Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Executing a certain action for all elements in an Enumerable<T>

A quick-and-easy way to get this is:

Names.ToList().ForEach(e => ...);

You are looking for the ever-elusive ForEach that currently only exists on the List generic collection. There are many discussions online about whether Microsoft should or should not add this as a LINQ method. Currently, you have to roll your own:

public static void ForEach<T>(this IEnumerable<T> value, Action<T> action)
{
  foreach (T item in value)
  {
    action(item);
  }
}

While the All() method provides similar abilities, it's use-case is for performing a predicate test on every item rather than an action. Of course, it can be persuaded to perform other tasks but this somewhat changes the semantics and would make it harder for others to interpret your code (i.e. is this use of All() for a predicate test or an action?).


Disclaimer: This post no longer resembles my original answer, but rather incorporates the some seven years experience I've gained since. I made the edit because this is a highly-viewed question and none of the existing answers really covered all the angles. If you want to see my original answer, it's available in the revision history for this post.


The first thing to understand here is C# linq operations like Select(), All(), Where(), etc, have their roots in functional programming. The idea was to bring some of the more useful and approachable parts of functional programming to the .Net world. This is important, because a key tenet of functional programming is for operations to be free of side effects. It's hard to understate this. However, in the case of ForEach()/each(), side effects are the entire purpose of the operation. Adding each() or ForEach() is not just outside the functional programming scope of the other linq operators, but in direct opposition to them.

But I understand this feels unsatisfying. It may help explain why ForEach() was omitted from the framework, but fails to address the real issue at hand. You have a real problem you need to solve. Why should all this ivory tower philosophy get in the way of something that might be genuinely useful?

Eric Lippert, who was on the C# design team at the time, can help us out here. He recommends using a traditional foreach loop:

[ForEach()] adds zero new representational power to the language. Doing this lets you rewrite this perfectly clear code:

foreach(Foo foo in foos){ statement involving foo; }

into this code:

foos.ForEach(foo=>{ statement involving foo; });

His point is, when you look closely at your syntax options, you don't gain anything new from a ForEach() extension versus a traditional foreach loop. I partially disagree. Imagine you have this:

 foreach(var item in Some.Long(and => possibly)
                         .Complicated(set => ofLINQ)
                         .Expression(to => evaluate))
 {
     // now do something
 } 

This code obfuscates meaning, because it separates the foreach keyword from the operations in the loop. It also lists the loop command prior to the operations that define the sequence on which the loop operates. It feels much more natural to want to have those operations come first, and then have the the loop command at the end of the query definition. Also, the code is just ugly. It seems like it would be much nicer to be able to write this:

Some.Long(and => possibly)
   .Complicated(set => ofLINQ)
   .Expression(to => evaluate)
   .ForEach(item => 
{
    // now do something
});

However, even here, I eventually came around to Eric's point of view. I realized code like you see above is calling out for an additional variable. If you have a complicated set of LINQ expressions like that, you can add valuable information to your code by first assigning the result of the LINQ expression to a new variable:

var queryForSomeThing = Some.Long(and => possibly)
                        .Complicated(set => ofLINQ)
                        .Expressions(to => evaluate);
foreach(var item in queryForSomeThing)
{
    // now do something
}

This code feels more natural. It puts the foreach keyword back next to the rest of the loop, and after the query definition. Most of all, the variable name can add new information that will be helpful to future programmers trying to understand the purpose of the LINQ query. Again, we see the desired ForEach() operator really added no new expressive power to the language.

However, we are still missing two features of a hypothetical ForEach() extension method:

  1. It's not composable. I can't add a further .Where() or GroupBy() or OrderBy() after a foreach loop inline with the rest of the code, without creating a new statement.
  2. It's not lazy. These operations happen immediately. It doesn't allow me to, say, have a form where a user chooses an operation as one field in a larger screen that is not acted on until the user presses a command button. This form might allow the user to change their mind before executing the command. This is perfectly normal (easy even) with a LINQ query, but not as simple with a foreach.

(FWIW, most naive .ForEach() implementations also have these issues. But it's possible to craft one without them.)

You could, of course, make your own ForEach() extension method. Several other answers have implementations of this method already; it's not all that complicated. However, I feel like it's unnecessary. There's already an existing method that fits what we want to do from both semantic and operational standpoints. Both of the missing features above can be addressed by use of the existing Select() operation.

Select() fits the kind of transformation or projection described by both of the examples above. Keep in mind, though, that I would still avoid creating side effects. The call to Select() should return either new objects or projections from the originals. This can sometimes be aided through the use of an anonymous type or dynamic object (if and only if necessary). If you need the results to persist in, say, an original list variable, you can always call .ToList() and assign it back to your original variable. I'll add here that I prefer working with IEnumerable<T> variables as much as possible over more concrete types.

myList = myList.Select(item => new SomeType(item.value1, item.value2 *4)).ToList();

In summary:

  1. Just stick with foreach most of the time.
  2. When foreach really won't do (which probably isn't as often as you think), use Select()
  3. When you need to use Select(), you can still generally avoid (program-visible) side effects, possibly by projecting to an anonymous type.
  4. Avoid the crutch of calling ToList(). You don't need it as much as you might think, and it can have significant negative consequence for performance and memory use.

Unfortunately there is no built-in way to do this in the current version of LINQ. The framework team neglected to add a .ForEach extension method. There's a good discussion about this going on right now on the following blog.

http://blogs.msdn.com/kirillosenkov/archive/2009/01/31/foreach.aspx

It's rather easy to add one though.

public static void ForEach<T>(this IEnumerable<T> enumerable, Action<T> action) {
  foreach ( var cur in enumerable ) {
    action(cur);
  }
}