How to return an nvarchar(max) in a CLR UDF?

Oh, whatever, I found the answer myself:

/// <summary>
/// Replaces each occurrence of sPattern in sInput with sReplace. This is done 
/// with the CLR: 
/// new RegEx(sPattern, RegexOptions.Multiline).Replace(sInput, sReplace). 
/// The result of the replacement is the return value.
/// </summary>
[SqlFunction(IsDeterministic = true)]
[return: SqlFacet(MaxSize = -1)]
public static  SqlString FRegexReplace([SqlFacet(MaxSize = -1)]string sInput, 
       string sPattern, string sReplace)
{
    return new Regex(sPattern, RegexOptions.Multiline).Replace(sInput, sReplace);
}

The idea is to hint to SQL Server that the input and return values are not the default nvarchar(4000), but have a different size.

I learned a new trick regarding attributes: They can be added to the parameters as well as the method itself (quite obvious), but also to the return value with the [return: AttributeName(Parameter=Value, ...)] Syntax.

See also How to create CLR stored procedure with Nvarchar(max) parameter where you'll discover how/why you really should use the SqlChars data type. See Handling Large Object (LOB) Parameters in the CLR in MSDN.