use of com.facebook.presto.hive.HiveErrorCode.HIVE_WRITER_OPEN_ERROR in project presto by prestodb.
the class RcFileFileWriterFactory method createFileWriter.
@Override
public Optional<HiveFileWriter> createFileWriter(Path path, List<String> inputColumnNames, StorageFormat storageFormat, Properties schema, JobConf configuration, ConnectorSession session, Optional<EncryptionInformation> encryptionInformation) {
if (!HiveSessionProperties.isRcfileOptimizedWriterEnabled(session)) {
return Optional.empty();
}
if (!RCFileOutputFormat.class.getName().equals(storageFormat.getOutputFormat())) {
return Optional.empty();
}
RcFileEncoding rcFileEncoding;
if (LazyBinaryColumnarSerDe.class.getName().equals(storageFormat.getSerDe())) {
rcFileEncoding = new BinaryRcFileEncoding();
} else if (ColumnarSerDe.class.getName().equals(storageFormat.getSerDe())) {
rcFileEncoding = createTextVectorEncoding(schema, hiveStorageTimeZone);
} else {
return Optional.empty();
}
Optional<String> codecName = Optional.ofNullable(configuration.get(FileOutputFormat.COMPRESS_CODEC));
// existing tables and partitions may have columns in a different order than the writer is providing, so build
// an index to rearrange columns in the proper order
List<String> fileColumnNames = Splitter.on(',').trimResults().omitEmptyStrings().splitToList(schema.getProperty(META_TABLE_COLUMNS, ""));
List<Type> fileColumnTypes = toHiveTypes(schema.getProperty(META_TABLE_COLUMN_TYPES, "")).stream().map(hiveType -> hiveType.getType(typeManager)).collect(toList());
int[] fileInputColumnIndexes = fileColumnNames.stream().mapToInt(inputColumnNames::indexOf).toArray();
try {
FileSystem fileSystem = hdfsEnvironment.getFileSystem(session.getUser(), path, configuration);
OutputStream outputStream = fileSystem.create(path);
Optional<Supplier<RcFileDataSource>> validationInputFactory = Optional.empty();
if (HiveSessionProperties.isRcfileOptimizedWriterValidate(session)) {
validationInputFactory = Optional.of(() -> {
try {
return new HdfsRcFileDataSource(path.toString(), fileSystem.open(path), fileSystem.getFileStatus(path).getLen(), stats);
} catch (IOException e) {
throw new PrestoException(HIVE_WRITE_VALIDATION_FAILED, e);
}
});
}
Callable<Void> rollbackAction = () -> {
fileSystem.delete(path, false);
return null;
};
return Optional.of(new RcFileFileWriter(outputStream, rollbackAction, rcFileEncoding, fileColumnTypes, codecName, fileInputColumnIndexes, ImmutableMap.<String, String>builder().put(HiveMetadata.PRESTO_VERSION_NAME, nodeVersion.toString()).put(MetastoreUtil.PRESTO_QUERY_ID_NAME, session.getQueryId()).build(), validationInputFactory));
} catch (Exception e) {
throw new PrestoException(HIVE_WRITER_OPEN_ERROR, "Error creating RCFile file", e);
}
}
use of com.facebook.presto.hive.HiveErrorCode.HIVE_WRITER_OPEN_ERROR in project presto by prestodb.
the class ParquetFileWriterFactory method createFileWriter.
@Override
public Optional<HiveFileWriter> createFileWriter(Path path, List<String> inputColumnNames, StorageFormat storageFormat, Properties schema, JobConf conf, ConnectorSession session, Optional<EncryptionInformation> encryptionInformation) {
if (!isParquetOptimizedWriterEnabled(session)) {
return Optional.empty();
}
if (!MapredParquetOutputFormat.class.getName().equals(storageFormat.getOutputFormat())) {
return Optional.empty();
}
ParquetWriterOptions parquetWriterOptions = ParquetWriterOptions.builder().setMaxPageSize(getParquetWriterPageSize(session)).setMaxBlockSize(getParquetWriterBlockSize(session)).build();
CompressionCodecName compressionCodecName = getCompression(conf);
List<String> fileColumnNames = Splitter.on(',').trimResults().omitEmptyStrings().splitToList(schema.getProperty(META_TABLE_COLUMNS, ""));
List<Type> fileColumnTypes = toHiveTypes(schema.getProperty(META_TABLE_COLUMN_TYPES, "")).stream().map(hiveType -> hiveType.getType(typeManager)).collect(toList());
int[] fileInputColumnIndexes = fileColumnNames.stream().mapToInt(inputColumnNames::indexOf).toArray();
try {
FileSystem fileSystem = hdfsEnvironment.getFileSystem(session.getUser(), path, conf);
Callable<Void> rollbackAction = () -> {
fileSystem.delete(path, false);
return null;
};
return Optional.of(new ParquetFileWriter(fileSystem.create(path), rollbackAction, fileColumnNames, fileColumnTypes, parquetWriterOptions, fileInputColumnIndexes, compressionCodecName));
} catch (IOException e) {
throw new PrestoException(HIVE_WRITER_OPEN_ERROR, "Error creating Parquet file", e);
}
}
use of com.facebook.presto.hive.HiveErrorCode.HIVE_WRITER_OPEN_ERROR in project presto by prestodb.
the class OrcFileWriterFactory method createFileWriter.
@Override
public Optional<HiveFileWriter> createFileWriter(Path path, List<String> inputColumnNames, StorageFormat storageFormat, Properties schema, JobConf configuration, ConnectorSession session, Optional<EncryptionInformation> encryptionInformation) {
if (!HiveSessionProperties.isOrcOptimizedWriterEnabled(session)) {
return Optional.empty();
}
OrcEncoding orcEncoding;
if (OrcOutputFormat.class.getName().equals(storageFormat.getOutputFormat())) {
orcEncoding = ORC;
} else if (com.facebook.hive.orc.OrcOutputFormat.class.getName().equals(storageFormat.getOutputFormat())) {
orcEncoding = DWRF;
} else {
return Optional.empty();
}
CompressionKind compression = getCompression(schema, configuration, orcEncoding);
// existing tables and partitions may have columns in a different order than the writer is providing, so build
// an index to rearrange columns in the proper order
List<String> fileColumnNames = Splitter.on(',').trimResults().omitEmptyStrings().splitToList(schema.getProperty(META_TABLE_COLUMNS, ""));
List<Type> fileColumnTypes = toHiveTypes(schema.getProperty(META_TABLE_COLUMN_TYPES, "")).stream().map(hiveType -> hiveType.getType(typeManager)).collect(toList());
int[] fileInputColumnIndexes = fileColumnNames.stream().mapToInt(inputColumnNames::indexOf).toArray();
try {
FileSystem fileSystem = hdfsEnvironment.getFileSystem(session.getUser(), path, configuration);
DataSink dataSink = createDataSink(session, fileSystem, path);
Optional<Supplier<OrcDataSource>> validationInputFactory = Optional.empty();
if (HiveSessionProperties.isOrcOptimizedWriterValidate(session)) {
validationInputFactory = Optional.of(() -> {
try {
return new HdfsOrcDataSource(new OrcDataSourceId(path.toString()), fileSystem.getFileStatus(path).getLen(), getOrcMaxMergeDistance(session), getOrcMaxBufferSize(session), getOrcStreamBufferSize(session), false, fileSystem.open(path), readStats);
} catch (IOException e) {
throw new PrestoException(HIVE_WRITE_VALIDATION_FAILED, e);
}
});
}
Callable<Void> rollbackAction = () -> {
fileSystem.delete(path, false);
return null;
};
Optional<DwrfWriterEncryption> dwrfWriterEncryption = createDwrfEncryption(encryptionInformation, fileColumnNames, fileColumnTypes);
return Optional.of(new OrcFileWriter(dataSink, rollbackAction, orcEncoding, fileColumnNames, fileColumnTypes, compression, orcFileWriterConfig.toOrcWriterOptionsBuilder().withFlushPolicy(DefaultOrcWriterFlushPolicy.builder().withStripeMinSize(getOrcOptimizedWriterMinStripeSize(session)).withStripeMaxSize(getOrcOptimizedWriterMaxStripeSize(session)).withStripeMaxRowCount(getOrcOptimizedWriterMaxStripeRows(session)).build()).withDictionaryMaxMemory(getOrcOptimizedWriterMaxDictionaryMemory(session)).withMaxStringStatisticsLimit(getOrcStringStatisticsLimit(session)).withIgnoreDictionaryRowGroupSizes(isExecutionBasedMemoryAccountingEnabled(session)).withDwrfStripeCacheEnabled(isDwrfWriterStripeCacheEnabled(session)).withDwrfStripeCacheMaxSize(getDwrfWriterStripeCacheeMaxSize(session)).build(), fileInputColumnIndexes, ImmutableMap.<String, String>builder().put(HiveMetadata.PRESTO_VERSION_NAME, nodeVersion.toString()).put(MetastoreUtil.PRESTO_QUERY_ID_NAME, session.getQueryId()).build(), hiveStorageTimeZone, validationInputFactory, getOrcOptimizedWriterValidateMode(session), stats, dwrfEncryptionProvider, dwrfWriterEncryption));
} catch (IOException e) {
throw new PrestoException(HIVE_WRITER_OPEN_ERROR, "Error creating " + orcEncoding + " file. " + e.getMessage(), e);
}
}
Aggregations