Find a list of non-duplicate numbers in a collection. Can this be done more efficient/shorter code?

static IEnumerable<T> FindUniqueNumbersInCollection<T>(ICollection<T> value)
{
    Dictionary<T, byte> hash = new Dictionary<T, byte>();

    foreach (T 开发者_JAVA技巧val in value)
    {
        if (hash.ContainsKey(val)) { hash[val] = 1; continue; }
        hash.Add(val, 0);
    }

    List<T> unique = new List<T>();

    foreach (KeyValuePair<T, byte> kvp in hash)
    {
        if (kvp.Value == 0) unique.Add(kvp.Key);

    }

    return unique;
}

---

Edit:

I think this is finally correct. :)

var dict = new Dictionary<T, bool>();
foreach (var v in value)
    dict[v] = dict.ContainsKey(v);
foreach (var pair in dict)
    if (!pair.Value)
        yield return pair.Key;  //Algorithmically, as fast as possible

Or if you'd like some LINQ:

var dict = new Dictionary<T, bool>();
foreach (var v in value)
    dict[v] = dict.ContainsKey(v);
return dict.Keys.Where(k => !dict[k]);  //*COULD* be slower on lots of collisions

Or even

var dict = new Dictionary<T, bool>();
foreach (var v in value)
    dict[v] = dict.ContainsKey(v);
return dict.Where(p => !p.Value).Select(p => p.Key); //"Fastest".. but not really

I wouldn't say it's "more efficient" than yours (it's really not -- they're pretty much the same), but it's definitely shorter.

---

If you want overkill efficiency (at the cost of accuracy), you can always use a Bloom Filter!

---

Since you're returning an IEnumerable<T>, you could

return hash.Where(kvp => kvp.Value == 0).Select(kvp => kvp.Key);

instead of the second loop to perform that iteration lazily.

---

The following code will do this (in one statement).

var unqiue = new[] { 1, 1, 2, 3, 3, 4 }
    .GroupBy(x => x)             // groups the elements using the element value as the key
    .Where(x => x.Count() == 1)  // restricts to only those groups having a count of 1
    .Select(x => x.Key);         // selects the key from each of those groups

(You will need to have using System.Linq; in your code file.)

Although your ultimate goal is efficiency, I would use the simplest code possible (which I think this is) and then optimise when the profiler tells me to.