Adam Anderson

Posts

Mar 23, 2023
Pipenv And Azure Private Package Indexes
Integrating Pipenv and Azure Private Package Indexes
Jul 21, 2021
Windows Terminal setup
Windows Terminal

Windows Terminal is a modern feature-rich open source terminal from Microsoft that runs many shells, including PowerShell. The easiest way to install it is from the Microsoft Store or via Chocolately with the command choco install microsoft-windows-terminal.
Jun 15, 2021
Running unit tests automatically before pushing
Something that I have been using lately is a git pre-push hook to build and run our unit test suite before pushing. I like that I can ensure that I do not push any commits that break the build or unit tests and that I get fast feedback.

May 25, 2021

Linq Any() performance compared to Count() > 0

I often see .Count() > 0 or .Count > 0 to check if there are any elements in some collection. Linq provides an Any() method to check if there are any elements. Any() sure reads better but does it come with any downsides? Inspired by a twitter post on using Any() and benchmarking it, I decided to run some of my own tests using BenchmarkDotNet.

The benchmarks that were run compare the use of these two methods, comparing enumerables and lists with two sizes, 1000 and 1000.

BenchmarkDotNet=v0.13.0, OS=Windows 10.0.19042.985 (20H2/October2020Update)
Intel Core i7-8650U CPU 1.90GHz (Kaby Lake R), 1 CPU, 8 logical and 4 physical cores
.NET SDK=5.0.203
  [Host]     : .NET 5.0.6 (5.0.621.22011), X64 RyuJIT
  DefaultJob : .NET 5.0.6 (5.0.621.22011), X64 RyuJIT

Method	N	Mean	Error	StdDev	Median
Enumerable_Any	1000	12.7297 ns	0.2874 ns	0.2400 ns	12.7911 ns
Enumerable_CountGreaterThanZero	1000	83,476.7798 ns	1,015.1761 ns	949.5963 ns	83,264.1113 ns
List_Any	1000	5.6846 ns	0.1155 ns	0.1080 ns	5.6549 ns
List_CountGreaterThanZero	1000	0.0248 ns	0.0268 ns	0.0358 ns	0.0000 ns
Enumerable_Any	10000	12.5423 ns	0.1560 ns	0.1382 ns	12.5044 ns
Enumerable_CountGreaterThanZero	10000	835,227.9053 ns	5,795.2965 ns	4,524.5863 ns	834,107.7637 ns
List_Any	10000	5.5781 ns	0.1236 ns	0.1423 ns	5.5443 ns
List_CountGreaterThanZero	10000	0.0106 ns	0.0153 ns	0.0143 ns	0.0000 ns

Enumerable performance

It’s pretty safe to say that Count() > 0 on an enumerable is pretty slow, but why is that? Enumerable.Count() needs to iterate over each element and count them. Let’s first take a look at Any().

public static bool Any<TSource>(this IEnumerable<TSource> source)
{
    if (source == null) ThrowHelper.ThrowArgumentNullException(ExceptionArgument.source);

    if (source is ICollection<TSource> collectionoft) return collectionoft.Count != 0;
    else if (source is IIListProvider<TSource> listProv)
    {
        // Note that this check differs from the corresponding check in
        // Count (whereas otherwise this method parallels it).  If the count
        // can't be retrieved cheaply, that likely means we'd need to iterate
        // through the entire sequence in order to get the count, and in that
        // case, we'll generally be better off falling through to the logic
        // below that only enumerates at most a single element.
        int count = listProv.GetCount(onlyIfCheap: true);
        if (count >= 0) return count != 0; 
    }
    else if (source is ICollection collection) return collection.Count != 0;

    using (IEnumerator<TSource> e = source.GetEnumerator())
    {
        return e.MoveNext();
    }
}

Here we can see that there are a lot of conditionals to find the fastest way to determine if there are any elements. The last bit of code is what gets executed in our benchmark. There is a single iteration e.MoveNext() and a Boolean return. This doesn’t depend on the size, it’s O(1). Lets now take a look at Count().

public static int Count<TSource>(this IEnumerable<TSource> source) {
    if (source == null) throw Error.ArgumentNull("source");
    ICollection<TSource> collectionoft = source as ICollection<TSource>;
    if (collectionoft != null) return collectionoft.Count;
    ICollection collection = source as ICollection;
    if (collection != null) return collection.Count;
    
    int count = 0;
    using (IEnumerator<TSource> e = source.GetEnumerator()) {
        checked {
            while (e.MoveNext()) count++;
        }
    }
    return count;
}

Here we see that the iteration is called for every element while (e.MoveNext()), a counter incremented, and the count returned. Looking back at the benchmark data and comparing Enumerable_CountGreaterThanZero with N = 1,000 and N = 10,000 the runtime is a direct function of the size, it’s O(n).

So if you want to know if there are any elements in an Enumerable, use Any(). If you use Resharper, it will provide a hint to use Any() instead of Count() > 0

alt text

List performance

When it comes to lists, calling count Count is near instantaneous. Any() takes longer, however, it’s still fast. Both are O(1).

The Any() method, is the same method that is called on the enumerable since list implements IEnumerable. Referring to the code for Any(), we can see that it finds the fastest way to check if there are any elements. Since List is an IColleciton, it executes .Count != 0 for this check.

Notice that Count isn’t a method, it’s a property. In lists, it doesn’t need to iterate over every element to determine how many there are of them, but instead lists keep track of how many elements there are and it returns that integer.

// Read-only property describing how many elements are in the List.
public int Count => _size;

The performance difference between Any() and Count > 0 is the type checking overhead in Any(). The performance is almost negligible however and it is always advised to avoid premature optimization and to prefer legibility. So unless you are running this in a hot path, calling it very heavily, Any() would be preferred.

Benchmark Source Code

The benchmark source code used for this comparison is included in it’s entirety here.

LinqAnyBenchmark.csproj

<Project Sdk="Microsoft.NET.Sdk">

  <PropertyGroup>
    <OutputType>Exe</OutputType>
    <TargetFramework>net5.0</TargetFramework>
  </PropertyGroup>

  <ItemGroup>
    <PackageReference Include="BenchmarkDotNet" Version="0.13.0" />
  </ItemGroup>

</Project>

Benchmarks.cs

namespace LinqAnyBenchmark
{
    using System;
    using System.Collections.Generic;
    using System.Linq;
    using BenchmarkDotNet.Attributes;

    [MemoryDiagnoser]

    public class Benchmark
    {
        private IEnumerable<Guid> enumerableGuids;
        private List<Guid> listGuids;

        [Params(1000, 10000)]
        public int N;

        [GlobalSetup]
        public void Setup()
        {
            enumerableGuids =
                Enumerable
                    .Range(0, N)
                    .Select(_ => Guid.NewGuid());
            listGuids =
                Enumerable
                    .Range(0, N)
                    .Select(_ => Guid.NewGuid())
                    .ToList();
        }

        [Benchmark]
        public bool Enumerable_Any() => enumerableGuids.Any();

        [Benchmark]
        public bool Enumerable_CountGreaterThanZero() => enumerableGuids.Count() > 0;

        [Benchmark]
        public bool List_Any() => listGuids.Any();

        [Benchmark]
        public bool List_CountGreaterThanZero() => listGuids.Count > 0;
    }
}

Program.cs

namespace LinqAnyBenchmark
{
    using BenchmarkDotNet.Running;

    class Program
    {
        static void Main(string[] args) => BenchmarkRunner.Run<Benchmark>();
    }
}

Posts

Pipenv And Azure Private Package Indexes

Integrating Pipenv and Azure Private Package Indexes

Windows Terminal setup

Windows Terminal

Running unit tests automatically before pushing

Linq Any() performance compared to Count() > 0

Enumerable performance

List performance

Benchmark Source Code