Examples with Type io.trino.spi.type.Type used on opensource projects

Example 76 with Type

use of io.trino.spi.type.Type in project trino by trinodb.

the class HiveTypeTranslator method toTypeSignature.

public static TypeSignature toTypeSignature(TypeInfo typeInfo, HiveTimestampPrecision timestampPrecision) {
    switch(typeInfo.getCategory()) {
        case PRIMITIVE:
            Type primitiveType = fromPrimitiveType((PrimitiveTypeInfo) typeInfo, timestampPrecision);
            if (primitiveType == null) {
                break;
            }
            return primitiveType.getTypeSignature();
        case MAP:
            MapTypeInfo mapTypeInfo = (MapTypeInfo) typeInfo;
            return mapType(toTypeSignature(mapTypeInfo.getMapKeyTypeInfo(), timestampPrecision), toTypeSignature(mapTypeInfo.getMapValueTypeInfo(), timestampPrecision));
        case LIST:
            ListTypeInfo listTypeInfo = (ListTypeInfo) typeInfo;
            TypeSignature elementType = toTypeSignature(listTypeInfo.getListElementTypeInfo(), timestampPrecision);
            return arrayType(typeParameter(elementType));
        case STRUCT:
            StructTypeInfo structTypeInfo = (StructTypeInfo) typeInfo;
            List<TypeInfo> fieldTypes = structTypeInfo.getAllStructFieldTypeInfos();
            List<String> fieldNames = structTypeInfo.getAllStructFieldNames();
            if (fieldTypes.size() != fieldNames.size()) {
                throw new TrinoException(HiveErrorCode.HIVE_INVALID_METADATA, format("Invalid Hive struct type: %s", typeInfo));
            }
            return rowType(Streams.zip(// TODO: This is a hack. Trino engine should be able to handle identifiers in a case insensitive way where necessary.
            fieldNames.stream().map(s -> s.toLowerCase(Locale.US)), fieldTypes.stream().map(type -> toTypeSignature(type, timestampPrecision)), TypeSignatureParameter::namedField).collect(Collectors.toList()));
        case UNION:
            // Use a row type to represent a union type in Hive for reading
            UnionTypeInfo unionTypeInfo = (UnionTypeInfo) typeInfo;
            List<TypeInfo> unionObjectTypes = unionTypeInfo.getAllUnionObjectTypeInfos();
            ImmutableList.Builder<TypeSignatureParameter> typeSignatures = ImmutableList.builder();
            typeSignatures.add(namedField("tag", TINYINT.getTypeSignature()));
            for (int i = 0; i < unionObjectTypes.size(); i++) {
                TypeInfo unionObjectType = unionObjectTypes.get(i);
                typeSignatures.add(namedField("field" + i, toTypeSignature(unionObjectType, timestampPrecision)));
            }
            return rowType(typeSignatures.build());
    }
    throw new TrinoException(NOT_SUPPORTED, format("Unsupported Hive type: %s", typeInfo));
}

Example 77 with Type

use of io.trino.spi.type.Type in project trino by trinodb.

the class HiveBlockEncodingSerde method readType.

@Override
public Type readType(SliceInput sliceInput) {
    requireNonNull(sliceInput, "sliceInput is null");
    TypeId id = TypeId.of(readLengthPrefixedString(sliceInput));
    for (Type type : TYPES) {
        if (type.getTypeId().equals(id)) {
            return type;
        }
    }
    throw new IllegalArgumentException("Type not found: " + id);
}

Example 78 with Type

use of io.trino.spi.type.Type in project trino by trinodb.

the class MetastoreHiveStatisticsProvider method createZeroStatistics.

private TableStatistics createZeroStatistics(Map<String, ColumnHandle> columns, Map<String, Type> columnTypes) {
    TableStatistics.Builder result = TableStatistics.builder();
    result.setRowCount(Estimate.of(0));
    columns.forEach((columnName, columnHandle) -> {
        Type columnType = columnTypes.get(columnName);
        verifyNotNull(columnType, "columnType is missing for column: %s", columnName);
        ColumnStatistics.Builder columnStatistics = ColumnStatistics.builder();
        columnStatistics.setNullsFraction(Estimate.of(0));
        columnStatistics.setDistinctValuesCount(Estimate.of(0));
        if (hasDataSize(columnType)) {
            columnStatistics.setDataSize(Estimate.of(0));
        }
        result.setColumnStatistics(columnHandle, columnStatistics.build());
    });
    return result.build();
}

Example 79 with Type

use of io.trino.spi.type.Type in project trino by trinodb.

the class AbstractTestHive method setup.

protected final void setup(String databaseName, HiveConfig hiveConfig, HiveMetastore hiveMetastore, HdfsEnvironment hdfsConfiguration) {
    setupHive(databaseName);
    metastoreClient = hiveMetastore;
    hdfsEnvironment = hdfsConfiguration;
    HivePartitionManager partitionManager = new HivePartitionManager(hiveConfig);
    locationService = new HiveLocationService(hdfsEnvironment);
    JsonCodec<PartitionUpdate> partitionUpdateCodec = JsonCodec.jsonCodec(PartitionUpdate.class);
    metadataFactory = new HiveMetadataFactory(new CatalogName("hive"), HiveMetastoreFactory.ofInstance(metastoreClient), hdfsEnvironment, partitionManager, 10, 10, 10, false, false, false, true, true, false, 1000, Optional.empty(), true, TESTING_TYPE_MANAGER, NOOP_METADATA_PROVIDER, locationService, partitionUpdateCodec, newFixedThreadPool(2), heartbeatService, TEST_SERVER_VERSION, (session, tableHandle) -> {
        if (!tableHandle.getTableName().contains("apply_redirection_tester")) {
            return Optional.empty();
        }
        return Optional.of(new TableScanRedirectApplicationResult(new CatalogSchemaTableName("hive", databaseName, "mock_redirection_target"), ImmutableMap.of(), TupleDomain.all()));
    }, ImmutableSet.of(new PartitionsSystemTableProvider(partitionManager, TESTING_TYPE_MANAGER), new PropertiesSystemTableProvider()), metastore -> new NoneHiveMaterializedViewMetadata() {

        @Override
        public Optional<ConnectorMaterializedViewDefinition> getMaterializedView(ConnectorSession session, SchemaTableName viewName) {
            if (!viewName.getTableName().contains("materialized_view_tester")) {
                return Optional.empty();
            }
            return Optional.of(new ConnectorMaterializedViewDefinition("dummy_view_sql", Optional.empty(), Optional.empty(), Optional.empty(), ImmutableList.of(new ConnectorMaterializedViewDefinition.Column("abc", TypeId.of("type"))), Optional.empty(), Optional.of("alice"), ImmutableMap.of()));
        }
    }, SqlStandardAccessControlMetadata::new, NO_REDIRECTIONS, TableInvalidationCallback.NOOP);
    transactionManager = new HiveTransactionManager(metadataFactory);
    splitManager = new HiveSplitManager(transactionManager, partitionManager, new NamenodeStats(), hdfsEnvironment, new CachingDirectoryLister(hiveConfig), directExecutor(), new CounterStat(), 100, hiveConfig.getMaxOutstandingSplitsSize(), hiveConfig.getMinPartitionBatchSize(), hiveConfig.getMaxPartitionBatchSize(), hiveConfig.getMaxInitialSplits(), hiveConfig.getSplitLoaderConcurrency(), hiveConfig.getMaxSplitsPerSecond(), false, TESTING_TYPE_MANAGER);
    pageSinkProvider = new HivePageSinkProvider(getDefaultHiveFileWriterFactories(hiveConfig, hdfsEnvironment), hdfsEnvironment, PAGE_SORTER, HiveMetastoreFactory.ofInstance(metastoreClient), new GroupByHashPageIndexerFactory(JOIN_COMPILER, BLOCK_TYPE_OPERATORS), TESTING_TYPE_MANAGER, getHiveConfig(), locationService, partitionUpdateCodec, new TestingNodeManager("fake-environment"), new HiveEventClient(), getHiveSessionProperties(hiveConfig), new HiveWriterStats());
    pageSourceProvider = new HivePageSourceProvider(TESTING_TYPE_MANAGER, hdfsEnvironment, hiveConfig, getDefaultHivePageSourceFactories(hdfsEnvironment, hiveConfig), getDefaultHiveRecordCursorProviders(hiveConfig, hdfsEnvironment), new GenericHiveRecordCursorProvider(hdfsEnvironment, hiveConfig), Optional.empty());
    nodePartitioningProvider = new HiveNodePartitioningProvider(new TestingNodeManager("fake-environment"), TESTING_TYPE_MANAGER);
}

Example 80 with Type

use of io.trino.spi.type.Type in project trino by trinodb.

the class AbstractTestHive method doTestBucketSortedTables.

private void doTestBucketSortedTables(SchemaTableName table) throws IOException {
    int bucketCount = 3;
    int expectedRowCount = 0;
    try (Transaction transaction = newTransaction()) {
        ConnectorSession session = newSession();
        ConnectorMetadata metadata = transaction.getMetadata();
        // begin creating the table
        ConnectorTableMetadata tableMetadata = new ConnectorTableMetadata(table, ImmutableList.<ColumnMetadata>builder().add(new ColumnMetadata("id", VARCHAR)).add(new ColumnMetadata("value_asc", VARCHAR)).add(new ColumnMetadata("value_desc", BIGINT)).add(new ColumnMetadata("ds", VARCHAR)).build(), ImmutableMap.<String, Object>builder().put(STORAGE_FORMAT_PROPERTY, RCBINARY).put(PARTITIONED_BY_PROPERTY, ImmutableList.of("ds")).put(BUCKETED_BY_PROPERTY, ImmutableList.of("id")).put(BUCKET_COUNT_PROPERTY, bucketCount).put(SORTED_BY_PROPERTY, ImmutableList.builder().add(new SortingColumn("value_asc", ASCENDING)).add(new SortingColumn("value_desc", DESCENDING)).build()).buildOrThrow());
        ConnectorOutputTableHandle outputHandle = metadata.beginCreateTable(session, tableMetadata, Optional.empty(), NO_RETRIES);
        // write the data
        ConnectorPageSink sink = pageSinkProvider.createPageSink(transaction.getTransactionHandle(), session, outputHandle);
        List<Type> types = tableMetadata.getColumns().stream().map(ColumnMetadata::getType).collect(toList());
        ThreadLocalRandom random = ThreadLocalRandom.current();
        for (int i = 0; i < 50; i++) {
            MaterializedResult.Builder builder = MaterializedResult.resultBuilder(session, types);
            for (int j = 0; j < 1000; j++) {
                builder.row(sha256().hashLong(random.nextLong()).toString(), "test" + random.nextInt(100), random.nextLong(100_000), "2018-04-01");
                expectedRowCount++;
            }
            sink.appendPage(builder.build().toPage());
        }
        HdfsContext context = new HdfsContext(session);
        // verify we have enough temporary files per bucket to require multiple passes
        Path stagingPathRoot;
        if (isTemporaryStagingDirectoryEnabled(session)) {
            stagingPathRoot = new Path(getTemporaryStagingDirectoryPath(session).replace("${USER}", context.getIdentity().getUser()));
        } else {
            stagingPathRoot = getStagingPathRoot(outputHandle);
        }
        assertThat(listAllDataFiles(context, stagingPathRoot)).filteredOn(file -> file.contains(".tmp-sort.")).size().isGreaterThan(bucketCount * getHiveConfig().getMaxOpenSortFiles() * 2);
        // finish the write
        Collection<Slice> fragments = getFutureValue(sink.finish());
        // verify there are no temporary files
        for (String file : listAllDataFiles(context, stagingPathRoot)) {
            assertThat(file).doesNotContain(".tmp-sort.");
        }
        // finish creating table
        metadata.finishCreateTable(session, outputHandle, fragments, ImmutableList.of());
        transaction.commit();
    }
    // verify that bucket files are sorted
    try (Transaction transaction = newTransaction()) {
        ConnectorMetadata metadata = transaction.getMetadata();
        ConnectorSession session = newSession();
        metadata.beginQuery(session);
        ConnectorTableHandle tableHandle = getTableHandle(metadata, table);
        List<ColumnHandle> columnHandles = ImmutableList.copyOf(metadata.getColumnHandles(session, tableHandle).values());
        // verify local sorting property
        ConnectorTableProperties properties = metadata.getTableProperties(newSession(ImmutableMap.of("propagate_table_scan_sorting_properties", true, "bucket_execution_enabled", false)), tableHandle);
        Map<String, Integer> columnIndex = indexColumns(columnHandles);
        assertEquals(properties.getLocalProperties(), ImmutableList.of(new SortingProperty<>(columnHandles.get(columnIndex.get("value_asc")), ASC_NULLS_FIRST), new SortingProperty<>(columnHandles.get(columnIndex.get("value_desc")), DESC_NULLS_LAST)));
        assertThat(metadata.getTableProperties(newSession(), tableHandle).getLocalProperties()).isEmpty();
        List<ConnectorSplit> splits = getAllSplits(tableHandle, transaction, session);
        assertThat(splits).hasSize(bucketCount);
        int actualRowCount = 0;
        for (ConnectorSplit split : splits) {
            try (ConnectorPageSource pageSource = pageSourceProvider.createPageSource(transaction.getTransactionHandle(), session, split, tableHandle, columnHandles, DynamicFilter.EMPTY)) {
                String lastValueAsc = null;
                long lastValueDesc = -1;
                while (!pageSource.isFinished()) {
                    Page page = pageSource.getNextPage();
                    if (page == null) {
                        continue;
                    }
                    for (int i = 0; i < page.getPositionCount(); i++) {
                        Block blockAsc = page.getBlock(1);
                        Block blockDesc = page.getBlock(2);
                        assertFalse(blockAsc.isNull(i));
                        assertFalse(blockDesc.isNull(i));
                        String valueAsc = VARCHAR.getSlice(blockAsc, i).toStringUtf8();
                        if (lastValueAsc != null) {
                            assertGreaterThanOrEqual(valueAsc, lastValueAsc);
                            if (valueAsc.equals(lastValueAsc)) {
                                long valueDesc = BIGINT.getLong(blockDesc, i);
                                if (lastValueDesc != -1) {
                                    assertLessThanOrEqual(valueDesc, lastValueDesc);
                                }
                                lastValueDesc = valueDesc;
                            } else {
                                lastValueDesc = -1;
                            }
                        }
                        lastValueAsc = valueAsc;
                        actualRowCount++;
                    }
                }
            }
        }
        assertThat(actualRowCount).isEqualTo(expectedRowCount);
    }
}