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>
2025-12-29 09:18:01 +01:00 · 2025-02-25 18:42:32 +02:00
parent ff353fc57c
commit e24bd67658
10 changed files with 417 additions and 23 deletions
--- a/Benchmarks/Text/RuneExtensions/Encode.cs
+++ b/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 };
+        }
+    }
+}