Lambdas in C++0x, Part II

Let's recall the syntax of a lambda expression, from the first part of this post:

  [C](P) M -> R { body }(Q)

  C = capture specification
  P = parameter specification
  M = mutable specification
  R = return type specification
  Q = actual call parameters

Lambda expressions are locally defined functions, also known as anonymous functions. A minimal example could look something like this:

int main()
{
    auto f = []() { std::cout << "hello"; };
    f();
    return 0;
}

The lambda can also be invoked as it is defined, on-the-fly:

[]() { std::cout << "hello"; }();

Here is a lambda that takes an int as argument:

auto fn = [](int x) { std::cout << "x = " << x << "\n"; };
fn(123);

Side note: I tend to use the terms parameter and argument more or less interchangeably. To make things clear, we should, however, make a distinction between the two:

• A parameter is a variable declared in the prototype or declaration of a function.
• The argument is the value that is actually passed to the function.

A return value can be specified, using the trailing return type specification (→ R):

auto fn = [](int x) -> int { return x + 1; };
int n = fn(1);
std::cout << "->" << n << "\n";

The auto keyword saves us some work when the compiler is able to infer the type of an expression automatically. Alternatively, we could write:

std::tr1::function<int(int)> fn = [](int x) -> int { return x + 1; };

In this post, I will continue to examine the capture specification and mutable keyword.

Type Inference in C++: The Return of Auto

With the new C++ standard (C++0x), the auto keyword has returned to the language, like a phoenix rising from the ashes – with new and completely different meaning.

With this new auto, we can ask the compiler to deduce the type of a declared variable from the expression used to initialize it. This feature is known as type inference, and is found in many functional programming languages. Type inference (or type deduction) introduces some of the advantages of dynamic type systems, without incurring the performance penalty and various other negative side-effects associated with relaxed type safety, such as increased risk of run time failure.

A very basic, and not very useful example could look something like:

int x = 1;
auto y = x + 1;

The important thing to note here is that auto is a keyword and not a type. This is to say that, unlike in a dynamically typed language, we can’t assign a value of a different type once the variable has been declared. The type of y (int) must be known at compile time, we just leave to the compiler to figure out what it is.