Module org.elasticsearch.server

Package org.elasticsearch.search.fetch

Class FetchPhase

java.lang.Object

org.elasticsearch.search.fetch.FetchPhase

public final class FetchPhase
extends Object

Fetch phase of a search request, used to fetch the actual top matching documents to be returned to the client, identified after reducing all the matches returned by the query phase Supports both traditional mode (all results in memory) and streaming mode (results sent in chunks).

Constructor Summary

Constructors

Constructor	Description
FetchPhase(List<FetchSubPhase> fetchSubPhases)

Method Summary

Modifier and Type	Method	Description
void	execute(SearchContext context, int[] docIdsToLoad, RankDocShardInfo rankDocs)	Executes the fetch phase without memory checking or streaming.
void	execute(SearchContext context, int[] docIdsToLoad, RankDocShardInfo rankDocs, IntConsumer memoryChecker)	Executes the fetch phase with an optional caller-supplied memory tracking callback and no streaming
void	execute(SearchContext context, int[] docIdsToLoad, RankDocShardInfo rankDocs, IntConsumer memoryChecker, FetchPhaseResponseChunk.Writer writer, Integer maxInFlightChunks, ActionListener<Void> buildListener, ActionListener<Void> listener)	Executes the fetch phase with an optional caller-supplied memory tracking callback and optional streaming.

Methods inherited from class java.lang.Object

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

Constructor Details

FetchPhase

public FetchPhase(List<FetchSubPhase> fetchSubPhases)

Method Details

execute

public void execute(SearchContext context,
 int[] docIdsToLoad,
 RankDocShardInfo rankDocs)

Executes the fetch phase without memory checking or streaming.

Parameters:

context - the search context

docIdsToLoad - document IDs to fetch

rankDocs - ranking information

execute

public void execute(SearchContext context,
 int[] docIdsToLoad,
 RankDocShardInfo rankDocs,
 @Nullable
 IntConsumer memoryChecker)

Executes the fetch phase with an optional caller-supplied memory tracking callback and no streaming

Parameters:

context - the search context

docIdsToLoad - document IDs to fetch

rankDocs - ranking information

memoryChecker - optional callback for memory tracking, may be null

execute

public void execute(SearchContext context,
 int[] docIdsToLoad,
 RankDocShardInfo rankDocs,
 @Nullable
 IntConsumer memoryChecker,
 @Nullable
 FetchPhaseResponseChunk.Writer writer,
 @Nullable
 Integer maxInFlightChunks,
 @Nullable
 ActionListener<Void> buildListener,
 ActionListener<Void> listener)

Executes the fetch phase with an optional caller-supplied memory tracking callback and optional streaming.

When writer is null (non-streaming), all hits are accumulated in memory and returned at once. When writer is provided (streaming), hits are emitted in chunks to reduce peak memory usage. In streaming mode, the final completion may be delayed by transport-level acknowledgements, but the fetch build completion is signaled as soon as the fetch work has finished.

Parameters:

context - the search context

docIdsToLoad - document IDs to fetch

rankDocs - ranking information

memoryChecker - optional callback for memory tracking, may be null

writer - optional chunk writer for streaming mode, may be null

buildListener - optional listener invoked when all SearchHit objects have been constructed (and, in streaming mode, serialized into chunks and dispatched to the writer). In non-streaming mode this fires immediately after the hits are built, just like listener. In streaming mode this fires before chunk ACKs arrive, allowing the caller to release shard resources (e.g. close the SearchContext) without waiting for network acknowledgements.

listener - final completion listener. In streaming mode this is invoked only after all chunks are ACKed; in non-streaming mode it is invoked immediately after hits are built.

Throws:

TaskCancelledException - if the task is cancelled