Interestingly enough I tried both of these in LinqPad and the variant using group from Dmitry Gribkov by appears to be quicker. (also the final distinct is not required as the result is already distinct.
My (somewhat simple) code was:
public class Pair
{
public int id {get;set;}
public string Arb {get;set;}
}
void Main()
{
var theList = new List<Pair>();
var randomiser = new Random();
for (int count = 1; count < 10000; count++)
{
theList.Add(new Pair
{
id = randomiser.Next(1, 50),
Arb = "not used"
});
}
var timer = new Stopwatch();
timer.Start();
var distinct = theList.GroupBy(c => c.id).Select(p => p.First().id);
timer.Stop();
Debug.WriteLine(timer.Elapsed);
timer.Start();
var otherDistinct = theList.Select(p => p.id).Distinct();
timer.Stop();
Debug.WriteLine(timer.Elapsed);
}