Examples with OrcReaderOptions com.facebook.presto.orc.OrcReaderOptions used on opensource projects

Example 1 with OrcReaderOptions

use of com.facebook.presto.orc.OrcReaderOptions in project presto by prestodb.

the class OrcStorageManager method getRowsFromUuid.

Optional<BitSet> getRowsFromUuid(FileSystem fileSystem, Optional<UUID> deltaShardUuid) {
    if (!deltaShardUuid.isPresent()) {
        return Optional.empty();
    }
    try (OrcDataSource dataSource = openShard(fileSystem, deltaShardUuid.get(), defaultReaderAttributes)) {
        OrcAggregatedMemoryContext systemMemoryUsage = new RaptorOrcAggregatedMemoryContext();
        OrcReader reader = new OrcReader(dataSource, ORC, orcFileTailSource, new StorageStripeMetadataSource(), new RaptorOrcAggregatedMemoryContext(), new OrcReaderOptions(defaultReaderAttributes.getMaxMergeDistance(), defaultReaderAttributes.getTinyStripeThreshold(), HUGE_MAX_READ_BLOCK_SIZE, defaultReaderAttributes.isZstdJniDecompressionEnabled()), false, NO_ENCRYPTION, DwrfKeyProvider.EMPTY, new RuntimeStats());
        if (reader.getFooter().getNumberOfRows() >= Integer.MAX_VALUE) {
            throw new IOException("File has too many rows");
        }
        try (OrcBatchRecordReader recordReader = reader.createBatchRecordReader(ImmutableMap.of(0, BIGINT), OrcPredicate.TRUE, DEFAULT_STORAGE_TIMEZONE, systemMemoryUsage, INITIAL_BATCH_SIZE)) {
            BitSet bitSet = new BitSet();
            while (recordReader.nextBatch() > 0) {
                Block block = recordReader.readBlock(0);
                for (int i = 0; i < block.getPositionCount(); i++) {
                    bitSet.set(toIntExact(block.getLong(i)));
                }
            }
            return Optional.of(bitSet);
        }
    } catch (IOException | RuntimeException e) {
        throw new PrestoException(RAPTOR_ERROR, "Failed to read file: " + deltaShardUuid, e);
    }
}

Example 2 with OrcReaderOptions

use of com.facebook.presto.orc.OrcReaderOptions in project presto by prestodb.

the class OrcStorageManager method computeShardStats.

private List<ColumnStats> computeShardStats(FileSystem fileSystem, Path file) {
    try (OrcDataSource dataSource = orcDataEnvironment.createOrcDataSource(fileSystem, file, defaultReaderAttributes)) {
        OrcReader reader = new OrcReader(dataSource, ORC, orcFileTailSource, stripeMetadataSourceFactory, new RaptorOrcAggregatedMemoryContext(), new OrcReaderOptions(defaultReaderAttributes.getMaxMergeDistance(), defaultReaderAttributes.getTinyStripeThreshold(), HUGE_MAX_READ_BLOCK_SIZE, defaultReaderAttributes.isZstdJniDecompressionEnabled()), false, NO_ENCRYPTION, DwrfKeyProvider.EMPTY, new RuntimeStats());
        ImmutableList.Builder<ColumnStats> list = ImmutableList.builder();
        for (ColumnInfo info : getColumnInfo(reader)) {
            computeColumnStats(reader, info.getColumnId(), info.getType(), typeManager).ifPresent(list::add);
        }
        return list.build();
    } catch (IOException e) {
        throw new PrestoException(RAPTOR_ERROR, "Failed to read file: " + file, e);
    }
}

Example 3 with OrcReaderOptions

use of com.facebook.presto.orc.OrcReaderOptions in project presto by prestodb.

the class OrcStorageManager method getPageSource.

@Override
public ConnectorPageSource getPageSource(HdfsContext hdfsContext, HiveFileContext hiveFileContext, UUID shardUuid, Optional<UUID> deltaShardUuid, boolean tableSupportsDeltaDelete, OptionalInt bucketNumber, List<Long> columnIds, List<Type> columnTypes, TupleDomain<RaptorColumnHandle> effectivePredicate, ReaderAttributes readerAttributes, OptionalLong transactionId, Optional<Map<String, Type>> allColumnTypes) {
    FileSystem fileSystem = orcDataEnvironment.getFileSystem(hdfsContext);
    OrcDataSource dataSource = openShard(fileSystem, shardUuid, readerAttributes);
    OrcAggregatedMemoryContext systemMemoryUsage = new RaptorOrcAggregatedMemoryContext();
    try {
        OrcReader reader = new OrcReader(dataSource, ORC, orcFileTailSource, stripeMetadataSourceFactory, new RaptorOrcAggregatedMemoryContext(), new OrcReaderOptions(readerAttributes.getMaxMergeDistance(), readerAttributes.getTinyStripeThreshold(), HUGE_MAX_READ_BLOCK_SIZE, readerAttributes.isZstdJniDecompressionEnabled()), hiveFileContext.isCacheable(), NO_ENCRYPTION, DwrfKeyProvider.EMPTY, new RuntimeStats());
        Map<Long, Integer> indexMap = columnIdIndex(reader.getColumnNames());
        ImmutableMap.Builder<Integer, Type> includedColumns = ImmutableMap.builder();
        ImmutableList.Builder<Integer> columnIndexes = ImmutableList.builder();
        for (int i = 0; i < columnIds.size(); i++) {
            long columnId = columnIds.get(i);
            if (isHiddenColumn(columnId)) {
                columnIndexes.add(toSpecialIndex(columnId));
                continue;
            }
            Integer index = indexMap.get(columnId);
            if (index == null) {
                columnIndexes.add(NULL_COLUMN);
            } else {
                columnIndexes.add(index);
                includedColumns.put(index, toOrcFileType(columnTypes.get(i), typeManager));
            }
        }
        OrcPredicate predicate = getPredicate(effectivePredicate, indexMap);
        StorageTypeConverter storageTypeConverter = new StorageTypeConverter(typeManager);
        OrcBatchRecordReader recordReader = reader.createBatchRecordReader(storageTypeConverter.toStorageTypes(includedColumns.build()), predicate, DEFAULT_STORAGE_TIMEZONE, systemMemoryUsage, INITIAL_BATCH_SIZE);
        Optional<ShardRewriter> shardRewriter = Optional.empty();
        if (transactionId.isPresent()) {
            checkState(allColumnTypes.isPresent());
            if (reader.getFooter().getNumberOfRows() >= Integer.MAX_VALUE) {
                throw new PrestoException(RAPTOR_ERROR, "File has too many rows, failed to read file: " + shardUuid);
            }
            shardRewriter = Optional.of(createShardRewriter(hdfsContext, fileSystem, transactionId.getAsLong(), bucketNumber, shardUuid, toIntExact(reader.getFooter().getNumberOfRows()), deltaShardUuid, tableSupportsDeltaDelete, allColumnTypes.get()));
        }
        return new OrcUpdatablePageSource(shardRewriter, recordReader, new OrcPageSource(recordReader, dataSource, columnIds, columnTypes, columnIndexes.build(), shardUuid, bucketNumber, systemMemoryUsage, new DeltaShardLoader(deltaShardUuid, tableSupportsDeltaDelete, this, fileSystem)));
    } catch (IOException | RuntimeException e) {
        closeQuietly(dataSource);
        throw new PrestoException(RAPTOR_ERROR, "Failed to create page source for shard " + shardUuid, e);
    } catch (Throwable t) {
        closeQuietly(dataSource);
        throw t;
    }
}

Example 4 with OrcReaderOptions

use of com.facebook.presto.orc.OrcReaderOptions in project presto by prestodb.

the class OrcBatchPageSourceFactory method createOrcPageSource.

public static ConnectorPageSource createOrcPageSource(OrcEncoding orcEncoding, HdfsEnvironment hdfsEnvironment, String sessionUser, Configuration configuration, Path path, long start, long length, long fileSize, List<HiveColumnHandle> columns, boolean useOrcColumnNames, TupleDomain<HiveColumnHandle> effectivePredicate, DateTimeZone hiveStorageTimeZone, TypeManager typeManager, StandardFunctionResolution functionResolution, DataSize maxBufferSize, DataSize streamBufferSize, boolean lazyReadSmallRanges, boolean orcBloomFiltersEnabled, FileFormatDataSourceStats stats, int domainCompactionThreshold, OrcFileTailSource orcFileTailSource, StripeMetadataSourceFactory stripeMetadataSourceFactory, HiveFileContext hiveFileContext, OrcReaderOptions orcReaderOptions, Optional<EncryptionInformation> encryptionInformation, DwrfEncryptionProvider dwrfEncryptionProvider) {
    checkArgument(domainCompactionThreshold >= 1, "domainCompactionThreshold must be at least 1");
    OrcDataSource orcDataSource;
    try {
        FSDataInputStream inputStream = hdfsEnvironment.getFileSystem(sessionUser, path, configuration).openFile(path, hiveFileContext);
        orcDataSource = new HdfsOrcDataSource(new OrcDataSourceId(path.toString()), fileSize, orcReaderOptions.getMaxMergeDistance(), maxBufferSize, streamBufferSize, lazyReadSmallRanges, inputStream, stats);
    } catch (Exception e) {
        if (nullToEmpty(e.getMessage()).trim().equals("Filesystem closed") || e instanceof FileNotFoundException) {
            throw new PrestoException(HIVE_CANNOT_OPEN_SPLIT, e);
        }
        throw new PrestoException(HIVE_CANNOT_OPEN_SPLIT, splitError(e, path, start, length), e);
    }
    OrcAggregatedMemoryContext systemMemoryUsage = new HiveOrcAggregatedMemoryContext();
    try {
        DwrfKeyProvider dwrfKeyProvider = new ProjectionBasedDwrfKeyProvider(encryptionInformation, columns, useOrcColumnNames, path);
        OrcReader reader = new OrcReader(orcDataSource, orcEncoding, orcFileTailSource, stripeMetadataSourceFactory, new HiveOrcAggregatedMemoryContext(), orcReaderOptions, hiveFileContext.isCacheable(), dwrfEncryptionProvider, dwrfKeyProvider, hiveFileContext.getStats());
        List<HiveColumnHandle> physicalColumns = getPhysicalHiveColumnHandles(columns, useOrcColumnNames, reader.getTypes(), path);
        ImmutableMap.Builder<Integer, Type> includedColumns = ImmutableMap.builder();
        ImmutableList.Builder<ColumnReference<HiveColumnHandle>> columnReferences = ImmutableList.builder();
        for (HiveColumnHandle column : physicalColumns) {
            if (column.getColumnType() == REGULAR) {
                Type type = typeManager.getType(column.getTypeSignature());
                includedColumns.put(column.getHiveColumnIndex(), type);
                columnReferences.add(new ColumnReference<>(column, column.getHiveColumnIndex(), type));
            }
        }
        if (!physicalColumns.isEmpty() && physicalColumns.stream().allMatch(hiveColumnHandle -> hiveColumnHandle.getColumnType() == AGGREGATED)) {
            return new AggregatedOrcPageSource(physicalColumns, reader.getFooter(), typeManager, functionResolution);
        }
        OrcPredicate predicate = new TupleDomainOrcPredicate<>(effectivePredicate, columnReferences.build(), orcBloomFiltersEnabled, Optional.of(domainCompactionThreshold));
        OrcBatchRecordReader recordReader = reader.createBatchRecordReader(includedColumns.build(), predicate, start, length, hiveStorageTimeZone, systemMemoryUsage, INITIAL_BATCH_SIZE);
        return new OrcBatchPageSource(recordReader, reader.getOrcDataSource(), physicalColumns, typeManager, systemMemoryUsage, stats, hiveFileContext.getStats());
    } catch (Exception e) {
        try {
            orcDataSource.close();
        } catch (IOException ignored) {
        }
        if (e instanceof PrestoException) {
            throw (PrestoException) e;
        }
        String message = splitError(e, path, start, length);
        if (e.getClass().getSimpleName().equals("BlockMissingException")) {
            throw new PrestoException(HIVE_MISSING_DATA, message, e);
        }
        throw new PrestoException(HIVE_CANNOT_OPEN_SPLIT, message, e);
    }
}

Example 5 with OrcReaderOptions

use of com.facebook.presto.orc.OrcReaderOptions in project presto by prestodb.

the class IcebergPageSourceProvider method createBatchOrcPageSource.

private static ConnectorPageSource createBatchOrcPageSource(HdfsEnvironment hdfsEnvironment, String user, Configuration configuration, Path path, long start, long length, boolean isCacheable, List<IcebergColumnHandle> regularColumns, TypeManager typeManager, TupleDomain<IcebergColumnHandle> effectivePredicate, OrcReaderOptions options, OrcEncoding orcEncoding, DataSize maxBufferSize, DataSize streamBufferSize, boolean lazyReadSmallRanges, boolean orcBloomFiltersEnabled, int domainCompactionThreshold, OrcFileTailSource orcFileTailSource, StripeMetadataSourceFactory stripeMetadataSourceFactory, FileFormatDataSourceStats stats, Optional<EncryptionInformation> encryptionInformation, DwrfEncryptionProvider dwrfEncryptionProvider) {
    OrcDataSource orcDataSource = null;
    try {
        ExtendedFileSystem fileSystem = hdfsEnvironment.getFileSystem(user, path, configuration);
        FileStatus fileStatus = fileSystem.getFileStatus(path);
        long fileSize = fileStatus.getLen();
        long modificationTime = fileStatus.getModificationTime();
        HiveFileContext hiveFileContext = new HiveFileContext(true, NO_CACHE_CONSTRAINTS, Optional.empty(), Optional.of(fileSize), modificationTime, false);
        FSDataInputStream inputStream = hdfsEnvironment.doAs(user, () -> fileSystem.openFile(path, hiveFileContext));
        orcDataSource = new HdfsOrcDataSource(new OrcDataSourceId(path.toString()), fileSize, options.getMaxMergeDistance(), maxBufferSize, streamBufferSize, lazyReadSmallRanges, inputStream, stats);
        // Todo: pass real columns to ProjectionBasedDwrfKeyProvider instead of ImmutableList.of()
        DwrfKeyProvider dwrfKeyProvider = new ProjectionBasedDwrfKeyProvider(encryptionInformation, ImmutableList.of(), true, path);
        RuntimeStats runtimeStats = new RuntimeStats();
        OrcReader reader = new OrcReader(orcDataSource, orcEncoding, orcFileTailSource, stripeMetadataSourceFactory, new HiveOrcAggregatedMemoryContext(), options, isCacheable, dwrfEncryptionProvider, dwrfKeyProvider, runtimeStats);
        List<HiveColumnHandle> physicalColumnHandles = new ArrayList<>(regularColumns.size());
        ImmutableMap.Builder<Integer, Type> includedColumns = ImmutableMap.builder();
        ImmutableList.Builder<TupleDomainOrcPredicate.ColumnReference<HiveColumnHandle>> columnReferences = ImmutableList.builder();
        List<IcebergOrcColumn> fileOrcColumns = getFileOrcColumns(reader);
        Map<Integer, IcebergOrcColumn> fileOrcColumnByIcebergId = fileOrcColumns.stream().filter(orcColumn -> orcColumn.getAttributes().containsKey(ORC_ICEBERG_ID_KEY)).collect(toImmutableMap(orcColumn -> Integer.parseInt(orcColumn.getAttributes().get(ORC_ICEBERG_ID_KEY)), orcColumn -> IcebergOrcColumn.copy(orcColumn).setIcebergColumnId(Optional.of(Integer.parseInt(orcColumn.getAttributes().get(ORC_ICEBERG_ID_KEY))))));
        Map<String, IcebergOrcColumn> fileOrcColumnsByName = uniqueIndex(fileOrcColumns, orcColumn -> orcColumn.getColumnName().toLowerCase(ENGLISH));
        int nextMissingColumnIndex = fileOrcColumnsByName.size();
        for (IcebergColumnHandle column : regularColumns) {
            IcebergOrcColumn icebergOrcColumn;
            boolean isExcludeColumn = false;
            if (fileOrcColumnByIcebergId.isEmpty()) {
                icebergOrcColumn = fileOrcColumnsByName.get(column.getName());
            } else {
                icebergOrcColumn = fileOrcColumnByIcebergId.get(column.getId());
                if (icebergOrcColumn == null) {
                    // Cannot get orc column from 'fileOrcColumnByIcebergId', which means SchemaEvolution may have happened, so we get orc column by column name.
                    icebergOrcColumn = fileOrcColumnsByName.get(column.getName());
                    if (icebergOrcColumn != null) {
                        isExcludeColumn = true;
                    }
                }
            }
            if (icebergOrcColumn != null) {
                HiveColumnHandle columnHandle = new HiveColumnHandle(// Todo: using orc file column name
                column.getName(), toHiveType(column.getType()), column.getType().getTypeSignature(), icebergOrcColumn.getOrcColumnId(), icebergOrcColumn.getColumnType(), Optional.empty(), Optional.empty());
                physicalColumnHandles.add(columnHandle);
                // Skip SchemaEvolution column
                if (!isExcludeColumn) {
                    includedColumns.put(columnHandle.getHiveColumnIndex(), typeManager.getType(columnHandle.getTypeSignature()));
                    columnReferences.add(new TupleDomainOrcPredicate.ColumnReference<>(columnHandle, columnHandle.getHiveColumnIndex(), typeManager.getType(columnHandle.getTypeSignature())));
                }
            } else {
                physicalColumnHandles.add(new HiveColumnHandle(column.getName(), toHiveType(column.getType()), column.getType().getTypeSignature(), nextMissingColumnIndex++, REGULAR, Optional.empty(), Optional.empty()));
            }
        }
        TupleDomain<HiveColumnHandle> hiveColumnHandleTupleDomain = effectivePredicate.transform(column -> {
            IcebergOrcColumn icebergOrcColumn;
            if (fileOrcColumnByIcebergId.isEmpty()) {
                icebergOrcColumn = fileOrcColumnsByName.get(column.getName());
            } else {
                icebergOrcColumn = fileOrcColumnByIcebergId.get(column.getId());
                if (icebergOrcColumn == null) {
                    // Cannot get orc column from 'fileOrcColumnByIcebergId', which means SchemaEvolution may have happened, so we get orc column by column name.
                    icebergOrcColumn = fileOrcColumnsByName.get(column.getName());
                }
            }
            return new HiveColumnHandle(column.getName(), toHiveType(column.getType()), column.getType().getTypeSignature(), // Note: the HiveColumnHandle.hiveColumnIndex starts from '0' while the IcebergColumnHandle.id starts from '1'
            icebergOrcColumn != null ? icebergOrcColumn.getOrcColumnId() : column.getId() - 1, icebergOrcColumn != null ? icebergOrcColumn.getColumnType() : REGULAR, Optional.empty(), Optional.empty());
        });
        OrcPredicate predicate = new TupleDomainOrcPredicate<>(hiveColumnHandleTupleDomain, columnReferences.build(), orcBloomFiltersEnabled, Optional.of(domainCompactionThreshold));
        OrcAggregatedMemoryContext systemMemoryUsage = new HiveOrcAggregatedMemoryContext();
        OrcBatchRecordReader recordReader = reader.createBatchRecordReader(includedColumns.build(), predicate, start, length, UTC, systemMemoryUsage, INITIAL_BATCH_SIZE);
        return new OrcBatchPageSource(recordReader, orcDataSource, physicalColumnHandles, typeManager, systemMemoryUsage, stats, runtimeStats);
    } catch (Exception e) {
        if (orcDataSource != null) {
            try {
                orcDataSource.close();
            } catch (IOException ignored) {
            }
        }
        if (e instanceof PrestoException) {
            throw (PrestoException) e;
        }
        String message = format("Error opening Iceberg split %s (offset=%s, length=%s): %s", path, start, length, e.getMessage());
        if (e instanceof BlockMissingException) {
            throw new PrestoException(ICEBERG_MISSING_DATA, message, e);
        }
        throw new PrestoException(ICEBERG_CANNOT_OPEN_SPLIT, message, e);
    }
}