public static string RemoveSpecialCharacters(string str){
    return str.replaceAll("[^A-Za-z0-9_\\\\.]", "");
}

我建议创建一个简单的查找表，您可以在静态构造函数中初始化它，以将任何字符组合设置为有效。这让您可以进行快速、单一的检查。

edit

另外，为了提高速度，您需要将StringBuilder的容量初始化为输入字符串的长度。这将避免重新分配。这两种方法结合起来会给你速度和灵活性。

另一个编辑

我认为编译器可能会优化它，但作为风格和效率的问题，我建议foreach而不是for。

我觉得不错。我要做的唯一改进是用字符串的长度初始化StringBuilder。

StringBuilder sb = new StringBuilder(str.Length);

我不确定这是最有效的方法，但对我来说很有效

 Public Function RemoverTildes(stIn As String) As String
    Dim stFormD As String = stIn.Normalize(NormalizationForm.FormD)
    Dim sb As New StringBuilder()

    For ich As Integer = 0 To stFormD.Length - 1
        Dim uc As UnicodeCategory = CharUnicodeInfo.GetUnicodeCategory(stFormD(ich))
        If uc <> UnicodeCategory.NonSpacingMark Then
            sb.Append(stFormD(ich))
        End If
    Next
    Return (sb.ToString().Normalize(NormalizationForm.FormC))
End Function

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();
}

我想知道基于regx的替代品(可能是编译的)是否更快。必须测试，有人发现这是~5倍慢。

除此之外，您应该使用预期的长度初始化StringBuilder，这样在中间字符串增长时就不必复制它。

一个好的数字是原始字符串的长度，或者稍低一些(取决于函数输入的性质)。

最后，您可以使用一个查找表(范围0..127)来确定一个字符是否被接受。