如果担心速度问题，可以使用指针编辑现有字符串。您可以固定字符串并获取指向它的指针，然后在每个字符上运行for循环，用替换字符覆盖每个无效字符。这将是非常有效的，并且不需要分配任何新的字符串内存。为了使用指针，你还需要用不安全选项来编译你的模块，并在方法头中添加“不安全”修饰符。

static void Main(string[] args)
{
    string str = "string!$%with^&*invalid!!characters";
    Console.WriteLine( str ); //print original string
    FixMyString( str, ' ' );
    Console.WriteLine( str ); //print string again to verify that it has been modified
    Console.ReadLine(); //pause to leave command prompt open
}


public static unsafe void FixMyString( string str, char replacement_char )
{
    fixed (char* p_str = str)
    {
        char* c = p_str; //temp pointer, since p_str is read-only
        for (int i = 0; i < str.Length; i++, c++) //loop through each character in string, advancing the character pointer as well
            if (!IsValidChar(*c)) //check whether the current character is invalid
                (*c) = replacement_char; //overwrite character in existing string with replacement character
    }
}

public static bool IsValidChar( char c )
{
    return (c >= '0' && c <= '9') || (c >= 'A' && c <= 'Z') || (c >= 'a' && c <= 'z') || (c == '.' || c == '_');
    //return char.IsLetterOrDigit( c ) || c == '.' || c == '_'; //this may work as well
}

对于S&G, linq化的方式:

var original = "(*^%foo)(@)&^@#><>?:\":';=-+_";
var valid = new char[] { 
    'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 
    'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z', 'A', 'B', 'C', 'D', 
    'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 
    'T', 'U', 'V', 'W', 'X', 'Y', 'Z', '1', '2', '3', '4', '5', '6', '7', '8', 
    '9', '0', '.', '_' };
var result = string.Join("",
    (from x in original.ToCharArray() 
     where valid.Contains(x) select x.ToString())
        .ToArray());

然而，我不认为这是最有效的方法。

正则表达式如下所示:

public string RemoveSpecialChars(string input)
{
    return Regex.Replace(input, @"[^0-9a-zA-Z\._]", string.Empty);
}

但如果性能非常重要，我建议你在选择“正则表达式路径”之前做一些基准测试……

I had to do something similar for work, but in my case I had to filter all that is not a letter, number or whitespace (but you could easily modify it to your needs). The filtering is done client-side in JavaScript, but for security reasons I am also doing the filtering server-side. Since I can expect most of the strings to be clean, I would like to avoid copying the string unless I really need to. This let my to the implementation below, which should perform better for both clean and dirty strings.

public static string EnsureOnlyLetterDigitOrWhiteSpace(string input)
{
    StringBuilder cleanedInput = null;
    for (var i = 0; i < input.Length; ++i)
    {
        var currentChar = input[i];
        var charIsValid = char.IsLetterOrDigit(currentChar) || char.IsWhiteSpace(currentChar);

        if (charIsValid)
        {
            if(cleanedInput != null)
                cleanedInput.Append(currentChar);
        }
        else
        {
            if (cleanedInput != null) continue;
            cleanedInput = new StringBuilder();
            if (i > 0)
                cleanedInput.Append(input.Substring(0, i));
        }
    }

    return cleanedInput == null ? input : cleanedInput.ToString();
}

另一种试图通过减少分配来提高性能的方法，特别是在多次调用此函数的情况下。

它之所以有效，是因为可以保证结果不会比输入长，因此可以在不在内存中创建额外副本的情况下传递输入和输出。因此，您不能使用stackalloc来创建缓冲区数组，因为这需要从缓冲区中复制一个副本。

public static string RemoveSpecialCharacters(this string str)
{
    return RemoveSpecialCharacters(str.AsSpan()).ToString();
}

public static ReadOnlySpan<char> RemoveSpecialCharacters(this ReadOnlySpan<char> str)
{
    Span<char> buffer = new char[str.Length];
    int idx = 0;

    foreach (char c in str)
    {
        if (char.IsLetterOrDigit(c))
        {
            buffer[idx] = c;
            idx++;
        }
    }

    return buffer.Slice(0, idx);
}

如果你使用的是动态字符列表，LINQ可以提供一个更快更优雅的解决方案:

public static string RemoveSpecialCharacters(string value, char[] specialCharacters)
{
    return new String(value.Except(specialCharacters).ToArray());
}

我将这种方法与之前的两种“快速”方法(发行版编译)进行了比较:

字符数组解决方案由LukeH - 427毫秒 StringBuilder解决方案- 429毫秒 LINQ(这个答案)- 98毫秒

注意，算法略有修改-字符作为数组传入，而不是硬编码，这可能会有轻微的影响(即/其他解决方案将有一个内部for循环来检查字符数组)。

如果我使用LINQ where子句切换到硬编码的解决方案，结果是:

字符数组解决方案- 7ms StringBuilder解决方案- 22ms LINQ - 60毫秒

如果您计划编写一个更通用的解决方案，而不是硬编码字符列表，那么可能值得考虑LINQ或经过修改的方法。LINQ绝对能给你简洁、高可读性的代码——甚至比Regex更好。