Java Code Examples for org.apache.kudu.client.CreateTableOptions#setRangePartitionColumns()

public CreateTableOptions getCreateTableOptions() {
    CreateTableOptions options = new CreateTableOptions();
    if(replicas!=null){
        options.setNumReplicas(replicas);
    }
    if(hasColummns()) {
        List<String> rangeKeys = new ArrayList<>();
        List<String> hashKeys = new ArrayList<>();
        for(KuduColumnInfo column : columns){
            if(column.isRangeKey()){
                rangeKeys.add(column.name());
            }
            if(column.isHashKey()){
                hashKeys.add(column.name());
            }
        }
        options.setRangePartitionColumns(rangeKeys);
        options.addHashPartitions(hashKeys, replicas*2);
    }

    return options;
}
 

public CreateTableOptions getCreateTableOptions() {
    CreateTableOptions options = new CreateTableOptions();
    if(replicas!=null){
        options.setNumReplicas(replicas);
    }
    if(hasColummns()) {
        List<String> rangeKeys = new ArrayList<>();
        List<String> hashKeys = new ArrayList<>();
        for(KuduColumnInfo column : columns){
            if(column.isRangeKey()){
                rangeKeys.add(column.name());
            }
            if(column.isHashKey()){
                hashKeys.add(column.name());
            }
        }
        options.setRangePartitionColumns(rangeKeys);
        options.addHashPartitions(hashKeys, replicas*2);
    }

    return options;
}
 

/**
 * The {@code tagsets} table is range partitioned on the {@code id} column.
 * Because the table is essentially a linear probe hash table, it must be able
 * to be scanned in PK order, so hash partitioning is not possible. Since the
 * tagset IDs are effectively random, setting split points at even intervals
 * over the ID range gives good protection against hotspotting.
 * @param options the create options
 * @param numTabletServers the number of tablet servers
 * @return the tagset table create options
 */
static CreateTableOptions tagsetsCreateTableOptions(CreateOptions options,
                                                    int numTabletServers) {
  CreateTableOptions create = new CreateTableOptions();
  create.setNumReplicas(options.getNumReplicas());

  create.setRangePartitionColumns(ImmutableList.of("id"));

  int numTablets = options.getNumTagsetsTablets(numTabletServers);
  long interval = (1L << 32) / numTablets;
  for (int i = 1; i < numTablets; i++) {
    PartialRow split = TAGSETS_SCHEMA.newPartialRow();
    split.addInt(TAGSETS_ID_INDEX, (int) (Integer.MIN_VALUE + i * interval));
    create.addSplitRow(split);
  }
  return create;
}
 

private CreateTableOptions buildCreateTableOptions(Schema schema, Map<String, Object> properties)
{
    CreateTableOptions options = new CreateTableOptions();

    RangePartitionDefinition rangePartitionDefinition = null;
    PartitionDesign partitionDesign = KuduTableProperties.getPartitionDesign(properties);
    if (partitionDesign.getHash() != null) {
        for (HashPartitionDefinition partition : partitionDesign.getHash()) {
            options.addHashPartitions(partition.getColumns(), partition.getBuckets());
        }
    }
    if (partitionDesign.getRange() != null) {
        rangePartitionDefinition = partitionDesign.getRange();
        options.setRangePartitionColumns(rangePartitionDefinition.getColumns());
    }

    List<RangePartition> rangePartitions = KuduTableProperties.getRangePartitions(properties);
    if (rangePartitionDefinition != null && !rangePartitions.isEmpty()) {
        for (RangePartition rangePartition : rangePartitions) {
            PartialRow lower = KuduTableProperties.toRangeBoundToPartialRow(schema, rangePartitionDefinition, rangePartition.getLower());
            PartialRow upper = KuduTableProperties.toRangeBoundToPartialRow(schema, rangePartitionDefinition, rangePartition.getUpper());
            options.addRangePartition(lower, upper);
        }
    }

    Optional<Integer> numReplicas = KuduTableProperties.getNumReplicas(properties);
    numReplicas.ifPresent(options::setNumReplicas);

    return options;
}
 

private CreateTableOptions buildCreateTableOptions(Schema schema, Map<String, Object> properties) {
    CreateTableOptions options = new CreateTableOptions();

    RangePartitionDefinition rangePartitionDefinition = null;
    Optional<PartitionDesign> optPartitionDesign = KuduTableProperties.getPartitionDesign(properties);
    if (optPartitionDesign.isPresent()) {
        PartitionDesign partitionDesign = optPartitionDesign.get();
        if (partitionDesign.getHash() != null) {
            for (HashPartitionDefinition partition : partitionDesign.getHash()) {
                options.addHashPartitions(partition.getColumns(), partition.getBuckets());
            }
        }
        if (partitionDesign.getRange() != null) {
            rangePartitionDefinition = partitionDesign.getRange();
            options.setRangePartitionColumns(rangePartitionDefinition.getColumns());
        }
    } else {
        String firstColumn = schema.getColumnByIndex(0).getName();
        options.setRangePartitionColumns(Collections.singletonList(firstColumn));
    }

    List<RangePartition> rangePartitions = KuduTableProperties.getRangePartitions(properties);
    if (rangePartitionDefinition != null && !rangePartitions.isEmpty()) {
        for (RangePartition rangePartition: rangePartitions) {
            PartialRow lower = KuduTableProperties.toRangeBoundToPartialRow(schema, rangePartitionDefinition, rangePartition.getLower());
            PartialRow upper = KuduTableProperties.toRangeBoundToPartialRow(schema, rangePartitionDefinition, rangePartition.getUpper());
            options.addRangePartition(lower, upper);
        }
    }


    Optional<Integer> numReplicas = KuduTableProperties.getNumReplicas(properties);
    numReplicas.ifPresent(options::setNumReplicas);

    return options;
}
 

/**
 * The {@code metrics} table is hash partitioned on the {@code metric} and
 * {@code tagset_id} columns, and range partitioned on the {@code time}
 * column. The hash partitioning allows writes and scans at the current time
 * to be evenly distributed over the cluster. Range partitioning on time
 * allows whole tablets to be pruned based on the time constraint, and allows
 * old data to be dropped if desired.
 * @param options the create options
 * @param numTabletServers the number of tablet servers
 * @return the tags table create options
 */
static CreateTableOptions metricsCreateTableOptions(CreateOptions options,
                                                    int numTabletServers) {
  CreateTableOptions create = new CreateTableOptions();
  create.setNumReplicas(options.getNumReplicas());
  create.addHashPartitions(ImmutableList.of("metric", "tagset_id"),
                           options.getNumMetricsHashBuckets(numTabletServers));
  create.setRangePartitionColumns(ImmutableList.of("time"));
  for (Long time : options.getMetricsSplits()) {
    PartialRow split = METRICS_SCHEMA.newPartialRow();
    split.addLong(METRICS_TIME_INDEX, time);
    create.addSplitRow(split);
  }
  return create;
}
 

/**
 * The {@code tags} table is hash partitioned on the {@code key} and
 * {@code value} columns, which allows a scan of a ({@code key}, {@code value})
 * pair to be fully satisfied by a single tablet, and gives good protection
 * against hotspotting.
 * @param options the create options
 * @param numTabletServers the number of tablet servers
 * @return the tags table create options
 */
static CreateTableOptions tagsCreateTableOptions(CreateOptions options,
                                                 int numTabletServers) {
  CreateTableOptions create = new CreateTableOptions();
  create.setNumReplicas(options.getNumReplicas());
  create.addHashPartitions(ImmutableList.of("key", "value"),
                           options.getNumTagsTablets(numTabletServers));
  create.setRangePartitionColumns(ImmutableList.<String>of());
  return create;
}