{"id":6143,"date":"2021-05-19T20:33:29","date_gmt":"2021-05-19T20:33:29","guid":{"rendered":"https:\/\/www.modernescpp.com\/index.php\/class-templates\/"},"modified":"2021-05-19T20:33:29","modified_gmt":"2021-05-19T20:33:29","slug":"class-templates","status":"publish","type":"post","link":"https:\/\/www.modernescpp.com\/index.php\/class-templates\/","title":{"rendered":"Class Templates"},"content":{"rendered":"

A function template represents a family of functions. Accordingly, a class template represents a family of classes. Today, I want to introduce class templates.<\/p>\n

<\/p>\n

 \"ClassTemplate\"<\/p>\n

Defining a class template is straightforward.<\/p>\n<\/p>\n

Definition of a Class Template<\/h2>\n

Assume you have a class Array<\/code> that should become a class template.<\/p>\n

<\/p>\n

\n
class<\/span> Array<\/span>{\r\n \r\n public:<\/span>\r\n<\/span>    int<\/span> getSize() const<\/span> {\r\n        return<\/span> 10<\/span>;\r\n    }\r\n\r\n private:<\/span>\r\n    int<\/span> elem[10<\/span>];\r\n};\r\n<\/pre>\n<\/div>\n
 <\/p>\n
The class Array<\/code> holds a C-array of int with length 10. The type of the C-array and its length are obvious generalization points. Let’s make a class template by introducing a type parameter T<\/code> and a non-type parameter N<\/code> and playing with them.<\/p>\n
<\/p>\n
 <\/p>\n
<\/p>\n
\n
\/\/ arrayClassTemplate.cpp<\/span>\r\n\r\n#include <cstddef>                     <\/span>\/\/ (1)<\/span>\r\n#include <iostream><\/span>\r\n#include <string><\/span>\r\n\r\ntemplate<\/span> <<\/span>typename<\/span> T, std::<\/span>size_t<\/span> N><\/span>   \/\/ (2)<\/span>\r\nclass<\/span> Array<\/span>{\r\n\r\n public:<\/span>\r\n<\/span>    std::size_t<\/span> getSize() const<\/span> {\r\n        return<\/span> N;\r\n    }\r\n\r\nprivate:<\/span>\r\n    T elem[N];\r\n};\r\n\r\nint<\/span> main<\/span>() {\r\n\r\n    std::<\/span>cout <<<\/span> '\\n'<\/span>;\r\n\r\n    Array<<\/span>int<\/span>, 100<\/span>><\/span> intArr;             \/\/ (3)<\/span>\r\n    std::<\/span>cout <<<\/span> \"intArr.getSize(): \"<\/span> <<<\/span> intArr.getSize() <<<\/span> '\\n'<\/span>;\r\n\r\n    Array<<\/span>std::<\/span>string, 5<\/span>><\/span> strArr;       \/\/ (4)<\/span>\r\n    std::<\/span>cout <<<\/span> \"strArr.getSize(): \"<\/span> <<<\/span> strArr.getSize() <<<\/span> '\\n'<\/span>;\r\n\r\n    Array<<\/span>Array<<\/span>int<\/span>, 3<\/span>><\/span>, 25<\/span>><\/span> intArrArr; \/\/ (5)<\/span>\r\n    std::<\/span>cout <<<\/span> \"intArrArr.getSize(): \"<\/span> <<<\/span> intArrArr.getSize() <<<\/span> '\\n'<\/span>;\r\n\r\n    std::<\/span>cout <<<\/span> '\\n'<\/span>;\r\n\r\n}\r\n<\/pre>\n<\/div>\n
 <\/p>\n
The Array<\/code> is parametrized by its type and its size. For size, I used the unsigned integer type std::size_t<\/code> (2) that can store the maximum size. To use std::size_t<\/code>I have to include the header <cstddef> <\/code>(1). So far, the Array<\/code> can be instantiated with an int<\/code> (3) with a std::string<\/code> (4), and with an Array<int, 3><\/code> (5). The following screenshot shows the output of the program.<\/p>\n
\"arrayClassTemplate\"<\/p>\n
You can define the member functions of a template inside and outside the class template.<\/p>\n
Definitions of Member Functions<\/h3>\n
Defining the member function inside the class template <\/code>is straightforward.<\/p>\n
<\/p>\n
\n
template<\/span> <<\/span>typename<\/span> T, std::<\/span>size_t<\/span> N><\/span>   <\/span>\r\nclass<\/span> Array<\/span>{\r\n\r\n public:<\/span>\r\n    std::size_t<\/span> getSize() const<\/span> {\r\n        return<\/span> N;\r\n    }\r\n\r\nprivate:<\/span>\r\n    T elem[N];\r\n};\r\n<\/pre>\n<\/div>\n
 <\/p>\n
When you define the member functions outside the class, you have to specify that it is a template, and you have to specify the full type qualification of the class template. Here is the modified class template Array<\/code>:<\/p>\n
 <\/p>\n
<\/p>\n
\n
template<\/span> <<\/span>typename<\/span> T, std::<\/span>size_t<\/span> N><\/span> \r\nclass<\/span> Array<\/span>{\r\n\r\n public:<\/span>\r\n    std::sizt_<\/span> getSize() const<\/span>;\r\n\r\nprivate:<\/span>\r\n    T elem[N];\r\n};\r\n\r\ntemplate<\/span> <<\/span>typename<\/span> T, std::<\/span>size_t<\/span> N><\/span>   \/\/ (1)<\/span>\r\nstd::size_t<\/span> Array<<\/span>T, N>::<\/span>getSize() const<\/span> {\r\n    return<\/span> N;\r\n}\r\n<\/pre>\n<\/div>\n
 <\/p>\n
(1) is the member function getSize<\/code> of the Array<\/code>, defined outside the class. Defining the member function outside the class template becomes challenging if the member function itself is a template.<\/p>\n
Member Functions as Templates<\/h3>\n
A typical example of a generic member function is a templated assignment operator. The reason is straightforward. You want to assign an Array<T, N><\/code> to an Array<T2, N2><\/code> if T<\/code> is assignable to T2<\/code> and both arrays have the same size.<\/p>\n
Assigning an Array<float, 5><\/code> to an Array<double, 5><\/code> is not valid because both arrays have different types.<\/p>\n
 <\/p>\n
<\/p>\n
\n
\/\/ arrayAssignmentError.cpp<\/span>\r\n\r\n#include <cstddef>                     <\/span>\r\n#include <iostream><\/span>\r\n#include <string><\/span>\r\n\r\ntemplate<\/span> <<\/span>typename<\/span> T, std::<\/span>size_t<\/span> N><\/span>   \r\nclass<\/span> Array<\/span>{\r\n\r\n public:<\/span>\r\n    std::size_t<\/span> getSize() const<\/span> {\r\n        return<\/span> N;\r\n    }\r\n\r\nprivate:<\/span>\r\n    T elem[N];\r\n};\r\n\r\nint<\/span> main<\/span>() {\r\n\r\n    std::<\/span>cout <<<\/span> '\\n'<\/span>;\r\n\r\n    Array<<\/span>float<\/span>, 5<\/span>><\/span> floatArr;  \r\n    Array<<\/span>float<\/span>, 5<\/span>><\/span> floatArr2;\r\n    \r\n    floatArr2 =<\/span> floatArr;             \/\/ (1)<\/span>\r\n     \r\n    \r\n    Array<<\/span>double<\/span>, 5<\/span>><\/span> doubleArr;       \r\n    doubleArr =<\/span> floatArr;             \/\/ (2)<\/span>\r\n    \r\n    \r\n}\r\n<\/pre>\n<\/div>\n
 <\/p>\n
Assigning floatArr<\/code> to floatArr2<\/code> (1) is valid because both arrays have the same type. Assigning floatArr<\/code> to doubleArr<\/code> is an error (2) because both classes have different types. The compiler consequently complains that there is no conversion from Array<float, 5><\/code> to an Array<double, 5>. <\/code><\/p>\n
 \"arrayAssignmentError\"<\/p>\n
Here is a naive implementation of the class Array that supports the assignment of two arrays of the same length. The C-array elem<\/code> is intentionally public.<\/p>\n
 <\/p>\n
<\/p>\n
\n
template<\/span> <<\/span>typename<\/span> T, std::<\/span>size_t<\/span> N><\/span>   \r\nclass<\/span> Array<\/span>{\r\n\r\n public:<\/span>\r\n    template<\/span> <<\/span>typename<\/span> T2><\/span>\r\n    Array<<\/span>T, N>&<\/span> operator<\/span> =<\/span> (const<\/span> Array<<\/span>T2, N>&<\/span> arr) {\r\n        std::<\/span>copy(std::<\/span>begin(arr.elem), std::<\/span>end(arr.elem), std::<\/span>begin(elem));\r\n        return<\/span> *<\/span>this<\/span>;\r\n    }\r\n    std::size_t<\/span> getSize() const<\/span> {\r\n        return<\/span> N;\r\n    }\r\n    T elem[N];\r\n    \r\n};\r\n<\/pre>\n<\/div>\n
 <\/p>\n
The assignment operator Array<T, N>& operator = (const Array<T2, N>& arr)<\/code> accepts arrays that could vary in the underlying type but could not vary in length.  Before I show the code in action, I want to improve it.<\/p>\n
Friendship<\/h4>\n
To make elem<\/code> private, it must be a friend of the class.<\/p>\n
<\/p>\n
\n
template<\/span> <<\/span>typename<\/span> T, std::<\/span>size_t<\/span> N><\/span>   \r\nclass<\/span> Array<\/span>{\r\n\r\n public:<\/span>\r\n    template<\/span> <<\/span>typename<\/span> T2><\/span>\r\n    Array<<\/span>T, N>&<\/span> operator<\/span> =<\/span> (const<\/span> Array<<\/span>T2, N>&<\/span> arr) {\r\n        std::<\/span>copy(std::<\/span>begin(arr.elem), std::<\/span>end(arr.elem), std::<\/span>begin(elem));\r\n        return<\/span> *<\/span>this<\/span>;\r\n    }\r\n    template<\/span><<\/span>typename<\/span>, std::<\/span>size_t<\/span>><\/span> friend<\/span> class<\/span> Array<\/span>;          \/\/ (1)<\/span>\r\n    std::size_t<\/span> getSize() const<\/span> {\r\n        return<\/span> N;\r\n    }\r\n private:<\/span>\r\n    T elem[N];\r\n    \r\n};\r\n<\/pre>\n<\/div>\n
 <\/p>\n
The line template<typename, std::size_t> friend class Array <\/code>(1) declares all instances of Array to friends.<\/p>\n
Member Function defined Outside The Class<\/h4>\n
Defining the generic member function outside the class is quite a job.<\/p>\n
<\/p>\n
\n
template<\/span> <<\/span>typename<\/span> T, std::<\/span>size_t<\/span> N><\/span>   \r\nclass<\/span> Array<\/span>{\r\n\r\n public:<\/span>\r\n    template<\/span> <<\/span>typename<\/span> T2><\/span>\r\n    Array<<\/span>T, N>&<\/span> operator<\/span> =<\/span> (const<\/span> Array<<\/span>T2, N>&<\/span> arr);\r\n    template<\/span><<\/span>typename<\/span>, std::<\/span>size_t<\/span>><\/span> friend<\/span> class<\/span> Array<\/span>;\r\n    std::size_t<\/span> getSize() const<\/span>;\r\n private:<\/span>\r\n    T elem[N];\r\n    \r\n};\r\n\r\ntemplate<\/span> <<\/span>typename<\/span> T, std::<\/span>size_t<\/span> N><\/span> \r\nstd::size_t<\/span> Array<<\/span>T, N>::<\/span>getSize() const<\/span> { return<\/span> N; }\r\n\r\ntemplate<\/span><<\/span>typename<\/span> T, std::<\/span>size_t<\/span> N>                \/\/ (1)<\/span><\/span>\r\ntemplate<\/span><<\/span>typename<\/span> T2><\/span>\r\nArray<<\/span>T, N>&<\/span> Array<<\/span>T, N>::<\/span>operator<\/span> =<\/span> (const<\/span> Array<<\/span>T2, N>&<\/span> arr) {\r\n    std::<\/span>copy(std::<\/span>begin(arr.elem), std::<\/span>end(arr.elem), std::<\/span>begin(elem));\r\n    return<\/span> *<\/span>this<\/span>;\r\n}\r\n<\/pre>\n<\/div>\n
 <\/p>\n
In this case, you define a generic member function (1) outside the class body; you have to specify that the class and the member functions are templates. Additionally, you have to specify the full type qualification of the generic member function. So far, the assignment operator is used for types T<\/code> and T2<\/code> that is not convertible. Invoking the assignment operator with non-convertible types gives an ugly error message. I should fix this.<\/p>\n
Requirements on the Type Parameters<\/h4>\n
The requirements can be formulated with the type traits library<\/a> and static_assert<\/code> (C++11), or with concepts (C++20). Here are the two variations of the generic assignment operator:<\/p>\n