What is the difference between lambdas and delegates in the .NET Framework?

I get asked this question a lot and I thought I'd solicit some input on how to best describe the difference.


Solution 1:

They are actually two very different things. "Delegate" is actually the name for a variable that holds a reference to a method or a lambda, and a lambda is a method without a permanent name.

Lambdas are very much like other methods, except for a couple subtle differences.

  1. A normal method is defined in a "statement" and tied to a permanent name, whereas a lambda is defined "on the fly" in an "expression" and has no permanent name.
  2. Some lambdas can be used with .NET expression trees, whereas methods cannot.

A delegate is defined like this:

delegate Int32 BinaryIntOp(Int32 x, Int32 y);

A variable of type BinaryIntOp can have either a method or a labmda assigned to it, as long as the signature is the same: two Int32 arguments, and an Int32 return.

A lambda might be defined like this:

BinaryIntOp sumOfSquares = (a, b) => a*a + b*b;

Another thing to note is that although the generic Func and Action types are often considered "lambda types", they are just like any other delegates. The nice thing about them is that they essentially define a name for any type of delegate you might need (up to 4 parameters, though you can certainly add more of your own). So if you are using a wide variety of delegate types, but none more than once, you can avoid cluttering your code with delegate declarations by using Func and Action.

Here is an illustration of how Func and Action are "not just for lambdas":

Int32 DiffOfSquares(Int32 x, Int32 y)
{
  return x*x - y*y;
}

Func<Int32, Int32, Int32> funcPtr = DiffOfSquares;

Another useful thing to know is that delegate types (not methods themselves) with the same signature but different names will not be implicitly casted to each other. This includes the Func and Action delegates. However if the signature is identical, you can explicitly cast between them.

Going the extra mile.... In C# functions are flexible, with the use of lambdas and delegates. But C# does not have "first-class functions". You can use a function's name assigned to a delegate variable to essentially create an object representing that function. But it's really a compiler trick. If you start a statement by writing the function name followed by a dot (i.e. try to do member access on the function itself) you'll find there are no members there to reference. Not even the ones from Object. This prevents the programmer from doing useful (and potentially dangerous of course) things such as adding extension methods that can be called on any function. The best you can do is extend the Delegate class itself, which is surely also useful, but not quite as much.

Update: Also see Karg's answer illustrating the difference between anonymous delegates vs. methods & lambdas.

Update 2: James Hart makes an important, though very technical, note that lambdas and delegates are not .NET entities (i.e. the CLR has no concept of a delegate or lambda), but rather they are framework and language constructs.

Solution 2:

The question is a little ambiguous, which explains the wide disparity in answers you're getting.

You actually asked what the difference is between lambdas and delegates in the .NET framework; that might be one of a number of things. Are you asking:

  • What is the difference between lambda expressions and anonymous delegates in the C# (or VB.NET) language?

  • What is the difference between System.Linq.Expressions.LambdaExpression objects and System.Delegate objects in .NET 3.5?

  • Or something somewhere between or around those extremes?

Some people seem to be trying to give you the answer to the question 'what is the difference between C# Lambda expressions and .NET System.Delegate?', which doesn't make a whole lot of sense.

The .NET framework does not in itself understand the concepts of anonymous delegates, lambda expressions, or closures - those are all things defined by language specifications. Think about how the C# compiler translates the definition of an anonymous method into a method on a generated class with member variables to hold closure state; to .NET, there's nothing anonymous about the delegate; it's just anonymous to the C# programmer writing it. That's equally true of a lambda expression assigned to a delegate type.

What .NET DOES understand is the idea of a delegate - a type that describes a method signature, instances of which represent either bound calls to specific methods on specific objects, or unbound calls to a particular method on a particular type that can be invoked against any object of that type, where said method adheres to the said signature. Such types all inherit from System.Delegate.

.NET 3.5 also introduces the System.Linq.Expressions namespace, which contains classes for describing code expressions - and which can also therefore represent bound or unbound calls to methods on particular types or objects. LambdaExpression instances can then be compiled into actual delegates (whereby a dynamic method based on the structure of the expression is codegenned, and a delegate pointer to it is returned).

In C# you can produce instances of System.Expressions.Expression types by assigning a lambda expression to a variable of said type, which will produce the appropriate code to construct the expression at runtime.

Of course, if you were asking what the difference is between lambda expressions and anonymous methods in C#, after all, then all this is pretty much irelevant, and in that case the primary difference is brevity, which leans towards anonymous delegates when you don't care about parameters and don't plan on returning a value, and towards lambdas when you want type inferenced parameters and return types.

And lambda expressions support expression generation.