Examples with ConnectorPageSource io.trino.spi.connector.ConnectorPageSource used on opensource projects

Example 36 with ConnectorPageSource

use of io.trino.spi.connector.ConnectorPageSource in project trino by trinodb.

the class OrcPageSourceFactory method createPageSource.

@Override
public Optional<ReaderPageSource> createPageSource(Configuration configuration, ConnectorSession session, Path path, long start, long length, long estimatedFileSize, Properties schema, List<HiveColumnHandle> columns, TupleDomain<HiveColumnHandle> effectivePredicate, Optional<AcidInfo> acidInfo, OptionalInt bucketNumber, boolean originalFile, AcidTransaction transaction) {
    if (!isDeserializerClass(schema, OrcSerde.class)) {
        return Optional.empty();
    }
    List<HiveColumnHandle> readerColumnHandles = columns;
    Optional<ReaderColumns> readerColumns = projectBaseColumns(columns);
    if (readerColumns.isPresent()) {
        readerColumnHandles = readerColumns.get().get().stream().map(HiveColumnHandle.class::cast).collect(toUnmodifiableList());
    }
    ConnectorPageSource orcPageSource = createOrcPageSource(hdfsEnvironment, session.getIdentity(), configuration, path, start, length, estimatedFileSize, readerColumnHandles, columns, isUseOrcColumnNames(session), isFullAcidTable(Maps.fromProperties(schema)), effectivePredicate, legacyTimeZone, orcReaderOptions.withMaxMergeDistance(getOrcMaxMergeDistance(session)).withMaxBufferSize(getOrcMaxBufferSize(session)).withStreamBufferSize(getOrcStreamBufferSize(session)).withTinyStripeThreshold(getOrcTinyStripeThreshold(session)).withMaxReadBlockSize(getOrcMaxReadBlockSize(session)).withLazyReadSmallRanges(getOrcLazyReadSmallRanges(session)).withNestedLazy(isOrcNestedLazy(session)).withBloomFiltersEnabled(isOrcBloomFiltersEnabled(session)), acidInfo, bucketNumber, originalFile, transaction, stats);
    return Optional.of(new ReaderPageSource(orcPageSource, readerColumns));
}

Example 37 with ConnectorPageSource

use of io.trino.spi.connector.ConnectorPageSource in project trino by trinodb.

the class ParquetTester method assertMaxReadBytes.

void assertMaxReadBytes(List<ObjectInspector> objectInspectors, Iterable<?>[] writeValues, Iterable<?>[] readValues, List<String> columnNames, List<Type> columnTypes, Optional<MessageType> parquetSchema, DataSize maxReadBlockSize) throws Exception {
    CompressionCodecName compressionCodecName = UNCOMPRESSED;
    HiveSessionProperties hiveSessionProperties = new HiveSessionProperties(new HiveConfig().setHiveStorageFormat(HiveStorageFormat.PARQUET).setUseParquetColumnNames(false), new OrcReaderConfig(), new OrcWriterConfig(), new ParquetReaderConfig().setMaxReadBlockSize(maxReadBlockSize), new ParquetWriterConfig());
    ConnectorSession session = TestingConnectorSession.builder().setPropertyMetadata(hiveSessionProperties.getSessionProperties()).build();
    try (TempFile tempFile = new TempFile("test", "parquet")) {
        JobConf jobConf = new JobConf();
        jobConf.setEnum(COMPRESSION, compressionCodecName);
        jobConf.setBoolean(ENABLE_DICTIONARY, true);
        jobConf.setEnum(WRITER_VERSION, PARQUET_1_0);
        writeParquetColumn(jobConf, tempFile.getFile(), compressionCodecName, createTableProperties(columnNames, objectInspectors), getStandardStructObjectInspector(columnNames, objectInspectors), getIterators(writeValues), parquetSchema, false);
        Iterator<?>[] expectedValues = getIterators(readValues);
        try (ConnectorPageSource pageSource = fileFormat.createFileFormatReader(session, HDFS_ENVIRONMENT, tempFile.getFile(), columnNames, columnTypes)) {
            assertPageSource(columnTypes, expectedValues, pageSource, Optional.of(getParquetMaxReadBlockSize(session).toBytes()));
            assertFalse(stream(expectedValues).allMatch(Iterator::hasNext));
        }
    }
}

Example 38 with ConnectorPageSource

use of io.trino.spi.connector.ConnectorPageSource in project trino by trinodb.

the class DeltaLakePageSourceProvider method createPageSource.

@Override
public ConnectorPageSource createPageSource(ConnectorTransactionHandle transaction, ConnectorSession session, ConnectorSplit connectorSplit, ConnectorTableHandle connectorTable, List<ColumnHandle> columns, DynamicFilter dynamicFilter) {
    DeltaLakeSplit split = (DeltaLakeSplit) connectorSplit;
    DeltaLakeTableHandle table = (DeltaLakeTableHandle) connectorTable;
    // We reach here when we could not prune the split using file level stats, table predicate
    // and the dynamic filter in the coordinator during split generation. The file level stats
    // in DeltaLakeSplit#filePredicate could help to prune this split when a more selective dynamic filter
    // is available now, without having to access parquet file footer for row-group stats.
    // We avoid sending DeltaLakeSplit#splitPredicate to workers by using table.getPredicate() here.
    TupleDomain<DeltaLakeColumnHandle> filteredSplitPredicate = TupleDomain.intersect(ImmutableList.of(table.getNonPartitionConstraint(), split.getStatisticsPredicate(), dynamicFilter.getCurrentPredicate().transformKeys(DeltaLakeColumnHandle.class::cast)));
    if (filteredSplitPredicate.isNone()) {
        return new EmptyPageSource();
    }
    List<DeltaLakeColumnHandle> deltaLakeColumns = columns.stream().map(DeltaLakeColumnHandle.class::cast).collect(toImmutableList());
    Map<String, Optional<String>> partitionKeys = split.getPartitionKeys();
    List<DeltaLakeColumnHandle> regularColumns = deltaLakeColumns.stream().filter(column -> column.getColumnType() == REGULAR).collect(toImmutableList());
    List<HiveColumnHandle> hiveColumnHandles = regularColumns.stream().map(DeltaLakeColumnHandle::toHiveColumnHandle).collect(toImmutableList());
    Path path = new Path(split.getPath());
    HdfsContext hdfsContext = new HdfsContext(session);
    TupleDomain<HiveColumnHandle> parquetPredicate = getParquetTupleDomain(filteredSplitPredicate.simplify(domainCompactionThreshold));
    if (table.getWriteType().isPresent()) {
        return new DeltaLakeUpdatablePageSource(table, deltaLakeColumns, partitionKeys, split.getPath(), split.getFileSize(), split.getFileModifiedTime(), session, executorService, hdfsEnvironment, hdfsContext, parquetDateTimeZone, parquetReaderOptions, parquetPredicate, typeManager, updateResultJsonCodec);
    }
    ReaderPageSource pageSource = ParquetPageSourceFactory.createPageSource(path, split.getStart(), split.getLength(), split.getFileSize(), hiveColumnHandles, parquetPredicate, true, hdfsEnvironment, hdfsEnvironment.getConfiguration(hdfsContext, path), session.getIdentity(), parquetDateTimeZone, fileFormatDataSourceStats, parquetReaderOptions.withMaxReadBlockSize(getParquetMaxReadBlockSize(session)).withUseColumnIndex(isParquetUseColumnIndex(session)));
    verify(pageSource.getReaderColumns().isEmpty(), "All columns expected to be base columns");
    return new DeltaLakePageSource(deltaLakeColumns, partitionKeys, pageSource.get(), split.getPath(), split.getFileSize(), split.getFileModifiedTime());
}

Example 39 with ConnectorPageSource

use of io.trino.spi.connector.ConnectorPageSource in project trino by trinodb.

the class TestHivePageSink method writeTestFile.

private static long writeTestFile(HiveConfig config, HiveMetastore metastore, String outputPath) {
    HiveTransactionHandle transaction = new HiveTransactionHandle(false);
    HiveWriterStats stats = new HiveWriterStats();
    ConnectorPageSink pageSink = createPageSink(transaction, config, metastore, new Path("file:///" + outputPath), stats);
    List<LineItemColumn> columns = getTestColumns();
    List<Type> columnTypes = columns.stream().map(LineItemColumn::getType).map(TestHivePageSink::getHiveType).map(hiveType -> hiveType.getType(TESTING_TYPE_MANAGER)).collect(toList());
    PageBuilder pageBuilder = new PageBuilder(columnTypes);
    int rows = 0;
    for (LineItem lineItem : new LineItemGenerator(0.01, 1, 1)) {
        rows++;
        if (rows >= NUM_ROWS) {
            break;
        }
        pageBuilder.declarePosition();
        for (int i = 0; i < columns.size(); i++) {
            LineItemColumn column = columns.get(i);
            BlockBuilder blockBuilder = pageBuilder.getBlockBuilder(i);
            switch(column.getType().getBase()) {
                case IDENTIFIER:
                    BIGINT.writeLong(blockBuilder, column.getIdentifier(lineItem));
                    break;
                case INTEGER:
                    INTEGER.writeLong(blockBuilder, column.getInteger(lineItem));
                    break;
                case DATE:
                    DATE.writeLong(blockBuilder, column.getDate(lineItem));
                    break;
                case DOUBLE:
                    DOUBLE.writeDouble(blockBuilder, column.getDouble(lineItem));
                    break;
                case VARCHAR:
                    createUnboundedVarcharType().writeSlice(blockBuilder, Slices.utf8Slice(column.getString(lineItem)));
                    break;
                default:
                    throw new IllegalArgumentException("Unsupported type " + column.getType());
            }
        }
    }
    Page page = pageBuilder.build();
    pageSink.appendPage(page);
    getFutureValue(pageSink.finish());
    File outputDir = new File(outputPath);
    List<File> files = ImmutableList.copyOf(outputDir.listFiles((dir, name) -> !name.endsWith(".crc")));
    File outputFile = getOnlyElement(files);
    long length = outputFile.length();
    ConnectorPageSource pageSource = createPageSource(transaction, config, outputFile);
    List<Page> pages = new ArrayList<>();
    while (!pageSource.isFinished()) {
        Page nextPage = pageSource.getNextPage();
        if (nextPage != null) {
            pages.add(nextPage.getLoadedPage());
        }
    }
    MaterializedResult expectedResults = toMaterializedResult(getHiveSession(config), columnTypes, ImmutableList.of(page));
    MaterializedResult results = toMaterializedResult(getHiveSession(config), columnTypes, pages);
    assertEquals(results, expectedResults);
    assertEquals(round(stats.getInputPageSizeInBytes().getAllTime().getMax()), page.getRetainedSizeInBytes());
    return length;
}

Example 40 with ConnectorPageSource

use of io.trino.spi.connector.ConnectorPageSource in project trino by trinodb.

the class TestNodeLocalDynamicSplitPruning method testDynamicPartitionPruning.

@Test
public void testDynamicPartitionPruning() throws IOException {
    HiveConfig config = new HiveConfig();
    HiveTransactionHandle transaction = new HiveTransactionHandle(false);
    try (TempFile tempFile = new TempFile()) {
        ConnectorPageSource emptyPageSource = createTestingPageSource(transaction, config, tempFile.file(), getDynamicFilter(getTupleDomainForPartitionSplitPruning()));
        assertEquals(emptyPageSource.getClass(), EmptyPageSource.class);
        ConnectorPageSource nonEmptyPageSource = createTestingPageSource(transaction, config, tempFile.file(), getDynamicFilter(getNonSelectivePartitionTupleDomain()));
        assertEquals(nonEmptyPageSource.getClass(), HivePageSource.class);
    }
}