Garbage collection on startup

src/ZoneTree.UnitTests/FixedSizeKeyAndValueTests.cs

@@ -104,6 +104,42 @@ public void IntNullableIntDeleteTest()
         Assert.That(data.ContainsKey(3), Is.True);
     }
 
+    [Test]
+    public void IntStringGarbageCollectionTest()
+    {
+        var dataPath = "data/IntStringGarbageCollectionTest";
+        if (Directory.Exists(dataPath))
+            Directory.Delete(dataPath, true);
+
+        // load and populate tree
+        {
+            using var data = new ZoneTreeFactory<int, string>()
+                .SetDataDirectory(dataPath)
+                .OpenOrCreate();
+            data.TryAtomicAdd(1, "1");
+            data.TryAtomicAdd(2, "2");
+            data.TryAtomicAdd(3, "3");
+            data.TryDelete(2);
+            Assert.That(data.ContainsKey(1), Is.True);
+            Assert.That(data.ContainsKey(2), Is.False);
+            Assert.That(data.ContainsKey(3), Is.True);
+            Assert.That(data.Maintenance.MutableSegment.Length, Is.EqualTo(3));
+        }
+
+        // reload tree and check the length
+        for (var i = 0; i < 3; ++i)
+        {
+            using var data = new ZoneTreeFactory<int, string>()
+                .Configure(options => options.EnableSingleSegmentGarbageCollection = true)
+                .SetDataDirectory(dataPath)
+                .Open();
+            Assert.That(data.ContainsKey(1), Is.True);
+            Assert.That(data.ContainsKey(2), Is.False);
+            Assert.That(data.ContainsKey(3), Is.True);
+            Assert.That(data.Maintenance.MutableSegment.Length, Is.EqualTo(2));
+        }
+    }
+
     [TestCase(true)]
     [TestCase(false)]
     public void StringIntTreeTest(bool useSparseArray)

src/ZoneTree/Core/ZoneTreeLoader.cs

@@ -6,6 +6,7 @@
 using Tenray.ZoneTree.Segments.InMemory;
 using Tenray.ZoneTree.Segments.MultiPart;
 using Tenray.ZoneTree.Segments.NullDisk;
+using Tenray.ZoneTree.WAL;
 
 namespace Tenray.ZoneTree.Core;
 
@@ -145,11 +146,17 @@ void ValidateSegmentOrder()
         }
     }
 
-    void LoadMutableSegment(long maximumOpIndex)
+    IWriteAheadLog<TKey, TValue> LoadMutableSegment(long maximumOpIndex,
+        bool collectGarbage)
     {
         var loader = new MutableSegmentLoader<TKey, TValue>(Options);
         MutableSegment = loader
-            .LoadMutableSegment(ZoneTreeMeta.MutableSegment, maximumOpIndex);
+            .LoadMutableSegment(
+                ZoneTreeMeta.MutableSegment,
+                maximumOpIndex,
+                collectGarbage,
+                out var wal);
+        return wal;
     }
 
     long LoadReadOnlySegments()
@@ -225,13 +232,29 @@ void SetMaximumId()
         maximumId = bs.Count > 0 ? bs.Max() : 0;
         SetMaximumSegmentId(maximumId);
     }
+
     public ZoneTree<TKey, TValue> LoadZoneTree()
     {
         LoadZoneTreeMeta();
         LoadZoneTreeMetaWAL();
         SetMaximumId();
         var maximumOpIndex = LoadReadOnlySegments();
-        LoadMutableSegment(maximumOpIndex);
+        bool collectGarbage = Options.EnableSingleSegmentGarbageCollection && !ZoneTreeMeta.HasDiskSegment && ReadOnlySegments.Count == 0;
+        var mutableSegmentWal = LoadMutableSegment(maximumOpIndex, collectGarbage);
+        if (collectGarbage)
+        {
+            var len = MutableSegment.Length;
+            var keys = new TKey[len];
+            var values = new TValue[len];
+            var iterator = MutableSegment.GetSeekableIterator();
+            var i = 0;
+            while (iterator.Next())
+            {
+                keys[i] = iterator.CurrentKey;
+                values[i++] = iterator.CurrentValue;
+            }
+            mutableSegmentWal.ReplaceWriteAheadLog(keys, values, true);
+        }
         LoadDiskSegment();
         LoadBottomSegments();
         var zoneTree = new ZoneTree<TKey, TValue>(Options, ZoneTreeMeta,

src/ZoneTree/Core/ZoneTreeMeta.cs

@@ -21,7 +21,7 @@ public sealed class ZoneTreeMeta
     public int DiskSegmentMaxItemCount { get; set; } = 20_000_000;
 
     public WriteAheadLogOptions WriteAheadLogOptions { get; set; }
-    
+
     public DiskSegmentOptions DiskSegmentOptions { get; set; }
 
     public long MutableSegment { get; set; }
@@ -31,4 +31,6 @@ public sealed class ZoneTreeMeta
     public long DiskSegment { get; set; }
 
     public IReadOnlyList<long> BottomSegments { get; set; }
+
+    public bool HasDiskSegment => DiskSegment != 0 && BottomSegments?.Count > 0;
 }

src/ZoneTree/Directory.Build.props

@@ -5,8 +5,8 @@
     <Authors>Ahmed Yasin Koculu</Authors>
     <PackageId>ZoneTree</PackageId>
     <Title>ZoneTree</Title>
-    <ProductVersion>1.6.4.0</ProductVersion>
-    <Version>1.6.4.0</Version>
+    <ProductVersion>1.6.5.0</ProductVersion>
+    <Version>1.6.5.0</Version>
     <Authors>Ahmed Yasin Koculu</Authors>
     <AssemblyTitle>ZoneTree</AssemblyTitle>
     <Description>ZoneTree is a persistent, high-performance, transactional, ACID-compliant ordered key-value database for NET. It can operate in memory or on local/cloud storage.</Description>

src/ZoneTree/Options/ZoneTreeOptions.cs

@@ -256,6 +256,13 @@ public void Validate()
     /// </summary>
     public DeleteValueConfigurationValidation DeleteValueConfigurationValidation { get; set; }
 
+    /// <summary>
+    /// If the ZoneTree contains only a single segment (which is the mutable segment),
+    /// there is an opportunity to perform a hard delete of the soft deleted values.
+    /// If enabled, the tree performs garbage collection on load if it is applicable.
+    /// </summary>
+    public bool EnableSingleSegmentGarbageCollection { get; set; }
+
     /// <summary>
     /// Creates default delete delegates for nullable types.
     /// </summary>

src/ZoneTree/Segments/InMemory/MutableSegment.cs

@@ -68,7 +68,8 @@ public sealed class MutableSegment<TKey, TValue> : IMutableSegment<TKey, TValue>
         ZoneTreeOptions<TKey, TValue> options,
         IReadOnlyList<TKey> keys,
         IReadOnlyList<TValue> values,
-        long nextOpIndex)
+        long nextOpIndex,
+        bool collectGarbage)
     {
         SegmentId = segmentId;
         WriteAheadLog = wal;
@@ -85,7 +86,16 @@ public sealed class MutableSegment<TKey, TValue> : IMutableSegment<TKey, TValue>
 
         MarkValueDeleted = options.MarkValueDeleted;
         MutableSegmentMaxItemCount = options.MutableSegmentMaxItemCount;
-        LoadLogEntries(keys, values);
+        if (collectGarbage)
+        {
+            // If there isn't any disk segment and readonly segment,
+            // it is safe to hard delete the soft deleted values.
+            LoadLogEntriesWithGarbageCollection(keys, values);
+        }
+        else
+        {
+            LoadLogEntries(keys, values);
+        }
     }
 
     void LoadLogEntries(IReadOnlyList<TKey> keys, IReadOnlyList<TValue> values)
@@ -101,6 +111,29 @@ void LoadLogEntries(IReadOnlyList<TKey> keys, IReadOnlyList<TValue> values)
         }
     }
 
+    void LoadLogEntriesWithGarbageCollection(
+        IReadOnlyList<TKey> keys,
+        IReadOnlyList<TValue> values)
+    {
+        var distinctKeys =
+            new BTree<TKey, byte>(Options.Comparer, Collections.BTree.Lock.BTreeLockMode.NoLock);
+
+        var isValueDeleted = Options.IsValueDeleted;
+        for (var i = keys.Count - 1; i >= 0; --i)
+        {
+            var key = keys[i];
+            if (distinctKeys.ContainsKey(in key))
+                continue;
+            var value = values[i];
+            distinctKeys.Upsert(in key, 1, out _);
+            if (isValueDeleted(in value))
+            {
+                continue;
+            }
+            BTree.Upsert(in key, in value, out var _);
+        }
+    }
+
     public bool ContainsKey(in TKey key)
     {
         return BTree.ContainsKey(key);

src/ZoneTree/Segments/InMemory/MutableSegmentLoader.cs

@@ -1,6 +1,7 @@
 using Tenray.ZoneTree.Exceptions;
 using Tenray.ZoneTree.Core;
 using Tenray.ZoneTree.Options;
+using Tenray.ZoneTree.WAL;
 
 namespace Tenray.ZoneTree.Segments.InMemory;
 
@@ -14,9 +15,13 @@ public sealed class MutableSegmentLoader<TKey, TValue>
         Options = options;
     }
 
-    public IMutableSegment<TKey, TValue> LoadMutableSegment(long segmentId, long maximumOpIndex)
+    public IMutableSegment<TKey, TValue> LoadMutableSegment(
+        long segmentId,
+        long maximumOpIndex,
+        bool collectGarbage,
+        out IWriteAheadLog<TKey, TValue> wal)
     {
-        var wal = Options.WriteAheadLogProvider
+        wal = Options.WriteAheadLogProvider
             .GetOrCreateWAL(
                 segmentId,
                 ZoneTree<TKey, TValue>.SegmentWalCategory,
@@ -40,8 +45,13 @@ public sealed class MutableSegmentLoader<TKey, TValue>
             }
         }
         maximumOpIndex = Math.Max(result.MaximumOpIndex, maximumOpIndex);
-        return new MutableSegment<TKey, TValue>
-            (segmentId, wal, Options, result.Keys,
-            result.Values, maximumOpIndex + 1);
+        return new MutableSegment<TKey, TValue>(
+            segmentId,
+            wal,
+            Options,
+            result.Keys,
+            result.Values,
+            maximumOpIndex + 1,
+            collectGarbage);
     }
 }

src/ZoneTree/ZoneTreeFactory.cs

@@ -475,6 +475,20 @@ void FillValueSerializer()
                 new ByteArraySerializer() as ISerializer<TValue>;
     }
 
+    void LoadInitialSparseArrays(ZoneTree<TKey, TValue> zoneTree)
+    {
+        if (InitialSparseArrayLength <= 1)
+            return;
+
+        var t1 = Task.Run(() =>
+            zoneTree.Maintenance.DiskSegment.InitSparseArray(InitialSparseArrayLength));
+        Parallel.ForEach(zoneTree.Maintenance.BottomSegments, (bs) =>
+        {
+            bs.InitSparseArray(InitialSparseArrayLength);
+        });
+        t1.Wait();
+    }
+
     /// <summary>
     /// Opens or creates a ZoneTree.
     /// </summary>
@@ -487,13 +501,7 @@ void FillValueSerializer()
         if (loader.ZoneTreeMetaExists)
         {
             var zoneTree = loader.LoadZoneTree();
-            var t1 = Task.Run(() =>
-                zoneTree.Maintenance.DiskSegment.InitSparseArray(InitialSparseArrayLength));
-            Parallel.ForEach(zoneTree.Maintenance.BottomSegments, (bs) =>
-            {
-                bs.InitSparseArray(InitialSparseArrayLength);
-            });
-            t1.Wait();
+            LoadInitialSparseArrays(zoneTree);
             return zoneTree;
         }
         return new ZoneTree<TKey, TValue>(Options);
@@ -526,7 +534,9 @@ void FillValueSerializer()
         var loader = new ZoneTreeLoader<TKey, TValue>(Options);
         if (!loader.ZoneTreeMetaExists)
             throw new DatabaseNotFoundException();
-        return loader.LoadZoneTree();
+        var zoneTree = loader.LoadZoneTree();
+        LoadInitialSparseArrays(zoneTree);
+        return zoneTree;
     }
 
     /// <summary>