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>
2025-12-27 00:07:58 +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/GetEncodingLength.cs
+++ b/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;
+    }
+}