Checking if a list is empty with LINQ

Question

What's the "best" (taking both speed and readability into account) way to determine if a list is empty? Even if the list is of type IEnumerable<T> and doesn't have a Count property.

Right now I'm tossing up between this:

if (myList.Count() == 0) { ... }

and this:

if (!myList.Any()) { ... }

My guess is that the second option is faster, since it'll come back with a result as soon as it sees the first item, whereas the second option (for an IEnumerable) will need to visit every item to return the count.

That being said, does the second option look as readable to you? Which would you prefer? Or can you think of a better way to test for an empty list?

Edit @lassevk's response seems to be the most logical, coupled with a bit of runtime checking to use a cached count if possible, like this:

public static bool IsEmpty<T>(this IEnumerable<T> list)
{
    if (list is ICollection<T>) return ((ICollection<T>)list).Count == 0;

    return !list.Any();
}

Answer 1

You could do this:

public static Boolean IsEmpty<T>(this IEnumerable<T> source)
{
    if (source == null)
        return true; // or throw an exception
    return !source.Any();
}

Edit: Note that simply using the .Count method will be fast if the underlying source actually has a fast Count property. A valid optimization above would be to detect a few base types and simply use the .Count property of those, instead of the .Any() approach, but then fall back to .Any() if no guarantee can be made.

Answer 2

I just wrote up a quick test, try this:

 IEnumerable<Object> myList = new List<Object>();

 Stopwatch watch = new Stopwatch();

 int x;

 watch.Start();
 for (var i = 0; i <= 1000000; i++)
 {
    if (myList.Count() == 0) x = i; 
 }
 watch.Stop();

 Stopwatch watch2 = new Stopwatch();

 watch2.Start();
 for (var i = 0; i <= 1000000; i++)
 {
     if (!myList.Any()) x = i;
 }
 watch2.Stop();

 Console.WriteLine("myList.Count() = " + watch.ElapsedMilliseconds.ToString());
 Console.WriteLine("myList.Any() = " + watch2.ElapsedMilliseconds.ToString());
 Console.ReadLine();

The second is almost three times slower :)

Trying the stopwatch test again with a Stack or array or other scenarios it really depends on the type of list it seems - because they prove Count to be slower.

So I guess it depends on the type of list you're using!

(Just to point out, I put 2000+ objects in the List and count was still faster, opposite with other types)

Answer 3

@Crucible's profiling code yields some interesting results for different implementations of the list.

I like @lassevk's flexible extension-method approach. Certainly myList.IsEmpty() is the most readable syntax.

I'll leave this question open for a while to garner more opinions. I think it's an interesting question.

Answer 4

The second option is much quicker if you have multiple items.

• Any() returns as soon as 1 item is found.
• Count() has to keep going through the entire list.

For instance suppose the enumeration had 1000 items.

• Any() would check the first one, then return true.
• Count() would return 1000 after traversing the entire enumeration.

This is potentially worse if you use one of the predicate overrides - Count() still has to check every single item, even it there is only one match.

You get used to using the Any one - it does make sense and is readable.

One caveat - if you have a List, rather than just an IEnumerable then use that list's Count property.

Answer 5

@Keith The differences between Any() and Count() seem clear, but @crucible's profiling code seems to indicate that Count() is faster for certain implementations of IEnumerable. For List I can't get Any() to give a faster result than Count() until the list size is up in the thousands of items.

LINQ itself must be doing some serious optimization around the Count() method somehow.

Answer 6

LINQ itself must be doing some serious optimization around the Count() method somehow.

Does this surprise you? I imagine that for IList implementations, Count simply reads the number of elements directly while Any has to query the IEnumerable.GetEnumerator method, create an instance and call MoveNext at least once.

/EDIT @Matt:

I can only assume that the Count() extension method for IEnumerable is doing something like this:

Yes, of course it does. This is what I meant. Actually, it uses ICollection instead of IList but the result is the same.

Answer 7

@Konrad what surprises me is that in my tests, I'm passing the list into a method that accepts IEnumerable<T>, so the runtime can't optimize it by calling the Count() extension method for IList<T>.

I can only assume that the Count() extension method for IEnumerable is doing something like this:

public static int Count<T>(this IEnumerable<T> list)
{
    if (list is IList<T>) return ((IList<T>)list).Count;

    int i = 0;
    foreach (var t in list) i++;
    return i;
}

... in other words, a bit of runtime optimization for the special case of IList<T>.

/EDIT @Konrad +1 mate - you're right about it more likely being on ICollection<T>.

Answer 8

This extension method works for me:

public static bool IsEmpty<T>(this IEnumerable<T> enumerable)
{
    try
    {
        enumerable.First();
        return false;
    }
    catch (InvalidOperationException)
    {
        return true;
    }
}

Answer 9

Ok, so what about this one?

public static bool IsEmpty<T>(this IEnumerable<T> enumerable)
{
    return !enumerable.GetEnumerator().MoveNext();
}

EDIT: I've just realized that someone has sketched this solution already. It was mentioned that the Any() method will do this, but why not do it yourself? Regards

Answer 10

Another idea:

if(enumerable.FristOrDefault() != null)

However I like the Any() approach more.

Answer 11

List.Count is O(1) according to Microsoft's documentation:
http://msdn.microsoft.com/en-us/library/27b47ht3.aspx

so just use List.Count == 0 it's much faster than a query

This is because it has a data member called Count which is updated any time something is added or removed from the list, so when you call List.Count it doesn't have to iterate through every element to get it, it just returns the data member.

Answer 12

I would make one small addition to the code you seem to have settled on: check also for ICollection, as this is implemented even by some non-obsolete generic classes as well (i.e., Queue<T> and Stack<T>). I would also use as instead of is as it's more idiomatic and has been shown to be faster.

public static bool IsEmpty<T>(this IEnumerable<T> list)
{
    if (list == null)
    {
        throw new ArgumentNullException("list");
    }

    var genericCollection = list as ICollection<T>;
    if (genericCollection != null)
    {
        return genericCollection.Count == 0;
    }

    var nonGenericCollection = list as ICollection;
    if (nonGenericCollection != null)
    {
        return nonGenericCollection.Count == 0;
    }

    return !list.Any();
}