org.apache.flink.table.planner.plan.metadata

Class FlinkRelMetadataQuery

java.lang.Object

org.apache.calcite.rel.metadata.RelMetadataQueryBase

org.apache.calcite.rel.metadata.RelMetadataQuery

org.apache.flink.table.planner.plan.metadata.FlinkRelMetadataQuery

public class FlinkRelMetadataQuery
extends org.apache.calcite.rel.metadata.RelMetadataQuery

A RelMetadataQuery that defines extended metadata handler in Flink, e.g ColumnInterval, ColumnNullCount.

Field Summary

Fields inherited from class org.apache.calcite.rel.metadata.RelMetadataQueryBase

map, metadataProvider, THREAD_PROVIDERS

Method Summary

Modifier and Type	Method and Description
org.apache.flink.table.planner.plan.trait.FlinkRelDistribution	flinkDistribution(org.apache.calcite.rel.RelNode rel) Returns the FlinkRelDistribution statistic.
org.apache.flink.table.planner.plan.stats.ValueInterval	getColumnInterval(org.apache.calcite.rel.RelNode rel, int index) Returns the FlinkMetadata.ColumnInterval statistic.
Double	getColumnNullCount(org.apache.calcite.rel.RelNode rel, int index) Returns the null count of the given column.
Double	getColumnOriginNullCount(org.apache.calcite.rel.RelNode rel, int index) Returns origin null count of the given column.
org.apache.flink.table.planner.plan.stats.ValueInterval	getFilteredColumnInterval(org.apache.calcite.rel.RelNode rel, int columnIndex, int filterArg) Returns the FlinkMetadata.ColumnInterval of the given column under the given filter argument.
org.apache.flink.table.planner.plan.trait.RelModifiedMonotonicity	getRelModifiedMonotonicity(org.apache.calcite.rel.RelNode rel) Returns the RelModifiedMonotonicity statistic.
RelWindowProperties	getRelWindowProperties(org.apache.calcite.rel.RelNode rel) Returns the RelWindowProperties statistic.
org.apache.calcite.util.ImmutableBitSet	getUniqueGroups(org.apache.calcite.rel.RelNode rel, org.apache.calcite.util.ImmutableBitSet columns) Returns the (minimum) unique groups of the given columns.
Set<org.apache.calcite.util.ImmutableBitSet>	getUpsertKeys(org.apache.calcite.rel.RelNode rel) Determines the set of upsert minimal keys for this expression.
Set<org.apache.calcite.util.ImmutableBitSet>	getUpsertKeysInKeyGroupRange(org.apache.calcite.rel.RelNode rel, int[] partitionKeys) Determines the set of upsert minimal keys in a single key group range, which means can ignore exchange by partition keys.
static FlinkRelMetadataQuery	instance() Returns an instance of FlinkRelMetadataQuery.
static FlinkRelMetadataQuery	reuseOrCreate(org.apache.calcite.rel.metadata.RelMetadataQuery mq) Reuse input metadataQuery instance if it could cast to FlinkRelMetadataQuery class, or create one if not.

Methods inherited from class org.apache.calcite.rel.metadata.RelMetadataQuery

areColumnsUnique, areColumnsUnique, areRowsUnique, areRowsUnique, collations, cumulativeMemoryWithinPhase, cumulativeMemoryWithinPhaseSplit, distribution, getAllPredicates, getAverageColumnSizes, getAverageColumnSizesNotNull, getAverageRowSize, getColumnOrigin, getColumnOrigins, getCumulativeCost, getDistinctRowCount, getDistribution, getExpressionLineage, getLowerBoundCost, getMaxRowCount, getMinRowCount, getNodeTypes, getNonCumulativeCost, getPercentageOriginalRows, getPopulationSize, getPulledUpPredicates, getRowCount, getSelectivity, getTableOrigin, getTableReferences, getUniqueKeys, getUniqueKeys, isPhaseTransition, isVisibleInExplain, memory, splitCount

Methods inherited from class org.apache.calcite.rel.metadata.RelMetadataQueryBase

clearCache, handler, initialHandler, revise, revise

Methods inherited from class java.lang.Object

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

Method Detail

instance

public static FlinkRelMetadataQuery instance()

Returns an instance of FlinkRelMetadataQuery. It ensures that cycles do not occur while computing metadata.

reuseOrCreate

public static FlinkRelMetadataQuery reuseOrCreate(org.apache.calcite.rel.metadata.RelMetadataQuery mq)

Reuse input metadataQuery instance if it could cast to FlinkRelMetadataQuery class, or create one if not.

Parameters:

mq - metadataQuery which try to reuse

Returns:

a FlinkRelMetadataQuery instance

getColumnInterval

public org.apache.flink.table.planner.plan.stats.ValueInterval getColumnInterval(org.apache.calcite.rel.RelNode rel,
                                                                                 int index)

Returns the FlinkMetadata.ColumnInterval statistic.

Parameters:

rel - the relational expression

index - the index of the given column

Returns:

the interval of the given column of a specified relational expression. Returns null if interval cannot be estimated, Returns EmptyValueInterval if column values does not contains any value except for null.

getFilteredColumnInterval

public org.apache.flink.table.planner.plan.stats.ValueInterval getFilteredColumnInterval(org.apache.calcite.rel.RelNode rel,
                                                                                         int columnIndex,
                                                                                         int filterArg)

Returns the FlinkMetadata.ColumnInterval of the given column under the given filter argument.

Parameters:

rel - the relational expression

columnIndex - the index of the given column

filterArg - the index of the filter argument

Returns:

the interval of the given column of a specified relational expression. Returns null if interval cannot be estimated, Returns EmptyValueInterval if column values does not contains any value except for null.

getColumnNullCount

public Double getColumnNullCount(org.apache.calcite.rel.RelNode rel,
                                 int index)

Returns the null count of the given column.

Parameters:

rel - the relational expression

index - the index of the given column

Returns:

the null count of the given column if can be estimated, else return null.

getColumnOriginNullCount

public Double getColumnOriginNullCount(org.apache.calcite.rel.RelNode rel,
                                       int index)

Returns origin null count of the given column.

Parameters:

rel - the relational expression

index - the index of the given column

Returns:

the null count of the given column if can be estimated, else return null.

getUniqueGroups

public org.apache.calcite.util.ImmutableBitSet getUniqueGroups(org.apache.calcite.rel.RelNode rel,
                                                               org.apache.calcite.util.ImmutableBitSet columns)

Returns the (minimum) unique groups of the given columns.

Parameters:

rel - the relational expression

columns - the given columns in a specified relational expression. The given columns should not be null.

Returns:

the (minimum) unique columns which should be a sub-collection of the given columns, and should not be null or empty. If none unique columns can be found, return the given columns.

flinkDistribution

public org.apache.flink.table.planner.plan.trait.FlinkRelDistribution flinkDistribution(org.apache.calcite.rel.RelNode rel)

Returns the FlinkRelDistribution statistic.

Parameters:

rel - the relational expression

Returns:

description of how the rows in the relational expression are physically distributed

getRelModifiedMonotonicity

public org.apache.flink.table.planner.plan.trait.RelModifiedMonotonicity getRelModifiedMonotonicity(org.apache.calcite.rel.RelNode rel)

Returns the RelModifiedMonotonicity statistic.

Parameters:

rel - the relational expression

Returns:

the monotonicity for the corresponding RelNode

getRelWindowProperties

public RelWindowProperties getRelWindowProperties(org.apache.calcite.rel.RelNode rel)

Returns the RelWindowProperties statistic.

Parameters:

rel - the relational expression

Returns:

the window properties for the corresponding RelNode

getUpsertKeys

public Set<org.apache.calcite.util.ImmutableBitSet> getUpsertKeys(org.apache.calcite.rel.RelNode rel)

Determines the set of upsert minimal keys for this expression. A key is represented as an ImmutableBitSet, where each bit position represents a 0-based output column ordinal.

Different from the unique keys: In distributed streaming computing, one record may be divided into RowKind.UPDATE_BEFORE and RowKind.UPDATE_AFTER. If a key changing join is connected downstream, the two records will be divided into different tasks, resulting in disorder. In this case, the downstream cannot rely on the order of the original key. So in this case, it has unique keys in the traditional sense, but it doesn't have upsert keys.

Returns:

set of keys, or null if this information cannot be determined (whereas empty set indicates definitely no keys at all)

getUpsertKeysInKeyGroupRange

public Set<org.apache.calcite.util.ImmutableBitSet> getUpsertKeysInKeyGroupRange(org.apache.calcite.rel.RelNode rel,
                                                                                 int[] partitionKeys)

Determines the set of upsert minimal keys in a single key group range, which means can ignore exchange by partition keys.

Some optimizations can rely on this ability to do upsert in a single key group range.