Fix elemMatch queries to use array field indexes

nitrite/src/main/java/org/dizitart/no2/filters/ElementMatchFilter.java

@@ -20,9 +20,11 @@
 import org.dizitart.no2.collection.NitriteId;
 import org.dizitart.no2.common.tuples.Pair;
 import org.dizitart.no2.exceptions.FilterException;
+import org.dizitart.no2.index.IndexMap;
 
 import java.lang.reflect.Array;
 import java.util.ArrayList;
+import java.util.HashSet;
 import java.util.List;
 import java.util.Set;
 import java.util.regex.Matcher;
@@ -35,13 +37,19 @@
  * @author Anindya Chatterjee
  * @since 1.0
  */
-class ElementMatchFilter extends NitriteFilter {
-    private final String field;
+class ElementMatchFilter extends ComparableFilter {
     private final Filter elementFilter;
 
     ElementMatchFilter(String field, Filter elementFilter) {
+        super(field, null);
         this.elementFilter = elementFilter;
-        this.field = field;
+    }
+
+    @Override
+    public Comparable<?> getComparable() {
+        // ElementMatchFilter doesn't use the comparable value directly
+        // It delegates to the inner filter for index operations
+        return null;
     }
 
     @Override
@@ -56,7 +64,7 @@ public boolean apply(Pair<NitriteId, Document> element) {
         }
 
         Document document = element.getSecond();
-        Object fieldValue = document.get(field);
+        Object fieldValue = document.get(getField());
         if (fieldValue == null) {
             return false;
         }
@@ -77,9 +85,98 @@ public boolean apply(Pair<NitriteId, Document> element) {
         }
     }
 
+    @Override
+    public List<?> applyOnIndex(IndexMap indexMap) {
+        // If the element filter is a ComparableFilter, we can use the index
+        // Since arrays are indexed by individual elements, we can directly
+        // apply the inner filter on the index
+        if (elementFilter instanceof ComparableFilter) {
+            return ((ComparableFilter) elementFilter).applyOnIndex(indexMap);
+        }
+
+        // For other filter types (AND, OR, NOT with comparable filters),
+        // we need to handle them differently
+        if (elementFilter instanceof AndFilter) {
+            return applyAndFilterOnIndex((AndFilter) elementFilter, indexMap);
+        } else if (elementFilter instanceof OrFilter) {
+            return applyOrFilterOnIndex((OrFilter) elementFilter, indexMap);
+        }
+
+        // If we can't use index, return empty list to trigger collection scan
+        return new ArrayList<>();
+    }
+
+    private List<?> applyAndFilterOnIndex(AndFilter andFilter, IndexMap indexMap) {
+        // For AND filters, we need to check if all filters are comparable
+        // and if so, apply them sequentially (intersection)
+        List<Filter> filters = andFilter.getFilters();
+        List<?> result = null;
+
+        for (Filter filter : filters) {
+            if (filter instanceof ComparableFilter) {
+                List<?> filterResult = ((ComparableFilter) filter).applyOnIndex(indexMap);
+                if (result == null) {
+                    result = filterResult;
+                } else {
+                    // Intersection of results
+                    result = intersect(result, filterResult);
+                }
+                if (result.isEmpty()) {
+                    return result; // Short-circuit if no matches
+                }
+            } else {
+                // If any filter is not comparable, we can't use index
+                return new ArrayList<>();
+            }
+        }
+
+        return result != null ? result : new ArrayList<>();
+    }
+
+    private List<?> applyOrFilterOnIndex(OrFilter orFilter, IndexMap indexMap) {
+        // For OR filters, we union the results from each comparable filter
+        List<Filter> filters = orFilter.getFilters();
+        Set<Object> resultSet = new HashSet<>();
+
+        for (Filter filter : filters) {
+            if (filter instanceof ComparableFilter) {
+                List<?> filterResult = ((ComparableFilter) filter).applyOnIndex(indexMap);
+                if (filterResult != null && !filterResult.isEmpty()) {
+                    resultSet.addAll(filterResult);
+                }
+            } else {
+                // If any filter is not comparable, we can't use index
+                return new ArrayList<>();
+            }
+        }
+
+        return new ArrayList<>(resultSet);
+    }
+
+    private List<?> intersect(List<?> list1, List<?> list2) {
+        if (list1 == null || list1.isEmpty() || list2 == null || list2.isEmpty()) {
+            return new ArrayList<>();
+        }
+
+        // Convert the second list to a set for O(1) lookup
+        Set<Object> set2 = new HashSet<>(list2);
+        List<Object> result = new ArrayList<>();
+
+        for (Object item : list1) {
+            if (item != null && set2.contains(item)) {
+                result.add(item);
+            }
+        }
+        // Explicitly handle intersection of null values
+        if (list1.contains(null) && list2.contains(null)) {
+            result.add(null);
+        }
+        return result;
+    }
+
     @Override
     public String toString() {
-        return "elemMatch(" + field + " : " + elementFilter.toString() + ")";
+        return "elemMatch(" + getField() + " : " + elementFilter.toString() + ")";
     }
 
     @SuppressWarnings("rawtypes")

nitrite/src/test/java/org/dizitart/no2/integration/collection/CollectionFindBySingleFieldIndexTest.java

@@ -20,6 +20,7 @@
 import com.github.javafaker.Faker;
 import org.dizitart.no2.collection.Document;
 import org.dizitart.no2.collection.DocumentCursor;
+import org.dizitart.no2.collection.FindPlan;
 import org.dizitart.no2.collection.NitriteCollection;
 import org.dizitart.no2.common.SortOrder;
 import org.dizitart.no2.exceptions.FilterException;
@@ -627,4 +628,220 @@ public void testSortByIndexAscendingLessThan() {
 
         assertArrayEquals(nonIndexedResult, indexedResult);
     }
+
+    @Test
+    public void testFindByArrayFieldIndexWithElemMatch() {
+        // Create a collection with array field
+        NitriteCollection userCollection = db.getCollection("users");
+
+        // Insert a larger dataset (15k documents as mentioned in the issue)
+        for (int i = 0; i < 15000; i++) {
+            Document doc = Document.createDocument("name", "user" + i)
+                .put("emails", new String[]{"user" + i + "@example.com", "user" + i + "@test.com"});
+            userCollection.insert(doc);
+        }
+
+        // Add a specific test document
+        userCollection.insert(Document.createDocument("name", "testuser")
+            .put("emails", new String[]{"test@gmail.com", "test@example.com"}));
+
+        // Measure query time WITHOUT index
+        long startWithoutIndex = System.nanoTime();
+        DocumentCursor cursorWithoutIndex = userCollection.find(
+            where("emails").elemMatch(org.dizitart.no2.filters.FluentFilter.$.eq("test@gmail.com")));
+        long withoutIndexCount = cursorWithoutIndex.size();
+        long endWithoutIndex = System.nanoTime();
+        long timeWithoutIndex = (endWithoutIndex - startWithoutIndex) / 1_000_000;
+
+        assertEquals(1, withoutIndexCount);
+
+        // Verify collection scan is used when no index exists (no index descriptor)
+        FindPlan planWithoutIndex = cursorWithoutIndex.getFindPlan();
+        assertNull("Index descriptor should be null when no index exists", 
+            planWithoutIndex.getIndexDescriptor());
+
+        // Create index on emails field
+        userCollection.createIndex(IndexOptions.indexOptions(IndexType.NON_UNIQUE), "emails");
+
+        // Measure query time WITH index
+        long startWithIndex = System.nanoTime();
+        DocumentCursor cursorWithIndex = userCollection.find(
+            where("emails").elemMatch(org.dizitart.no2.filters.FluentFilter.$.eq("test@gmail.com")));
+        long withIndexCount = cursorWithIndex.size();
+        long endWithIndex = System.nanoTime();
+        long timeWithIndex = (endWithIndex - startWithIndex) / 1_000_000;
+
+        assertEquals(1, withIndexCount);
+
+        // Verify index is actually being used by checking the find plan
+        FindPlan planWithIndex = cursorWithIndex.getFindPlan();
+        assertNotNull("Index scan filter should not be null when index exists", 
+            planWithIndex.getIndexScanFilter());
+        assertNotNull("Index descriptor should not be null when index is used", 
+            planWithIndex.getIndexDescriptor());
+
+        // With index should be significantly faster
+        System.out.println("ElemMatch query on 15k documents:");
+        System.out.println("  Time without index: " + timeWithoutIndex + " ms");
+        System.out.println("  Time with index: " + timeWithIndex + " ms");
+        System.out.println("  Speedup: " + (timeWithoutIndex > 0 ? (timeWithoutIndex / (double) Math.max(1, timeWithIndex)) : "N/A") + "x");
+
+        // Assert that index provides significant improvement (at least 2x faster)
+        // This is a conservative check - actual improvement should be much higher
+        assertTrue("Index should provide significant performance improvement", 
+            timeWithIndex < timeWithoutIndex || timeWithIndex < 100);
+    }
+
+    @Test
+    public void testFindByArrayFieldIndexWithElemMatchComplexFilter() {
+        // Create a collection with array field
+        NitriteCollection productCollection = db.getCollection("products");
+
+        // Insert documents with array of scores
+        for (int i = 0; i < 1000; i++) {
+            Document doc = Document.createDocument("name", "product" + i)
+                .put("scores", new Integer[]{i, i + 10, i + 20});
+            productCollection.insert(doc);
+        }
+
+        // Create index on scores field
+        productCollection.createIndex(IndexOptions.indexOptions(IndexType.NON_UNIQUE), "scores");
+
+        // Test 1: Query with elemMatch using gt filter
+        DocumentCursor cursor = productCollection.find(
+            where("scores").elemMatch(org.dizitart.no2.filters.FluentFilter.$.gt(995)));
+
+        // Verify index is used
+        FindPlan findPlan = cursor.getFindPlan();
+        assertNotNull("Index scan filter should be used for gt query", findPlan.getIndexScanFilter());
+        assertNotNull("Index descriptor should be present", findPlan.getIndexDescriptor());
+
+        // Should find products where at least one score is > 995
+        assertTrue("Should find products with scores > 995", cursor.size() > 0);
+
+        // Test 2: Query with elemMatch using lt filter
+        cursor = productCollection.find(
+            where("scores").elemMatch(org.dizitart.no2.filters.FluentFilter.$.lt(5)));
+
+        // Verify index is used
+        findPlan = cursor.getFindPlan();
+        assertNotNull("Index scan filter should be used for lt query", findPlan.getIndexScanFilter());
+        assertNotNull("Index descriptor should be present", findPlan.getIndexDescriptor());
+
+        // Should find products where at least one score is < 5
+        assertTrue("Should find products with scores < 5", cursor.size() > 0);
+
+        // Test 3: Query with elemMatch using gte filter
+        cursor = productCollection.find(
+            where("scores").elemMatch(org.dizitart.no2.filters.FluentFilter.$.gte(500)));
+
+        findPlan = cursor.getFindPlan();
+        assertNotNull("Index scan filter should be used for gte query", findPlan.getIndexScanFilter());
+        assertTrue("Should find products with scores >= 500", cursor.size() > 0);
+
+        // Test 4: Query with elemMatch using lte filter
+        cursor = productCollection.find(
+            where("scores").elemMatch(org.dizitart.no2.filters.FluentFilter.$.lte(500)));
+
+        findPlan = cursor.getFindPlan();
+        assertNotNull("Index scan filter should be used for lte query", findPlan.getIndexScanFilter());
+        assertTrue("Should find products with scores <= 500", cursor.size() > 0);
+    }
+
+    @Test
+    public void testElemMatchWithNonUniqueIndex() {
+        // Test that elemMatch works with non-unique index
+        NitriteCollection tagCollection = db.getCollection("tags");
+
+        // Insert documents with tag arrays (some tags are common)
+        for (int i = 0; i < 500; i++) {
+            Document doc = Document.createDocument("id", i)
+                .put("tags", new String[]{"tag" + i, "category" + (i % 10), "item" + i});
+            tagCollection.insert(doc);
+        }
+
+        // Create non-unique index on tags field (since there are duplicate values)
+        tagCollection.createIndex(IndexOptions.indexOptions(IndexType.NON_UNIQUE), "tags");
+
+        // Query with elemMatch
+        DocumentCursor cursor = tagCollection.find(
+            where("tags").elemMatch(org.dizitart.no2.filters.FluentFilter.$.eq("tag100")));
+
+        // Verify index is used
+        FindPlan findPlan = cursor.getFindPlan();
+        assertNotNull("Index scan filter should be used", 
+            findPlan.getIndexScanFilter());
+        assertNotNull("Index descriptor should be present", 
+            findPlan.getIndexDescriptor());
+        assertEquals("Should find exactly one document", 1, cursor.size());
+
+        // Query for a common category tag (should find multiple)
+        cursor = tagCollection.find(
+            where("tags").elemMatch(org.dizitart.no2.filters.FluentFilter.$.eq("category5")));
+
+        findPlan = cursor.getFindPlan();
+        assertNotNull("Index should be used for common values too", 
+            findPlan.getIndexScanFilter());
+        assertEquals("Should find all documents with category5", 50, cursor.size());
+    }
+
+    @Test
+    public void testElemMatchIndexPerformanceComparison() {
+        // This test explicitly measures and compares performance
+        NitriteCollection perfCollection = db.getCollection("performance");
+
+        // Insert a meaningful dataset
+        for (int i = 0; i < 10000; i++) {
+            Document doc = Document.createDocument("id", i)
+                .put("values", new Integer[]{i, i * 2, i * 3});
+            perfCollection.insert(doc);
+        }
+
+        // Add a unique test value that only appears once
+        perfCollection.insert(Document.createDocument("id", 99999)
+            .put("values", new Integer[]{77777, 88888, 99999}));
+
+        // Test WITHOUT index
+        long startNoIndex = System.nanoTime();
+        DocumentCursor noIndexCursor = perfCollection.find(
+            where("values").elemMatch(org.dizitart.no2.filters.FluentFilter.$.eq(99999)));
+        long noIndexCount = noIndexCursor.size();
+        long endNoIndex = System.nanoTime();
+        long timeNoIndex = (endNoIndex - startNoIndex) / 1_000_000;
+
+        // Verify no index was used (no index descriptor)
+        FindPlan noIndexPlan = noIndexCursor.getFindPlan();
+        assertNull("Index descriptor should be null without index", 
+            noIndexPlan.getIndexDescriptor());
+        assertEquals(1, noIndexCount);
+
+        // Create index
+        perfCollection.createIndex(IndexOptions.indexOptions(IndexType.NON_UNIQUE), "values");
+
+        // Test WITH index
+        long startWithIndex = System.nanoTime();
+        DocumentCursor withIndexCursor = perfCollection.find(
+            where("values").elemMatch(org.dizitart.no2.filters.FluentFilter.$.eq(99999)));
+        long withIndexCount = withIndexCursor.size();
+        long endWithIndex = System.nanoTime();
+        long timeWithIndex = (endWithIndex - startWithIndex) / 1_000_000;
+
+        // Verify index was used
+        FindPlan withIndexPlan = withIndexCursor.getFindPlan();
+        assertNotNull("Index scan filter should be used with index", 
+            withIndexPlan.getIndexScanFilter());
+        assertNotNull("Index descriptor should be present", 
+            withIndexPlan.getIndexDescriptor());
+        assertEquals(1, withIndexCount);
+
+        System.out.println("Performance comparison for elemMatch on 10k documents:");
+        System.out.println("  Without index: " + timeNoIndex + " ms");
+        System.out.println("  With index: " + timeWithIndex + " ms");
+        System.out.println("  Improvement: " + 
+            (timeNoIndex > 0 ? String.format("%.1fx", timeNoIndex / (double) Math.max(1, timeWithIndex)) : "N/A"));
+
+        // Index should provide measurable improvement
+        assertTrue("Index should improve performance or complete very quickly", 
+            timeWithIndex < timeNoIndex || timeWithIndex < 100);
+    }
 }