com.bigdata.btree.proc

Class AbstractLocalSplitResultAggregator<R>

java.lang.Object

com.bigdata.btree.proc.AbstractLocalSplitResultAggregator<R>

All Implemented Interfaces:

IResultHandler<R,R>

public abstract class AbstractLocalSplitResultAggregator<R>
extends Object
implements IResultHandler<R,R>

Aggregator base class collects the individual results in an internal ordered map and assembles the final result when it is requested from the individual results. With this approach there is no overhead or contention when the results are being produced in parallel and they can be combined efficiently within a single thread in getResult().

Note: This implementation assumes that there is one element of the result for each key in the original request. It places the Split-wise results into a total ordering over the Splits and then delegates to a concrete implementation to build the aggregated results out of the ordered pairs of (Split, partial-result).

Author:

bryan

Constructor Summary

Constructors

Constructor and Description
AbstractLocalSplitResultAggregator(int size)

Method Summary

Methods

Modifier and Type	Method and Description
void	aggregate(R result, Split split) Method is invoked for each result and is responsible for combining the results in whatever manner is meaningful for the procedure.
R	getResult() Return the aggregated results as an implementation dependent object.
protected abstract R	newResult(int size, SplitValuePair<Split,R>[] a) Build the aggregated result by aggregate the individual results in the given order.

Methods inherited from class java.lang.Object

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

Constructor Detail

AbstractLocalSplitResultAggregator

public AbstractLocalSplitResultAggregator(int size)

Parameters:

size - The #of elements in the request (which is the same as the cardinality of the aggregated result).

Method Detail

aggregate

public void aggregate(R result,
             Split split)

Description copied from interface: IResultHandler

Method is invoked for each result and is responsible for combining the results in whatever manner is meaningful for the procedure. Implementations of this method MUST be thread-safe since the procedure MAY be applied in parallel when it spans more than one index partition.

Specified by:

aggregate in interface IResultHandler<R,R>

Parameters:

result - The result from applying the procedure to a single index partition.

split - The Split that generated that result.

getResult

public R getResult()

Description copied from interface: IResultHandler

Return the aggregated results as an implementation dependent object.

Specified by:

getResult in interface IResultHandler<R,R>

newResult

protected abstract R newResult(int size,
          SplitValuePair<Split,R>[] a)

Build the aggregated result by aggregate the individual results in the given order.

Parameters:

size - The number of keys in the original request.

a - An array of SplitValuePairs that is ordered on Split.fromIndex.

Returns:

The aggregated result.