org.apache.druid.query

Interface QueryRunnerFactory<T,QueryType extends Query<T>>

All Known Implementing Classes:

DataSourceMetadataQueryRunnerFactory, GroupByQueryRunnerFactory, ScanQueryRunnerFactory, SearchQueryRunnerFactory, SegmentMetadataQueryRunnerFactory, TimeBoundaryQueryRunnerFactory, TimeseriesQueryRunnerFactory, TopNQueryRunnerFactory, WindowOperatorQueryQueryRunnerFactory

public interface QueryRunnerFactory<T,QueryType extends Query<T>>

An interface that defines the nitty gritty implementation details of a Query on a Segment

Method Summary

Modifier and Type	Method and Description
QueryRunner<T>	createRunner(Segment segment) Given a specific segment, this method will create a QueryRunner.
QueryToolChest<T,QueryType>	getToolchest() Provides access to the QueryToolChest for this specific Query type.
default QueryRunner<T>	mergeRunners(ExecutorService queryExecutor, Iterable<QueryRunner<T>> queryRunners) Deprecated. Use mergeRunners(QueryProcessingPool, Iterable) instead. Runners generated with createRunner() and combined into an Iterable in (time,shardId) order are passed along to this method with an ExecutorService. The method should then return a QueryRunner that, when asked, will use the ExecutorService to run the base QueryRunners in some fashion. The vast majority of the time, this should be implemented with ChainedExecutionQueryRunner: return new ChainedExecutionQueryRunner<>(queryExecutor, toolChest.getOrdering(), queryWatcher, queryRunners); Which will allow for parallel execution up to the maximum number of processing threads allowed.
QueryRunner<T>	mergeRunners(QueryProcessingPool queryProcessingPool, Iterable<QueryRunner<T>> queryRunners) Runners generated with createRunner() and combined into an Iterable in (time,shardId) order are passed along to this method with an QueryProcessingPool.

Method Detail

createRunner

QueryRunner<T> createRunner(Segment segment)

Given a specific segment, this method will create a QueryRunner. The QueryRunner, when asked, will generate a Sequence of results based on the given segment. This is the meat of the Query processing and is where the results are actually generated. Everything else is just merging and reduction logic.

Parameters:

segment - The segment to process

Returns:

A QueryRunner that, when asked, will generate a Sequence of results based on the given segment

mergeRunners

@Deprecated
default QueryRunner<T> mergeRunners(ExecutorService queryExecutor,
                                                Iterable<QueryRunner<T>> queryRunners)

Deprecated. Use mergeRunners(QueryProcessingPool, Iterable) instead. Runners generated with createRunner() and combined into an Iterable in (time,shardId) order are passed along to this method with an ExecutorService. The method should then return a QueryRunner that, when asked, will use the ExecutorService to run the base QueryRunners in some fashion. The vast majority of the time, this should be implemented with ChainedExecutionQueryRunner: return new ChainedExecutionQueryRunner<>(queryExecutor, toolChest.getOrdering(), queryWatcher, queryRunners); Which will allow for parallel execution up to the maximum number of processing threads allowed.

Parameters:

queryExecutor - ExecutorService to be used for parallel processing

queryRunners - Individual QueryRunner objects that produce some results

Returns:

a QueryRunner that, when asked, will use the ExecutorService to run the base QueryRunner collection.

mergeRunners

QueryRunner<T> mergeRunners(QueryProcessingPool queryProcessingPool,
                            Iterable<QueryRunner<T>> queryRunners)

Runners generated with createRunner() and combined into an Iterable in (time,shardId) order are passed along to this method with an QueryProcessingPool. The method should then return a QueryRunner that, when asked, will use the QueryProcessingPool to run the base QueryRunners in some fashion. The vast majority of the time, this should be implemented with ChainedExecutionQueryRunner: return new ChainedExecutionQueryRunner<>(queryProcessingPool, toolChest.getOrdering(), queryWatcher, queryRunners); Which will allow for parallel execution up to the maximum number of processing threads allowed. Unlike mergeRunners(ExecutorService, Iterable), this method takes a QueryProcessingPool instead which allows custom implementations for prioritize query execution on segments.

Parameters:

queryProcessingPool - QueryProcessingPool to be used for parallel processing

queryRunners - Individual QueryRunner objects that produce some results

Returns:

a QueryRunner that, when asked, will use the ExecutorService to run the base QueryRunner collection.

getToolchest

QueryToolChest<T,QueryType> getToolchest()

Provides access to the QueryToolChest for this specific Query type.

Returns:

an instance of the toolchest for this specific query type.