The yield Statement
yield is an overlooked, yet powerful feature.
yield basics
yield places values into an IEnumerable object.
To further understand what yield does, here’s an example with trace statements between each yield statement:
//:SimpleYield.cs
using System;
using System.Collections.Generic;
class SimpleYield
{
static IEnumerable GetValues(){
Console.WriteLine("Returning 1");
yield return 1;
Console.WriteLine("Returning 2");
yield return 2;
Console.WriteLine("Returning 3");
yield return 3;
}
static void Main(){
//Iterate over YieldMethod's IEnumerable
foreach(int i in GetValues())
Console.WriteLine(i);
}
}/*Output:
Returning 1
1
Returning 2
2
Returning 3
3
*///:~
GetValues( )’s yield statements return hard-coded integer values. Main( ) iterates these values, printing each. Notice how GetValue( )’s trace statements interlace with Main( )’s.
A yield statement comes in two different forms, yield return and yield break. yield return places the evaluated expression as the current value of the IEnumerable. yield break marks the end of an iterator:
//:YieldBreak.cs
using System;
using System.Collections.Generic;
class YieldBreak{
static IEnumerable GetValues(){
Console.WriteLine("Returning 1");
yield return 1;
Console.WriteLine("Returning 2");
yield return 2;
Console.WriteLine("Breaking");
yield break; //Break Here
Console.WriteLine("Returning 3");
yield return 3;
}
static void Main(){
foreach(int i in GetValues())
Console.WriteLine(i);
}
}/*Output:
Returning 1
1
Returning 2
2
Breaking
*///:~
yield break prevents remaining code from executing. It reports back to the foreach that there are no values remaining in the IEnumerable. C#’s compiler reports a warning about any unreachable code following yield break.
Exercise 1: Use Visual Studio’s debugger to step through (F11) SimpleYield.cs. Follow the code path as the foreach executes.
Exercise 2: Create a GetValue( ) method that returns a random number of integer values. Use yield return and yield break.
///: Exercise2.cs
using System;
using System.Collections.Generic;
class Exercise2{
public static IEnumerable GetValues(){
Random rand = new Random();
int current = 0;
while (true){
current = rand.Next(20);
if (current == 17)
yield break;
yield return current;
}
}
public static void Main(){
foreach (int i in GetValues()){
Console.WriteLine(i);
}
}
}
///:~
yield Benefits
One strong feature of yield is its ability to defer processing. It delays any calculation until absolutely necessary. Deferred processing makes more responsive programs. yield spreads wait time amongst all iterations by creating values only when necessary:
///: DeferredProcessingTime.cs
using System;
using System.Collections.Generic;
using System.Threading;
class DeferredProcessingTime{
public static IEnumerable CalculateAtOnce(){
int[] intArray = new int[10];
for (int i = 0; i < 10; i++){
Thread.Sleep(1000);
intArray[i] = i;
}
return intArray;
}
public static IEnumerable DeferredCalculate(){
for (int i = 0; i < 10; i++){
Thread.Sleep(1000);
yield return i;
}
}
public static void Main(){
Console.WriteLine("Calculate at once");
foreach (int i in CalculateAtOnce()) //Ten seconds before printing
Console.Write(i);
Console.WriteLine("\nUsing Yield");
foreach (int z in DeferredCalculate())
Console.Write(z); //One second between each write
}
}/*Output:
Calculate at once
0123456789
Using Yield
0123456789
*///:~
The above example emulates an algorithm that takes a long time to process by using Thread.Sleep( ) before adding each value.
Both foreachs over CalculateAtOnce( ) and DefferedCalculate( ) print the same results to the console. However the execution behavior is noticeably different. CalculateAtOnce( )’s foreach waits for 10 seconds and prints 0 through 9. DeferredCalculate( )’s foreach prints each value per second.
Another benefit of yield’s deferred processing is that it does not require a collection stored in memory. yield can iterate over a large set of data without consuming large amounts of memory. Take the following example. You iterate over a set of data with 10 million integer values. Without yield, the entire collection of values must be returned. With yield, each integer value is returned when requested.
///:DeferredProcessingMemory.cs
using System;
using System.Collections.Generic;
using System.Threading;
class DeferredProcessingMemory{
public static IEnumerable CalculateAtOnce(){
int[] intArray = new int[10000000];
for (int i = 0; i < 10000000; i++)
intArray[i] = i;
return intArray;
}
public static IEnumerable DeferredCalculate(){
for (int i = 0; i < 10000000; i++)
yield return i;
}
public static void Main(){
Console.WriteLine("Using Yield");
IEnumerable deferred = DeferredCalculate();
Console.ReadLine(); //Memory usage: ~2MB
Console.WriteLine("Calculating at once");
IEnumerable atOnce = CalculateAtOnce();
Console.ReadLine(); //Memory usage: ~40MB
}
}/*Output:
Using Yield
Calculating at once
*///:~
Start the Windows Task Manager (Ctrl+Shift+Esc) before running this example. Find your process, and notice the amount of memory usage from using DeferredCalculate( ). Press enter and see the drastic increase that CalculateAtOnce( ) makes; intArray’s 10 million in-memory values are the cause.
yield gives potential to handle data sets of any size without concern for memory limitations.
Exercise 3: Use yield to iterate over an infinite data set of random integers.
///:Exercise3.cs
using System;
using System.Collections.Generic;
class Exercise3{
public static Random rand = new Random();
public static IEnumerable GetValues(){
while(true)
yield return rand.Next();
}
public static void Main(){
foreach(int i in GetValues())
Console.WriteLine(i);
}
}
///:~
Under the Hood
Many C# keywords involve the generation of MSIL code, methods, or even classes. yield is no exception.
Red Gate’s .NET Reflector shows the generated C# from using the yield statement. The disassembled assembly of SimpleYield.cs contains an extra class implementing IEnumerable.
public class GeneratedClass : IEnumerable, IEnumerable,
IEnumerator, IEnumerator, IDisposable{
//...CUT...
public bool MoveNext(){
switch (this.state){
case 0:
this.state = -1;
Console.WriteLine("Returning 1");
this.current = 1;
this.state = 1;
return true;
case 1:
this.state = -1;
Console.WriteLine("Returning 2");
this.current = 2;
this.state = 2;
return true;
case 2:
this.state = -1;
Console.WriteLine("Returning 3");
this.current = 3;
this.state = 3;
return true;
case 3:
this.state = -1;
break;
}
return false;
}
//...CUT...
}
The compiler generates a switch statement to handle executing the code written in your original yielding method. Only one case statement is used for each time MoveNext( ) is called. This is how the deferred processing actually occurs.
GetValues( ) is also changed to use the GeneratedClass( ) in place of the original code.
private static IEnumerable GetValues(){
return new GeneratedClass(-2);
}
The -2 in the new GeneratedClass( ) call is used to set the initial state. The state’s value is changed when MoveNext( ) is called. Beginning each case, state is set to -1 to prevent further execution if an exception occurs. Just before a case returns, state is assigned the value of the next case.
Exercise 4: Write an implementation of yield’s generated code and execute a foreach over it.
//:Exercise4.cs
using System;
using System.Collections.Generic;
using System.Collections;
class Exercise4 : IEnumerable,IEnumerable,
IEnumerator,IEnumerator, IDisposable{
string current = "";
int state;
public IEnumerator GetEnumerator(){
return this;
}
IEnumerator IEnumerable.GetEnumerator(){
return GetEnumerator();
}
public string Current{
get { return current; }
}
public void Dispose(){}
object IEnumerator.Current{
get { return Current; }
}
public bool MoveNext(){
switch(state){
case 0:
state = -1;
current = "One";
state = 1;
return true;
case 1:
state = -1;
current = "Two";
state = 2;
return true;
case 2:
state = -3;
current = "Three";
state = 3;
return true;
}
return false;
}
public void Reset(){
throw new NotImplementedException();
}
public static void Main(){
foreach(string s in new Exercise4())
Console.WriteLine(s);
}
}/*Output:
One
Two
Three
*///:~
