Package org.apache.druid.query.rowsandcols.semantic

Class DefaultClusteredGroupPartitioner

java.lang.Object

org.apache.druid.query.rowsandcols.semantic.DefaultClusteredGroupPartitioner

All Implemented Interfaces:

ClusteredGroupPartitioner

public class DefaultClusteredGroupPartitioner
extends Object
implements ClusteredGroupPartitioner

Constructor Summary

Constructors

Constructor	Description
DefaultClusteredGroupPartitioner(RowsAndColumns rac)

Method Summary

Modifier and Type	Method	Description
int[]	computeBoundaries(List<String> columns)	Computes and returns a list of contiguous boundaries for independent groups.
ArrayList<RowsAndColumns>	partitionOnBoundaries(List<String> partitionColumns)	Semantically equivalent to computeBoundaries, but returns a list of RowsAndColumns objects instead of just boundary positions.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

DefaultClusteredGroupPartitioner

public DefaultClusteredGroupPartitioner(RowsAndColumns rac)

Method Detail

computeBoundaries

public int[] computeBoundaries(List<String> columns)

Description copied from interface: ClusteredGroupPartitioner

Computes and returns a list of contiguous boundaries for independent groups. All rows in a specific grouping should have the same values for the identified columns. Additionally, as this is assuming it is dealing with clustered data, there should only be a single entry in the return value for a given set of values of the columns.

Note that implementations are not expected to do any validation that the data is pre-clustered. There is no expectation that an implementation will identify that the same cluster existed non-contiguously. It is up to the caller to ensure that data is clustered correctly before invoking this method.

Specified by:

computeBoundaries in interface ClusteredGroupPartitioner

Parameters:

columns - the columns to partition on

Returns:

an int[] representing the start (inclusive) and stop (exclusive) offsets of boundaries. Boundaries are contiguous, so the stop of the previous boundary is the start of the subsequent one.

partitionOnBoundaries

public ArrayList<RowsAndColumns> partitionOnBoundaries(List<String> partitionColumns)

Description copied from interface: ClusteredGroupPartitioner

Semantically equivalent to computeBoundaries, but returns a list of RowsAndColumns objects instead of just boundary positions. This is useful as it allows the concrete implementation to return RowsAndColumns objects that are aware of the internal representation of the data and thus can provide optimized implementations of other semantic interfaces as the "child" RowsAndColumns are used

Specified by:

partitionOnBoundaries in interface ClusteredGroupPartitioner

Parameters:

partitionColumns - the columns to partition on

Returns:

a list of RowsAndColumns representing the data grouped by the partition columns.