I'd like to do the equivalent of the following in LINQ, but I can't figure out how:
IEnumerable<Item> items = GetItems();
items.ForEach(i => i.DoStuff());
What is the real syntax?
I'd like to do the equivalent of the following in LINQ, but I can't figure out how:
IEnumerable<Item> items = GetItems();
items.ForEach(i => i.DoStuff());
What is the real syntax?
There is no ForEach extension for IEnumerable; only for List<T>. So you could do
items.ToList().ForEach(i => i.DoStuff());
Alternatively, write your own ForEach extension method:
public static void ForEach<T>(this IEnumerable<T> enumeration, Action<T> action)
{
foreach(T item in enumeration)
{
action(item);
}
}
Fredrik has provided the fix, but it may be worth considering why this isn't in the framework to start with. I believe the idea is that the LINQ query operators should be side-effect-free, fitting in with a reasonably functional way of looking at the world. Clearly ForEach is exactly the opposite - a purely side-effect-based construct.
That's not to say this is a bad thing to do - just thinking about the philosophical reasons behind the decision.
So many answers, yet ALL fail to pinpoint one very significant problem with a custom generic ForEach extension: Performance! And more specifically, memory usage and GC.
Consider the sample below. Targeting .NET Framework 4.7.2 or .NET Core 3.1.401, configuration is Release and platform is Any CPU.
public static class Enumerables
{
public static void ForEach<T>(this IEnumerable<T> @this, Action<T> action)
{
foreach (T item in @this)
{
action(item);
}
}
}
class Program
{
private static void NoOp(int value) {}
static void Main(string[] args)
{
var list = Enumerable.Range(0, 10).ToList();
for (int i = 0; i < 1000000; i++)
{
// WithLinq(list);
// WithoutLinqNoGood(list);
WithoutLinq(list);
}
}
private static void WithoutLinq(List<int> list)
{
foreach (var item in list)
{
NoOp(item);
}
}
private static void WithLinq(IEnumerable<int> list) => list.ForEach(NoOp);
private static void WithoutLinqNoGood(IEnumerable<int> enumerable)
{
foreach (var item in enumerable)
{
NoOp(item);
}
}
}
At a first glance, all three variants should perform equally well. However, when the ForEach extension method is called many, many times, you will end up with garbage that implies a costly GC. In fact, having this ForEach extension method on a hot path has been proven to totally kill performance in our loop-intensive application.
Similarly, the weakly typed foreach loop will also produce garbage, but it will still be faster and less memory-intensive than the ForEach extension (which also suffers from a delegate allocation).
For a strongly typed foreach the compiler is able to use any optimized enumerator (e.g. value based) of a class, whereas a generic ForEach extension must fall back to a generic enumerator which will be allocated on each run. Furthermore, the actual delegate will also imply an additional allocation.
You would get similar bad results with the WithoutLinqNoGood method. There, the argument is of type IEnumerable<int> instead of List<int> implying the same type of enumerator allocation.
Below are the relevant differences in IL. A value based enumerator is certainly preferable!
IL_0001: callvirt instance class
[mscorlib]System.Collections.Generic.IEnumerator`1<!0>
class [mscorlib]System.Collections.Generic.IEnumerable`1<!!T>::GetEnumerator()
vs
IL_0001: callvirt instance valuetype
[mscorlib]System.Collections.Generic.List`1/Enumerator<!0>
class [mscorlib]System.Collections.Generic.List`1<int32>::GetEnumerator()
The OP asked how to call ForEach() on an IEnumerable<T>. The original answer clearly shows how it can be done. Sure you can do it, but then again; my answer clearly shows that you shouldn't.
Verified the same behavior when targeting .NET Core 3.1.401 (compiling with Visual Studio 16.7.2).
MoreLinq has IEnumerable<T>.ForEach and a ton of other useful extensions. It's probably not worth taking the dependency just for ForEach, but there's a lot of useful stuff in there.
I respectually disagree with the notion that link extension methods should be side-effect free (not only because they aren't, any delegate can perform side effects).
Consider the following:
public class Element {}
public Enum ProcessType
{
This = 0, That = 1, SomethingElse = 2
}
public class Class1
{
private Dictionary<ProcessType, Action<Element>> actions =
new Dictionary<ProcessType,Action<Element>>();
public Class1()
{
actions.Add( ProcessType.This, DoThis );
actions.Add( ProcessType.That, DoThat );
actions.Add( ProcessType.SomethingElse, DoSomethingElse );
}
// Element actions:
// This example defines 3 distict actions
// that can be applied to individual elements,
// But for the sake of the argument, make
// no assumption about how many distict
// actions there may, and that there could
// possibly be many more.
public void DoThis( Element element )
{
// Do something to element
}
public void DoThat( Element element )
{
// Do something to element
}
public void DoSomethingElse( Element element )
{
// Do something to element
}
public void Apply( ProcessType processType, IEnumerable<Element> elements )
{
Action<Element> action = null;
if( ! actions.TryGetValue( processType, out action ) )
throw new ArgumentException("processType");
foreach( element in elements )
action(element);
}
}
What the example shows is really just a kind of late-binding that allows one invoke one of many possible actions having side-effects on a sequence of elements, without having to write a big switch construct to decode the value that defines the action and translate it into its corresponding method.
To stay fluent one can use such a trick:
GetItems()
.Select(i => new Action(i.DoStuf)))
.Aggregate((a, b) => a + b)
.Invoke();