Fix NPE for when fields are missing in searchNearestVectors (#13195)

Related to: #13162 Since this is unreleased, no changelog entry is necessary.
apache · Mar 20, 2024 · d5f8853 · d5f8853
1 parent c78533c
commit d5f8853
Show file tree

Hide file tree

Showing 2 changed files with 40 additions and 0 deletions.
diff --git a/lucene/core/src/java/org/apache/lucene/index/LeafReader.java b/lucene/core/src/java/org/apache/lucene/index/LeafReader.java
@@ -246,6 +246,9 @@ public final PostingsEnum postings(Term term) throws IOException {
   public final TopDocs searchNearestVectors(
       String field, float[] target, int k, Bits acceptDocs, int visitedLimit) throws IOException {
     FieldInfo fi = getFieldInfos().fieldInfo(field);
+    if (fi == null || fi.getVectorDimension() == 0) {
+      return TopDocsCollector.EMPTY_TOPDOCS;
+    }
     FloatVectorValues floatVectorValues = getFloatVectorValues(fi.name);
     if (floatVectorValues == null) {
       return TopDocsCollector.EMPTY_TOPDOCS;
@@ -287,6 +290,9 @@ public final TopDocs searchNearestVectors(
   public final TopDocs searchNearestVectors(
       String field, byte[] target, int k, Bits acceptDocs, int visitedLimit) throws IOException {
     FieldInfo fi = getFieldInfos().fieldInfo(field);
+    if (fi == null || fi.getVectorDimension() == 0) {
+      return TopDocsCollector.EMPTY_TOPDOCS;
+    }
     ByteVectorValues byteVectorValues = getByteVectorValues(fi.name);
     if (byteVectorValues == null) {
       return TopDocsCollector.EMPTY_TOPDOCS;

diff --git a/lucene/core/src/test/org/apache/lucene/codecs/perfield/TestPerFieldKnnVectorsFormat.java b/lucene/core/src/test/org/apache/lucene/codecs/perfield/TestPerFieldKnnVectorsFormat.java
@@ -71,6 +71,40 @@ protected Codec getCodec() {
     return codec;
   }
 
+  public void testMissingFieldReturnsNoResults() throws IOException {
+    try (Directory directory = newDirectory()) {
+      IndexWriterConfig iwc = newIndexWriterConfig(new MockAnalyzer(random()));
+      iwc.setCodec(
+          new AssertingCodec() {
+            @Override
+            public KnnVectorsFormat getKnnVectorsFormatForField(String field) {
+              return TestUtil.getDefaultKnnVectorsFormat();
+            }
+          });
+      try (IndexWriter iwriter = new IndexWriter(directory, iwc)) {
+        Document doc = new Document();
+        doc.add(newTextField("id", "1", Field.Store.YES));
+        iwriter.addDocument(doc);
+      }
+
+      try (IndexReader ireader = DirectoryReader.open(directory)) {
+        LeafReader reader = ireader.leaves().get(0).reader();
+        TopDocs hits =
+            reader.searchNearestVectors(
+                "missing_field",
+                new float[] {1, 2, 3},
+                10,
+                reader.getLiveDocs(),
+                Integer.MAX_VALUE);
+        assertEquals(0, hits.scoreDocs.length);
+        hits =
+            reader.searchNearestVectors(
+                "id", new float[] {1, 2, 3}, 10, reader.getLiveDocs(), Integer.MAX_VALUE);
+        assertEquals(0, hits.scoreDocs.length);
+      }
+    }
+  }
+
   public void testTwoFieldsTwoFormats() throws IOException {
     try (Directory directory = newDirectory()) {
       // we don't use RandomIndexWriter because it might add more values than we expect !!!!1