Fix Rule-based auto tagging label resolving logic (opensearch-project#19599)

Signed-off-by: Ruirui Zhang <mariazrr@amazon.com>

Author: ruai0511

modules/autotagging-commons/common/src/main/java/org/opensearch/rule/storage/AttributeValueStore.java

@@ -11,10 +11,8 @@
 import org.opensearch.rule.MatchLabel;
 
 import java.util.ArrayList;
-import java.util.HashSet;
 import java.util.List;
 import java.util.Optional;
-import java.util.Set;
 
 /**
  * This interface provides apis to store Rule attribute values
@@ -56,8 +54,8 @@ default List<MatchLabel<V>> getMatches(K key) {
      * Returns the values that exactly match the given key.
      * @param key the key to look up
      */
-    default Set<V> getExactMatch(K key) {
-        return new HashSet<>();
+    default List<MatchLabel<V>> getExactMatch(K key) {
+        return new ArrayList<>();
     }
 
     /**

modules/autotagging-commons/common/src/main/java/org/opensearch/rule/storage/DefaultAttributeValueStore.java

@@ -12,7 +12,6 @@
 import org.opensearch.rule.MatchLabel;
 
 import java.util.ArrayList;
-import java.util.Collections;
 import java.util.HashSet;
 import java.util.List;
 import java.util.Optional;
@@ -80,12 +79,12 @@ public Optional<V> get(K key) {
     }
 
     @Override
-    public Set<V> getExactMatch(K key) {
+    public List<MatchLabel<V>> getExactMatch(K key) {
         readLock.lock();
         try {
-            Set<V> results = new HashSet<>();
-            results.addAll(trie.getOrDefault(key, Collections.emptySet()));
-            results.addAll(trie.getOrDefault("", Collections.emptySet()));
+            List<MatchLabel<V>> results = new ArrayList<>();
+            addMatches(results, trie.get(key), 1f);
+            addMatches(results, trie.get(""), 0f);
             return results;
         } finally {
             readLock.unlock();
@@ -100,13 +99,10 @@ public List<MatchLabel<V>> getMatches(String key) {
             StringBuilder prefixBuilder = new StringBuilder(key);
 
             for (int i = key.length(); i >= 0; i--) {
-                String prefix = prefixBuilder.toString();
-                Set<V> value = trie.get(prefix);
-                if (value != null && !value.isEmpty()) {
-                    float matchScore = (float) prefixBuilder.length() / key.length();
-                    for (V label : value) {
-                        results.add(new MatchLabel<>(label, matchScore));
-                    }
+                Set<V> values = trie.get(prefixBuilder.toString());
+                if (values != null && !values.isEmpty()) {
+                    float score = (float) prefixBuilder.length() / key.length();
+                    addMatches(results, values, score);
                 }
                 if (!prefixBuilder.isEmpty()) {
                     prefixBuilder.deleteCharAt(prefixBuilder.length() - 1);
@@ -118,6 +114,15 @@ public List<MatchLabel<V>> getMatches(String key) {
         }
     }
 
+    private void addMatches(List<MatchLabel<V>> results, Set<V> values, float score) {
+        if (values == null || values.isEmpty()) {
+            return;
+        }
+        for (V label : values) {
+            results.add(new MatchLabel<>(label, score));
+        }
+    }
+
     @Override
     public void clear() {
         writeLock.lock();

modules/autotagging-commons/common/src/test/java/org/opensearch/rule/storage/AttributeValueStoreTests.java

@@ -174,5 +174,8 @@ public int size() {
         List<MatchLabel<String>> result = store.getMatches("foo");
         assertNotNull(result);
         assertTrue(result.isEmpty());
+        List<MatchLabel<String>> exactMatches = store.getExactMatch("foo");
+        assertNotNull(exactMatches);
+        assertTrue(exactMatches.isEmpty());
     }
 }

release-notes/opensearch.release-notes-3.3.0.md

@@ -111,6 +111,7 @@ Compatible with OpenSearch and OpenSearch Dashboards version 3.3.0
 * Remove unnecessary iteration per-shard in request cache cleanup ([#19263](https://github.com/opensearch-project/OpenSearch/pull/19263))
 * Fix derived field rewrite to handle range queries ([#19496](https://github.com/opensearch-project/OpenSearch/pull/19496))
 * Fix incorrect rewriting of terms query with more than two consecutive whole numbers ([#19587](https://github.com/opensearch-project/OpenSearch/pull/19587))
+* Fix Rule-based auto tagging label resolving logic ([#19599](https://github.com/opensearch-project/OpenSearch/pull/19599))
 
 ### Dependencies
 * Bump `com.gradleup.shadow:shadow-gradle-plugin` from 8.3.5 to 8.3.9 ([#19400](https://github.com/opensearch-project/OpenSearch/pull/19400))