Examples with ConnectorSession io.prestosql.spi.connector.ConnectorSession used on opensource projects

Example 1 with ConnectorSession

use of io.prestosql.spi.connector.ConnectorSession in project hetu-core by openlookeng.

the class TpchIndexMetadata method resolveIndex.

@Override
public Optional<ConnectorResolvedIndex> resolveIndex(ConnectorSession session, ConnectorTableHandle tableHandle, Set<ColumnHandle> indexableColumns, Set<ColumnHandle> outputColumns, TupleDomain<ColumnHandle> tupleDomain) {
    TpchTableHandle tpchTableHandle = (TpchTableHandle) tableHandle;
    // Keep the fixed values that don't overlap with the indexableColumns
    // Note: technically we could more efficiently utilize the overlapped columns, but this way is simpler for now
    Map<ColumnHandle, NullableValue> fixedValues = TupleDomain.extractFixedValues(tupleDomain).orElse(ImmutableMap.of()).entrySet().stream().filter(entry -> !indexableColumns.contains(entry.getKey())).filter(// strip nulls since meaningless in index join lookups
    entry -> !entry.getValue().isNull()).collect(Collectors.toMap(Map.Entry::getKey, Map.Entry::getValue));
    // determine all columns available for index lookup
    Set<String> lookupColumnNames = ImmutableSet.<String>builder().addAll(handleToNames(ImmutableList.copyOf(indexableColumns))).addAll(handleToNames(ImmutableList.copyOf(fixedValues.keySet()))).build();
    // do we have an index?
    if (!indexedData.getIndexedTable(tpchTableHandle.getTableName(), tpchTableHandle.getScaleFactor(), lookupColumnNames).isPresent()) {
        return Optional.empty();
    }
    TupleDomain<ColumnHandle> filteredTupleDomain = tupleDomain;
    if (!tupleDomain.isNone()) {
        filteredTupleDomain = TupleDomain.withColumnDomains(Maps.filterKeys(tupleDomain.getDomains().get(), not(in(fixedValues.keySet()))));
    }
    TpchIndexHandle indexHandle = new TpchIndexHandle(tpchTableHandle.getTableName(), tpchTableHandle.getScaleFactor(), lookupColumnNames, TupleDomain.fromFixedValues(fixedValues));
    return Optional.of(new ConnectorResolvedIndex(indexHandle, filteredTupleDomain));
}

Example 2 with ConnectorSession

use of io.prestosql.spi.connector.ConnectorSession in project hetu-core by openlookeng.

the class CarbondataPageSourceProvider method createPageSource.

@Override
public ConnectorPageSource createPageSource(ConnectorTransactionHandle transactionHandle, ConnectorSession session, ConnectorSplit split, ConnectorTableHandle table, List<ColumnHandle> columns) {
    HiveSplit carbonSplit = Types.checkType(((HiveSplitWrapper) (split)).getSplits().get(0), HiveSplit.class, "split is not class HiveSplit");
    this.queryId = carbonSplit.getSchema().getProperty("queryId");
    if (this.queryId == null) {
        // Fall back to hive pagesource.
        return super.createPageSource(transactionHandle, session, split, table, columns);
    }
    try {
        hdfsEnvironment.getFileSystem(new HdfsEnvironment.HdfsContext(session, carbonSplit.getDatabase()), new Path(carbonSplit.getSchema().getProperty("tablePath")));
    } catch (IOException e) {
        throw new PrestoException(GENERIC_INTERNAL_ERROR, "Failed to get file system: " + e.getMessage());
    }
    return hdfsEnvironment.doAs(session.getUser(), () -> {
        Configuration configuration = this.hdfsEnvironment.getConfiguration(new HdfsEnvironment.HdfsContext(session, carbonSplit.getDatabase(), carbonSplit.getTable()), new Path(carbonSplit.getSchema().getProperty("tablePath")));
        CarbonTable carbonTable = getCarbonTable(carbonSplit, configuration);
        /* So that CarbonTLS can access it */
        ThreadLocalSessionInfo.setConfigurationToCurrentThread(configuration);
        boolean isFullACID = isFullAcidTable(Maps.fromProperties(carbonSplit.getSchema()));
        boolean isDirectVectorFill = (carbonTableReader.config.getPushRowFilter() == null) || carbonTableReader.config.getPushRowFilter().equalsIgnoreCase("false") || columns.stream().anyMatch(c -> c.getColumnName().equalsIgnoreCase(CarbonCommonConstants.CARBON_IMPLICIT_COLUMN_TUPLEID));
        return new CarbondataPageSource(carbonTable, queryId, carbonSplit, columns, table, configuration, isDirectVectorFill, isFullACID, session.getUser(), hdfsEnvironment);
    });
}

Example 3 with ConnectorSession

use of io.prestosql.spi.connector.ConnectorSession in project hetu-core by openlookeng.

the class RcFileFileWriterFactory method createFileWriter.

@Override
public Optional<HiveFileWriter> createFileWriter(Path path, List<String> inputColumnNames, StorageFormat storageFormat, Properties schema, JobConf configuration, ConnectorSession session, Optional<AcidOutputFormat.Options> acidOptions, Optional<HiveACIDWriteType> acidWriteType) {
    if (!RCFileOutputFormat.class.getName().equals(storageFormat.getOutputFormat())) {
        return Optional.empty();
    }
    RcFileEncoding rcFileEncoding;
    if (LazyBinaryColumnarSerDe.class.getName().equals(storageFormat.getSerDe())) {
        rcFileEncoding = new BinaryRcFileEncoding(timeZone);
    } else if (ColumnarSerDe.class.getName().equals(storageFormat.getSerDe())) {
        rcFileEncoding = RcFilePageSourceFactory.createTextVectorEncoding(schema);
    } else {
        return Optional.empty();
    }
    Optional<String> codecName = Optional.ofNullable(configuration.get(FileOutputFormat.COMPRESS_CODEC));
    // existing tables and partitions may have columns in a different order than the writer is providing, so build
    // an index to rearrange columns in the proper order
    List<String> fileColumnNames = getColumnNames(schema);
    List<Type> fileColumnTypes = getColumnTypes(schema).stream().map(hiveType -> hiveType.getType(typeManager)).collect(toList());
    int[] fileInputColumnIndexes = fileColumnNames.stream().mapToInt(inputColumnNames::indexOf).toArray();
    try {
        FileSystem fileSystem = hdfsEnvironment.getFileSystem(session.getUser(), path, configuration);
        OutputStream outputStream = fileSystem.create(path);
        Optional<Supplier<RcFileDataSource>> validationInputFactory = Optional.empty();
        if (HiveSessionProperties.isRcfileOptimizedWriterValidate(session)) {
            validationInputFactory = Optional.of(() -> {
                try {
                    return new HdfsRcFileDataSource(path.toString(), fileSystem.open(path), fileSystem.getFileStatus(path).getLen(), stats);
                } catch (IOException e) {
                    throw new PrestoException(HiveErrorCode.HIVE_WRITE_VALIDATION_FAILED, e);
                }
            });
        }
        Callable<Void> rollbackAction = () -> {
            fileSystem.delete(path, false);
            return null;
        };
        return Optional.of(new RcFileFileWriter(outputStream, rollbackAction, rcFileEncoding, fileColumnTypes, codecName, fileInputColumnIndexes, ImmutableMap.<String, String>builder().put(HiveMetadata.PRESTO_VERSION_NAME, nodeVersion.toString()).put(HiveMetadata.PRESTO_QUERY_ID_NAME, session.getQueryId()).build(), validationInputFactory));
    } catch (Exception e) {
        throw new PrestoException(HiveErrorCode.HIVE_WRITER_OPEN_ERROR, "Error creating RCFile file", e);
    }
}

Example 4 with ConnectorSession

use of io.prestosql.spi.connector.ConnectorSession in project hetu-core by openlookeng.

the class OrcFileWriterFactory method createFileWriter.

@Override
public Optional<HiveFileWriter> createFileWriter(Path path, List<String> inputColumnNames, StorageFormat storageFormat, Properties schema, JobConf configuration, ConnectorSession session, Optional<AcidOutputFormat.Options> acidOptions, Optional<HiveACIDWriteType> acidWriteType) {
    if (!OrcOutputFormat.class.getName().equals(storageFormat.getOutputFormat())) {
        return Optional.empty();
    }
    CompressionKind compression = getCompression(schema, configuration);
    // existing tables and partitions may have columns in a different order than the writer is providing, so build
    // an index to rearrange columns in the proper order
    List<String> fileColumnNames = getColumnNames(schema);
    List<Type> fileColumnTypes = getColumnTypes(schema).stream().map(hiveType -> hiveType.getType(typeManager)).collect(toList());
    List<Type> dataFileColumnTypes = fileColumnTypes;
    int[] fileInputColumnIndexes = fileColumnNames.stream().mapToInt(inputColumnNames::indexOf).toArray();
    Optional<HiveFileWriter> deleteDeltaWriter = Optional.empty();
    if (AcidUtils.isTablePropertyTransactional(schema) && !AcidUtils.isInsertOnlyTable(schema)) {
        ImmutableList<String> orcFileColumnNames = ImmutableList.of(OrcPageSourceFactory.ACID_COLUMN_OPERATION, OrcPageSourceFactory.ACID_COLUMN_ORIGINAL_TRANSACTION, OrcPageSourceFactory.ACID_COLUMN_BUCKET, OrcPageSourceFactory.ACID_COLUMN_ROW_ID, OrcPageSourceFactory.ACID_COLUMN_CURRENT_TRANSACTION, OrcPageSourceFactory.ACID_COLUMN_ROW_STRUCT);
        ImmutableList.Builder<RowType.Field> fieldsBuilder = ImmutableList.builder();
        for (int i = 0; i < fileColumnNames.size(); i++) {
            fieldsBuilder.add(new RowType.Field(Optional.of(fileColumnNames.get(i)), fileColumnTypes.get(i)));
        }
        ImmutableList<Type> orcFileColumnTypes = ImmutableList.of(INTEGER, BIGINT, INTEGER, BIGINT, BIGINT, RowType.from(fieldsBuilder.build()));
        fileColumnNames = orcFileColumnNames;
        fileColumnTypes = orcFileColumnTypes;
        if (acidWriteType.isPresent() && acidWriteType.get() == HiveACIDWriteType.UPDATE) {
            AcidOutputFormat.Options deleteOptions = acidOptions.get().clone().writingDeleteDelta(true);
            Path deleteDeltaPath = AcidUtils.createFilename(path.getParent().getParent(), deleteOptions);
            deleteDeltaWriter = createFileWriter(deleteDeltaPath, inputColumnNames, storageFormat, schema, configuration, session, Optional.of(deleteOptions), Optional.of(HiveACIDWriteType.DELETE));
        }
    }
    try {
        FileSystem fileSystem = hdfsEnvironment.getFileSystem(session.getUser(), path, configuration);
        OrcDataSink orcDataSink = createOrcDataSink(session, fileSystem, path);
        Optional<Supplier<OrcDataSource>> validationInputFactory = Optional.empty();
        if (HiveSessionProperties.isOrcOptimizedWriterValidate(session)) {
            validationInputFactory = Optional.of(() -> {
                try {
                    FileStatus fileStatus = fileSystem.getFileStatus(path);
                    return new HdfsOrcDataSource(new OrcDataSourceId(path.toString()), fileStatus.getLen(), HiveSessionProperties.getOrcMaxMergeDistance(session), HiveSessionProperties.getOrcMaxBufferSize(session), HiveSessionProperties.getOrcStreamBufferSize(session), false, fileSystem.open(path), readStats, fileStatus.getModificationTime());
                } catch (IOException e) {
                    throw new PrestoException(HiveErrorCode.HIVE_WRITE_VALIDATION_FAILED, e);
                }
            });
        }
        Callable<Void> rollbackAction = () -> {
            fileSystem.delete(path, false);
            return null;
        };
        return Optional.of(new OrcFileWriter(orcDataSink, rollbackAction, fileColumnNames, fileColumnTypes, dataFileColumnTypes, compression, orcWriterOptions.withStripeMinSize(HiveSessionProperties.getOrcOptimizedWriterMinStripeSize(session)).withStripeMaxSize(HiveSessionProperties.getOrcOptimizedWriterMaxStripeSize(session)).withStripeMaxRowCount(HiveSessionProperties.getOrcOptimizedWriterMaxStripeRows(session)).withDictionaryMaxMemory(HiveSessionProperties.getOrcOptimizedWriterMaxDictionaryMemory(session)).withMaxStringStatisticsLimit(HiveSessionProperties.getOrcStringStatisticsLimit(session)), writeLegacyVersion, fileInputColumnIndexes, ImmutableMap.<String, String>builder().put(HiveMetadata.PRESTO_VERSION_NAME, nodeVersion.toString()).put(HiveMetadata.PRESTO_QUERY_ID_NAME, session.getQueryId()).put("hive.acid.version", String.valueOf(AcidUtils.OrcAcidVersion.ORC_ACID_VERSION)).build(), validationInputFactory, HiveSessionProperties.getOrcOptimizedWriterValidateMode(session), stats, acidOptions, acidWriteType, deleteDeltaWriter, path));
    } catch (IOException e) {
        throw new PrestoException(HiveErrorCode.HIVE_WRITER_OPEN_ERROR, "Error creating ORC file", e);
    }
}

Example 5 with ConnectorSession

use of io.prestosql.spi.connector.ConnectorSession in project hetu-core by openlookeng.

the class SyncPartitionMetadataProcedure method doSyncPartitionMetadata.

private void doSyncPartitionMetadata(ConnectorSession session, String schemaName, String tableName, String mode) {
    SyncMode syncMode = toSyncMode(mode);
    HdfsContext hdfsContext = new HdfsContext(session, schemaName, tableName);
    HiveIdentity identity = new HiveIdentity(session);
    SemiTransactionalHiveMetastore metastore = ((HiveMetadata) hiveMetadataFactory.get()).getMetastore();
    SchemaTableName schemaTableName = new SchemaTableName(schemaName, tableName);
    Table table = metastore.getTable(identity, schemaName, tableName).orElseThrow(() -> new TableNotFoundException(schemaTableName));
    if (table.getPartitionColumns().isEmpty()) {
        throw new PrestoException(INVALID_PROCEDURE_ARGUMENT, "Table is not partitioned: " + schemaTableName);
    }
    Path tableLocation = new Path(table.getStorage().getLocation());
    Set<String> partitionsToAdd;
    Set<String> partitionsToDrop;
    try {
        FileSystem fileSystem = hdfsEnvironment.getFileSystem(hdfsContext, tableLocation);
        List<String> partitionsInMetastore = metastore.getPartitionNames(identity, schemaName, tableName).orElseThrow(() -> new TableNotFoundException(schemaTableName));
        List<String> partitionsInFileSystem = listDirectory(fileSystem, fileSystem.getFileStatus(tableLocation), table.getPartitionColumns(), table.getPartitionColumns().size()).stream().map(fileStatus -> fileStatus.getPath().toUri()).map(uri -> tableLocation.toUri().relativize(uri).getPath()).collect(toImmutableList());
        // partitions in file system but not in metastore
        partitionsToAdd = difference(partitionsInFileSystem, partitionsInMetastore);
        // partitions in metastore but not in file system
        partitionsToDrop = difference(partitionsInMetastore, partitionsInFileSystem);
    } catch (IOException e) {
        throw new PrestoException(HIVE_FILESYSTEM_ERROR, e);
    }
    syncPartitions(partitionsToAdd, partitionsToDrop, syncMode, metastore, session, table);
}