use of io.trino.plugin.hive.HiveTimestampPrecision.NANOSECONDS in project trino by trinodb.
the class HiveMetadata method getTableHandleForOptimize.
private Optional<ConnectorTableExecuteHandle> getTableHandleForOptimize(ConnectorSession session, ConnectorTableHandle tableHandle, Map<String, Object> executeProperties, RetryMode retryMode) {
// TODO lots of that is copied from beginInsert; rafactoring opportunity
if (!isNonTransactionalOptimizeEnabled(session)) {
// post-optimize data files duplicate rows will be left in table and manual cleanup from user will be required.
throw new TrinoException(NOT_SUPPORTED, "OPTIMIZE procedure must be explicitly enabled via " + NON_TRANSACTIONAL_OPTIMIZE_ENABLED + " session property");
}
if (retryMode != NO_RETRIES) {
throw new TrinoException(NOT_SUPPORTED, "OPTIMIZE procedure is not supported with query retries enabled");
}
HiveTableHandle hiveTableHandle = (HiveTableHandle) tableHandle;
SchemaTableName tableName = hiveTableHandle.getSchemaTableName();
Table table = metastore.getTable(tableName.getSchemaName(), tableName.getTableName()).orElseThrow(() -> new TableNotFoundException(tableName));
checkTableIsWritable(table, writesToNonManagedTablesEnabled);
for (Column column : table.getDataColumns()) {
if (!isWritableType(column.getType())) {
throw new TrinoException(NOT_SUPPORTED, format("Optimizing Hive table %s with column type %s not supported", tableName, column.getType()));
}
}
if (isTransactionalTable(table.getParameters())) {
throw new TrinoException(NOT_SUPPORTED, format("Optimizing transactional Hive table %s is not supported", tableName));
}
if (table.getStorage().getBucketProperty().isPresent()) {
throw new TrinoException(NOT_SUPPORTED, format("Optimizing bucketed Hive table %s is not supported", tableName));
}
// TODO forcing NANOSECONDS precision here so we do not loose data. In future we may be smarter; options:
// - respect timestamp_precision but recognize situation when rounding occurs, and fail query
// - detect data's precision and maintain it
List<HiveColumnHandle> columns = hiveColumnHandles(table, typeManager, NANOSECONDS).stream().filter(columnHandle -> !columnHandle.isHidden()).collect(toImmutableList());
HiveStorageFormat tableStorageFormat = extractHiveStorageFormat(table);
Optional.ofNullable(table.getParameters().get(SKIP_HEADER_COUNT_KEY)).map(Integer::parseInt).ifPresent(headerSkipCount -> {
if (headerSkipCount > 1) {
throw new TrinoException(NOT_SUPPORTED, format("Optimizing Hive table %s with value of %s property greater than 1 is not supported", tableName, SKIP_HEADER_COUNT_KEY));
}
});
if (table.getParameters().containsKey(SKIP_FOOTER_COUNT_KEY)) {
throw new TrinoException(NOT_SUPPORTED, format("Optimizing Hive table %s with %s property not supported", tableName, SKIP_FOOTER_COUNT_KEY));
}
LocationHandle locationHandle = locationService.forOptimize(metastore, session, table);
DataSize fileSizeThreshold = (DataSize) executeProperties.get("file_size_threshold");
return Optional.of(new HiveTableExecuteHandle(OptimizeTableProcedure.NAME, Optional.empty(), Optional.of(fileSizeThreshold.toBytes()), tableName.getSchemaName(), tableName.getTableName(), columns, metastore.generatePageSinkMetadata(tableName), locationHandle, table.getStorage().getBucketProperty(), tableStorageFormat, // TODO: test with multiple partitions using different storage format
tableStorageFormat, NO_ACID_TRANSACTION, retryMode != NO_RETRIES));
}
Aggregations