I think you're approaching this incorrectly. You're not sorting strings, you're sorting structured objects that are misrepresented as strings (somebody aptly named this antipattern "stringly typed"). Your requirements show that you know this structure, yet it's not represented in the datastructure List<string[]>, and that's making your life hard. You should parse that structure into a real type (struct or class), and then sort that.
enum PrefixCode { MW1, FW, DN, MWSTX1CK, MWSTX2FF, }
enum TheseLetters { Q, J, C, E, I, A, }
struct CardRecord : IComparable<CardRecord> {
public readonly PrefixCode Code;
public readonly TheseLetters Letter;
public readonly uint Number;
public CardRecord(string input) {
Code = ParseEnum<PrefixCode>(ref input);
Letter = ParseEnum<TheseLetters>(ref input);
Number = uint.Parse(input);
}
static T ParseEnum<T>(ref string input) { //assumes non-overlapping prefixes
foreach(T val in Enum.GetValues(typeof(T))) {
if(input.StartsWith(val.ToString())) {
input = input.Substring(val.ToString().Length);
return val;
}
}
throw new InvalidOperationException("Failed to parse: "+input);
}
public int CompareTo(CardRecord other) {
var codeCmp = Code.CompareTo(other.Code);
if (codeCmp!=0) return codeCmp;
var letterCmp = Letter.CompareTo(other.Letter);
if (letterCmp!=0) return letterCmp;
return Number.CompareTo(other.Number);
}
public override string ToString() {
return Code.ToString() + Letter + Number.ToString("00");
}
}
A program using the above to process your example might then be:
static class Program {
static void Main() {
var inputStrings = new []{ "MW1E10", "MWSTX2FFI06", "FWQ02", "MW1Q04", "MW1Q05",
"FWI01", "MWSTX2FFA01", "DNC03", "MWSTX1CKC02", "MWSTX2FFI03", "MW1I06" };
var outputStrings = inputStrings
.Select(s => new CardRecord(s))
.OrderBy(c => c)
.Select(c => c.ToString());
Console.WriteLine(string.Join("\n", outputStrings));
}
}
This generates the same ordering as in your example. In real code, I'd recommend you name the types according to what they represent, and not, for example, TheseLetters.
This solution - with a real parse step - is superior because it's almost certain that you'll want to do more with this data at some point, and this allows you to actually access the components of the data easily. Furthermore, it's comprehensible to a future maintainer since the reason behind the ordering is somewhat clear. By contrast, if you chose to do complex string-based processing it's often very hard to understand what's going on (especially if it's part of a larger program, and not a tiny example as here).
Making new types is cheap. If your method's return value doesn't quite "fit" in an existing type, just make a new one, even if that means 1000's of types.