Linq: GroupBy vs Distinct

Question

I've been trying to get a Linq query to return distinct values from a collection. I've found two ways to go about it; either use GroupBy or Distinct. I know that Distinct was made for the job but I have to implement IEquatable on the object.

I tried GroupBy and that worked just fine. I want to know if using Distinct vs GroupBy has a distinct performance advantage.

Answer 1

Distinct() will compare entire objects in collection (for reference types you need GetHashCode and Equals to be overridden). It will enumerate items and just add them to set. Simple and fast. Something like:

Set<TSource> set = new Set<TSource>(comparer);

foreach (TSource tSource in source)
{
     if (!set.Add(tSource))
          continue;

     yield return tSource;
}

GroupBy() allows you to group object by some key. In this case keys will be compared. It will need to execute key selector lambda for each item in collection. Also it will need to create grouping for each distinct key and add each item in collection to its group:

Func<TSource, TElement> elementSelector = x => x;

<TKey, TElement> lookup = new Lookup<TKey, TElement>(comparer);
foreach (TSource tSource in source)
{
     TKey key = keySelector(tSource);

     // simplified pseudo-code
     if (!lookup.Contains(key))
          lookup.Add(new Grouping(key)); 

     lookup[key].Add(elementSelector(tSource));
}

foreach(IGrouping<TKey, TElement> grouping in lookup)
    yield return grouping;

So, I think GroupBy() is not that fast as simple Distict().