Iterators
Almost every program you write will have some need to iterate over a collection. You'll write code that examines every item in a collection.
You'll also create iterator methods, which are methods that produce an iterator for the elements of that class. An iterator is an object that traverses a container, particularly lists. Iterators can be used for:
- Performing an action on each item in a collection.
- Enumerating a custom collection.
- Extending LINQ or other libraries.
- Creating a data pipeline where data flows efficiently through iterator methods.
The C# language provides features for both generating and consuming sequences. These sequences can be produced and consumed synchronously or asynchronously. This article provides an overview of those features.
Iterating with foreach
Enumerating a collection is simple: The foreach keyword enumerates a collection, executing the embedded statement once for each element in the collection:
foreach (var item in collection)
{
Console.WriteLine(item?.ToString());
}
That's all. To iterate over all the contents of a collection, the foreach statement is all you need. The foreach statement isn't magic, though. It relies on two generic interfaces defined in the .NET core library to generate the code necessary to iterate a collection: IEnumerable<T> and IEnumerator<T>. This mechanism is explained in more detail below.
Both of these interfaces also have non-generic counterparts: IEnumerable and IEnumerator. The generic versions are preferred for modern code.
When a sequence is generated asynchronously, you can use the await foreach statement to asynchronously consume the sequence:
await foreach (var item in asyncSequence)
{
Console.WriteLine(item?.ToString());
}
When a sequence is an System.Collections.Generic.IEnumerable<T>, you use foreach. When a sequence is an System.Collections.Generic.IAsyncEnumerable<T>, you use await foreach. In the latter case, the sequence is generated asynchronously.
Enumeration sources with iterator methods
Another great feature of the C# language enables you to build methods that create a source for an enumeration. These methods are referred to as iterator methods. An iterator method defines how to generate the objects in a sequence when requested. You use the yield return contextual keywords to define an iterator method.
You could write this method to produce the sequence of integers from 0 through 9:
public IEnumerable<int> GetSingleDigitNumbers()
{
yield return 0;
yield return 1;
yield return 2;
yield return 3;
yield return 4;
yield return 5;
yield return 6;
yield return 7;
yield return 8;
yield return 9;
}
The code above shows distinct yield return statements to highlight the fact that you can use multiple discrete yield return statements in an iterator method. You can (and often do) use other language constructs to simplify the code of an iterator method. The method definition below produces the exact same sequence of numbers:
public IEnumerable<int> GetSingleDigitNumbersLoop()
{
int index = 0;
while (index < 10)
yield return index++;
}
You don't have to decide one or the other. You can have as many yield return statements as necessary to meet the needs of your method:
public IEnumerable<int> GetSetsOfNumbers()
{
int index = 0;
while (index < 10)
yield return index++;
yield return 50;
index = 100;
while (index < 110)
yield return index++;
}
All of these preceding examples would have an asynchronous counterpart. In each case, you'd replace the return type of IEnumerable<T> with an IAsyncEnumerable<T>. For example, the previous example would have the following asynchronous version:
public async IAsyncEnumerable<int> GetSetsOfNumbersAsync()
{
int index = 0;
while (index < 10)
yield return index++;
await Task.Delay(500);
yield return 50;
await Task.Delay(500);
index = 100;
while (index < 110)
yield return index++;
}
That's the syntax for both synchronous and asynchronous iterators. Let's consider a real world example. Imagine you're on an IoT project and the device sensors generate a very large stream of data. To get a feel for the data, you might write a method that samples every Nth data element. This small iterator method does the trick:
public static IEnumerable<T> Sample<T>(this IEnumerable<T> sourceSequence, int interval)
{
int index = 0;
foreach (T item in sourceSequence)
{
if (index++ % interval == 0)
yield return item;
}
}
If reading from the IoT device produces an asynchronous sequence, you'd modify the method as the following method shows:
public static async IAsyncEnumerable<T> Sample<T>(this IAsyncEnumerable<T> sourceSequence, int interval)
{
int index = 0;
await foreach (T item in sourceSequence)
{
if (index++ % interval == 0)
yield return item;
}
}
There's one important restriction on iterator methods: you can't have both a return statement and a yield return statement in the same method. The following code won't compile:
public IEnumerable<int> GetSingleDigitNumbers()
{
int index = 0;
while (index < 10)
yield return index++;
yield return 50;
// generates a compile time error:
var items = new int[] {100, 101, 102, 103, 104, 105, 106, 107, 108, 109 };
return items;
}
This restriction normally isn't a problem. You have a choice of either using yield return throughout the method, or separating the original method into multiple methods, some using return, and some using yield return.
You can modify the last method slightly to use yield return everywhere:
public IEnumerable<int> GetFirstDecile()
{
int index = 0;
while (index < 10)
yield return index++;
yield return 50;
var items = new int[] {100, 101, 102, 103, 104, 105, 106, 107, 108, 109 };
foreach (var item in items)
yield return item;
}
Sometimes, the right answer is to split an iterator method into two different methods. One that uses return, and a second that uses yield return. Consider a situation where you might want to return an empty collection, or the first five odd numbers, based on a boolean argument. You could write that as these two methods:
public IEnumerable<int> GetSingleDigitOddNumbers(bool getCollection)
{
if (getCollection == false)
return new int[0];
else
return IteratorMethod();
}
private IEnumerable<int> IteratorMethod()
{
int index = 0;
while (index < 10)
{
if (index % 2 == 1)
yield return index;
index++;
}
}
Look at the methods above. The first uses the standard return statement to return either an empty collection, or the iterator created by the second method. The second method uses the yield return statement to create the requested sequence.
Deeper dive into foreach
The foreach statement expands into a standard idiom that uses the IEnumerable<T> and IEnumerator<T> interfaces to iterate across all elements of a collection. It also minimizes errors developers make by not properly managing resources.
The compiler translates the foreach loop shown in the first example into something similar to this construct:
IEnumerator<int> enumerator = collection.GetEnumerator();
while (enumerator.MoveNext())
{
var item = enumerator.Current;
Console.WriteLine(item.ToString());
}
The exact code generated by the compiler is more complicated, and handles situations where the object returned by GetEnumerator() implements the IDisposable interface. The full expansion generates code more like this:
{
var enumerator = collection.GetEnumerator();
try
{
while (enumerator.MoveNext())
{
var item = enumerator.Current;
Console.WriteLine(item.ToString());
}
}
finally
{
// dispose of enumerator.
}
}
The compiler translates the first asynchronous sample into something similar to this construct:
{
var enumerator = collection.GetAsyncEnumerator();
try
{
while (await enumerator.MoveNextAsync())
{
var item = enumerator.Current;
Console.WriteLine(item.ToString());
}
}
finally
{
// dispose of async enumerator.
}
}
The manner in which the enumerator is disposed of depends on the characteristics of the type of enumerator. In the general synchronous case, the finally clause expands to:
finally
{
(enumerator as IDisposable)?.Dispose();
}
The general asynchronous case expands to:
finally
{
if (enumerator is IAsyncDisposable asyncDisposable)
await asyncDisposable.DisposeAsync();
}
However, if the type of enumerator is a sealed type and there's no implicit conversion from the type of enumerator to IDisposable or IAsyncDisposable, the finally clause expands to an empty block:
finally
{
}
If there's an implicit conversion from the type of enumerator to IDisposable, and enumerator is a non-nullable value type, the finally clause expands to:
finally
{
((IDisposable)enumerator).Dispose();
}
Thankfully, you don't need to remember all these details. The foreach statement handles all those nuances for you. The compiler will generate the correct code for any of these constructs.
