com.bigdata.bop.joinGraph.rto

Class EdgeSample

java.lang.Object

com.bigdata.bop.joinGraph.rto.SampleBase

com.bigdata.bop.joinGraph.rto.EdgeSample

public class EdgeSample
extends SampleBase

A sample produced by a cutoff join.

Field Summary

Fields

Modifier and Type	Field and Description
long	adjCard The adjusted cardinality estimate for the cutoff join (this is outputCount as adjusted for a variety of edge conditions).
long	estRead Estimated tuples read if the operator were fully executed.
double	f The ratio of the #of input samples consumed to the #of output samples generated (the join hit ratio or scale factor).
int	inputCount The #of binding sets out of the source sample vertex sample which were consumed.
long	outputCount The #of binding sets generated before the join was cutoff.
SampleBase	sourceSample The source sample used to compute the cutoff join.
long	sumRangeCount The sum of the fast range count for each access path tested.
long	tuplesRead The #of tuples read from the access path when processing the cutoff join.

Fields inherited from class com.bigdata.bop.joinGraph.rto.SampleBase

estCard, estimateEnum, limit

Constructor Summary

Constructors

Constructor and Description
EdgeSample(SampleBase sourceSample, int inputCount, long tuplesRead, long sumRangeCount, long outputCount, long adjustedCard, double f, long estCard, long estRead, int limit, EstimateEnum estimateEnum, IBindingSet[] sample) Create an object which encapsulates a sample of an edge.

Method Summary

Methods

Modifier and Type	Method and Description
protected void	toString(StringBuilder sb) Hook for extending SampleBase.toString().

Methods inherited from class com.bigdata.bop.joinGraph.rto.SampleBase

getSample, isExact, isUnderflow, toString

Methods inherited from class java.lang.Object

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

Field Detail

sourceSample

public final SampleBase sourceSample

The source sample used to compute the cutoff join.

inputCount

public final int inputCount

The #of binding sets out of the source sample vertex sample which were consumed.

tuplesRead

public final long tuplesRead

The #of tuples read from the access path when processing the cutoff join.

outputCount

public final long outputCount

The #of binding sets generated before the join was cutoff.

Note: If the outputCount is zero then this is a good indicator that there is an error in the query such that the join will not select anything. This is not 100%, merely indicative.

adjCard

public final long adjCard

The adjusted cardinality estimate for the cutoff join (this is outputCount as adjusted for a variety of edge conditions).

f

public final double f

The ratio of the #of input samples consumed to the #of output samples generated (the join hit ratio or scale factor).

sumRangeCount

public final long sumRangeCount

The sum of the fast range count for each access path tested.

Note: We use pipeline joins to sample cutoff joins so there will be one access path read for each solution in. However, a hash join could be used when the operator is fully executed. The hash join will have one access path on which we read for all input solutions and the range count of the access path will be larger since the access path will be less constrained.

estRead

public final long estRead

Estimated tuples read if the operator were fully executed. This is in contrast to SampleBase.estCard, which is the estimated output cardinality if the operator were fully executed. TODO The actual IOs depend on the join type (hash join versus pipeline join) and whether or not the file has index order (segment versus journal). A hash join will read once on the AP. A pipeline join will read once per input solution. A key-range read on a segment uses multi-block IO while a key-range read on a journal uses random IO. Also, remote access path reads are more expensive than sharded or hash partitioned access path reads in scale-out.

Constructor Detail

EdgeSample

public EdgeSample(SampleBase sourceSample,
          int inputCount,
          long tuplesRead,
          long sumRangeCount,
          long outputCount,
          long adjustedCard,
          double f,
          long estCard,
          long estRead,
          int limit,
          EstimateEnum estimateEnum,
          IBindingSet[] sample)

Create an object which encapsulates a sample of an edge.

Parameters:

sourceSample - The input sample.

limit - The limit used to sample the edge (this is the limit on the #of solutions generated by the cutoff join used when this sample was taken).

inputCount - The #of binding sets out of the source sample vertex sample which were consumed.

tuplesRead - The #of tuples read from the access path when processing the cutoff join.

outputCount - The #of binding sets generated before the join was cutoff.

adjustedCard - The adjusted cardinality estimate for the cutoff join (this is outputCount as adjusted for a variety of edge conditions).

Method Detail

toString

protected void toString(StringBuilder sb)

Description copied from class: SampleBase

Hook for extending SampleBase.toString().

Overrides:

toString in class SampleBase

Parameters:

sb - The buffer into which the implementation can write additional information.