fix: Optimizes GetString to eliminate allocations (#2154)

### Summary

* Optimized GetString by eliminating the intermediate string allocation.

** Note **: Encoding.GetChars() is still really inefficient, especially when strings are not aligned or have regular ascii/unicode characters. Thankfully we generally don't have to worry about these odd edge cases, but if they happen then GetChars can allocate hundreds of bytes.


This is a benchmark for just the related changes. NonSpecial are just ascii characters, while the other tests include control codes.
```cs
| Method                         | Mean     | Error   | StdDev  | Gen0   | Allocated |
|------------------------------- |---------:|--------:|--------:|-------:|----------:|
| GetString                      | 306.7 ns | 5.96 ns | 6.86 ns | 0.0124 |     200 B |
| GetStringNotSpecial            | 257.0 ns | 4.83 ns | 4.52 ns | 0.0114 |     184 B |
| GetStringSpanHelpers           | 166.4 ns | 3.26 ns | 3.48 ns |      - |         - |
| GetStringSpanHelpersNotSpecial | 137.3 ns | 1.57 ns | 1.22 ns |      - |         - |
```
This commit is contained in:
Kamron Batman 2025-04-12 12:42:15 -07:00 committed by GitHub
parent a94814a7ef
commit 872de8d095
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 41 additions and 18 deletions

View file

@ -1,6 +1,6 @@
/*************************************************************************
* ModernUO *
* Copyright 2019-2023 - ModernUO Development Team *
* Copyright 2019-2025 - ModernUO Development Team *
* Email: hi@modernuo.com *
* File: TextEncoding.cs *
* *
@ -16,6 +16,7 @@
using System;
using System.Runtime.CompilerServices;
using System.Text;
using Server.Buffers;
namespace Server.Text;
@ -114,33 +115,55 @@ public static class TextEncoding
_ => 1
};
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static bool IsSafeChar(ushort c) => c is >= 0x20 and < 0xFFFE;
public static string GetString(ReadOnlySpan<byte> span, Encoding encoding, bool safeString = false)
{
string s = encoding.GetString(span);
if (!safeString)
{
return s;
return encoding.GetString(span);
}
ReadOnlySpan<char> chars = s.AsSpan();
var charCount = encoding.GetMaxCharCount(span.Length);
using var sb = new ValueStringBuilder(stackalloc char[256]);
var hasDoneAnyReplacements = false;
char[] rentedChars = null;
Span<char> chars = charCount <= 256
? stackalloc char[charCount]
: rentedChars = STArrayPool<char>.Shared.Rent(charCount);
for (int i = 0, last = 0; i < chars.Length; i++)
try
{
if (!IsSafeChar(chars[i]) && i == chars.Length - 1)
var length = encoding.GetChars(span, chars);
chars = chars[..length];
var index = chars.IndexOfAnyExceptInRange((char)0x20, (char)0xFFFD);
if (index == -1)
{
hasDoneAnyReplacements = true;
sb.Append(chars.Slice(last, i - last));
last = i + 1; // Skip the unsafe char
return new string(chars);
}
using var sb = charCount <= 256
? new ValueStringBuilder(stackalloc char[charCount])
: ValueStringBuilder.Create(charCount);
while (index != -1)
{
sb.Append(chars[..index]);
chars = chars[(index + 1)..];
index = chars.IndexOfAnyExceptInRange((char)0x20, (char)0xFFFD);
}
if (chars.Length > 0)
{
sb.Append(chars);
}
return sb.ToString();
}
finally
{
if (rentedChars != null)
{
STArrayPool<char>.Shared.Return(rentedChars);
}
}
return !hasDoneAnyReplacements ? s : sb.ToString();
}
}