edu.umich.verdict.relation

Class ApproxFilteredRelation

java.lang.Object

edu.umich.verdict.relation.Relation

edu.umich.verdict.relation.ApproxRelation

edu.umich.verdict.relation.ApproxFilteredRelation

public class ApproxFilteredRelation
extends ApproxRelation

Field Summary

Fields inherited from class edu.umich.verdict.relation.ApproxRelation

partitionSizeAlias

Fields inherited from class edu.umich.verdict.relation.Relation

alias, approximate, availableJoinTypes, subquery, vc

Constructor Summary

Constructors

Constructor and Description
ApproxFilteredRelation(VerdictContext vc, ApproxRelation source, Cond cond)

Method Summary

Modifier and Type	Method and Description
double	cost()
boolean	equals(ApproxRelation o)
Cond	getFilter()
ApproxRelation	getSource()
ExactRelation	rewriteForPointEstimate()
ExactRelation	rewriteWithPartition() Internal method for ApproxRelation.rewriteWithSubsampledErrorBounds().
protected List<String>	sampleColumns() Returns a set of columns on which a sample is created.
String	sampleType() Returns an effective sample type of this relation.
double	samplingProbability() rough sampling probability, which is obtained from the sampling params.
protected List<Expr>	samplingProbabilityExprsFor(FuncExpr f) Computes an appropriate sampling probability for a particular aggregate function.
protected Map<TableUniqueName,String>	tableSubstitution() Pairs of original table name and a sample table name.
protected String	toStringWithIndent(String indent)

Methods inherited from class edu.umich.verdict.relation.ApproxRelation

agg, agg, avg, confidenceIntervalMultiplier, count, countDistinct, exprWithTableNamesSubstituted, groupby, groupby, limit, orderby, rewrite, rewriteWithBootstrappedErrorBounds, rewriteWithSubsampledErrorBounds, sourceTableName, sum, toSql, toString

Methods inherited from class edu.umich.verdict.relation.Relation

areMatchingUniverseSamples, avgValue, collect, collectAsString, collectDataFrame, collectResultSet, countDistinctValue, countValue, errorBoundColumn, exprsInSelectElems, genColumnAlias, genTableAlias, getAlias, getTempTableName, getTempTableName, getVerdictContext, isApproximate, isSubquery, partitionColumnName, prettyfySql, samplingProbabilityColumnName, setAlias, setSubquery, sumValue

Methods inherited from class java.lang.Object

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

Constructor Detail

ApproxFilteredRelation

public ApproxFilteredRelation(VerdictContext vc,
                              ApproxRelation source,
                              Cond cond)

Method Detail

getSource

public ApproxRelation getSource()

getFilter

public Cond getFilter()

rewriteForPointEstimate

public ExactRelation rewriteForPointEstimate()

Specified by:

rewriteForPointEstimate in class ApproxRelation

rewriteWithPartition

public ExactRelation rewriteWithPartition()

Description copied from class: ApproxRelation

Internal method for ApproxRelation.rewriteWithSubsampledErrorBounds().

Specified by:

rewriteWithPartition in class ApproxRelation

Returns:

samplingProbabilityExprsFor

protected List<Expr> samplingProbabilityExprsFor(FuncExpr f)

Description copied from class: ApproxRelation

Computes an appropriate sampling probability for a particular aggregate function. For uniform random sample, returns the ratio between the sample table and the original table. For universe sample, if the aggregate function is COUNT, AVG, SUM, returns the ratio between the sample table and the original table. if the aggregate function is COUNT-DISTINCT, returns the sampling probability. For stratified sample, this method returns the sampling probability only for the joined tables. Verdict sample rules. For COUNT, AVG, and SUM, uniform random samples, universe samples, stratified samples, or no samples can be used. For COUNT-DISTINCT, universe sample, stratified samples, or no samples can be used. For stratified samples, the distinct number of groups is assumed to be limited. Verdict join rules. (uniform, uniform) -> uniform (uniform, stratified) -> stratified (uniform, universe) -> uniform (uniform, no sample) -> uniform (stratified, stratified) -> stratified (stratified, universe) -> no allowed (stratified, no sample) -> stratified (universe, universe) -> universe (only when the columns on which samples are built coincide) (universe, no sample) -> universe

Specified by:

samplingProbabilityExprsFor in class ApproxRelation

Returns:

tableSubstitution

protected Map<TableUniqueName,String> tableSubstitution()

Description copied from class: ApproxRelation

Pairs of original table name and a sample table name. This function does not inspect subqueries. The substitution expression is an alias (thus, string type).

Specified by:

tableSubstitution in class ApproxRelation

Returns:

sampleType

public String sampleType()

Description copied from class: ApproxRelation

Returns an effective sample type of this relation.

Specified by:

sampleType in class ApproxRelation

Returns:

One of "uniform", "universe", "stratified", "nosample".

cost

public double cost()

Specified by:

cost in class ApproxRelation

sampleColumns

protected List<String> sampleColumns()

Description copied from class: ApproxRelation

Returns a set of columns on which a sample is created. Only meaningful for stratified and universe samples.

Specified by:

sampleColumns in class ApproxRelation

Returns:

toStringWithIndent

protected String toStringWithIndent(String indent)

Specified by:

toStringWithIndent in class ApproxRelation

equals

public boolean equals(ApproxRelation o)

Specified by:

equals in class ApproxRelation

samplingProbability

public double samplingProbability()

Description copied from class: ApproxRelation

rough sampling probability, which is obtained from the sampling params.

Specified by:

samplingProbability in class ApproxRelation

Returns: