What is the collection equivalent of a multi-dimensional array?

Question

I've got a group of data that looks like this:

001 001 One
001 002 Two
001 003 Three

002 001 One
002 002 Two
002 003 Three

...

Now, certainly, I could create an array of string[x][y] = z, but this array has to be resizable, and i'd prefer to use the string representations of the indexers than convert to numeric. The reason is that i will need to look up the data by string, and i don't see the point in needless string->number conversions.

My first thought was this:

Dictionary<string, Dictionary<string, string>> data;

data = new Dictionary<string, Dictionary<string, string>>();

Dictionary<string, string> subdata = Dictionary<string, string>();

subdata.Add(key, string);
data.add(key2, subdata);

and this works, but is somewhat cumbersome. It also feels wrong and kludgy and not particularly efficient.

So what's the best way to store this sort of data in a collection?

I also thought of creating my own collection class, but I'd rather not if I don't have to. I'd rather just use the existing tools.

Answer 1

This is pretty common request, and most people end up writing some variation of a Tuple class. If you're using ASP.Net, you can utilize the Triple class that's already available, otherwise, write something like:

public class Tuple<T, T2, T3>
{
    public Tuple(T first, T2 second, T3 third)

    {
        First = first;
        Second = second;
        Third = third;
    }

    public T First { get; set; }
    public T2 Second { get; set; }
    public T3 Third { get; set; }

}

There's a generic three-tuple class, so you can create a new List<Tuple<string, string, string>>() and create your tuples and add them. Expand on that basic class with some indexing functionality and you're up up and away.

Edit: A list with a dictionary doesn't seem like the correct approach, because each dictionary is only holding one value. There is no multi-entry relationship between the key and values - there is simply one multi-part key and one associated value. The data is equivalent to a database row (or tuple!).

Edit2: Here's an indexable list class you could use for convenience.

    public class MyTupleList : List<Tuple<string, string, string>>
    {
        public Tuple<string, string, string> this[string first, string second]
        {
            get
            {
                return (this.Find(x => x.First == first && x.Second == second));
            }
            set
            {
                this[first, second] = value;
            }
        }
    }

Answer 2

I think this really depends on what you are modelling here. If you're planning to use an object-oriented approach, you shouldn't be thinking of these as arbitrary items inside a data structure.

I'm guessing from looking at this that the first two columns are serving as a "key" for the other items. Define a simple struct, and create a dictionary of like so:

struct Key {
   public int Val1 { get; set; }
   public int Val2 { get; set; }
}

....

Dictionary<Key, string> values;

Obviously Key and the items inside it should be mapped to something closer to what you are representing.

Answer 3

Given a suitable Pair<A,B> class*, left as an exercise for the reader, you could use a Dictionary<Pair<string, string>, string>.

^{* A class with equality and hash code overrides, nothing terribly hard.}

Answer 4

Would a List<List<T>> work for you? Still kludgy, but better than dictionaries IMHO.

---

EDIT: What about a Dictionary<string,string> and mapping the two keys to a single string?

var data = new Dictionary<string,string>(StringComparer.Ordinal);

data[GetKey("002", "001")] = "One";

with

string GetKey(string a, string b) {
    return a + "\0" + b;
}

Answer 5

List<List<string>> is really your best bet in this case. But I agree, it's kludgy. Personally, I would create a custom class that implements a two-dimensional indexer and maybe use a List<List<T>> internally.

For example:

public class DynamicTwoDimensonalArray<T>
{
  private List<List<T>> Items = new List<List<T>>();

  public T this[int i1, int i2]
  {
    get
    {
      return Items[i1][i2];
    }
    set
    {
      Items[i1][i2] = value;
    }
  }  
}

This is a basic idea to get you going; clearly the setter needs to deal with bounds issues. But it's a start.

Edit:

No. As I said, I would prefer to index them by string. And they may not always be sequential (might have a missing number in the middle). - Mystere Man

Hmm... this is interesting. If that's the case, your best bet would be to create some sort of concatenation of the combination of the two indexers and use that as the key in a single-level dictionary. I would still use a custom class to make using the indexing easier. For example:

public class TwoDimensionalDictionary
{
  private Dictionary<string, string> Items = new Dictionary<string, string>();

  public string this[string i1, string i2]
  {
    get
    {
      // insert null checks here
      return Items[BuildKey(i1, i2)];
    }
    set
    {
      Items[BuildKey(i1, i2)] = value;
    }
  }
  public string BuildKey(string i1, string i2)
  {
    return "I1: " + i1 + " I2: " + i2;
  }  
}

Answer 6

If you are ever going to need to find z by given (x,y) (and not, for example, find all y by given x), then use this:

Dictionary<KeyValuePair<string, string>, string>

Otherwise, your dictionary is fine as is.