use of io.trino.spi.connector.ColumnMetadata in project trino by trinodb.
the class TestTransactionLogAccess method testSnapshotsAreConsistent.
@Test
public void testSnapshotsAreConsistent() throws Exception {
String tableName = "person";
File tempDir = Files.createTempDir();
File tableDir = new File(tempDir, tableName);
File transactionLogDir = new File(tableDir, TRANSACTION_LOG_DIRECTORY);
transactionLogDir.mkdirs();
File resourceDir = new File(getClass().getClassLoader().getResource("databricks/person/_delta_log").toURI());
copyTransactionLogEntry(0, 12, resourceDir, transactionLogDir);
Files.copy(new File(resourceDir, LAST_CHECKPOINT_FILENAME), new File(transactionLogDir, LAST_CHECKPOINT_FILENAME));
setupTransactionLogAccess(tableName, new Path(tableDir.toURI()));
List<AddFileEntry> expectedDataFiles = transactionLogAccess.getActiveFiles(tableSnapshot, SESSION);
copyTransactionLogEntry(12, 14, resourceDir, transactionLogDir);
Set<String> newDataFiles = ImmutableSet.of("age=28/part-00000-40dd1707-1d42-4328-a59a-21f5c945fe60.c000.snappy.parquet", "age=29/part-00000-3794c463-cb0c-4beb-8d07-7cc1e3b5920f.c000.snappy.parquet");
TableSnapshot updatedTableSnapshot = transactionLogAccess.loadSnapshot(new SchemaTableName("schema", tableName), new Path(tableDir.toURI()), SESSION);
List<AddFileEntry> allDataFiles = transactionLogAccess.getActiveFiles(updatedTableSnapshot, SESSION);
List<AddFileEntry> dataFilesWithFixedVersion = transactionLogAccess.getActiveFiles(tableSnapshot, SESSION);
for (String newFilePath : newDataFiles) {
assertTrue(allDataFiles.stream().anyMatch(entry -> entry.getPath().equals(newFilePath)));
assertTrue(dataFilesWithFixedVersion.stream().noneMatch(entry -> entry.getPath().equals(newFilePath)));
}
assertEquals(expectedDataFiles.size(), dataFilesWithFixedVersion.size());
List<ColumnMetadata> columns = extractSchema(transactionLogAccess.getMetadataEntry(tableSnapshot, SESSION).get(), TESTING_TYPE_MANAGER);
for (int i = 0; i < expectedDataFiles.size(); i++) {
AddFileEntry expected = expectedDataFiles.get(i);
AddFileEntry actual = dataFilesWithFixedVersion.get(i);
assertEquals(expected.getPath(), actual.getPath());
assertEquals(expected.getPartitionValues(), actual.getPartitionValues());
assertEquals(expected.getSize(), actual.getSize());
assertEquals(expected.getModificationTime(), actual.getModificationTime());
assertEquals(expected.isDataChange(), actual.isDataChange());
assertEquals(expected.getTags(), actual.getTags());
assertTrue(expected.getStats().isPresent());
assertTrue(actual.getStats().isPresent());
for (ColumnMetadata column : columns) {
DeltaLakeColumnHandle columnHandle = new DeltaLakeColumnHandle(column.getName(), column.getType(), REGULAR);
assertEquals(expected.getStats().get().getMinColumnValue(columnHandle), actual.getStats().get().getMinColumnValue(columnHandle));
assertEquals(expected.getStats().get().getMaxColumnValue(columnHandle), actual.getStats().get().getMaxColumnValue(columnHandle));
assertEquals(expected.getStats().get().getNullCount(columnHandle.getName()), actual.getStats().get().getNullCount(columnHandle.getName()));
assertEquals(expected.getStats().get().getNumRecords(), actual.getStats().get().getNumRecords());
}
}
}
use of io.trino.spi.connector.ColumnMetadata in project trino by trinodb.
the class CreateTableTask method internalExecute.
@VisibleForTesting
ListenableFuture<Void> internalExecute(CreateTable statement, Session session, List<Expression> parameters, Consumer<Output> outputConsumer) {
checkArgument(!statement.getElements().isEmpty(), "no columns for table");
Map<NodeRef<Parameter>, Expression> parameterLookup = parameterExtractor(statement, parameters);
QualifiedObjectName tableName = createQualifiedObjectName(session, statement, statement.getName());
Optional<TableHandle> tableHandle = plannerContext.getMetadata().getTableHandle(session, tableName);
if (tableHandle.isPresent()) {
if (!statement.isNotExists()) {
throw semanticException(TABLE_ALREADY_EXISTS, statement, "Table '%s' already exists", tableName);
}
return immediateVoidFuture();
}
CatalogName catalogName = getRequiredCatalogHandle(plannerContext.getMetadata(), session, statement, tableName.getCatalogName());
LinkedHashMap<String, ColumnMetadata> columns = new LinkedHashMap<>();
Map<String, Object> inheritedProperties = ImmutableMap.of();
boolean includingProperties = false;
for (TableElement element : statement.getElements()) {
if (element instanceof ColumnDefinition) {
ColumnDefinition column = (ColumnDefinition) element;
String name = column.getName().getValue().toLowerCase(Locale.ENGLISH);
Type type;
try {
type = plannerContext.getTypeManager().getType(toTypeSignature(column.getType()));
} catch (TypeNotFoundException e) {
throw semanticException(TYPE_NOT_FOUND, element, "Unknown type '%s' for column '%s'", column.getType(), column.getName());
}
if (type.equals(UNKNOWN)) {
throw semanticException(COLUMN_TYPE_UNKNOWN, element, "Unknown type '%s' for column '%s'", column.getType(), column.getName());
}
if (columns.containsKey(name)) {
throw semanticException(DUPLICATE_COLUMN_NAME, column, "Column name '%s' specified more than once", column.getName());
}
if (!column.isNullable() && !plannerContext.getMetadata().getConnectorCapabilities(session, catalogName).contains(NOT_NULL_COLUMN_CONSTRAINT)) {
throw semanticException(NOT_SUPPORTED, column, "Catalog '%s' does not support non-null column for column name '%s'", catalogName.getCatalogName(), column.getName());
}
Map<String, Object> columnProperties = columnPropertyManager.getProperties(catalogName, column.getProperties(), session, plannerContext, accessControl, parameterLookup, true);
columns.put(name, ColumnMetadata.builder().setName(name).setType(type).setNullable(column.isNullable()).setComment(column.getComment()).setProperties(columnProperties).build());
} else if (element instanceof LikeClause) {
LikeClause likeClause = (LikeClause) element;
QualifiedObjectName originalLikeTableName = createQualifiedObjectName(session, statement, likeClause.getTableName());
if (plannerContext.getMetadata().getCatalogHandle(session, originalLikeTableName.getCatalogName()).isEmpty()) {
throw semanticException(CATALOG_NOT_FOUND, statement, "LIKE table catalog '%s' does not exist", originalLikeTableName.getCatalogName());
}
RedirectionAwareTableHandle redirection = plannerContext.getMetadata().getRedirectionAwareTableHandle(session, originalLikeTableName);
TableHandle likeTable = redirection.getTableHandle().orElseThrow(() -> semanticException(TABLE_NOT_FOUND, statement, "LIKE table '%s' does not exist", originalLikeTableName));
QualifiedObjectName likeTableName = redirection.getRedirectedTableName().orElse(originalLikeTableName);
if (!tableName.getCatalogName().equals(likeTableName.getCatalogName())) {
String message = "CREATE TABLE LIKE across catalogs is not supported";
if (!originalLikeTableName.equals(likeTableName)) {
message += format(". LIKE table '%s' redirected to '%s'.", originalLikeTableName, likeTableName);
}
throw semanticException(NOT_SUPPORTED, statement, message);
}
TableMetadata likeTableMetadata = plannerContext.getMetadata().getTableMetadata(session, likeTable);
Optional<LikeClause.PropertiesOption> propertiesOption = likeClause.getPropertiesOption();
if (propertiesOption.isPresent() && propertiesOption.get() == LikeClause.PropertiesOption.INCLUDING) {
if (includingProperties) {
throw semanticException(NOT_SUPPORTED, statement, "Only one LIKE clause can specify INCLUDING PROPERTIES");
}
includingProperties = true;
inheritedProperties = likeTableMetadata.getMetadata().getProperties();
}
try {
accessControl.checkCanSelectFromColumns(session.toSecurityContext(), likeTableName, likeTableMetadata.getColumns().stream().map(ColumnMetadata::getName).collect(toImmutableSet()));
} catch (AccessDeniedException e) {
throw new AccessDeniedException("Cannot reference columns of table " + likeTableName);
}
if (propertiesOption.orElse(EXCLUDING) == INCLUDING) {
try {
accessControl.checkCanShowCreateTable(session.toSecurityContext(), likeTableName);
} catch (AccessDeniedException e) {
throw new AccessDeniedException("Cannot reference properties of table " + likeTableName);
}
}
likeTableMetadata.getColumns().stream().filter(column -> !column.isHidden()).forEach(column -> {
if (columns.containsKey(column.getName().toLowerCase(Locale.ENGLISH))) {
throw semanticException(DUPLICATE_COLUMN_NAME, element, "Column name '%s' specified more than once", column.getName());
}
columns.put(column.getName().toLowerCase(Locale.ENGLISH), column);
});
} else {
throw new TrinoException(GENERIC_INTERNAL_ERROR, "Invalid TableElement: " + element.getClass().getName());
}
}
Map<String, Object> properties = tablePropertyManager.getProperties(catalogName, statement.getProperties(), session, plannerContext, accessControl, parameterLookup, true);
accessControl.checkCanCreateTable(session.toSecurityContext(), tableName, properties);
Set<String> specifiedPropertyKeys = statement.getProperties().stream().map(property -> property.getName().getValue()).collect(toImmutableSet());
Map<String, Object> finalProperties = combineProperties(specifiedPropertyKeys, properties, inheritedProperties);
ConnectorTableMetadata tableMetadata = new ConnectorTableMetadata(tableName.asSchemaTableName(), ImmutableList.copyOf(columns.values()), finalProperties, statement.getComment());
try {
plannerContext.getMetadata().createTable(session, tableName.getCatalogName(), tableMetadata, statement.isNotExists());
} catch (TrinoException e) {
// connectors are not required to handle the ignoreExisting flag
if (!e.getErrorCode().equals(ALREADY_EXISTS.toErrorCode()) || !statement.isNotExists()) {
throw e;
}
}
outputConsumer.accept(new Output(tableName.getCatalogName(), tableName.getSchemaName(), tableName.getObjectName(), Optional.of(tableMetadata.getColumns().stream().map(column -> new OutputColumn(new Column(column.getName(), column.getType().toString()), ImmutableSet.of())).collect(toImmutableList()))));
return immediateVoidFuture();
}
use of io.trino.spi.connector.ColumnMetadata in project trino by trinodb.
the class LogicalPlanner method getInsertPlan.
private RelationPlan getInsertPlan(Analysis analysis, Table table, Query query, TableHandle tableHandle, List<ColumnHandle> insertColumns, Optional<TableLayout> newTableLayout, Optional<WriterTarget> materializedViewRefreshWriterTarget) {
TableMetadata tableMetadata = metadata.getTableMetadata(session, tableHandle);
Map<NodeRef<LambdaArgumentDeclaration>, Symbol> lambdaDeclarationToSymbolMap = buildLambdaDeclarationToSymbolMap(analysis, symbolAllocator);
RelationPlanner planner = new RelationPlanner(analysis, symbolAllocator, idAllocator, lambdaDeclarationToSymbolMap, plannerContext, Optional.empty(), session, ImmutableMap.of());
RelationPlan plan = planner.process(query, null);
ImmutableList.Builder<Symbol> builder = ImmutableList.builder();
for (int i = 0; i < plan.getFieldMappings().size(); i++) {
if (!plan.getDescriptor().getFieldByIndex(i).isHidden()) {
builder.add(plan.getFieldMappings().get(i));
}
}
List<Symbol> visibleFieldMappings = builder.build();
Map<String, ColumnHandle> columns = metadata.getColumnHandles(session, tableHandle);
Assignments.Builder assignments = Assignments.builder();
boolean supportsMissingColumnsOnInsert = metadata.supportsMissingColumnsOnInsert(session, tableHandle);
ImmutableList.Builder<ColumnMetadata> insertedColumnsBuilder = ImmutableList.builder();
for (ColumnMetadata column : tableMetadata.getColumns()) {
if (column.isHidden()) {
continue;
}
Symbol output = symbolAllocator.newSymbol(column.getName(), column.getType());
int index = insertColumns.indexOf(columns.get(column.getName()));
if (index < 0) {
if (supportsMissingColumnsOnInsert) {
continue;
}
Expression cast = new Cast(new NullLiteral(), toSqlType(column.getType()));
assignments.put(output, cast);
insertedColumnsBuilder.add(column);
} else {
Symbol input = visibleFieldMappings.get(index);
Type tableType = column.getType();
Type queryType = symbolAllocator.getTypes().get(input);
if (queryType.equals(tableType) || typeCoercion.isTypeOnlyCoercion(queryType, tableType)) {
assignments.put(output, input.toSymbolReference());
} else {
Expression cast = noTruncationCast(input.toSymbolReference(), queryType, tableType);
assignments.put(output, cast);
}
insertedColumnsBuilder.add(column);
}
}
ProjectNode projectNode = new ProjectNode(idAllocator.getNextId(), plan.getRoot(), assignments.build());
List<ColumnMetadata> insertedColumns = insertedColumnsBuilder.build();
List<Field> fields = insertedColumns.stream().map(column -> Field.newUnqualified(column.getName(), column.getType())).collect(toImmutableList());
Scope scope = Scope.builder().withRelationType(RelationId.anonymous(), new RelationType(fields)).build();
plan = new RelationPlan(projectNode, scope, projectNode.getOutputSymbols(), Optional.empty());
plan = planner.addRowFilters(table, plan, failIfPredicateIsNotMeet(metadata, session, PERMISSION_DENIED, AccessDeniedException.PREFIX + "Cannot insert row that does not match to a row filter"), node -> {
Scope accessControlScope = analysis.getAccessControlScope(table);
// hidden fields are not accessible in insert
return Scope.builder().like(accessControlScope).withRelationType(accessControlScope.getRelationId(), accessControlScope.getRelationType().withOnlyVisibleFields()).build();
});
List<String> insertedTableColumnNames = insertedColumns.stream().map(ColumnMetadata::getName).collect(toImmutableList());
String catalogName = tableHandle.getCatalogName().getCatalogName();
TableStatisticsMetadata statisticsMetadata = metadata.getStatisticsCollectionMetadataForWrite(session, catalogName, tableMetadata.getMetadata());
if (materializedViewRefreshWriterTarget.isPresent()) {
return createTableWriterPlan(analysis, plan.getRoot(), plan.getFieldMappings(), materializedViewRefreshWriterTarget.get(), insertedTableColumnNames, insertedColumns, newTableLayout, statisticsMetadata);
}
InsertReference insertTarget = new InsertReference(tableHandle, insertedTableColumnNames.stream().map(columns::get).collect(toImmutableList()));
return createTableWriterPlan(analysis, plan.getRoot(), plan.getFieldMappings(), insertTarget, insertedTableColumnNames, insertedColumns, newTableLayout, statisticsMetadata);
}
use of io.trino.spi.connector.ColumnMetadata in project trino by trinodb.
the class LogicalPlanner method createAnalyzePlan.
private RelationPlan createAnalyzePlan(Analysis analysis, Analyze analyzeStatement) {
TableHandle targetTable = analysis.getAnalyzeTarget().orElseThrow();
// Plan table scan
Map<String, ColumnHandle> columnHandles = metadata.getColumnHandles(session, targetTable);
ImmutableList.Builder<Symbol> tableScanOutputs = ImmutableList.builder();
ImmutableMap.Builder<Symbol, ColumnHandle> symbolToColumnHandle = ImmutableMap.builder();
ImmutableMap.Builder<String, Symbol> columnNameToSymbol = ImmutableMap.builder();
TableMetadata tableMetadata = metadata.getTableMetadata(session, targetTable);
for (ColumnMetadata column : tableMetadata.getColumns()) {
Symbol symbol = symbolAllocator.newSymbol(column.getName(), column.getType());
tableScanOutputs.add(symbol);
symbolToColumnHandle.put(symbol, columnHandles.get(column.getName()));
columnNameToSymbol.put(column.getName(), symbol);
}
TableStatisticsMetadata tableStatisticsMetadata = metadata.getStatisticsCollectionMetadata(session, targetTable.getCatalogName().getCatalogName(), tableMetadata.getMetadata());
TableStatisticAggregation tableStatisticAggregation = statisticsAggregationPlanner.createStatisticsAggregation(tableStatisticsMetadata, columnNameToSymbol.buildOrThrow());
StatisticAggregations statisticAggregations = tableStatisticAggregation.getAggregations();
List<Symbol> groupingSymbols = statisticAggregations.getGroupingSymbols();
PlanNode planNode = new StatisticsWriterNode(idAllocator.getNextId(), new AggregationNode(idAllocator.getNextId(), TableScanNode.newInstance(idAllocator.getNextId(), targetTable, tableScanOutputs.build(), symbolToColumnHandle.buildOrThrow(), false, Optional.empty()), statisticAggregations.getAggregations(), singleGroupingSet(groupingSymbols), ImmutableList.of(), AggregationNode.Step.SINGLE, Optional.empty(), Optional.empty()), new StatisticsWriterNode.WriteStatisticsReference(targetTable), symbolAllocator.newSymbol("rows", BIGINT), tableStatisticsMetadata.getTableStatistics().contains(ROW_COUNT), tableStatisticAggregation.getDescriptor());
return new RelationPlan(planNode, analysis.getScope(analyzeStatement), planNode.getOutputSymbols(), Optional.empty());
}
use of io.trino.spi.connector.ColumnMetadata in project trino by trinodb.
the class LogicalPlanner method createTableWriterPlan.
private RelationPlan createTableWriterPlan(Analysis analysis, PlanNode source, List<Symbol> symbols, WriterTarget target, List<String> columnNames, List<ColumnMetadata> columnMetadataList, Optional<TableLayout> writeTableLayout, TableStatisticsMetadata statisticsMetadata) {
Optional<PartitioningScheme> partitioningScheme = Optional.empty();
Optional<PartitioningScheme> preferredPartitioningScheme = Optional.empty();
if (writeTableLayout.isPresent()) {
List<Symbol> partitionFunctionArguments = new ArrayList<>();
writeTableLayout.get().getPartitionColumns().stream().mapToInt(columnNames::indexOf).mapToObj(symbols::get).forEach(partitionFunctionArguments::add);
List<Symbol> outputLayout = new ArrayList<>(symbols);
Optional<PartitioningHandle> partitioningHandle = writeTableLayout.get().getPartitioning();
if (partitioningHandle.isPresent()) {
partitioningScheme = Optional.of(new PartitioningScheme(Partitioning.create(partitioningHandle.get(), partitionFunctionArguments), outputLayout));
} else {
// empty connector partitioning handle means evenly partitioning on partitioning columns
preferredPartitioningScheme = Optional.of(new PartitioningScheme(Partitioning.create(FIXED_HASH_DISTRIBUTION, partitionFunctionArguments), outputLayout));
}
}
verify(columnNames.size() == symbols.size(), "columnNames.size() != symbols.size(): %s and %s", columnNames, symbols);
Map<String, Symbol> columnToSymbolMap = zip(columnNames.stream(), symbols.stream(), SimpleImmutableEntry::new).collect(toImmutableMap(Entry::getKey, Entry::getValue));
Set<Symbol> notNullColumnSymbols = columnMetadataList.stream().filter(column -> !column.isNullable()).map(ColumnMetadata::getName).map(columnToSymbolMap::get).collect(toImmutableSet());
if (!statisticsMetadata.isEmpty()) {
TableStatisticAggregation result = statisticsAggregationPlanner.createStatisticsAggregation(statisticsMetadata, columnToSymbolMap);
StatisticAggregations.Parts aggregations = result.getAggregations().createPartialAggregations(symbolAllocator, plannerContext);
// partial aggregation is run within the TableWriteOperator to calculate the statistics for
// the data consumed by the TableWriteOperator
// final aggregation is run within the TableFinishOperator to summarize collected statistics
// by the partial aggregation from all of the writer nodes
StatisticAggregations partialAggregation = aggregations.getPartialAggregation();
TableFinishNode commitNode = new TableFinishNode(idAllocator.getNextId(), new TableWriterNode(idAllocator.getNextId(), source, target, symbolAllocator.newSymbol("partialrows", BIGINT), symbolAllocator.newSymbol("fragment", VARBINARY), symbols, columnNames, notNullColumnSymbols, partitioningScheme, preferredPartitioningScheme, Optional.of(partialAggregation), Optional.of(result.getDescriptor().map(aggregations.getMappings()::get))), target, symbolAllocator.newSymbol("rows", BIGINT), Optional.of(aggregations.getFinalAggregation()), Optional.of(result.getDescriptor()));
return new RelationPlan(commitNode, analysis.getRootScope(), commitNode.getOutputSymbols(), Optional.empty());
}
TableFinishNode commitNode = new TableFinishNode(idAllocator.getNextId(), new TableWriterNode(idAllocator.getNextId(), source, target, symbolAllocator.newSymbol("partialrows", BIGINT), symbolAllocator.newSymbol("fragment", VARBINARY), symbols, columnNames, notNullColumnSymbols, partitioningScheme, preferredPartitioningScheme, Optional.empty(), Optional.empty()), target, symbolAllocator.newSymbol("rows", BIGINT), Optional.empty(), Optional.empty());
return new RelationPlan(commitNode, analysis.getRootScope(), commitNode.getOutputSymbols(), Optional.empty());
}
Aggregations