchemaTransformation);
        } else {
            readerIterator = reader.getRecordIterator(readSchema);
        }
        ThreadLocal<BinaryEncoder> encoderCache = new ThreadLocal<>();
        ThreadLocal<BinaryDecoder> decoderCache = new ThreadLocal<>();
        wrapper = new BoundedInMemoryExecutor(table.getConfig().getWriteBufferLimitBytes(), readerIterator, new UpdateHandler(mergeHandle), record -> {
            if (!externalSchemaTransformation) {
                return record;
            }
            return transformRecordBasedOnNewSchema(gReader, gWriter, encoderCache, decoderCache, (GenericRecord) record);
        }, table.getPreExecuteRunnable());
        wrapper.execute();
    } catch (Exception e) {
        throw new HoodieException(e);
    } finally {
        if (reader != null) {
            reader.close();
        }
        mergeHandle.close();
        if (null != wrapper) {
            wrapper.shutdownNow();
        }
    }
}

Example 15 with HoodieData

use of org.apache.hudi.common.data.HoodieData in project hudi by apache.

the class RunCompactionActionExecutor method execute.

@Override
public HoodieWriteMetadata<HoodieData<WriteStatus>> execute() {
    HoodieTimeline pendingCompactionTimeline = table.getActiveTimeline().filterPendingCompactionTimeline();
    compactor.preCompact(table, pendingCompactionTimeline, instantTime);
    HoodieWriteMetadata<HoodieData<WriteStatus>> compactionMetadata = new HoodieWriteMetadata<>();
    try {
        // generate compaction plan
        // should support configurable commit metadata
        HoodieCompactionPlan compactionPlan = CompactionUtils.getCompactionPlan(table.getMetaClient(), instantTime);
        HoodieData<WriteStatus> statuses = compactor.compact(context, compactionPlan, table, config, instantTime, compactionHandler);
        compactor.maybePersist(statuses, config);
        context.setJobStatus(this.getClass().getSimpleName(), "Preparing compaction metadata");
        List<HoodieWriteStat> updateStatusMap = statuses.map(WriteStatus::getStat).collectAsList();
        HoodieCommitMetadata metadata = new HoodieCommitMetadata(true);
        for (HoodieWriteStat stat : updateStatusMap) {
            metadata.addWriteStat(stat.getPartitionPath(), stat);
        }
        metadata.addMetadata(HoodieCommitMetadata.SCHEMA_KEY, config.getSchema());
        compactionMetadata.setWriteStatuses(statuses);
        compactionMetadata.setCommitted(false);
        compactionMetadata.setCommitMetadata(Option.of(metadata));
    } catch (IOException e) {
        throw new HoodieCompactionException("Could not compact " + config.getBasePath(), e);
    }
    return compactionMetadata;
}