Skip to content

[DiskBBQ] Fix index sorting on flush (2nd attempt) #145076

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
iverase merged 3 commits into from
Mar 27, 2026
Merged

[DiskBBQ] Fix index sorting on flush (2nd attempt) #145076

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6d124d7
[DiskBBQ] Fix index sorting on flush (2nd attempt)
iverase Mar 27, 2026
82d3438
Update docs/changelog/145076.yaml
iverase Mar 27, 2026
c5329f1
Merge branch 'main' into indexSortflush
iverase Mar 27, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions docs/changelog/145076.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
area: Vector Search
issues: []
pr: 145076
summary: "[DiskBBQ] Fix index sorting on flush (2nd attempt)"
type: bug
40 changes: 33 additions & 7 deletions server/src/main/java/org/elasticsearch/index/codec/vectors/diskbbq/IVFVectorsWriter.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@
import org.apache.lucene.codecs.KnnVectorsWriter;
import org.apache.lucene.codecs.hnsw.FlatFieldVectorsWriter;
import org.apache.lucene.codecs.hnsw.FlatVectorsWriter;
import org.apache.lucene.index.DocsWithFieldSet;
import org.apache.lucene.index.FieldInfo;
import org.apache.lucene.index.FloatVectorValues;
import org.apache.lucene.index.IndexFileNames;
Expand All @@ -37,6 +38,7 @@
import java.io.IOException;
import java.nio.ByteBuffer;
import java.nio.ByteOrder;
import java.util.AbstractList;
import java.util.ArrayList;
import java.util.List;
import java.util.function.Consumer;
Expand Down Expand Up @@ -265,14 +267,38 @@ private static KMeansFloatVectorValues getKMeansFloatVectorValues(
vectorTransform.accept(vectors);
if (vectors.size() == maxDoc && sortMap == null) {
return KMeansFloatVectorValues.build(vectors, null, fieldInfo.getVectorDimension());
} else if (sortMap == null) {
final DocIdSetIterator iterator = fieldVectorsWriter.getDocsWithFieldSet().iterator();
final int[] docIds = new int[vectors.size()];
for (int i = 0; i < docIds.length; i++) {
docIds[i] = iterator.nextDoc();
}
assert iterator.nextDoc() == NO_MORE_DOCS;
return KMeansFloatVectorValues.build(vectors, docIds, fieldInfo.getVectorDimension());
} else {
DocsWithFieldSet newDocsWithField = new DocsWithFieldSet();
final int[] ordMap = new int[fieldVectorsWriter.getDocsWithFieldSet().cardinality()]; // new ord to old ord
KnnVectorsWriter.mapOldOrdToNewOrd(fieldVectorsWriter.getDocsWithFieldSet(), sortMap, null, ordMap, newDocsWithField);
Copy link
Copy Markdown
Member

@benwtrent benwtrent Mar 27, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ah yeah! I forgot about this helper!

final DocIdSetIterator iterator = newDocsWithField.iterator();
final int[] docIds = new int[vectors.size()];
for (int i = 0; i < docIds.length; i++) {
docIds[i] = iterator.nextDoc();
}
assert iterator.nextDoc() == NO_MORE_DOCS;
List<float[]> orderedVectors = new AbstractList<>() {

@Override
public int size() {
return vectors.size();
}

@Override
public float[] get(int index) {
return vectors.get(ordMap[index]);
}
};
return KMeansFloatVectorValues.build(orderedVectors, docIds, fieldInfo.getVectorDimension());
}
final DocIdSetIterator iterator = fieldVectorsWriter.getDocsWithFieldSet().iterator();
final int[] docIds = new int[vectors.size()];
for (int i = 0; i < docIds.length; i++) {
docIds[i] = sortMap == null ? iterator.nextDoc() : sortMap.oldToNew(iterator.nextDoc());
}
assert iterator.nextDoc() == NO_MORE_DOCS;
return KMeansFloatVectorValues.build(vectors, docIds, fieldInfo.getVectorDimension());
}

/**
Expand Down
108 changes: 71 additions & 37 deletions ...va/org/elasticsearch/index/codec/vectors/diskbbq/es94/ES940DiskBBQVectorsFormatTests.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,9 +17,9 @@
import org.apache.lucene.codecs.perfield.PerFieldKnnVectorsFormat;
import org.apache.lucene.document.Document;
import org.apache.lucene.document.Field;
import org.apache.lucene.document.KeywordField;
import org.apache.lucene.document.KnnFloatVectorField;
import org.apache.lucene.document.SortedDocValuesField;
import org.apache.lucene.document.StringField;
import org.apache.lucene.index.CodecReader;
import org.apache.lucene.index.DirectoryReader;
import org.apache.lucene.index.IndexReader;
Expand All @@ -34,9 +34,11 @@
import org.apache.lucene.search.AcceptDocs;
import org.apache.lucene.search.DocIdSetIterator;
import org.apache.lucene.search.KnnCollector;
import org.apache.lucene.search.ScoreDoc;
import org.apache.lucene.search.Sort;
import org.apache.lucene.search.SortField;
import org.apache.lucene.search.TopDocs;
import org.apache.lucene.search.TopDocsCollector;
import org.apache.lucene.search.TopKnnCollector;
import org.apache.lucene.store.Directory;
import org.apache.lucene.tests.index.BaseKnnVectorsFormatTestCase;
Expand Down Expand Up @@ -398,25 +400,37 @@ private void doRestrictiveFilter(boolean dense) throws IOException {
int maxMatchingDocs = random().nextInt(1, 10);
int matchingDocs = 0;
int numDocs = random().nextInt(100, 3_000);
try (Directory dir = newDirectory(); IndexWriter w = new IndexWriter(dir, newIndexWriterConfig())) {
IndexWriterConfig iwc = newIndexWriterConfig();
if (random().nextBoolean()) {
iwc.setIndexSort(new Sort(new SortField("k", SortField.Type.STRING)));
}
try (Directory dir = newDirectory(); IndexWriter w = new IndexWriter(dir, iwc)) {
for (int i = 0; i < numDocs; i++) {
Document doc = new Document();
if (dense || rarely() == false) {
float[] vector = randomVector(dimensions);
doc.add(new KnnFloatVectorField("f", vector, VectorSimilarityFunction.EUCLIDEAN));
}
doc.add(new KeywordField("k", new BytesRef("B"), Field.Store.NO));
if (random().nextBoolean()) {
doc.add(new StringField("k", new BytesRef("A"), Field.Store.YES));
doc.add(new SortedDocValuesField("k", new BytesRef("A")));
} else {
doc.add(new StringField("k", new BytesRef("C"), Field.Store.YES));
doc.add(new SortedDocValuesField("k", new BytesRef("C")));
}
w.addDocument(doc);
if (matchingDocs < maxMatchingDocs && rarely()) {
matchingDocs++;
doc = new Document();
doc.add(new KnnFloatVectorField("f", randomVector(dimensions), VectorSimilarityFunction.EUCLIDEAN));
doc.add(new KeywordField("k", new BytesRef("A"), Field.Store.NO));
doc.add(new StringField("k", new BytesRef("B"), Field.Store.YES));
doc.add(new SortedDocValuesField("k", new BytesRef("B")));
w.addDocument(doc);
}
if (dense == false && rarely()) {
doc = new Document();
doc.add(new KeywordField("k", new BytesRef("A"), Field.Store.NO));
doc.add(new StringField("k", new BytesRef("B"), Field.Store.YES));
doc.add(new SortedDocValuesField("k", new BytesRef("B")));
w.addDocument(doc);
}
}
Expand All @@ -426,49 +440,69 @@ private void doRestrictiveFilter(boolean dense) throws IOException {
float[] vector = randomVector(dimensions);
Document doc = new Document();
doc.add(new KnnFloatVectorField("f", vector, VectorSimilarityFunction.EUCLIDEAN));
doc.add(new KeywordField("k", new BytesRef("A"), Field.Store.NO));
doc.add(new StringField("k", new BytesRef("B"), Field.Store.YES));
doc.add(new SortedDocValuesField("k", new BytesRef("B")));
w.addDocument(doc);
}
w.commit();
// force one leave
w.forceMerge(1);
if (random().nextBoolean()) {
// force one leave
w.forceMerge(1);
}
try (IndexReader reader = DirectoryReader.open(w)) {
LeafReader leafReader = getOnlyLeafReader(reader);
float[] vector = randomVector(dimensions);
// we might collect the same document twice because of soar assignments
KnnCollector collector;
if (random().nextBoolean()) {
collector = new TopKnnCollector(random().nextInt(2 * matchingDocs, 3 * matchingDocs), Integer.MAX_VALUE);
} else {
collector = new TopKnnCollector(
random().nextInt(2 * matchingDocs, 3 * matchingDocs),
Integer.MAX_VALUE,
new IVFKnnSearchStrategy(0.25f, null)
TopDocs[] topDocsArray = new TopDocs[reader.leaves().size()];
for (int i = 0; i < reader.leaves().size(); i++) {
LeafReaderContext context = reader.leaves().get(i);
LeafReader leafReader = context.reader();
float[] vector = randomVector(dimensions);
// we might collect the same document twice because of soar assignments
KnnCollector collector;
if (random().nextBoolean()) {
collector = new TopKnnCollector(random().nextInt(2 * matchingDocs, 3 * matchingDocs), Integer.MAX_VALUE);
} else {
collector = new TopKnnCollector(
random().nextInt(2 * matchingDocs, 3 * matchingDocs),
Integer.MAX_VALUE,
new IVFKnnSearchStrategy(0.25f, null)
);
}
if (leafReader.postings(new Term("k", new BytesRef("B"))) == null) {
topDocsArray[i] = TopDocsCollector.EMPTY_TOPDOCS;
continue;
}
leafReader.searchNearestVectors(
"f",
vector,
collector,
AcceptDocs.fromIteratorSupplier(
() -> leafReader.postings(new Term("k", new BytesRef("B"))),
leafReader.getLiveDocs(),
leafReader.maxDoc()
)
);
TopDocs leafTopDocs = collector.topDocs();
ScoreDoc[] adjusted = new ScoreDoc[leafTopDocs.scoreDocs.length];
for (int docIndex = 0; docIndex < leafTopDocs.scoreDocs.length; docIndex++) {
ScoreDoc scoreDoc = leafTopDocs.scoreDocs[docIndex];
adjusted[docIndex] = new ScoreDoc(scoreDoc.doc + context.docBase, scoreDoc.score);
}
topDocsArray[i] = new TopDocs(leafTopDocs.totalHits, adjusted);
// match no docs
leafReader.searchNearestVectors(
"f",
vector,
new TopKnnCollector(2, Integer.MAX_VALUE),
AcceptDocs.fromIteratorSupplier(DocIdSetIterator::empty, leafReader.getLiveDocs(), leafReader.maxDoc())
);
}
leafReader.searchNearestVectors(
"f",
vector,
collector,
AcceptDocs.fromIteratorSupplier(
() -> leafReader.postings(new Term("k", new BytesRef("A"))),
leafReader.getLiveDocs(),
leafReader.maxDoc()
)
);
TopDocs topDocs = collector.topDocs();
TopDocs topDocs = TopDocs.merge(2 * maxMatchingDocs, topDocsArray);
Set<Integer> uniqueDocIds = new HashSet<>();
for (int i = 0; i < topDocs.scoreDocs.length; i++) {
uniqueDocIds.add(topDocs.scoreDocs[i].doc);
Document document = reader.storedFields().document(topDocs.scoreDocs[i].doc);
assertThat(document.getField("k").binaryValue().utf8ToString(), equalTo("B"));
}
assertEquals(matchingDocs, uniqueDocIds.size());
// match no docs
leafReader.searchNearestVectors(
"f",
vector,
new TopKnnCollector(2, Integer.MAX_VALUE),
AcceptDocs.fromIteratorSupplier(DocIdSetIterator::empty, leafReader.getLiveDocs(), leafReader.maxDoc())
);
}
}
}
Expand Down
Loading
Loading