Package io.trino.operator.output

Class PagePartitioner

java.lang.Object

io.trino.operator.output.PagePartitioner

All Implemented Interfaces:

Closeable, AutoCloseable

public class PagePartitioner
extends Object
implements Closeable

Constructor Summary

Constructors

Constructor

Description

PagePartitioner(PartitionFunction partitionFunction, List<Integer> partitionChannels, List<Optional<NullableValue>> partitionConstants, boolean replicatesAnyRow, OptionalInt nullChannel, OutputBuffer outputBuffer, PagesSerdeFactory serdeFactory, List<Type> sourceTypes, io.airlift.units.DataSize maxMemory, PositionsAppenderFactory positionsAppenderFactory, Optional<io.airlift.slice.Slice> exchangeEncryptionKey, AggregatedMemoryContext aggregatedMemoryContext, boolean partitionProcessRleAndDictionaryBlocks)

Method Summary

Modifier and Type	Method	Description
void	close()
PartitionFunction	getPartitionFunction()
void	partitionPage(Page page, OperatorContext operatorContext)
void	partitionPageByColumn(Page page)
void	partitionPageByRow(Page page)
void	prepareForRelease(OperatorContext operatorContext)	Prepares this PagePartitioner for release to the pool by checking for dictionary mode appenders and either flattening them into direct appenders or forcing their current pages to flush to preserve a valuable dictionary encoded representation.

Methods inherited from class java.lang.Object

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

Constructor Details

PagePartitioner

public PagePartitioner(PartitionFunction partitionFunction,
 List<Integer> partitionChannels,
 List<Optional<NullableValue>> partitionConstants,
 boolean replicatesAnyRow,
 OptionalInt nullChannel,
 OutputBuffer outputBuffer,
 PagesSerdeFactory serdeFactory,
 List<Type> sourceTypes,
 io.airlift.units.DataSize maxMemory,
 PositionsAppenderFactory positionsAppenderFactory,
 Optional<io.airlift.slice.Slice> exchangeEncryptionKey,
 AggregatedMemoryContext aggregatedMemoryContext,
 boolean partitionProcessRleAndDictionaryBlocks)

Method Details

getPartitionFunction

public PartitionFunction getPartitionFunction()

partitionPage

public void partitionPage(Page page,
 OperatorContext operatorContext)

prepareForRelease

public void prepareForRelease(OperatorContext operatorContext)

Prepares this PagePartitioner for release to the pool by checking for dictionary mode appenders and either flattening them into direct appenders or forcing their current pages to flush to preserve a valuable dictionary encoded representation. This is done before release because we know that after reuse, the appenders will not observe any more inputs using the same dictionary.

When a PagePartitioner is released back to the PagePartitionerPool we don't know if it will ever be reused. If it is not reused, then we have no OperatorContext we can use to report the output size of the final flushed page, so instead we report the buffered bytes still in the partitioner after prepareForRelease(OperatorContext) as output bytes eagerly and record that amount in outputSizeReportedBeforeRelease. If the PagePartitioner is reused after having reported buffered bytes eagerly, we then have to subtract that same amount from the subsequent output bytes to avoid double counting them.

partitionPageByRow

public void partitionPageByRow(Page page)

partitionPageByColumn

public void partitionPageByColumn(Page page)

close

public void close()

Specified by:

close in interface AutoCloseable

Specified by:

close in interface Closeable