Question

Consider this code:

int size = 100 * 1000 * 1000;
var emu = Enumerable.Range(0, size);
var arr = Enumerable.Range(0, size).ToArray();

when I call emu.ElementAt(size-10) and arr.ElementAt(size-10) and measure the time the arr is much faster (the array is 0.0002s compared to IEnumerable 0.59s).

As I understand it, the extention method ElementAt() have the signature

public static TSource ElementAt<TSource>(this IEnumerable<TSource> source, int index)

and since the ‘source’ is a IEnumerable the logic carried out would be similar – opposed to what I see where the array is accessed directly.

Could someone please explain this 🙂

Answer 1

This is an optimization performed at execution-time. Although the call isn’t overloaded, it is able to check (using is or as) whether the source is actually an IList<T>. If it is, it’s able to go directly to the right element.

Various other calls do this – notable Count() which is optimised for ICollection<T> and (as of .NET 4) the nongeneric ICollection interface.

One of the downsides of extension methods is that all these optimizations have to be performed by the implementation itself – types can’t override anything to “opt in” to optimizing extension methods. That means the optimizations have to all be known by the original implementor 🙁