Skip to content

Apply HNSW merge optimizations when there are "no deleted vectors" instead of "no deleted documents"? #15429

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

Closed
kaivalnp wants to merge 2 commits into from
Closed

Apply HNSW merge optimizations when there are "no deleted vectors" instead of "no deleted documents"? #15429

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6969248
Apply HNSW merge optimizations when there are "no deleted vectors" in…
kaivalnp Nov 14, 2025
b58f570
Add CHANGES.txt entry
kaivalnp Nov 14, 2025
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
3 changes: 3 additions & 0 deletions lucene/CHANGES.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -189,6 +189,9 @@ Improvements

* GITHUB#15425: Refactoring internal HnswGraph.NodesIterator to avoid unneeded copying and sorting (Mike Sokolov)

* GITHUB#15429: Make the criteria to re-use per-segment HNSW graph information in merge optimizations more appropriate,
from "no deleted documents" to "no deleted vectors". (Kaival Parikh)

Optimizations
---------------------
* GITHUB#15140: Optimize TopScoreDocCollector with TernaryLongHeap for improved performance over Binary-LongHeap. (Ramakrishna Chilaka)
Expand Down
40 changes: 17 additions & 23 deletions lucene/core/src/java/org/apache/lucene/util/hnsw/IncrementalHnswGraphMerger.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,12 +24,11 @@
import java.util.List;
import org.apache.lucene.codecs.KnnVectorsReader;
import org.apache.lucene.codecs.hnsw.HnswGraphProvider;
import org.apache.lucene.index.ByteVectorValues;
import org.apache.lucene.index.FieldInfo;
import org.apache.lucene.index.FloatVectorValues;
import org.apache.lucene.index.KnnVectorValues;
import org.apache.lucene.index.MergeState;
import org.apache.lucene.internal.hppc.IntIntHashMap;
import org.apache.lucene.search.DocIdSetIterator;
import org.apache.lucene.util.BitSet;
import org.apache.lucene.util.Bits;
import org.apache.lucene.util.FixedBitSet;
Expand Down Expand Up @@ -73,31 +72,26 @@ public IncrementalHnswGraphMerger(
public IncrementalHnswGraphMerger addReader(
KnnVectorsReader reader, MergeState.DocMap docMap, Bits liveDocs) throws IOException {
numReaders++;
if (hasDeletes(liveDocs) || !(reader instanceof HnswGraphProvider)) {
if (!(reader instanceof HnswGraphProvider provider)) {
return this;
}
HnswGraph graph = ((HnswGraphProvider) reader).getGraph(fieldInfo.name);

HnswGraph graph = provider.getGraph(fieldInfo.name);
if (graph == null || graph.size() == 0) {
return this;
}

int candidateVectorCount = 0;
switch (fieldInfo.getVectorEncoding()) {
case BYTE -> {
ByteVectorValues byteVectorValues = reader.getByteVectorValues(fieldInfo.name);
if (byteVectorValues == null) {
return this;
}
candidateVectorCount = byteVectorValues.size();
}
case FLOAT32 -> {
FloatVectorValues vectorValues = reader.getFloatVectorValues(fieldInfo.name);
if (vectorValues == null) {
return this;
}
candidateVectorCount = vectorValues.size();
}
KnnVectorValues values =
switch (fieldInfo.getVectorEncoding()) {
case BYTE -> reader.getByteVectorValues(fieldInfo.name);
case FLOAT32 -> reader.getFloatVectorValues(fieldInfo.name);
};

if (values == null || hasDeletes(values.iterator(), liveDocs)) {
return this;
}

int candidateVectorCount = values.size();
graphReaders.add(new GraphReader(reader, docMap, candidateVectorCount));
return this;
}
Expand Down Expand Up @@ -192,13 +186,13 @@ public OnHeapHnswGraph merge(
return builder.build(maxOrd);
}

private static boolean hasDeletes(Bits liveDocs) {
private static boolean hasDeletes(DocIdSetIterator iterator, Bits liveDocs) throws IOException {
if (liveDocs == null) {
return false;
}

for (int i = 0; i < liveDocs.length(); i++) {
if (!liveDocs.get(i)) {
for (int doc = iterator.nextDoc(); doc != NO_MORE_DOCS; doc = iterator.nextDoc()) {
if (!liveDocs.get(doc)) {
return true;
}
}
Expand Down