use of io.trino.spi.connector.ConnectorSession in project trino by trinodb.
the class HiveMetadata method getInsertLayout.
@Override
public Optional<ConnectorTableLayout> getInsertLayout(ConnectorSession session, ConnectorTableHandle tableHandle) {
HiveTableHandle hiveTableHandle = (HiveTableHandle) tableHandle;
SchemaTableName tableName = hiveTableHandle.getSchemaTableName();
Table table = metastore.getTable(tableName.getSchemaName(), tableName.getTableName()).orElseThrow(() -> new TableNotFoundException(tableName));
if (table.getStorage().getBucketProperty().isPresent()) {
if (!isSupportedBucketing(table)) {
throw new TrinoException(NOT_SUPPORTED, "Cannot write to a table bucketed on an unsupported type");
}
} else // Note: we cannot use hiveTableHandle.isInAcidTransaction() here as transaction is not yet set in HiveTableHandle when getInsertLayout is called
if (isFullAcidTable(table.getParameters())) {
table = Table.builder(table).withStorage(storage -> storage.setBucketProperty(Optional.of(new HiveBucketProperty(ImmutableList.of(), HiveBucketing.BucketingVersion.BUCKETING_V2, 1, ImmutableList.of())))).build();
}
Optional<HiveBucketHandle> hiveBucketHandle = getHiveBucketHandle(session, table, typeManager);
List<Column> partitionColumns = table.getPartitionColumns();
if (hiveBucketHandle.isEmpty()) {
// return preferred layout which is partitioned by partition columns
if (partitionColumns.isEmpty()) {
return Optional.empty();
}
return Optional.of(new ConnectorTableLayout(partitionColumns.stream().map(Column::getName).collect(toImmutableList())));
}
HiveBucketProperty bucketProperty = table.getStorage().getBucketProperty().orElseThrow(() -> new NoSuchElementException("Bucket property should be set"));
if (!bucketProperty.getSortedBy().isEmpty() && !isSortedWritingEnabled(session)) {
throw new TrinoException(NOT_SUPPORTED, "Writing to bucketed sorted Hive tables is disabled");
}
ImmutableList.Builder<String> partitioningColumns = ImmutableList.builder();
hiveBucketHandle.get().getColumns().stream().map(HiveColumnHandle::getName).forEach(partitioningColumns::add);
partitionColumns.stream().map(Column::getName).forEach(partitioningColumns::add);
HivePartitioningHandle partitioningHandle = new HivePartitioningHandle(hiveBucketHandle.get().getBucketingVersion(), hiveBucketHandle.get().getTableBucketCount(), hiveBucketHandle.get().getColumns().stream().map(HiveColumnHandle::getHiveType).collect(toImmutableList()), OptionalInt.of(hiveBucketHandle.get().getTableBucketCount()), !partitionColumns.isEmpty() && isParallelPartitionedBucketedWrites(session));
return Optional.of(new ConnectorTableLayout(partitioningHandle, partitioningColumns.build()));
}
use of io.trino.spi.connector.ConnectorSession in project trino by trinodb.
the class PropertiesSystemTableProvider method getSystemTable.
@Override
public Optional<SystemTable> getSystemTable(HiveMetadata metadata, ConnectorSession session, SchemaTableName tableName) {
if (!PROPERTIES.matches(tableName)) {
return Optional.empty();
}
SchemaTableName sourceTableName = PROPERTIES.getSourceTableName(tableName);
Table table = metadata.getMetastore().getTable(sourceTableName.getSchemaName(), sourceTableName.getTableName()).orElseThrow(() -> new TableNotFoundException(tableName));
if (isDeltaLakeTable(table) || isIcebergTable(table)) {
return Optional.empty();
}
Map<String, String> sortedTableParameters = ImmutableSortedMap.copyOf(table.getParameters());
List<ColumnMetadata> columns = sortedTableParameters.keySet().stream().map(key -> new ColumnMetadata(key, VarcharType.VARCHAR)).collect(toImmutableList());
List<Type> types = columns.stream().map(ColumnMetadata::getType).collect(toImmutableList());
Iterable<List<Object>> propertyValues = ImmutableList.of(ImmutableList.copyOf(sortedTableParameters.values()));
return Optional.of(createSystemTable(new ConnectorTableMetadata(sourceTableName, columns), constraint -> new InMemoryRecordSet(types, propertyValues).cursor()));
}
use of io.trino.spi.connector.ConnectorSession in project trino by trinodb.
the class RcFileFileWriterFactory method createFileWriter.
@Override
public Optional<FileWriter> createFileWriter(Path path, List<String> inputColumnNames, StorageFormat storageFormat, Properties schema, JobConf configuration, ConnectorSession session, OptionalInt bucketNumber, AcidTransaction transaction, boolean useAcidSchema, WriterKind writerKind) {
if (!RCFileOutputFormat.class.getName().equals(storageFormat.getOutputFormat())) {
return Optional.empty();
}
RcFileEncoding rcFileEncoding;
if (LazyBinaryColumnarSerDe.class.getName().equals(storageFormat.getSerde())) {
rcFileEncoding = new BinaryRcFileEncoding(timeZone);
} else if (ColumnarSerDe.class.getName().equals(storageFormat.getSerde())) {
rcFileEncoding = createTextVectorEncoding(schema);
} 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 = getColumnNames(schema);
List<Type> fileColumnTypes = getColumnTypes(schema).stream().map(hiveType -> hiveType.getType(typeManager, getTimestampPrecision(session))).collect(toList());
int[] fileInputColumnIndexes = fileColumnNames.stream().mapToInt(inputColumnNames::indexOf).toArray();
try {
FileSystem fileSystem = hdfsEnvironment.getFileSystem(session.getIdentity(), path, configuration);
OutputStream outputStream = fileSystem.create(path, false);
Optional<Supplier<RcFileDataSource>> validationInputFactory = Optional.empty();
if (isRcfileOptimizedWriterValidate(session)) {
validationInputFactory = Optional.of(() -> {
try {
return new HdfsRcFileDataSource(path.toString(), fileSystem.open(path), fileSystem.getFileStatus(path).getLen(), stats);
} catch (IOException e) {
throw new TrinoException(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(PRESTO_VERSION_NAME, nodeVersion.toString()).put(PRESTO_QUERY_ID_NAME, session.getQueryId()).buildOrThrow(), validationInputFactory));
} catch (Exception e) {
throw new TrinoException(HIVE_WRITER_OPEN_ERROR, "Error creating RCFile file", e);
}
}
use of io.trino.spi.connector.ConnectorSession in project trino by trinodb.
the class SemiTransactionalHiveMetastore method dropDatabase.
public synchronized void dropDatabase(ConnectorSession session, String schemaName) {
Optional<Path> location = delegate.getDatabase(schemaName).orElseThrow(() -> new SchemaNotFoundException(schemaName)).getLocation().map(Path::new);
setExclusive((delegate, hdfsEnvironment) -> {
// If we see files in the schema location, don't delete it.
// If we see no files, request deletion.
// If we fail to check the schema location, behave according to fallback.
boolean deleteData = location.map(path -> {
HdfsContext context = new HdfsContext(session);
try (FileSystem fs = hdfsEnvironment.getFileSystem(context, path)) {
return !fs.listLocatedStatus(path).hasNext();
} catch (IOException | RuntimeException e) {
log.warn(e, "Could not check schema directory '%s'", path);
return deleteSchemaLocationsFallback;
}
}).orElse(deleteSchemaLocationsFallback);
delegate.dropDatabase(schemaName, deleteData);
});
}
use of io.trino.spi.connector.ConnectorSession in project trino by trinodb.
the class SyncPartitionMetadataProcedure method doSyncPartitionMetadata.
private void doSyncPartitionMetadata(ConnectorSession session, ConnectorAccessControl accessControl, String schemaName, String tableName, String mode, boolean caseSensitive) {
SyncMode syncMode = toSyncMode(mode);
HdfsContext hdfsContext = new HdfsContext(session);
SemiTransactionalHiveMetastore metastore = hiveMetadataFactory.create(session.getIdentity(), true).getMetastore();
SchemaTableName schemaTableName = new SchemaTableName(schemaName, tableName);
Table table = metastore.getTable(schemaName, tableName).orElseThrow(() -> new TableNotFoundException(schemaTableName));
if (table.getPartitionColumns().isEmpty()) {
throw new TrinoException(INVALID_PROCEDURE_ARGUMENT, "Table is not partitioned: " + schemaTableName);
}
if (syncMode == SyncMode.ADD || syncMode == SyncMode.FULL) {
accessControl.checkCanInsertIntoTable(null, new SchemaTableName(schemaName, tableName));
}
if (syncMode == SyncMode.DROP || syncMode == SyncMode.FULL) {
accessControl.checkCanDeleteFromTable(null, new SchemaTableName(schemaName, tableName));
}
Path tableLocation = new Path(table.getStorage().getLocation());
Set<String> partitionsToAdd;
Set<String> partitionsToDrop;
try {
FileSystem fileSystem = hdfsEnvironment.getFileSystem(hdfsContext, tableLocation);
List<String> partitionsInMetastore = metastore.getPartitionNames(schemaName, tableName).orElseThrow(() -> new TableNotFoundException(schemaTableName));
List<String> partitionsInFileSystem = listDirectory(fileSystem, fileSystem.getFileStatus(tableLocation), table.getPartitionColumns(), table.getPartitionColumns().size(), caseSensitive).stream().map(fileStatus -> fileStatus.getPath().toUri()).map(uri -> tableLocation.toUri().relativize(uri).getPath()).collect(toImmutableList());
// partitions in file system but not in metastore
partitionsToAdd = difference(partitionsInFileSystem, partitionsInMetastore);
// partitions in metastore but not in file system
partitionsToDrop = difference(partitionsInMetastore, partitionsInFileSystem);
} catch (IOException e) {
throw new TrinoException(HIVE_FILESYSTEM_ERROR, e);
}
syncPartitions(partitionsToAdd, partitionsToDrop, syncMode, metastore, session, table);
}
Aggregations