string-perf-patterns

SKILL: String Performance Patterns in .NET

When to apply

When writing or reviewing C# code that processes text content — especially log files, API responses, or search operations that may handle multi-MB strings.

Patterns to avoid

1. Chained .Replace() for line-ending normalization

// BAD — 2 intermediate full-size strings
var normalized = content.Replace("\r\n", "\n").Replace("\r", "\n");

Fix: Use a span-based line enumerator, or string.Create with a single-pass copy that normalizes as it goes. For search operations, consider handling mixed line endings during iteration instead of pre-normalizing.

2. Split() + Join() for tail/head operations

// BAD — allocates string[] of all lines just to get last N
var lines = content.Split('\n');
return string.Join('\n', lines[^tailLines..]);

Fix: Reverse-scan for Nth \n using ReadOnlySpan<char>:

var span = content.AsSpan();
int pos = span.Length;
for (int i = 0; i < tailLines && pos > 0; i++)
{
    pos = span[..pos].LastIndexOf('\n');
    if (pos < 0) { pos = 0; break; }
}
return pos > 0 ? content[(pos + 1)..] : content;

3. Substring allocation in pattern matching

// BAD — allocates new string every call
name.EndsWith(pattern[1..], StringComparison.OrdinalIgnoreCase);

Fix: Use spans:

name.AsSpan().EndsWith(pattern.AsSpan(1), StringComparison.OrdinalIgnoreCase);

4. Triple-iteration for categorization

// BAD — 3 passes over same list
Binlogs = files.Where(f => IsBinlog(f)).Select(Map).ToList();
TestResults = files.Where(f => IsTest(f)).Select(Map).ToList();
Other = files.Where(f => !IsBinlog(f) && !IsTest(f)).Select(Map).ToList();

Fix: Single-pass loop:

var binlogs = new List<FileInfo_>();
var testResults = new List<FileInfo_>();
var other = new List<FileInfo_>();
foreach (var f in files)
{
    var mapped = new FileInfo_ { Name = f.Name, Uri = f.Uri };
    if (IsBinlog(f.Name)) binlogs.Add(mapped);
    else if (IsTest(f.Name)) testResults.Add(mapped);
    else other.Add(mapped);
}