org.apache.hadoop.hive.ql.optimizer.calcite.correlation

Class HiveCorrelationInfo

java.lang.Object

org.apache.hadoop.hive.ql.optimizer.calcite.correlation.HiveCorrelationInfo

public class HiveCorrelationInfo
extends Object

HiveCorrelationInfo contains information gathered by visiting the RexSubQuery node

Field Summary

Fields

Modifier and Type	Field and Description
HiveAggregate	aggregateRel
Set<org.apache.calcite.rel.core.CorrelationId>	correlationIds
boolean	notFlag
org.apache.calcite.rex.RexSubQuery	rexSubQuery

Constructor Summary

Constructors

Constructor and Description
HiveCorrelationInfo(Set<org.apache.calcite.rel.core.CorrelationId> correlationIds, org.apache.calcite.rex.RexSubQuery rexSubQuery, HiveAggregate aggregateRel, boolean notFlag)

Method Summary

Modifier and Type	Method and Description
boolean	isCorrScalarQuery() isCorrScalarQuery returns true for special cases as specified in the HiveSubQueryRemoveRule rewrite methods for steps that need to be taken before writing the join operator.

Methods inherited from class java.lang.Object

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

Field Detail

rexSubQuery

public final org.apache.calcite.rex.RexSubQuery rexSubQuery

correlationIds

public final Set<org.apache.calcite.rel.core.CorrelationId> correlationIds

aggregateRel

public final HiveAggregate aggregateRel

notFlag

public final boolean notFlag

Constructor Detail

HiveCorrelationInfo

public HiveCorrelationInfo(Set<org.apache.calcite.rel.core.CorrelationId> correlationIds,
                           org.apache.calcite.rex.RexSubQuery rexSubQuery,
                           HiveAggregate aggregateRel,
                           boolean notFlag)

Method Detail

isCorrScalarQuery

public boolean isCorrScalarQuery()

isCorrScalarQuery returns true for special cases as specified in the HiveSubQueryRemoveRule rewrite methods for steps that need to be taken before writing the join operator. It will add an sq_count_check to ensure that there is a row. This logic was copied from QBSubQuery.java for HIVE-26736. The following comment was copied from there: Restriction.13.m :: In the case of an implied Group By on a correlated SubQuery, the SubQuery always returns 1 row. Following is special cases for different type of subqueries which have aggregate and implicit group by and are correlatd * SCALAR - This should return true since later in subquery remove rule we need to know about this case. * IN - always allowed, BUT returns true for cases with aggregate other than COUNT since later in subquery remove rule we need to know about this case. * NOT IN - always allow, but always return true because later subq remove rule will generate diff plan for this case