Examples with CarbonLocalMultiBlockSplit org.apache.carbondata.presto.impl.CarbonLocalMultiBlockSplit used on opensource projects

Example 1 with CarbonLocalMultiBlockSplit

use of org.apache.carbondata.presto.impl.CarbonLocalMultiBlockSplit in project carbondata by apache.

the class CarbondataSplitManager method getSplits.

@Override
public ConnectorSplitSource getSplits(ConnectorTransactionHandle transactionHandle, ConnectorSession session, ConnectorTableHandle tableHandle, SplitSchedulingStrategy splitSchedulingStrategy) {
    HiveTableHandle hiveTableHandle = (HiveTableHandle) tableHandle;
    SchemaTableName schemaTableName = hiveTableHandle.getSchemaTableName();
    carbonTableReader.setPrestoQueryId(session.getQueryId());
    // get table metadata
    SemiTransactionalHiveMetastore metastore = metastoreProvider.apply((HiveTransactionHandle) transactionHandle);
    Table table = metastore.getTable(new HiveIdentity(session), schemaTableName.getSchemaName(), schemaTableName.getTableName()).orElseThrow(() -> new TableNotFoundException(schemaTableName));
    if (!table.getStorage().getStorageFormat().getInputFormat().contains("carbon")) {
        return super.getSplits(transactionHandle, session, tableHandle, splitSchedulingStrategy);
    }
    // for hive metastore, get table location from catalog table's tablePath
    String location = table.getStorage().getSerdeParameters().get("tablePath");
    if (StringUtils.isEmpty(location)) {
        // file metastore case tablePath can be null, so get from location
        location = table.getStorage().getLocation();
    }
    List<PartitionSpec> filteredPartitions = new ArrayList<>();
    if (hiveTableHandle.getPartitionColumns().size() > 0 && hiveTableHandle.getPartitions().isPresent()) {
        List<String> colNames = hiveTableHandle.getPartitionColumns().stream().map(HiveColumnHandle::getName).collect(Collectors.toList());
        for (HivePartition partition : hiveTableHandle.getPartitions().get()) {
            filteredPartitions.add(new PartitionSpec(colNames, location + CarbonCommonConstants.FILE_SEPARATOR + partition.getPartitionId()));
        }
    }
    String queryId = System.nanoTime() + "";
    QueryStatistic statistic = new QueryStatistic();
    QueryStatisticsRecorder statisticRecorder = CarbonTimeStatisticsFactory.createDriverRecorder();
    statistic.addStatistics(QueryStatisticsConstants.BLOCK_ALLOCATION, System.currentTimeMillis());
    statisticRecorder.recordStatisticsForDriver(statistic, queryId);
    statistic = new QueryStatistic();
    carbonTableReader.setQueryId(queryId);
    TupleDomain<HiveColumnHandle> predicate = hiveTableHandle.getCompactEffectivePredicate();
    Configuration configuration = this.hdfsEnvironment.getConfiguration(new HdfsEnvironment.HdfsContext(session, schemaTableName.getSchemaName(), schemaTableName.getTableName()), new Path(location));
    configuration = carbonTableReader.updateS3Properties(configuration);
    for (Map.Entry<String, String> entry : table.getStorage().getSerdeParameters().entrySet()) {
        configuration.set(entry.getKey(), entry.getValue());
    }
    // set the hadoop configuration to thread local, so that FileFactory can use it.
    ThreadLocalSessionInfo.setConfigurationToCurrentThread(configuration);
    CarbonTableCacheModel cache = carbonTableReader.getCarbonCache(schemaTableName, location, configuration);
    Expression filters = PrestoFilterUtil.parseFilterExpression(predicate);
    try {
        List<CarbonLocalMultiBlockSplit> splits = carbonTableReader.getInputSplits(cache, filters, filteredPartitions, configuration);
        ImmutableList.Builder<ConnectorSplit> cSplits = ImmutableList.builder();
        long index = 0;
        for (CarbonLocalMultiBlockSplit split : splits) {
            index++;
            Properties properties = new Properties();
            for (Map.Entry<String, String> entry : table.getStorage().getSerdeParameters().entrySet()) {
                properties.setProperty(entry.getKey(), entry.getValue());
            }
            properties.setProperty("tablePath", cache.getCarbonTable().getTablePath());
            properties.setProperty("carbonSplit", split.getJsonString());
            properties.setProperty("queryId", queryId);
            properties.setProperty("index", String.valueOf(index));
            cSplits.add(new HiveSplit(schemaTableName.getSchemaName(), schemaTableName.getTableName(), schemaTableName.getTableName(), cache.getCarbonTable().getTablePath(), 0, 0, 0, 0, properties, new ArrayList<>(), getHostAddresses(split.getLocations()), OptionalInt.empty(), false, TableToPartitionMapping.empty(), Optional.empty(), false, Optional.empty()));
        }
        statisticRecorder.logStatisticsAsTableDriver();
        statistic.addStatistics(QueryStatisticsConstants.BLOCK_IDENTIFICATION, System.currentTimeMillis());
        statisticRecorder.recordStatisticsForDriver(statistic, queryId);
        statisticRecorder.logStatisticsAsTableDriver();
        return new FixedSplitSource(cSplits.build());
    } catch (Exception ex) {
        throw new RuntimeException(ex.getMessage(), ex);
    }
}

Example 2 with CarbonLocalMultiBlockSplit

use of org.apache.carbondata.presto.impl.CarbonLocalMultiBlockSplit in project carbondata by apache.

the class CarbondataSplitManager method getSplits.

public ConnectorSplitSource getSplits(ConnectorTransactionHandle transactionHandle, ConnectorSession session, ConnectorTableLayoutHandle layoutHandle, SplitSchedulingStrategy splitSchedulingStrategy) {
    HiveTableLayoutHandle layout = (HiveTableLayoutHandle) layoutHandle;
    SchemaTableName schemaTableName = layout.getSchemaTableName();
    carbonTableReader.setPrestoQueryId(session.getQueryId());
    // get table metadata
    SemiTransactionalHiveMetastore metastore = metastoreProvider.apply((HiveTransactionHandle) transactionHandle);
    Table table = metastore.getTable(schemaTableName.getSchemaName(), schemaTableName.getTableName()).orElseThrow(() -> new TableNotFoundException(schemaTableName));
    if (!table.getStorage().getStorageFormat().getInputFormat().contains("carbon")) {
        return super.getSplits(transactionHandle, session, layoutHandle, splitSchedulingStrategy);
    }
    // for hive metastore, get table location from catalog table's tablePath
    String location = table.getStorage().getSerdeParameters().get("tablePath");
    if (StringUtils.isEmpty(location)) {
        // file metastore case tablePath can be null, so get from location
        location = table.getStorage().getLocation();
    }
    List<PartitionSpec> filteredPartitions = new ArrayList<>();
    if (layout.getPartitionColumns().size() > 0 && layout.getPartitions().isPresent()) {
        List<String> colNames = layout.getPartitionColumns().stream().map(x -> ((HiveColumnHandle) x).getName()).collect(Collectors.toList());
        for (HivePartition partition : layout.getPartitions().get()) {
            filteredPartitions.add(new PartitionSpec(colNames, location + CarbonCommonConstants.FILE_SEPARATOR + partition.getPartitionId()));
        }
    }
    String queryId = System.nanoTime() + "";
    QueryStatistic statistic = new QueryStatistic();
    QueryStatisticsRecorder statisticRecorder = CarbonTimeStatisticsFactory.createDriverRecorder();
    statistic.addStatistics(QueryStatisticsConstants.BLOCK_ALLOCATION, System.currentTimeMillis());
    statisticRecorder.recordStatisticsForDriver(statistic, queryId);
    statistic = new QueryStatistic();
    carbonTableReader.setQueryId(queryId);
    TupleDomain<HiveColumnHandle> predicate = (TupleDomain<HiveColumnHandle>) layout.getCompactEffectivePredicate();
    Configuration configuration = this.hdfsEnvironment.getConfiguration(new HdfsEnvironment.HdfsContext(session, schemaTableName.getSchemaName(), schemaTableName.getTableName()), new Path(location));
    configuration = carbonTableReader.updateS3Properties(configuration);
    for (Map.Entry<String, String> entry : table.getStorage().getSerdeParameters().entrySet()) {
        configuration.set(entry.getKey(), entry.getValue());
    }
    // set the hadoop configuration to thread local, so that FileFactory can use it.
    ThreadLocalSessionInfo.setConfigurationToCurrentThread(configuration);
    CarbonTableCacheModel cache = carbonTableReader.getCarbonCache(schemaTableName, location, configuration);
    Expression filters = PrestoFilterUtil.parseFilterExpression(predicate);
    try {
        List<CarbonLocalMultiBlockSplit> splits = carbonTableReader.getInputSplits(cache, filters, filteredPartitions, configuration);
        ImmutableList.Builder<ConnectorSplit> cSplits = ImmutableList.builder();
        long index = 0;
        for (CarbonLocalMultiBlockSplit split : splits) {
            index++;
            Properties properties = new Properties();
            for (Map.Entry<String, String> entry : table.getStorage().getSerdeParameters().entrySet()) {
                properties.setProperty(entry.getKey(), entry.getValue());
            }
            properties.setProperty("tablePath", cache.getCarbonTable().getTablePath());
            properties.setProperty("carbonSplit", split.getJsonString());
            properties.setProperty("queryId", queryId);
            properties.setProperty("index", String.valueOf(index));
            cSplits.add(new HiveSplit(schemaTableName.getSchemaName(), schemaTableName.getTableName(), schemaTableName.getTableName(), "", 0, 0, 0, properties, new ArrayList(), getHostAddresses(split.getLocations()), OptionalInt.empty(), false, predicate, new HashMap<>(), Optional.empty(), false));
        }
        statisticRecorder.logStatisticsAsTableDriver();
        statistic.addStatistics(QueryStatisticsConstants.BLOCK_IDENTIFICATION, System.currentTimeMillis());
        statisticRecorder.recordStatisticsForDriver(statistic, queryId);
        statisticRecorder.logStatisticsAsTableDriver();
        return new FixedSplitSource(cSplits.build());
    } catch (Exception ex) {
        throw new RuntimeException(ex.getMessage(), ex);
    }
}