Package org.apache.lucene.codecs

Class KnnVectorsWriter

java.lang.Object

org.apache.lucene.codecs.KnnVectorsWriter

All Implemented Interfaces:

Closeable, AutoCloseable, Accountable

Direct Known Subclasses:

BufferingKnnVectorsWriter, FlatVectorsWriter, Lucene99HnswVectorsWriter

public abstract class KnnVectorsWriter
extends Object
implements Accountable, Closeable

Writes vectors to an index.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static final class	KnnVectorsWriter.MergedVectorValues	View over multiple vector values supporting iterator-style access via DocIdMerger.

Field Summary

Fields inherited from interface org.apache.lucene.util.Accountable

NULL_ACCOUNTABLE

Constructor Summary

Constructors

Modifier	Constructor	Description
protected	KnnVectorsWriter()	Sole constructor

Method Summary

Modifier and Type	Method	Description
abstract KnnFieldVectorsWriter<?>	addField(FieldInfo fieldInfo)	Add new field for indexing
abstract void	finish()	Called once at the end before close
abstract void	flush(int maxDoc, Sorter.DocMap sortMap)	Flush all buffered data on disk *
static void	mapOldOrdToNewOrd(DocsWithFieldSet oldDocIds, Sorter.DocMap sortMap, int[] old2NewOrd, int[] new2OldOrd, DocsWithFieldSet newDocsWithField)	Given old doc ids and an id mapping, maps old ordinal to new ordinal.
final void	merge(MergeState mergeState)	Merges the segment vectors for all fields using a two-phase strategy: Phase 1: Merge flat vectors for all fields by calling mergeOneField(FieldInfo, MergeState), collecting deferred work (runnables) for each field.
IORunnable	mergeOneField(FieldInfo fieldInfo, MergeState mergeState)	Merges vectors for a single field, returning a runnable for any deferred work (e.g., HNSW graph construction).

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.apache.lucene.util.Accountable

getChildResources, ramBytesUsed

Methods inherited from interface java.io.Closeable

close

Constructor Details

KnnVectorsWriter

protected KnnVectorsWriter()

Sole constructor

Method Details

addField

public abstract KnnFieldVectorsWriter<?> addField(FieldInfo fieldInfo)
                                           throws IOException

Add new field for indexing

Throws:

IOException

flush

public abstract void flush(int maxDoc,
 Sorter.DocMap sortMap)
                    throws IOException

Flush all buffered data on disk *

Throws:

IOException

finish

public abstract void finish()
                     throws IOException

Called once at the end before close

Throws:

IOException

mergeOneField

public IORunnable mergeOneField(FieldInfo fieldInfo,
 MergeState mergeState)
                         throws IOException

Merges vectors for a single field, returning a runnable for any deferred work (e.g., HNSW graph construction). The default implementation merges naively the vectors and returns null (no deferred work).

Subclasses should override this method may implement a two-phase merge strategy where flat vectors are written in the first phase and additional indexing structures (like HNSW graphs) are built in the second phase using the already-written flat vector data.

Parameters:

fieldInfo - the field to merge

mergeState - the merge state

Returns:

a runnable to execute in phase 2, or null if there is no deferred work

Throws:

IOException - if an I/O error occurs

merge

public final void merge(MergeState mergeState)
                 throws IOException

Merges the segment vectors for all fields using a two-phase strategy:

1. Phase 1: Merge flat vectors for all fields by calling mergeOneField(FieldInfo, MergeState), collecting deferred work (runnables) for each field.
2. Phase 2: Execute the deferred runnables (e.g., HNSW graph construction) using the flat vector data written in phase 1.

Throws:

IOException

mapOldOrdToNewOrd

public static void mapOldOrdToNewOrd(DocsWithFieldSet oldDocIds,
 Sorter.DocMap sortMap,
 int[] old2NewOrd,
 int[] new2OldOrd,
 DocsWithFieldSet newDocsWithField)
                              throws IOException

Given old doc ids and an id mapping, maps old ordinal to new ordinal. Note: this method return nothing and output are written to parameters

Parameters:

oldDocIds - the old or current document ordinals. Must not be null.

sortMap - the document sorting map for how to make the new ordinals. Must not be null.

old2NewOrd - int[] maps from old ord to new ord

new2OldOrd - int[] maps from new ord to old ord

newDocsWithField - set of new doc ids which has the value

Throws:

IOException