11.6 — Function templates

Let’s say you wanted to write a function to calculate the maximum of two numbers. You might do so like this:

int max(int x, int y)
{
    return (x < y) ? y : x;
    // Note: we use < instead of > because std::max uses <
}

While the caller can pass different values into the function, the type of the parameters is fixed, so the caller can only pass in int values. That means this function really only works well for integers (and types that can be promoted to int).

So what happens later when you want to find the max of two double values? Because C++ requires us to specify the type of all function parameters, the solution is to create a new overloaded version of max with parameters of type double:

double max(double x, double y)
{
    return (x < y) ? y: x;
}

Note that the code for the implementation of the double version of max is exactly the same as for the int version of max! In fact, this implementation works for many different types: including int, double, long, long double, and even new types that you’ve created yourself (which we’ll cover how to do in future lessons).

Having to create overloaded functions with the same implementation for each set of parameter types we want to support is a maintenance headache, a recipe for errors, and a clear violation of the DRY (don’t repeat yourself) principle. There’s a less-obvious challenge here as well: a programmer who wishes to use the max function may wish to call it with an argument type that the author of the max did not anticipate (and thus did not write an overloaded function for).

What we are really missing is some way to write a single version of max that can work with arguments of any type (even types that may not have been anticipated when the code for max was written). Normal functions are simply not up to the task here. Fortunately, C++ supports another feature that was designed specifically to solve this kind of problem.

Welcome to the world of C++ templates.

Introduction to C++ templates

In C++, the template system was designed to simplify the process of creating functions (or classes) that are able to work with different data types.

Instead of manually creating a bunch of mostly-identical functions or classes (one for each set of different types), we instead create a single template. Just like a normal definition, a template describes what a function or class looks like. Unlike a normal definition (where all types must be specified), in a template we can use one or more placeholder types. A placeholder type represents some type that is not known at the time the template is written, but that will be provided later.

Once a template is defined, the compiler can use the template to generate as many overloaded functions (or classes) as needed, each using different actual types!

The end result is the same -- we end up with a bunch of mostly-identical functions or classes (one for each set of different types). But we only have to create and maintain a single template, and the compiler does all the hard work for us.

Because the actual types aren’t determined until the template is used in a program (not when the template is written), the author of the template doesn’t have to try to anticipate all of the actual types that might be used. This means template code can be used with types that didn’t even exist when the template was written! We’ll see how this comes in handy later, when we start exploring the C++ standard library, which is absolutely full of template code!

In the rest of this lesson, we’ll introduce and explore how to create templates for functions, and describe how they work in more detail. We’ll save discussion of class templates until after we’ve covered what classes are.

Function templates

A function template is a function-like definition that is used to generate one or more overloaded functions, each with a different set of actual types. This is what will allow us to create functions that can work with many different types.

When we create our function template, we use placeholder types (also called type template parameters, or informally template types) for any parameter types, return types, or types used in the function body that we want to be specified later.

Function templates are something that is best taught by example, so let’s convert our normal max(int, int) function from the example above into a function template. It’s surprisingly easy, and we’ll explain what’s happening along the way.

Creating a templated max function

Here’s the int version of max again:

int max(int x, int y)
{
    return (x < y) ? y : x;
}

Note that we use type int three times in this function: once for parameter x, once for parameter y, and once for the return type of the function.

To create a function template, we’re going to do two things. First, we’re going to replace our specific types with type template parameters. In this case, because we have only one type that needs replacing (int), we only need one type template parameter (which we’ll call T):

Here’s our new function that uses a single template type:

T max(T x, T y) // won't compile because we haven't defined T
{
    return (x < y) ? y : x;
}

This is a good start -- however, it won’t compile because the compiler doesn’t know what T is! And this is still a normal function, not a function template.

Second, we’re going to tell the compiler that this is a function template, and that T is a type template parameter that is a placeholder for any type. This is done using what is called a template parameter declaration. The scope of a template parameter declaration is limited to the function template (or class template) that follows. Therefore, each function template (or class) needs its own template parameter declaration.

template <typename T> // this is the template parameter declaration
T max(T x, T y) // this is the function template definition for max<T>
{
    return (x < y) ? y : x;
}

In our template parameter declaration, we start with the keyword template, which tells the compiler that we’re creating a template. Next, we specify all of the template parameters that our template will use inside angled brackets (<>). For each type template parameter, we use the keyword typename or class, followed by the name of the type template parameter (e.g. T).

Believe it or not, we’re done! We have created a template version of our max function that can now accept arguments of different types.

Because this function template has one template type named T, we’ll refer to it as max<T>. In the next lesson, we’ll look at how we use our max<T> function template to generate one or more max() functions with parameters of different types.

Naming template parameters

Much like we often use a single letter for variable names used in trivial situations (e.g. x), it’s conventional to use a single capital letter (starting with T) when the template parameter is used in a trivial or obvious way. For example, in our max function template:

template <typename T>
T max(T x, T y)
{
    return (x < y) ? y : x;
}

We don’t need to give T a complex name, because it’s obviously just a placeholder type for the values being compared, and T can be any type that can be compared (such as int, double, or char, but not nullptr).

Our function templates will generally use this naming convention.

If a type template parameter has a non-obvious usage or specific requirements that must be met, there are two common conventions for such names:

• Starting with a capital letter (e.g. Allocator). The standard library uses this naming convention.
• Prefixed with a T, then starting with a capital letter (e.g. TAllocator). This makes it easier to see that the type is a type template parameter.

Which you choose is a matter of personal preference.

template< class T, class Compare >
const T& max( const T& a, const T& b, Compare comp ); // ignore the & for now, we'll cover these in a future lesson