Use direct UTF-8 paths for Char parsing and short string decoding

[EgorBo]

Route generic numeric UTF-8 Char parsing through the existing strict Rune-based parser. Reuse the sealed UTF-8 encoding's existing one-pass short-input decoder for span and validated array-slice GetString calls. Keep subclasses, custom encodings, and large inputs on their existing paths.

Benchmark

using System;
using System.Globalization;
using System.Numerics;
using System.Text;
using BenchmarkDotNet.Attributes;

[MemoryDiagnoser]
public class CharParsingBench
{
    private byte[] _input = null!;

    [Params("A", "\u00E9")]
    public string Text { get; set; } = null!;

    [GlobalSetup]
    public void Setup() => _input = Encoding.UTF8.GetBytes(Text);

    [Benchmark]
    public char NumericUtf8Char() => Parse<char>(_input);

    private static T Parse<T>(ReadOnlySpan<byte> input) where T : struct, INumberBase<T>
    {
        if (!T.TryParse(input, NumberStyles.Integer, CultureInfo.InvariantCulture, out T result))
        {
            throw new FormatException();
        }

        return result;
    }
}

[MemoryDiagnoser]
public class ShortUtf8Bench
{
    private byte[] _input = null!;

    [Params("Content-Length: ", "Gr\u00FC\u00DFe\u4E16\u754C")]
    public string Text { get; set; } = null!;

    [GlobalSetup]
    public void Setup() => _input = Encoding.UTF8.GetBytes(Text);

    [Benchmark]
    public string SpanGetString() => Encoding.UTF8.GetString(_input.AsSpan());

    [Benchmark]
    public string SliceGetString() => Encoding.UTF8.GetString(_input, 0, _input.Length);
}

Results from EgorBot/Benchmarks#652, comparing main against this PR. BenchmarkDotNet 0.16.0-preview.1; ARM64: Apple M4 / macOS 15.7.5; x64: AMD EPYC 9V45 / Ubuntu 24.04.4.

Method	Input	ARM64 main (ns)	ARM64 PR (ns)	Speedup	x64 main (ns)	x64 PR (ns)	Speedup
NumericUtf8Char	"A"	3.7331	0.7594	4.92x	4.9432	0.9233	5.35x
NumericUtf8Char	"é"	4.5533	1.2688	3.59x	5.6743	1.3445	4.22x
SpanGetString	"Content-Length: "	7.950	7.684	1.03x	9.383	8.307	1.13x
SliceGetString	"Content-Length: "	8.767	8.052	1.09x	10.285	9.263	1.11x
SpanGetString	"Grüße世界"	13.913	10.595	1.31x	19.876	12.137	1.64x
SliceGetString	"Grüße世界"	15.180	11.453	1.33x	20.389	12.853	1.59x

Speedup is main / PR. Char parsing is 3.59-5.35x faster; short-string decoding is 1.03-1.64x faster. Allocations are unchanged on both architectures: 0 B for Char parsing, 56 B for "Content-Length: ", and 40 B for "Grüße世界".

[azure-pipelines]

Azure Pipelines:
Successfully started running 3 pipeline(s).
13 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-text-encoding
See info in area-owners.md if you want to be subscribed.

[EgorBo]

@EgorBot -arm -amd

using System;
using System.Globalization;
using System.IO;
using System.Net;
using System.Numerics;
using System.Text;
using BenchmarkDotNet.Attributes;
using BenchmarkDotNet.Running;

[MemoryDiagnoser]
public class CharParsingBench
{
    private byte[] _input = null!;

    [Params("A", "\u00E9")]
    public string Text { get; set; } = null!;

    [GlobalSetup]
    public void Setup() => _input = Encoding.UTF8.GetBytes(Text);

    [Benchmark]
    public char NumericUtf8Char() => Parse<char>(_input);

    private static T Parse<T>(ReadOnlySpan<byte> input) where T : struct, INumberBase<T>
    {
        if (!T.TryParse(input, NumberStyles.Integer, CultureInfo.InvariantCulture, out T result))
        {
            throw new FormatException();
        }

        return result;
    }
}

[MemoryDiagnoser]
public class ShortUtf8Bench
{
    private byte[] _input = null!;

    [Params("Content-Length: ", "Gr\u00FC\u00DFe\u4E16\u754C")]
    public string Text { get; set; } = null!;

    [GlobalSetup]
    public void Setup() => _input = Encoding.UTF8.GetBytes(Text);

    [Benchmark]
    public string SpanGetString() => Encoding.UTF8.GetString(_input.AsSpan());

    [Benchmark]
    public string SliceGetString() => Encoding.UTF8.GetString(_input, 0, _input.Length);
}

[Copilot]

Copilot review overview

🟡 Changes recommended

The new paths need targeted tests, and both conditions unnecessarily type-check inputs that exceed the fast-path threshold.

Review effort: Balanced
Findings: 1 · 2

Open (3)

• Test INumberBase UTF-8 parsing override · New
• Check input length before runtime type test · New
• Check input length before sealed-type test · New

What changed in this PR

Optimizes UTF-8 character parsing and short-string decoding by reusing existing direct parsers and decoders.

Changes:

• Routes generic char UTF-8 parsing directly through Rune.
• Reuses the sealed UTF-8 short-input decoder for span and array-slice calls.
• Preserves existing paths for subclasses and larger inputs.

File	Description
Char.cs	Adds direct generic UTF-8 char parsing.
Encoding.cs	Adds the short-span UTF-8 fast path.
UTF8Encoding.cs	Adds the short array-slice fast path.
UTF8Encoding.Sealed.cs	Generalizes the existing short-input decoder to spans.