From the standpoint of a method which receives a parameter, the difference between ref and out is that C# requires that methods must write to every out parameter before returning, and must not do anything with such a parameter, other than passing it as an out parameter or writing to it, until it has been either passed as an out parameter to another method or written directly. Note that some other languages do not impose such requirements; a virtual or interface method which is declared in C# with an out parameter may be overridden in another language which does not impose any special restrictions on such parameters.

从调用者的角度来看，c#在很多情况下假设调用带有out形参的方法时，会导致传入的变量在未被读取之前就被写入。当调用用其他语言编写的方法时，这个假设可能不正确。例如:

struct MyStruct
{
   ...
   myStruct(IDictionary<int, MyStruct> d)
   {
     d.TryGetValue(23, out this);
   }
}

如果myDictionary标识了一个用c#以外的语言编写的dictionary <TKey,TValue>实现，即使MyStruct s = new MyStruct(myDictionary);看起来像一个任务，它可能会留下未修改的s。

注意用VB编写的构造函数。NET与c#中的方法不同，它不假设被调用的方法是否会修改任何out参数，并且无条件地清除所有字段。上面提到的奇怪行为不会发生在完全用VB或完全用c#编写的代码中，但是当用c#编写的代码调用用VB.NET编写的方法时就会发生。

对于那些以身作则的人(比如我)，以下是安东尼·科索夫所说的。

我创建了一些ref、out和其他例子来说明这一点。我并没有介绍最佳实践，只是举例来理解它们之间的差异。

https://gist.github.com/2upmedia/6d98a57b68d849ee7091

注意，在函数内部传递的引用参数是直接处理的。

例如,

    public class MyClass
    {
        public string Name { get; set; }
    }

    public void Foo()
    {
        MyClass myObject = new MyClass();
        myObject.Name = "Dog";
        Bar(myObject);
        Console.WriteLine(myObject.Name); // Writes "Dog".
    }

    public void Bar(MyClass someObject)
    {
        MyClass myTempObject = new MyClass();
        myTempObject.Name = "Cat";
        someObject = myTempObject;
    }

这会写狗，而不是猫。因此，您应该直接在someObject上工作。

除了允许你将别人的变量重新分配给类的不同实例，返回多个值等，使用ref或out可以让别人知道你需要从他们那里得到什么，以及你打算用他们提供的变量做什么

You don't need ref or out if all you're going to do is modify things inside the MyClass instance that is passed in the argument someClass. The calling method will see changes like someClass.Message = "Hello World" whether you use ref, out or nothing Writing someClass = new MyClass() inside myFunction(someClass) swaps out the object seen by the someClass in the scope of the myFunction method only. The calling method still knows about the original MyClass instance it created and passed to your method You need ref or out if you plan on swapping the someClass out for a whole new object and want the calling method to see your change Writing someClass = new MyClass() inside myFunction(out someClass) changes the object seen by the method that called myFunction

还有其他程序员

他们想知道你将如何处理他们的数据。假设您正在编写一个将被数百万开发人员使用的库。你想让他们知道当他们调用你的方法时你要对他们的变量做什么

Using ref makes a statement of "Pass a variable assigned to some value when you call my method. Be aware that I might change it out for something else entirely during the course of my method. Do not expect your variable to be pointing to the old object when I'm done" Using out makes a statement of "Pass a placeholder variable to my method. It doesn't matter whether it has a value or not; the compiler will force me to assign it to a new value. I absolutely guarantee that the object pointed to by your variable before you called my method, will be different by the time I'm done

顺便说一下，在c# 7.2中也有一个in修饰符

And that prevents the method from swapping out the passed in instance for a different instance. Think of it like saying to those millions of developers "pass me your original variable reference, and I promise not to swap your carefully crafted data out for something else". in has some peculiarities, and in some cases such as where an implicit conversion might be required to make your short compatible with an in int the compiler will temporarily make an int, widen your short to it, pass it by reference and finish up. It can do this because you've declared you're not going to mess with it.

微软对数值类型的.TryParse方法做到了这一点:

int i = 98234957;
bool success = int.TryParse("123", out i);

通过将参数标记为out他们在这里积极地声明"我们肯定会将你苦心制作的98234957值更改为其他值"

当然，对于像解析值类型这样的事情，它们有点不得不这样做，因为如果parse方法不允许将值类型替换为其他类型，那么它就不能很好地工作。但是想象一下在你创建的库中有一些虚构的方法:

public void PoorlyNamedMethod(out SomeClass x)

你可以看到它是一个out，因此你可以知道，如果你花了几个小时处理数字，创建一个完美的SomeClass:

SomeClass x = SpendHoursMakingMeAPerfectSomeClass();
//now give it to the library
PoorlyNamedMethod(out x);

那真是浪费时间，花那么多时间来做一节完美的课。它肯定会被丢弃，并被PoorlyNamedMethod取代

如果你想将参数作为引用传递，那么你应该在将参数传递给函数之前初始化它，否则编译器本身会显示错误。但是对于out形参，您不需要在将对象形参传递给方法之前初始化它。可以在调用方法本身中初始化对象。

: return语句只能用于从函数中返回一个值。但是，使用输出参数，可以从一个函数返回两个值。输出参数类似于引用参数，只是它们将数据传输出方法而不是传输到方法中。

下面的例子说明了这一点:

using System;

namespace CalculatorApplication
{
   class NumberManipulator
   {
      public void getValue(out int x )
      {
         int temp = 5;
         x = temp;
      }

      static void Main(string[] args)
      {
         NumberManipulator n = new NumberManipulator();
         /* local variable definition */
         int a = 100;

         Console.WriteLine("Before method call, value of a : {0}", a);

         /* calling a function to get the value */
         n.getValue(out a);

         Console.WriteLine("After method call, value of a : {0}", a);
         Console.ReadLine();

      }
   }
}

裁判: 引用形参是对变量内存位置的引用。与值参数不同，通过引用传递参数时，不会为这些参数创建新的存储位置。引用参数表示与提供给方法的实际参数相同的内存位置。

在c#中，使用ref关键字声明引用参数。下面的例子说明了这一点:

using System;
namespace CalculatorApplication
{
   class NumberManipulator
   {
      public void swap(ref int x, ref int y)
      {
         int temp;

         temp = x; /* save the value of x */
         x = y;   /* put y into x */
         y = temp; /* put temp into y */
       }

      static void Main(string[] args)
      {
         NumberManipulator n = new NumberManipulator();
         /* local variable definition */
         int a = 100;
         int b = 200;

         Console.WriteLine("Before swap, value of a : {0}", a);
         Console.WriteLine("Before swap, value of b : {0}", b);

         /* calling a function to swap the values */
         n.swap(ref a, ref b);

         Console.WriteLine("After swap, value of a : {0}", a);
         Console.WriteLine("After swap, value of b : {0}", b);

         Console.ReadLine();

      }
   }
}