Handle deduplication of join records

Author: fabien.menager

src/PhenX.EntityFrameworkCore.BulkInsert/Graph/EntityPairEqualityComparer.cs

@@ -0,0 +1,28 @@
+using System.Runtime.CompilerServices;
+
+namespace PhenX.EntityFrameworkCore.BulkInsert.Graph;
+
+/// <summary>
+/// Compares pairs of entity references for equality using reference equality.
+/// Used for deduplicating many-to-many join records.
+/// </summary>
+internal sealed class EntityPairEqualityComparer : IEqualityComparer<(object Left, object Right)>
+{
+    public static readonly EntityPairEqualityComparer Instance = new();
+
+    private EntityPairEqualityComparer() { }
+
+    public bool Equals((object Left, object Right) x, (object Left, object Right) y)
+    {
+        return ReferenceEquals(x.Left, y.Left) && ReferenceEquals(x.Right, y.Right);
+    }
+
+    public int GetHashCode((object Left, object Right) obj)
+    {
+        return HashCode.Combine(
+            RuntimeHelpers.GetHashCode(obj.Left),
+            RuntimeHelpers.GetHashCode(obj.Right)
+        );
+    }
+}
+

src/PhenX.EntityFrameworkCore.BulkInsert/Graph/GraphBulkInsertOrchestrator.cs

@@ -347,11 +347,19 @@ private async Task<int> InsertJoinRecords(
             // Get entity metadata for join type
             var joinEntityMetadata = graphMetadata.GetEntityMetadata(joinEntityType);
 
-            // Create join table entries
+            // Deduplicate join records by entity reference pairs
+            // Use a HashSet to track which (LeftEntity, RightEntity) pairs have been processed
+            var seenPairs = new HashSet<(object Left, object Right)>(EntityPairEqualityComparer.Instance);
             var joinEntities = new List<object>();
 
             foreach (var record in records)
             {
+                // Skip if this exact pair of entity instances has already been processed
+                if (!seenPairs.Add((record.LeftEntity, record.RightEntity)))
+                {
+                    continue;
+                }
+
                 // Get metadata for left and right entities
                 var leftMetadata = graphMetadata.GetEntityMetadata(record.LeftEntity.GetType());
                 var rightMetadata = graphMetadata.GetEntityMetadata(record.RightEntity.GetType());
@@ -415,7 +423,7 @@ private async Task<int> InsertJoinRecords(
 
             if (joinEntities.Count > 0)
             {
-                // Insert join entities
+                // Insert join entities (only unique ones)
                 await InsertJoinEntities(sync, context, joinEntityType, joinEntities, options, provider, ctk);
                 totalJoinRecordsInserted += joinEntities.Count;
             }