Add an additional length check to default FrozenDictionary and Frozen…

…Set (#92546) * Add an additional length check to FrozenDictionary and FrozenSet On construction of the collection, we compute an unsigned long which is effectively 64 boolean flags, each representing the presence of a key string of a particular length (mod 64). When reading from the collection, we can exit early if the key being tested does not map to a bit which has been switched on by the original computation. I believe this has similarities to how Bloom Filters work. This adds a relatively small cost on creation of the collection as small cost to each read operation. However it can speed up reads with certain data patterns especially when the difference between the maximum and minimum key length is large but there aren't many different lengths.
dotnet · Dec 11, 2023 · c28bec4 · c28bec4
1 parent 6de7549
commit c28bec4
Show file tree

Hide file tree

Showing 10 changed files with 72 additions and 30 deletions.
diff --git a/src/libraries/System.Collections.Immutable/src/System/Collections/Frozen/FrozenDictionary.cs b/src/libraries/System.Collections.Immutable/src/System/Collections/Frozen/FrozenDictionary.cs
@@ -161,10 +161,12 @@ private static FrozenDictionary<TKey, TValue> CreateFromDictionary<TKey, TValue>
 
                 // Calculate the minimum and maximum lengths of the strings in the dictionary. Several of the analyses need this.
                 int minLength = int.MaxValue, maxLength = 0;
+                ulong lengthFilter = 0;
                 foreach (string key in keys)
                 {
                     if (key.Length < minLength) minLength = key.Length;
                     if (key.Length > maxLength) maxLength = key.Length;
+                    lengthFilter |= (1UL << (key.Length % 64));
                 }
                 Debug.Assert(minLength >= 0 && maxLength >= minLength);
 
@@ -215,12 +217,12 @@ private static FrozenDictionary<TKey, TValue> CreateFromDictionary<TKey, TValue>
                     if (analysis.IgnoreCase)
                     {
                         frozenDictionary = analysis.AllAsciiIfIgnoreCase
-                            ? new OrdinalStringFrozenDictionary_FullCaseInsensitiveAscii<TValue>(keys, values, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff)
-                            : new OrdinalStringFrozenDictionary_FullCaseInsensitive<TValue>(keys, values, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff);
+                            ? new OrdinalStringFrozenDictionary_FullCaseInsensitiveAscii<TValue>(keys, values, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff, lengthFilter)
+                            : new OrdinalStringFrozenDictionary_FullCaseInsensitive<TValue>(keys, values, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff, lengthFilter);
                     }
                     else
                     {
-                        frozenDictionary = new OrdinalStringFrozenDictionary_Full<TValue>(keys, values, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff);
+                        frozenDictionary = new OrdinalStringFrozenDictionary_Full<TValue>(keys, values, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff, lengthFilter);
                     }
                 }
 

diff --git a/src/libraries/System.Collections.Immutable/src/System/Collections/Frozen/FrozenSet.cs b/src/libraries/System.Collections.Immutable/src/System/Collections/Frozen/FrozenSet.cs
@@ -109,10 +109,12 @@ private static FrozenSet<T> CreateFromSet<T>(HashSet<T> source)
 
                 // Calculate the minimum and maximum lengths of the strings in the set. Several of the analyses need this.
                 int minLength = int.MaxValue, maxLength = 0;
+                ulong lengthFilter = 0;
                 foreach (string s in entries)
                 {
                     if (s.Length < minLength) minLength = s.Length;
                     if (s.Length > maxLength) maxLength = s.Length;
+                    lengthFilter |= (1UL << (s.Length % 64));
                 }
                 Debug.Assert(minLength >= 0 && maxLength >= minLength);
 
@@ -163,12 +165,12 @@ private static FrozenSet<T> CreateFromSet<T>(HashSet<T> source)
                     if (analysis.IgnoreCase)
                     {
                         frozenSet = analysis.AllAsciiIfIgnoreCase
-                            ? new OrdinalStringFrozenSet_FullCaseInsensitiveAscii(entries, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff)
-                            : new OrdinalStringFrozenSet_FullCaseInsensitive(entries, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff);
+                            ? new OrdinalStringFrozenSet_FullCaseInsensitiveAscii(entries, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff, lengthFilter)
+                            : new OrdinalStringFrozenSet_FullCaseInsensitive(entries, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff, lengthFilter);
                     }
                     else
                     {
-                        frozenSet = new OrdinalStringFrozenSet_Full(entries, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff);
+                        frozenSet = new OrdinalStringFrozenSet_Full(entries, stringComparer, analysis.MinimumLength, analysis.MaximumLengthDiff, lengthFilter);
                     }
                 }
 

diff --git a/...llections.Immutable/src/System/Collections/Frozen/String/OrdinalStringFrozenDictionary.cs b/...llections.Immutable/src/System/Collections/Frozen/String/OrdinalStringFrozenDictionary.cs
@@ -64,6 +64,7 @@ internal OrdinalStringFrozenDictionary(
         private protected int HashCount { get; }
         private protected abstract bool Equals(string? x, string? y);
         private protected abstract int GetHashCode(string s);
+        private protected virtual bool CheckLengthQuick(string key) => true;
         private protected override string[] KeysCore => _keys;
         private protected override TValue[] ValuesCore => _values;
         private protected override Enumerator GetEnumeratorCore() => new Enumerator(_keys, _values);
@@ -74,20 +75,23 @@ private protected override ref readonly TValue GetValueRefOrNullRefCore(string k
         {
             if ((uint)(key.Length - _minimumLength) <= (uint)_maximumLengthDiff)
             {
-                int hashCode = GetHashCode(key);
-                _hashTable.FindMatchingEntries(hashCode, out int index, out int endIndex);
-
-                while (index <= endIndex)
+                if (CheckLengthQuick(key))
                 {
-                    if (hashCode == _hashTable.HashCodes[index])
+                    int hashCode = GetHashCode(key);
+                    _hashTable.FindMatchingEntries(hashCode, out int index, out int endIndex);
+
+                    while (index <= endIndex)
                     {
-                        if (Equals(key, _keys[index]))
+                        if (hashCode == _hashTable.HashCodes[index])
                         {
-                            return ref _values[index];
+                            if (Equals(key, _keys[index]))
+                            {
+                                return ref _values[index];
+                            }
                         }
-                    }
 
-                    index++;
+                        index++;
+                    }
                 }
             }
 

diff --git a/...ions.Immutable/src/System/Collections/Frozen/String/OrdinalStringFrozenDictionary_Full.cs b/...ions.Immutable/src/System/Collections/Frozen/String/OrdinalStringFrozenDictionary_Full.cs
@@ -7,14 +7,18 @@ namespace System.Collections.Frozen
 {
     internal sealed class OrdinalStringFrozenDictionary_Full<TValue> : OrdinalStringFrozenDictionary<TValue>
     {
+        private readonly ulong _lengthFilter;
+
         internal OrdinalStringFrozenDictionary_Full(
             string[] keys,
             TValue[] values,
             IEqualityComparer<string> comparer,
             int minimumLength,
-            int maximumLengthDiff)
+            int maximumLengthDiff,
+            ulong lengthFilter)
             : base(keys, values, comparer, minimumLength, maximumLengthDiff)
         {
+            _lengthFilter = lengthFilter;
         }
 
         // This override is necessary to force the jit to emit the code in such a way that it
@@ -24,5 +28,6 @@ internal OrdinalStringFrozenDictionary_Full(
 
         private protected override bool Equals(string? x, string? y) => string.Equals(x, y);
         private protected override int GetHashCode(string s) => Hashing.GetHashCodeOrdinal(s.AsSpan());
+        private protected override bool CheckLengthQuick(string key) => (_lengthFilter & (1UL << (key.Length % 64))) > 0;
     }
 }
diff --git a/...src/System/Collections/Frozen/String/OrdinalStringFrozenDictionary_FullCaseInsensitive.cs b/...src/System/Collections/Frozen/String/OrdinalStringFrozenDictionary_FullCaseInsensitive.cs
@@ -7,14 +7,18 @@ namespace System.Collections.Frozen
 {
     internal sealed class OrdinalStringFrozenDictionary_FullCaseInsensitive<TValue> : OrdinalStringFrozenDictionary<TValue>
     {
+        private readonly ulong _lengthFilter;
+
         internal OrdinalStringFrozenDictionary_FullCaseInsensitive(
             string[] keys,
             TValue[] values,
             IEqualityComparer<string> comparer,
             int minimumLength,
-            int maximumLengthDiff)
+            int maximumLengthDiff,
+            ulong lengthFilter)
             : base(keys, values, comparer, minimumLength, maximumLengthDiff)
         {
+            _lengthFilter = lengthFilter;
         }
 
         // This override is necessary to force the jit to emit the code in such a way that it
@@ -24,5 +28,6 @@ internal OrdinalStringFrozenDictionary_FullCaseInsensitive(
 
         private protected override bool Equals(string? x, string? y) => StringComparer.OrdinalIgnoreCase.Equals(x, y);
         private protected override int GetHashCode(string s) => Hashing.GetHashCodeOrdinalIgnoreCase(s.AsSpan());
+        private protected override bool CheckLengthQuick(string key) => (_lengthFilter & (1UL << (key.Length % 64))) > 0;
     }
 }
diff --git a/...ystem/Collections/Frozen/String/OrdinalStringFrozenDictionary_FullCaseInsensitiveAscii.cs b/...ystem/Collections/Frozen/String/OrdinalStringFrozenDictionary_FullCaseInsensitiveAscii.cs
@@ -7,14 +7,18 @@ namespace System.Collections.Frozen
 {
     internal sealed class OrdinalStringFrozenDictionary_FullCaseInsensitiveAscii<TValue> : OrdinalStringFrozenDictionary<TValue>
     {
+        private readonly ulong _lengthFilter;
+
         internal OrdinalStringFrozenDictionary_FullCaseInsensitiveAscii(
             string[] keys,
             TValue[] values,
             IEqualityComparer<string> comparer,
             int minimumLength,
-            int maximumLengthDiff)
+            int maximumLengthDiff,
+            ulong lengthFilter)
             : base(keys, values, comparer, minimumLength, maximumLengthDiff)
         {
+            _lengthFilter = lengthFilter;
         }
 
         // This override is necessary to force the jit to emit the code in such a way that it
@@ -24,5 +28,6 @@ internal OrdinalStringFrozenDictionary_FullCaseInsensitiveAscii(
 
         private protected override bool Equals(string? x, string? y) => StringComparer.OrdinalIgnoreCase.Equals(x, y);
         private protected override int GetHashCode(string s) => Hashing.GetHashCodeOrdinalIgnoreCaseAscii(s.AsSpan());
+        private protected override bool CheckLengthQuick(string key) => (_lengthFilter & (1UL << (key.Length % 64))) > 0;
     }
 }
diff --git a/...stem.Collections.Immutable/src/System/Collections/Frozen/String/OrdinalStringFrozenSet.cs b/...stem.Collections.Immutable/src/System/Collections/Frozen/String/OrdinalStringFrozenSet.cs
@@ -54,6 +54,7 @@ internal OrdinalStringFrozenSet(
         private protected int HashCount { get; }
         private protected abstract bool Equals(string? x, string? y);
         private protected abstract int GetHashCode(string s);
+        private protected virtual bool CheckLengthQuick(string key) => true;
         private protected override string[] ItemsCore => _items;
         private protected override Enumerator GetEnumeratorCore() => new Enumerator(_items);
         private protected override int CountCore => _hashTable.Count;
@@ -64,20 +65,23 @@ private protected override int FindItemIndex(string item)
             if (item is not null && // this implementation won't be used for null values
                 (uint)(item.Length - _minimumLength) <= (uint)_maximumLengthDiff)
             {
-                int hashCode = GetHashCode(item);
-                _hashTable.FindMatchingEntries(hashCode, out int index, out int endIndex);
-
-                while (index <= endIndex)
+                if (CheckLengthQuick(item))
                 {
-                    if (hashCode == _hashTable.HashCodes[index])
+                    int hashCode = GetHashCode(item);
+                    _hashTable.FindMatchingEntries(hashCode, out int index, out int endIndex);
+
+                    while (index <= endIndex)
                     {
-                        if (Equals(item, _items[index]))
+                        if (hashCode == _hashTable.HashCodes[index])
                         {
-                            return index;
+                            if (Equals(item, _items[index]))
+                            {
+                                return index;
+                            }
                         }
-                    }
 
-                    index++;
+                        index++;
+                    }
                 }
             }
 

diff --git a/...Collections.Immutable/src/System/Collections/Frozen/String/OrdinalStringFrozenSet_Full.cs b/...Collections.Immutable/src/System/Collections/Frozen/String/OrdinalStringFrozenSet_Full.cs
@@ -7,13 +7,17 @@ namespace System.Collections.Frozen
 {
     internal sealed class OrdinalStringFrozenSet_Full : OrdinalStringFrozenSet
     {
+        private readonly ulong _lengthFilter;
+
         internal OrdinalStringFrozenSet_Full(
             string[] entries,
             IEqualityComparer<string> comparer,
             int minimumLength,
-            int maximumLengthDiff)
+            int maximumLengthDiff,
+            ulong lengthFilter)
             : base(entries, comparer, minimumLength, maximumLengthDiff)
         {
+            _lengthFilter = lengthFilter;
         }
 
         // This override is necessary to force the jit to emit the code in such a way that it
@@ -23,5 +27,6 @@ internal OrdinalStringFrozenSet_Full(
 
         private protected override bool Equals(string? x, string? y) => string.Equals(x, y);
         private protected override int GetHashCode(string s) => Hashing.GetHashCodeOrdinal(s.AsSpan());
+        private protected override bool CheckLengthQuick(string key) => (_lengthFilter & (1UL << (key.Length % 64))) > 0;
     }
 }
diff --git a/...utable/src/System/Collections/Frozen/String/OrdinalStringFrozenSet_FullCaseInsensitive.cs b/...utable/src/System/Collections/Frozen/String/OrdinalStringFrozenSet_FullCaseInsensitive.cs
@@ -7,13 +7,17 @@ namespace System.Collections.Frozen
 {
     internal sealed class OrdinalStringFrozenSet_FullCaseInsensitive : OrdinalStringFrozenSet
     {
+        private readonly ulong _lengthFilter;
+
         internal OrdinalStringFrozenSet_FullCaseInsensitive(
             string[] entries,
             IEqualityComparer<string> comparer,
             int minimumLength,
-            int maximumLengthDiff)
+            int maximumLengthDiff,
+            ulong lengthFilter)
             : base(entries, comparer, minimumLength, maximumLengthDiff)
         {
+            _lengthFilter = lengthFilter;
         }
 
         // This override is necessary to force the jit to emit the code in such a way that it
@@ -23,5 +27,6 @@ internal OrdinalStringFrozenSet_FullCaseInsensitive(
 
         private protected override bool Equals(string? x, string? y) => StringComparer.OrdinalIgnoreCase.Equals(x, y);
         private protected override int GetHashCode(string s) => Hashing.GetHashCodeOrdinalIgnoreCase(s.AsSpan());
+        private protected override bool CheckLengthQuick(string key) => (_lengthFilter & (1UL << (key.Length % 64))) > 0;
     }
 }
diff --git a/...e/src/System/Collections/Frozen/String/OrdinalStringFrozenSet_FullCaseInsensitiveAscii.cs b/...e/src/System/Collections/Frozen/String/OrdinalStringFrozenSet_FullCaseInsensitiveAscii.cs
@@ -7,13 +7,17 @@ namespace System.Collections.Frozen
 {
     internal sealed class OrdinalStringFrozenSet_FullCaseInsensitiveAscii : OrdinalStringFrozenSet
     {
+        private readonly ulong _lengthFilter;
+
         internal OrdinalStringFrozenSet_FullCaseInsensitiveAscii(
             string[] entries,
             IEqualityComparer<string> comparer,
             int minimumLength,
-            int maximumLengthDiff)
+            int maximumLengthDiff,
+            ulong lengthFilter)
             : base(entries, comparer, minimumLength, maximumLengthDiff)
         {
+            _lengthFilter = lengthFilter;
         }
 
         // This override is necessary to force the jit to emit the code in such a way that it
@@ -23,5 +27,6 @@ internal OrdinalStringFrozenSet_FullCaseInsensitiveAscii(
 
         private protected override bool Equals(string? x, string? y) => StringComparer.OrdinalIgnoreCase.Equals(x, y);
         private protected override int GetHashCode(string s) => Hashing.GetHashCodeOrdinalIgnoreCaseAscii(s.AsSpan());
+        private protected override bool CheckLengthQuick(string key) => (_lengthFilter & (1UL << (key.Length % 64))) > 0;
     }
 }