Examples with HiveBasicStatistics io.trino.plugin.hive.HiveBasicStatistics used on opensource projects

Example 1 with HiveBasicStatistics

use of io.trino.plugin.hive.HiveBasicStatistics in project trino by trinodb.

the class FileHiveMetastore method getTableStatistics.

private synchronized PartitionStatistics getTableStatistics(String databaseName, String tableName) {
    Path tableMetadataDirectory = getTableMetadataDirectory(databaseName, tableName);
    TableMetadata tableMetadata = readSchemaFile(TABLE, tableMetadataDirectory, tableCodec).orElseThrow(() -> new TableNotFoundException(new SchemaTableName(databaseName, tableName)));
    checkVersion(tableMetadata.getWriterVersion());
    HiveBasicStatistics basicStatistics = getHiveBasicStatistics(tableMetadata.getParameters());
    Map<String, HiveColumnStatistics> columnStatistics = tableMetadata.getColumnStatistics();
    return new PartitionStatistics(basicStatistics, columnStatistics);
}

Example 2 with HiveBasicStatistics

use of io.trino.plugin.hive.HiveBasicStatistics in project trino by trinodb.

the class ThriftMetastoreUtil method getHiveBasicStatistics.

public static HiveBasicStatistics getHiveBasicStatistics(Map<String, String> parameters) {
    OptionalLong numFiles = parse(parameters.get(NUM_FILES));
    OptionalLong numRows = parse(parameters.get(NUM_ROWS));
    OptionalLong inMemoryDataSizeInBytes = parse(parameters.get(RAW_DATA_SIZE));
    OptionalLong onDiskDataSizeInBytes = parse(parameters.get(TOTAL_SIZE));
    return new HiveBasicStatistics(numFiles, numRows, inMemoryDataSizeInBytes, onDiskDataSizeInBytes);
}

Example 3 with HiveBasicStatistics

use of io.trino.plugin.hive.HiveBasicStatistics in project trino by trinodb.

the class ThriftHiveMetastore method getTableStatistics.

@Override
public PartitionStatistics getTableStatistics(HiveIdentity identity, Table table) {
    List<String> dataColumns = table.getSd().getCols().stream().map(FieldSchema::getName).collect(toImmutableList());
    HiveBasicStatistics basicStatistics = getHiveBasicStatistics(table.getParameters());
    Map<String, HiveColumnStatistics> columnStatistics = getTableColumnStatistics(identity, table.getDbName(), table.getTableName(), dataColumns, basicStatistics.getRowCount());
    return new PartitionStatistics(basicStatistics, columnStatistics);
}

Example 4 with HiveBasicStatistics

use of io.trino.plugin.hive.HiveBasicStatistics in project trino by trinodb.

the class MetastoreHiveStatisticsProvider method calculatePartitionsRowCount.

@VisibleForTesting
static Optional<PartitionsRowCount> calculatePartitionsRowCount(Collection<PartitionStatistics> statistics, int queriedPartitionsCount) {
    long[] rowCounts = statistics.stream().map(PartitionStatistics::getBasicStatistics).map(HiveBasicStatistics::getRowCount).filter(OptionalLong::isPresent).mapToLong(OptionalLong::getAsLong).peek(count -> verify(count >= 0, "count must be greater than or equal to zero")).toArray();
    int sampleSize = statistics.size();
    // Sample contains all the queried partitions, estimate avg normally
    if (rowCounts.length <= 2 || queriedPartitionsCount == sampleSize) {
        OptionalDouble averageRowsPerPartitionOptional = Arrays.stream(rowCounts).average();
        if (averageRowsPerPartitionOptional.isEmpty()) {
            return Optional.empty();
        }
        double averageRowsPerPartition = averageRowsPerPartitionOptional.getAsDouble();
        return Optional.of(new PartitionsRowCount(averageRowsPerPartition, averageRowsPerPartition * queriedPartitionsCount));
    }
    // Some partitions (e.g. __HIVE_DEFAULT_PARTITION__) may be outliers in terms of row count.
    // Excluding the min and max rowCount values from averageRowsPerPartition calculation helps to reduce the
    // possibility of errors in the extrapolated rowCount due to a couple of outliers.
    int minIndex = 0;
    int maxIndex = 0;
    long rowCountSum = rowCounts[0];
    for (int index = 1; index < rowCounts.length; index++) {
        if (rowCounts[index] < rowCounts[minIndex]) {
            minIndex = index;
        } else if (rowCounts[index] > rowCounts[maxIndex]) {
            maxIndex = index;
        }
        rowCountSum += rowCounts[index];
    }
    double averageWithoutOutliers = ((double) (rowCountSum - rowCounts[minIndex] - rowCounts[maxIndex])) / (rowCounts.length - 2);
    double rowCount = (averageWithoutOutliers * (queriedPartitionsCount - 2)) + rowCounts[minIndex] + rowCounts[maxIndex];
    return Optional.of(new PartitionsRowCount(averageWithoutOutliers, rowCount));
}

Example 5 with HiveBasicStatistics

use of io.trino.plugin.hive.HiveBasicStatistics in project trino by trinodb.

the class TestStatistics method testReduce.

@Test
public void testReduce() {
    assertThat(reduce(createEmptyStatistics(), createEmptyStatistics(), ADD)).isEqualTo(createEmptyStatistics());
    assertThat(reduce(createZeroStatistics(), createEmptyStatistics(), ADD)).isEqualTo(createEmptyStatistics());
    assertThat(reduce(createEmptyStatistics(), createZeroStatistics(), ADD)).isEqualTo(createEmptyStatistics());
    assertThat(reduce(createEmptyStatistics(), createEmptyStatistics(), SUBTRACT)).isEqualTo(createEmptyStatistics());
    assertThat(reduce(createZeroStatistics(), createEmptyStatistics(), SUBTRACT)).isEqualTo(createEmptyStatistics());
    assertThat(reduce(createEmptyStatistics(), createZeroStatistics(), SUBTRACT)).isEqualTo(createEmptyStatistics());
    assertThat(reduce(new HiveBasicStatistics(11, 9, 7, 5), new HiveBasicStatistics(1, 2, 3, 4), ADD)).isEqualTo(new HiveBasicStatistics(12, 11, 10, 9));
    assertThat(reduce(new HiveBasicStatistics(11, 9, 7, 5), new HiveBasicStatistics(1, 2, 3, 4), SUBTRACT)).isEqualTo(new HiveBasicStatistics(10, 7, 4, 1));
}