use of io.trino.spi.type.Type in project trino by trinodb.
the class AbstractTestHive method testApplyProjection.
@Test
public void testApplyProjection() throws Exception {
ColumnMetadata bigIntColumn0 = new ColumnMetadata("int0", BIGINT);
ColumnMetadata bigIntColumn1 = new ColumnMetadata("int1", BIGINT);
RowType oneLevelRowType = toRowType(ImmutableList.of(bigIntColumn0, bigIntColumn1));
ColumnMetadata oneLevelRow0 = new ColumnMetadata("onelevelrow0", oneLevelRowType);
RowType twoLevelRowType = toRowType(ImmutableList.of(oneLevelRow0, bigIntColumn0, bigIntColumn1));
ColumnMetadata twoLevelRow0 = new ColumnMetadata("twolevelrow0", twoLevelRowType);
List<ColumnMetadata> columnsForApplyProjectionTest = ImmutableList.of(bigIntColumn0, bigIntColumn1, oneLevelRow0, twoLevelRow0);
SchemaTableName tableName = temporaryTable("apply_projection_tester");
doCreateEmptyTable(tableName, ORC, columnsForApplyProjectionTest);
try (Transaction transaction = newTransaction()) {
ConnectorSession session = newSession();
ConnectorMetadata metadata = transaction.getMetadata();
ConnectorTableHandle tableHandle = getTableHandle(metadata, tableName);
List<ColumnHandle> columnHandles = metadata.getColumnHandles(session, tableHandle).values().stream().filter(columnHandle -> !((HiveColumnHandle) columnHandle).isHidden()).collect(toList());
assertEquals(columnHandles.size(), columnsForApplyProjectionTest.size());
Map<String, ColumnHandle> columnHandleMap = columnHandles.stream().collect(toImmutableMap(handle -> ((HiveColumnHandle) handle).getBaseColumnName(), Function.identity()));
// Emulate symbols coming from the query plan and map them to column handles
Map<String, ColumnHandle> columnHandlesWithSymbols = ImmutableMap.of("symbol_0", columnHandleMap.get("int0"), "symbol_1", columnHandleMap.get("int1"), "symbol_2", columnHandleMap.get("onelevelrow0"), "symbol_3", columnHandleMap.get("twolevelrow0"));
// Create variables for the emulated symbols
Map<String, Variable> symbolVariableMapping = columnHandlesWithSymbols.entrySet().stream().collect(toImmutableMap(Map.Entry::getKey, e -> new Variable(e.getKey(), ((HiveColumnHandle) e.getValue()).getBaseType())));
// Create dereference expressions for testing
FieldDereference symbol2Field0 = new FieldDereference(BIGINT, symbolVariableMapping.get("symbol_2"), 0);
FieldDereference symbol3Field0 = new FieldDereference(oneLevelRowType, symbolVariableMapping.get("symbol_3"), 0);
FieldDereference symbol3Field0Field0 = new FieldDereference(BIGINT, symbol3Field0, 0);
FieldDereference symbol3Field1 = new FieldDereference(BIGINT, symbolVariableMapping.get("symbol_3"), 1);
Map<String, ColumnHandle> inputAssignments;
List<ConnectorExpression> inputProjections;
Optional<ProjectionApplicationResult<ConnectorTableHandle>> projectionResult;
List<ConnectorExpression> expectedProjections;
Map<String, Type> expectedAssignments;
// Test projected columns pushdown to HiveTableHandle in case of all variable references
inputAssignments = getColumnHandlesFor(columnHandlesWithSymbols, ImmutableList.of("symbol_0", "symbol_1"));
inputProjections = ImmutableList.of(symbolVariableMapping.get("symbol_0"), symbolVariableMapping.get("symbol_1"));
expectedAssignments = ImmutableMap.of("symbol_0", BIGINT, "symbol_1", BIGINT);
projectionResult = metadata.applyProjection(session, tableHandle, inputProjections, inputAssignments);
assertProjectionResult(projectionResult, false, inputProjections, expectedAssignments);
// Empty result when projected column handles are same as those present in table handle
projectionResult = metadata.applyProjection(session, projectionResult.get().getHandle(), inputProjections, inputAssignments);
assertProjectionResult(projectionResult, true, ImmutableList.of(), ImmutableMap.of());
// Extra columns handles in HiveTableHandle should get pruned
projectionResult = metadata.applyProjection(session, ((HiveTableHandle) tableHandle).withProjectedColumns(ImmutableSet.copyOf(columnHandles)), inputProjections, inputAssignments);
assertProjectionResult(projectionResult, false, inputProjections, expectedAssignments);
// Test projection pushdown for dereferences
inputAssignments = getColumnHandlesFor(columnHandlesWithSymbols, ImmutableList.of("symbol_2", "symbol_3"));
inputProjections = ImmutableList.of(symbol2Field0, symbol3Field0Field0, symbol3Field1);
expectedAssignments = ImmutableMap.of("onelevelrow0#f_int0", BIGINT, "twolevelrow0#f_onelevelrow0#f_int0", BIGINT, "twolevelrow0#f_int0", BIGINT);
expectedProjections = ImmutableList.of(new Variable("onelevelrow0#f_int0", BIGINT), new Variable("twolevelrow0#f_onelevelrow0#f_int0", BIGINT), new Variable("twolevelrow0#f_int0", BIGINT));
projectionResult = metadata.applyProjection(session, tableHandle, inputProjections, inputAssignments);
assertProjectionResult(projectionResult, false, expectedProjections, expectedAssignments);
// Test reuse of virtual column handles
// Round-1: input projections [symbol_2, symbol_2.int0]. virtual handle is created for symbol_2.int0.
inputAssignments = getColumnHandlesFor(columnHandlesWithSymbols, ImmutableList.of("symbol_2"));
inputProjections = ImmutableList.of(symbol2Field0, symbolVariableMapping.get("symbol_2"));
projectionResult = metadata.applyProjection(session, tableHandle, inputProjections, inputAssignments);
expectedProjections = ImmutableList.of(new Variable("onelevelrow0#f_int0", BIGINT), symbolVariableMapping.get("symbol_2"));
expectedAssignments = ImmutableMap.of("onelevelrow0#f_int0", BIGINT, "symbol_2", oneLevelRowType);
assertProjectionResult(projectionResult, false, expectedProjections, expectedAssignments);
// Round-2: input projections [symbol_2.int0 and onelevelrow0#f_int0]. Virtual handle is reused.
Assignment newlyCreatedColumn = getOnlyElement(projectionResult.get().getAssignments().stream().filter(handle -> handle.getVariable().equals("onelevelrow0#f_int0")).collect(toList()));
inputAssignments = ImmutableMap.<String, ColumnHandle>builder().putAll(getColumnHandlesFor(columnHandlesWithSymbols, ImmutableList.of("symbol_2"))).put(newlyCreatedColumn.getVariable(), newlyCreatedColumn.getColumn()).buildOrThrow();
inputProjections = ImmutableList.of(symbol2Field0, new Variable("onelevelrow0#f_int0", BIGINT));
projectionResult = metadata.applyProjection(session, tableHandle, inputProjections, inputAssignments);
expectedProjections = ImmutableList.of(new Variable("onelevelrow0#f_int0", BIGINT), new Variable("onelevelrow0#f_int0", BIGINT));
expectedAssignments = ImmutableMap.of("onelevelrow0#f_int0", BIGINT);
assertProjectionResult(projectionResult, false, expectedProjections, expectedAssignments);
} finally {
dropTable(tableName);
}
}
use of io.trino.spi.type.Type in project trino by trinodb.
the class AbstractTestHiveFileFormats method createTestFileTrino.
public static FileSplit createTestFileTrino(String filePath, HiveStorageFormat storageFormat, HiveCompressionCodec compressionCodec, List<TestColumn> testColumns, ConnectorSession session, int numRows, HiveFileWriterFactory fileWriterFactory) {
// filter out partition keys, which are not written to the file
testColumns = testColumns.stream().filter(column -> !column.isPartitionKey()).collect(toImmutableList());
List<Type> types = testColumns.stream().map(TestColumn::getType).map(HiveType::valueOf).map(type -> type.getType(TESTING_TYPE_MANAGER)).collect(toList());
PageBuilder pageBuilder = new PageBuilder(types);
for (int rowNumber = 0; rowNumber < numRows; rowNumber++) {
pageBuilder.declarePosition();
for (int columnNumber = 0; columnNumber < testColumns.size(); columnNumber++) {
serializeObject(types.get(columnNumber), pageBuilder.getBlockBuilder(columnNumber), testColumns.get(columnNumber).getWriteValue(), testColumns.get(columnNumber).getObjectInspector(), false);
}
}
Page page = pageBuilder.build();
JobConf jobConf = new JobConf();
configureCompression(jobConf, compressionCodec);
Properties tableProperties = new Properties();
tableProperties.setProperty("columns", testColumns.stream().map(TestColumn::getName).collect(Collectors.joining(",")));
tableProperties.setProperty("columns.types", testColumns.stream().map(TestColumn::getType).collect(Collectors.joining(",")));
Optional<FileWriter> fileWriter = fileWriterFactory.createFileWriter(new Path(filePath), testColumns.stream().map(TestColumn::getName).collect(toList()), StorageFormat.fromHiveStorageFormat(storageFormat), tableProperties, jobConf, session, OptionalInt.empty(), NO_ACID_TRANSACTION, false, WriterKind.INSERT);
FileWriter hiveFileWriter = fileWriter.orElseThrow(() -> new IllegalArgumentException("fileWriterFactory"));
hiveFileWriter.appendRows(page);
hiveFileWriter.commit();
return new FileSplit(new Path(filePath), 0, new File(filePath).length(), new String[0]);
}
use of io.trino.spi.type.Type in project trino by trinodb.
the class AbstractTestHiveFileFormats method checkCursor.
protected void checkCursor(RecordCursor cursor, List<TestColumn> testColumns, int rowCount) {
List<Type> types = testColumns.stream().map(column -> column.getObjectInspector().getTypeName()).map(type -> HiveType.valueOf(type).getType(TESTING_TYPE_MANAGER)).collect(toImmutableList());
Map<Type, MethodHandle> distinctFromOperators = types.stream().distinct().collect(toImmutableMap(identity(), HiveTestUtils::distinctFromOperator));
for (int row = 0; row < rowCount; row++) {
assertTrue(cursor.advanceNextPosition());
for (int i = 0, testColumnsSize = testColumns.size(); i < testColumnsSize; i++) {
TestColumn testColumn = testColumns.get(i);
Type type = types.get(i);
Object fieldFromCursor = getFieldFromCursor(cursor, type, i);
if (fieldFromCursor == null) {
assertEquals(null, testColumn.getExpectedValue(), "Expected null for column " + testColumn.getName());
} else if (type instanceof DecimalType) {
DecimalType decimalType = (DecimalType) type;
fieldFromCursor = new BigDecimal((BigInteger) fieldFromCursor, decimalType.getScale());
assertEquals(fieldFromCursor, testColumn.getExpectedValue(), "Wrong value for column " + testColumn.getName());
} else if (testColumn.getObjectInspector().getTypeName().equals("float")) {
assertEquals((float) fieldFromCursor, (float) testColumn.getExpectedValue(), (float) EPSILON);
} else if (testColumn.getObjectInspector().getTypeName().equals("double")) {
assertEquals((double) fieldFromCursor, (double) testColumn.getExpectedValue(), EPSILON);
} else if (testColumn.getObjectInspector().getTypeName().equals("tinyint")) {
assertEquals(((Number) fieldFromCursor).byteValue(), testColumn.getExpectedValue());
} else if (testColumn.getObjectInspector().getTypeName().equals("smallint")) {
assertEquals(((Number) fieldFromCursor).shortValue(), testColumn.getExpectedValue());
} else if (testColumn.getObjectInspector().getTypeName().equals("int")) {
assertEquals(((Number) fieldFromCursor).intValue(), testColumn.getExpectedValue());
} else if (testColumn.getObjectInspector().getCategory() == Category.PRIMITIVE) {
assertEquals(fieldFromCursor, testColumn.getExpectedValue(), "Wrong value for column " + testColumn.getName());
} else {
Block expected = (Block) testColumn.getExpectedValue();
Block actual = (Block) fieldFromCursor;
boolean distinct = isDistinctFrom(distinctFromOperators.get(type), expected, actual);
assertFalse(distinct, "Wrong value for column: " + testColumn.getName());
}
}
}
assertFalse(cursor.advanceNextPosition());
}
use of io.trino.spi.type.Type in project trino by trinodb.
the class HiveWriteUtils method getField.
public static Object getField(DateTimeZone localZone, Type type, Block block, int position) {
if (block.isNull(position)) {
return null;
}
if (BOOLEAN.equals(type)) {
return type.getBoolean(block, position);
}
if (BIGINT.equals(type)) {
return type.getLong(block, position);
}
if (INTEGER.equals(type)) {
return toIntExact(type.getLong(block, position));
}
if (SMALLINT.equals(type)) {
return Shorts.checkedCast(type.getLong(block, position));
}
if (TINYINT.equals(type)) {
return SignedBytes.checkedCast(type.getLong(block, position));
}
if (REAL.equals(type)) {
return intBitsToFloat((int) type.getLong(block, position));
}
if (DOUBLE.equals(type)) {
return type.getDouble(block, position);
}
if (type instanceof VarcharType) {
return new Text(type.getSlice(block, position).getBytes());
}
if (type instanceof CharType) {
CharType charType = (CharType) type;
return new Text(padSpaces(type.getSlice(block, position), charType).toStringUtf8());
}
if (VARBINARY.equals(type)) {
return type.getSlice(block, position).getBytes();
}
if (DATE.equals(type)) {
return Date.ofEpochDay(toIntExact(type.getLong(block, position)));
}
if (type instanceof TimestampType) {
return getHiveTimestamp(localZone, (TimestampType) type, block, position);
}
if (type instanceof DecimalType) {
DecimalType decimalType = (DecimalType) type;
return getHiveDecimal(decimalType, block, position);
}
if (type instanceof ArrayType) {
Type elementType = ((ArrayType) type).getElementType();
Block arrayBlock = block.getObject(position, Block.class);
List<Object> list = new ArrayList<>(arrayBlock.getPositionCount());
for (int i = 0; i < arrayBlock.getPositionCount(); i++) {
list.add(getField(localZone, elementType, arrayBlock, i));
}
return unmodifiableList(list);
}
if (type instanceof MapType) {
Type keyType = ((MapType) type).getKeyType();
Type valueType = ((MapType) type).getValueType();
Block mapBlock = block.getObject(position, Block.class);
Map<Object, Object> map = new HashMap<>();
for (int i = 0; i < mapBlock.getPositionCount(); i += 2) {
map.put(getField(localZone, keyType, mapBlock, i), getField(localZone, valueType, mapBlock, i + 1));
}
return unmodifiableMap(map);
}
if (type instanceof RowType) {
List<Type> fieldTypes = type.getTypeParameters();
Block rowBlock = block.getObject(position, Block.class);
checkCondition(fieldTypes.size() == rowBlock.getPositionCount(), StandardErrorCode.GENERIC_INTERNAL_ERROR, "Expected row value field count does not match type field count");
List<Object> row = new ArrayList<>(rowBlock.getPositionCount());
for (int i = 0; i < rowBlock.getPositionCount(); i++) {
row.add(getField(localZone, fieldTypes.get(i), rowBlock, i));
}
return unmodifiableList(row);
}
throw new TrinoException(NOT_SUPPORTED, "unsupported type: " + type);
}
use of io.trino.spi.type.Type in project trino by trinodb.
the class SerDeUtils method serializeList.
private static Block serializeList(Type type, BlockBuilder builder, Object object, ListObjectInspector inspector) {
List<?> list = inspector.getList(object);
if (list == null) {
requireNonNull(builder, "builder is null").appendNull();
return null;
}
List<Type> typeParameters = type.getTypeParameters();
checkArgument(typeParameters.size() == 1, "list must have exactly 1 type parameter");
Type elementType = typeParameters.get(0);
ObjectInspector elementInspector = inspector.getListElementObjectInspector();
BlockBuilder currentBuilder;
if (builder != null) {
currentBuilder = builder.beginBlockEntry();
} else {
currentBuilder = elementType.createBlockBuilder(null, list.size());
}
for (Object element : list) {
serializeObject(elementType, currentBuilder, element, elementInspector);
}
if (builder != null) {
builder.closeEntry();
return null;
} else {
Block resultBlock = currentBuilder.build();
return resultBlock;
}
}
Aggregations