Rune extensions micro-optimizations (#3910)

* Add benchmarks for potentially optimizable RuneExtensions

* Add new RuneExtensions.DecodeSurrogatePair benchmark implementation

Avoids intermediate heap array allocations which is especially nice when the rune is not surrogate pair because then array heap allocations are completely avoided.

* Enable nullable reference types in RuneExtensions

* Make RuneExtensions.MaxUnicodeCodePoint readonly

Makes sure no one can accidentally change the value. Ideally would be const value.

* Optimize RuneExtensions.DecodeSurrogatePair

* Remove duplicate Rune.GetUnicodeCategory call

* Add new RuneExtensions.IsSurrogatePair benchmark implementation

Avoids intermediate heap allocations by using stack allocated buffer.

* Optimize RuneExtensions.IsSurrogatePair

* Add RuneExtensions.GetEncodingLength tests

* Optimize RuneExtensions.GetEncodingLength

* Optimize RuneExtensions.Encode

* Print encoding name in benchmark results

* Rename variable to better match return description

* Add RuneExtensions.EncodeSurrogatePair benchmark

---------

Co-authored-by: Tig <tig@users.noreply.github.com>

Benchmarks/Benchmarks.csproj

@@ -0,0 +1,20 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <OutputType>Exe</OutputType>
+    <TargetFramework>net8.0</TargetFramework>
+    <IsPackable>false</IsPackable>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <Nullable>enable</Nullable>
+    <RootNamespace>Terminal.Gui.$(MSBuildProjectName.Replace(" ", "_"))</RootNamespace>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="BenchmarkDotNet" Version="0.14.0" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <ProjectReference Include="..\Terminal.Gui\Terminal.Gui.csproj" />
+  </ItemGroup>
+
+</Project>

Benchmarks/Program.cs

@@ -0,0 +1,20 @@
+using BenchmarkDotNet.Configs;
+using BenchmarkDotNet.Running;
+
+namespace Terminal.Gui.Benchmarks;
+
+class Program
+{
+    static void Main (string [] args)
+    {
+        var config = DefaultConfig.Instance;
+
+        // Uncomment for faster but less accurate intermediate iteration.
+        // Final benchmarks should be run with at least the default run length.
+        //config = config.AddJob (BenchmarkDotNet.Jobs.Job.ShortRun);
+
+        BenchmarkSwitcher
+            .FromAssembly (typeof (Program).Assembly)
+            .Run(args, config);
+    }
+}

Benchmarks/Text/RuneExtensions/DecodeSurrogatePair.cs

@@ -0,0 +1,66 @@
+using System.Text;
+using BenchmarkDotNet.Attributes;
+using Tui = Terminal.Gui;
+
+namespace Terminal.Gui.Benchmarks.Text.RuneExtensions;
+
+/// <summary>
+/// Benchmarks for <see cref="Tui.RuneExtensions.DecodeSurrogatePair"/> performance fine-tuning.
+/// </summary>
+[MemoryDiagnoser]
+[BenchmarkCategory (nameof (Tui.RuneExtensions))]
+public class DecodeSurrogatePair
+{
+    /// <summary>
+    /// Benchmark for previous implementation.
+    /// </summary>
+    /// <param name="rune"></param>
+    /// <returns></returns>
+    [Benchmark]
+    [ArgumentsSource (nameof (DataSource))]
+    public char []? Previous (Rune rune)
+    {
+        _ = RuneToStringToCharArray (rune, out char []? chars);
+        return chars;
+    }
+
+    /// <summary>
+    /// Benchmark for current implementation.
+    /// 
+    /// Utilizes Rune methods that take Span argument avoiding intermediate heap array allocation when combined with stack allocated intermediate buffer.
+    /// When rune is not surrogate pair there will be no heap allocation.
+    /// 
+    /// Final surrogate pair array allocation cannot be avoided due to the current method signature design.
+    /// Changing the method signature, or providing an alternative method, to take a destination Span would allow further optimizations by allowing caller to reuse buffer for consecutive calls.
+    /// </summary>
+    [Benchmark (Baseline = true)]
+    [ArgumentsSource (nameof (DataSource))]
+    public char []? Current (Rune rune)
+    {
+        _ = Tui.RuneExtensions.DecodeSurrogatePair (rune, out char []? chars);
+        return chars;
+    }
+
+    /// <summary>
+    /// Previous implementation with intermediate string allocation.
+    /// 
+    /// The IsSurrogatePair implementation at the time had hidden extra string allocation so there were intermediate heap allocations even if rune is not surrogate pair.
+    /// </summary>
+    private static bool RuneToStringToCharArray (Rune rune, out char []? chars)
+    {
+        if (rune.IsSurrogatePair ())
+        {
+            chars = rune.ToString ().ToCharArray ();
+            return true;
+        }
+
+        chars = null;
+        return false;
+    }
+
+    public static IEnumerable<object> DataSource ()
+    {
+        yield return new Rune ('a');
+        yield return "𝔹".EnumerateRunes ().Single ();
+    }
+}

Benchmarks/Text/RuneExtensions/Encode.cs

@@ -0,0 +1,72 @@
+using System.Text;
+using BenchmarkDotNet.Attributes;
+using Tui = Terminal.Gui;
+
+namespace Terminal.Gui.Benchmarks.Text.RuneExtensions;
+
+/// <summary>
+/// Benchmarks for <see cref="Tui.RuneExtensions.Encode"/> performance fine-tuning.
+/// </summary>
+[MemoryDiagnoser]
+[BenchmarkCategory (nameof (Tui.RuneExtensions))]
+public class Encode
+{
+    /// <summary>
+    /// Benchmark for previous implementation.
+    /// </summary>
+    [Benchmark]
+    [ArgumentsSource (nameof (DataSource))]
+    public byte [] Previous (Rune rune, byte [] destination, int start, int count)
+    {
+        _ = StringEncodingGetBytes (rune, destination, start, count);
+        return destination;
+    }
+
+    /// <summary>
+    /// Benchmark for current implementation.
+    /// 
+    /// Avoids intermediate heap allocations with stack allocated intermediate buffer.
+    /// </summary>
+    [Benchmark (Baseline = true)]
+    [ArgumentsSource (nameof (DataSource))]
+    public byte [] Current (Rune rune, byte [] destination, int start, int count)
+    {
+        _ = Tui.RuneExtensions.Encode (rune, destination, start, count);
+        return destination;
+    }
+
+    /// <summary>
+    /// Previous implementation with intermediate byte array and string allocation.
+    /// </summary>
+    private static int StringEncodingGetBytes (Rune rune, byte [] dest, int start = 0, int count = -1)
+    {
+        byte [] bytes = Encoding.UTF8.GetBytes (rune.ToString ());
+        var length = 0;
+
+        for (var i = 0; i < (count == -1 ? bytes.Length : count); i++)
+        {
+            if (bytes [i] == 0)
+            {
+                break;
+            }
+
+            dest [start + i] = bytes [i];
+            length++;
+        }
+
+        return length;
+    }
+
+    public static IEnumerable<object []> DataSource ()
+    {
+        Rune[] runes = [ new Rune ('a'),"𝔞".EnumerateRunes().Single() ];
+
+        foreach (var rune in runes)
+        {
+            yield return new object [] { rune, new byte [16], 0, -1 };
+            yield return new object [] { rune, new byte [16], 8, -1 };
+            // Does not work in original implementation
+            //yield return new object [] { rune, new byte [16], 8, 8 };
+        }
+    }
+}

Benchmarks/Text/RuneExtensions/EncodeSurrogatePair.cs

@@ -0,0 +1,36 @@
+using System.Text;
+using BenchmarkDotNet.Attributes;
+using Tui = Terminal.Gui;
+
+namespace Terminal.Gui.Benchmarks.Text.RuneExtensions;
+
+/// <summary>
+/// Benchmarks for <see cref="Tui.RuneExtensions.EncodeSurrogatePair"/> performance fine-tuning.
+/// </summary>
+[MemoryDiagnoser]
+[BenchmarkCategory (nameof (Tui.RuneExtensions))]
+public class EncodeSurrogatePair
+{
+    /// <summary>
+    /// Benchmark for current implementation.
+    /// </summary>
+    [Benchmark (Baseline = true)]
+    [ArgumentsSource (nameof (DataSource))]
+    public Rune Current (char highSurrogate, char lowSurrogate)
+    {
+        _ = Tui.RuneExtensions.EncodeSurrogatePair (highSurrogate, lowSurrogate, out Rune rune);
+        return rune;
+    }
+
+    public static IEnumerable<object []> DataSource ()
+    {
+        string[] runeStrings = ["🍕", "🧠", "🌹"];
+        foreach (string symbol in runeStrings)
+        {
+            if (symbol is [char high, char low])
+            {
+                yield return [high, low];
+            }
+        }
+    }
+}

Benchmarks/Text/RuneExtensions/GetEncodingLength.cs

@@ -0,0 +1,74 @@
+using System.Text;
+using BenchmarkDotNet.Attributes;
+using Tui = Terminal.Gui;
+
+namespace Terminal.Gui.Benchmarks.Text.RuneExtensions;
+
+/// <summary>
+/// Benchmarks for <see cref="Tui.RuneExtensions.GetEncodingLength"/> performance fine-tuning.
+/// </summary>
+[MemoryDiagnoser]
+[BenchmarkCategory (nameof (Tui.RuneExtensions))]
+public class GetEncodingLength
+{
+    /// <summary>
+    /// Benchmark for previous implementation.
+    /// </summary>
+    [Benchmark]
+    [ArgumentsSource (nameof (DataSource))]
+    public int Previous (Rune rune, PrettyPrintedEncoding encoding)
+    {
+        return WithEncodingGetBytesArray (rune, encoding);
+    }
+
+    /// <summary>
+    /// Benchmark for current implementation.
+    /// </summary>
+    [Benchmark (Baseline = true)]
+    [ArgumentsSource (nameof (DataSource))]
+    public int Current (Rune rune, PrettyPrintedEncoding encoding)
+    {
+        return Tui.RuneExtensions.GetEncodingLength (rune, encoding);
+    }
+
+    /// <summary>
+    /// Previous implementation with intermediate byte array, string, and char array allocation.
+    /// </summary>
+    private static int WithEncodingGetBytesArray (Rune rune, Encoding? encoding = null)
+    {
+        encoding ??= Encoding.UTF8;
+        byte [] bytes = encoding.GetBytes (rune.ToString ().ToCharArray ());
+        var offset = 0;
+
+        if (bytes [^1] == 0)
+        {
+            offset++;
+        }
+
+        return bytes.Length - offset;
+    }
+
+    public static IEnumerable<object []> DataSource ()
+    {
+        PrettyPrintedEncoding[] encodings = [ new(Encoding.UTF8), new(Encoding.Unicode), new(Encoding.UTF32) ];
+        Rune[] runes = [ new Rune ('a'), "𝔹".EnumerateRunes ().Single () ];
+
+        foreach (var encoding in encodings)
+        {
+            foreach (Rune rune in runes)
+            {
+                yield return [rune, encoding];
+            }
+        }
+    }
+
+    /// <summary>
+    /// <see cref="System.Text.Encoding"/> wrapper to display proper encoding name in benchmark results.
+    /// </summary>
+    public record PrettyPrintedEncoding (Encoding Encoding)
+    {
+        public static implicit operator Encoding (PrettyPrintedEncoding ppe) => ppe.Encoding;
+
+        public override string ToString () => Encoding.HeaderName;
+    }
+}

Benchmarks/Text/RuneExtensions/IsSurrogatePair.cs

@@ -0,0 +1,50 @@
+using System.Text;
+using BenchmarkDotNet.Attributes;
+using Tui = Terminal.Gui;
+
+namespace Terminal.Gui.Benchmarks.Text.RuneExtensions;
+
+/// <summary>
+/// Benchmarks for <see cref="Tui.RuneExtensions.IsSurrogatePair"/> performance fine-tuning.
+/// </summary>
+[MemoryDiagnoser]
+[BenchmarkCategory (nameof (Tui.RuneExtensions))]
+public class IsSurrogatePair
+{
+    /// <summary>
+    /// Benchmark for previous implementation.
+    /// </summary>
+    /// <param name="rune"></param>
+    [Benchmark]
+    [ArgumentsSource (nameof (DataSource))]
+    public bool Previous (Rune rune)
+    {
+        return WithToString (rune);
+    }
+
+    /// <summary>
+    /// Benchmark for current implementation.
+    /// 
+    /// Avoids intermediate heap allocations by using stack allocated buffer.
+    /// </summary>
+    [Benchmark (Baseline = true)]
+    [ArgumentsSource (nameof (DataSource))]
+    public bool Current (Rune rune)
+    {
+        return Tui.RuneExtensions.IsSurrogatePair (rune);
+    }
+
+    /// <summary>
+    /// Previous implementation with intermediate string allocation.
+    /// </summary>
+    private static bool WithToString (Rune rune)
+    {
+        return char.IsSurrogatePair (rune.ToString (), 0);
+    }
+
+    public static IEnumerable<object> DataSource ()
+    {
+        yield return new Rune ('a');
+        yield return "𝔹".EnumerateRunes ().Single ();
+    }
+}