Generic Programming

3/20/22, 5:28 PM Function Templates - Generic Programming Templates in C++
(/learn)
Function Templates
In this lesson, we'll explore function templates in detail.
We'll cover the following
• Function Templates
• Passing Arguments in Function Templates

• Instantiation
• Overloading
Function Templates#
A function template will be defined by placing the keyword template
followed by type or non-type parameters in front of a concrete
function.
After that, you replace the concrete types or non-types with the type or non-
type parameters in the function.
The keyword class or typename declares the parameters.

The name T is usually used for the first parameter.
The parameters can be used in the body of the function.
Passing Arguments in Function Templates#

In the given code snippet, we’ll look at how we can call the initialized
variables with our template. Look at line 2, the function arguments x and y
in the function xchg must have the same type. By providing two type
parameters like in line 5, the types of arguments can be different. In line 9,
https://www.educative.io/courses/generic-templates-in-cpp/mEK9vPWE2qn 1/4
parameters like in line 5, the types of arguments can be different. In line 9,
you see a non-type template parameter N .
1 template <typename T>
2 void xchg(T& x , T& y){
3 ...
4
5 template <typename T, typename T1>
6 void add(T& x, T1& y){
7 ...
8
9 template <int N>
10 int nTimes(int n){
11 ...
Instantiation#
The process of substituting the template parameters for the template
arguments is called template instantiation.
The compiler:
Automatically creates an instance of the function template.

Will automatically create a function template if the template
parameters can be derived from the function arguments.
If the compiler cannot deduce the template arguments from the function
arguments, you will have to specify them explicitly.
2 void xchg(T& x, T& y){ ...
3
4 int a, b;
5 xchg(a, b);
6
7 template <int N>
8 int nTimes(int n){ ...
9
10 int n = 5;
11 nTimes<10>(n);
Overloading#
Function templates can be overloaded.
The following rules hold:
1. Templates do not support an automatic type conversion.
2. If a free function is better or equally as good as a function template that

already exists, the free function can be used.
3. You can explicitly specify the type of the function template.
func<type>(...)
4. You can specify that you are only interested in a specific instantiation of
a function template.
func<>(...)
To learn more about function templates, click here

(https://en.cppreference.com/w/cpp/language/function_template).
In the next lesson, we’ll look at the examples of function templates.
Back Next
Who Should Take this Course? - Examples
Mark as Completed
Report an Issue
3/20/22, 7:02 PM - Examples - Generic Programming Templates in C++
- Examples
Let's take a look at the examples of function templates in this lesson.
• Example 1: Templates in Functions
• Explanation
• Example 2: Overloading Function Templates
• Explanation
Example 1: Templates in Functions#

1 // templateFunctionsTemplates.cpp
2
3 #include <iostream>
4 #include <string>
5 #include <vector>
6
8 void xchg(T& x, T& y){
9 T t = x;
10 x =y;
11 y =t;
12 }
13
14 template <int N>
15 int nTimes(int n){
16 return N * n;
17 }
18
19 int main(){
20
21 std::cout << std::endl;
22
23 bool t true;
https://www.educative.io/courses/generic-templates-in-cpp/qZ1WEvv4nXD 1/6
23 bool t = true;
24 bool f = false;
25 std::cout << "(t, f): (" << t << ", " << f << ") "<< std::endl;
26 xchg(t, f);
27 std::cout << "(t, f): (" << t << ", " << f << ") "<< std::endl;
28
30
31 int int2011 = 2011;
32 int int2014 = 2014;
33 std::cout << "(int2011, int2014): (" << int2011 << ", " << int2014 << ") "<< std:
34 xchg(int2011, int2014);
35 std::cout << "(int2011, int2014): (" << int2011 << ", " << int2014 << ") "<< std:
36
38
39 std::string first{"first"};
40 std::string second{"second"};
41 std::cout << "(first, second): (" << first << ", " << second << ") "<< std::endl;
42 xchg(first, second);
43 std::cout << "(first, second): (" << first << ", " << second << ") "<< std::endl;
44
46 std::vector<int> intVec1{1, 2, 3, 4, 5};
47 std::vector<int> intVec2{5, 4, 3, 2, 1};
48
49 std::cout << "vec1: ";
50 for (auto v: intVec1)std::cout << v << " ";
51 std::cout << "\nvec2: ";
54 xchg(intVec1, intVec2);
55
56 std::cout << "vec1: ";
58 std::cout << "\nvec2: ";
61
62
63 std::cout << "\n\n";
64
65 std::cout << "nTimes<5>(10): " << nTimes<5>(10) << std::endl;
66 std::cout << "nTimes<10>(5): " << nTimes<10>(5) << std::endl;
67
69 }
Output 1.4s
(t, f): (1, 0)
(t, f): (0, 1)
(int2011, int2014): (2011, 2014)
(int2011, int2014): (2014, 2011)
(first, second): (first, second)
(first, second): (second, first)
Explanation#
In the example above, we’ve declared two function templates: xchg and
nTimes in lines 8 and 15. xchg swaps the values passed as arguments. The
only non-type, we use is N in the function templates nTimes . nTimes returns
the N times of the number passed n . We have initialized multiple instances
to check for functions in lines 31 and 32, lines 39 and 40, and lines 46 and 47.
Example 2: Overloading Function

Templates#
1 // templateFunctionsTemplatesOverloading.cpp
2
4
5 void xchg(int& x, int& y){ // 1
6 int t = x;
7 x = y;
8 y = t;
9 }
10
11
12 template <typename T> // 2
p yp //
13 void xchg(T& x, T& y){
14 T t = x;
15 x = y;
16 y = t;
17 }
18
19 template <typename T> // 3
20 void xchg(T& x, T& y, T& z){
21 xchg(x, y);
22 xchg(x, z);
23 }
24
25 int main(){
26
28
29 int intA = 5;
30 int intB = 10;
31 int intC = 20;
32
33 double doubleA = 5.5;
34 double doubleB = 10.0;
35
36 std::cout << "Before: " << intA << ", " << intB << std::endl;
37 xchg(intA, intB); // 1
38 std::cout << "After: " << intA << ", " << intB << std::endl;
39
41
42 std::cout << "Before: " << doubleA << ", " << doubleB << std::endl;
43 xchg(doubleA, doubleB); // 2
44 std::cout << "After: " << doubleA << ", " << doubleB << std::endl;
45
47
48 xchg<>(intA, intB); // explicit 2
49 xchg<int>(intA, intB); // explicit 2: xchg<int>
50 // xchg<double>(intA, intB); // ERROR explicit xchg<double>
51 (/learn)
52 std::cout << "Before: " << intA << ", " << intB << ", " << intC << std::endl;
53 xchg(intA, intB, intC); // 3
54 std::cout << "After: " << intA << ", " << intB << ", " << intC << std::endl;
55
57 }
Output 3.44s
Before: 5, 10
After: 10, 5
Before: 5.5, 10
After: 10, 5.5
Before: 10, 5, 20
After: 20, 10, 5
Explanation#
In the above example, we used the concept of function overloading by
calling xchg with different arguments passed to the function. We used the
xchg function with different data types by passing two arguments and three
arguments. In line 37, the non-template function is called, whereas, on all
other calls to xchg() , the template function is used. The call xchg<double>
(intA, intB) would be fine, when xchg would take its arguments by value.
In the next lesson, we’ll solve an exercise related to function templates.
Interviewing soon? We've partnered with Hired so that companies apply to

you instead of you applying to them. See how
(https://join.hired.com/educative/?
utm_source=educative&utm_medium=partner&utm_campaign=March_2022
Back Next
Function Templates - Exercise
Completed
Report an Issue
3/20/22, 5:30 PM - Exercise - Generic Programming Templates in C++
- Exercise (/learn)
In this lesson, we'll solve a challenge that attempts to compute the power of the expression.
• Problem Statement
Problem Statement#
You have to implement a function which calculates 210 in the program. Try
using templates to implement the function.
2
3 // Write your code here
4
5 int main() {
6 // Call function here
7 }
In the next lesson, we’ll look at the different ways to solve this exercise.
Back Next
- Examples - Solution
https://www.educative.io/courses/generic-templates-in-cpp/BnMO7PxOX6Q 1/2
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/BnMO7PxOX6Q 2/2
3/20/22, 5:31 PM - Solution - Generic Programming Templates in C++
- Solution
We'll learn different approaches to solve the previous challenge.
• Solution 1: Using Function Arguments
• Explanation
• Solution 2: Using Template Arguments
• Explanation
• Solution 3: Using Template Arguments and Function Arguments
• Explanation
Solution 1: Using Function

Arguments#
1 // power1.cpp
2
4
5 int power(int m, int n){
6 int r = 1;
7 for(int k=1; k<=n; ++k) r*= m;
8 return r;
9 }
10
11 int main(){
12 std::cout << power(2,10) << std::endl;
13 }
https://www.educative.io/courses/generic-templates-in-cpp/BnBZYxpxmQo 1/5
Explanation#
We’re using a for loop to compute the power. The loop runs a total of n
times by multiplying the number m with r for every iteration of the loop in
line 7.
To get a more in-depth insight into the above solution, click here
(https://godbolt.org/z/p0kKAZ). It shows how things are handled at the
assembler level.
The critical point of this example is that the function runs at runtime.
Solution 2: Using Template

Arguments#
1 // power2.cpp
2
4
5 template<int m, int n>
6 struct Power{
7 static int const value = m * Power<m,n-1>::value;
8 };
9
10 template<int m>
11 struct Power<m,0>{
12 static int const value = 1;
13 };
14
15 int main(){
16 std::cout << Power<2,10>::value << std::endl;
17 }
Explanation#
The call Power<2, 10>::value in line 16 triggers the recursive calculation.
First, the primary template in line 5 is called, then the Power<m, n-1>::value
in line 7 is executed. This expression instantiates recursively until the end
condition is met; n is equal to 0. Now, the boundary condition in line 12 is
applied, which returns 1. In the end, Power<2, 10>::value contains the
result.
To view how things are happening at the assembler level, click here
(https://godbolt.org/z/pjXrQm).
The critical point is that the calculation is done at compile-time.

Arguments and Function Arguments#
1 // power3 .cpp
2
4
5 template<int n>
6 int power(int m){
7 return m * power<n-1>(m);
8 }
9
10 template<>
11 int power<1>(int m){
12 return m;
13 }
14 (/learn)
15 template<>
16 int power<0>(int m){
17 return 1;
18 }
19
19
20 int main(){
21 std::cout << power<10>(2) << std::endl;
22 }
Explanation#
In the above code, the power function template exists in three variations.
First in the primary template in line 6. Second and third in the full
specializations for 1 and 0 in lines 11, and 16. The call power<10>(2) triggers
the recursive invocation of the primary template. The recursion ends with
the full specialization for 1. When you study the example carefully, you’ll see
that the full specialization for is not necessary in this case because the full
specialization for 0 is also a valid boundary condition.
When we invoke the power<10>(2) function, the argument in () brackets is

evaluated at runtime and the argument in <> brackets is evaluated at
compile-time. Therefore, we can say that the round brackets are run time
arguments and the angle brackets are compile-time arguments.
Let’s have a look at the assembler code and how they are managing this.
Click here (https://godbolt.org/z/wLujdp) to view the code.
In the lesson, we’ll learn about class templates in detail.
Back Next
- Exercise Class Templates

Mark as Completed
Report an Issue
- Solution
We'll learn different approaches to solve the previous challenge.
We'll(/learn)
cover the following
• Solution 1: Using Function Arguments
• Explanation
• Solution 2: Using Template Arguments
• Explanation
• Solution 3: Using Template Arguments and Function Arguments
• Explanation
Solution 1: Using Function

Arguments#
1 // power1.cpp
2
4
6 int r = 1;
7 for(int k=1; k<=n; ++k) r*= m;
8 return r;
9 }
10
11 int main(){
12 std::cout << power(2,10) << std::endl;
13 }
Explanation#
We’re using a for loop to compute the power. The loop runs a total of n
times by multiplying the number m with r for every iteration of the loop in
line 7.
To get a more in-depth insight into the above solution, click here
(https://godbolt.org/z/p0kKAZ). It shows how things are handled at the
assembler level.
The critical point of this example is that the function runs at runtime.

Arguments#
// power2.cpp
#include <iostream>
template<int m, int n>
struct Power{
static int const value = m * Power<m,n-1>::value;
};
template<int m>
struct Power<m,0>{
static int const value = 1;
};
int main(){
std::cout << Power<2,10>::value << std::endl;
Explanation#
The call Power<2, 10>::value in line 16 triggers the recursive calculation.
First, the primary template in line 5 is called, then the Power<m, n-1>::value
in line 7 is executed. This expression instantiates recursively until the end
condition is met; n is equal to 0. Now, the boundary condition in line 12 is
applied, which returns 1. In the end, Power<2, 10>::value contains the
result.
To view how things are happening at the assembler level, click here
(https://godbolt.org/z/pjXrQm).
The critical point is that the calculation is done at compile-time.

Arguments and Function Arguments#
// power3 .cpp
#include <iostream>
template<int n>
int power(int m){

return m * power<n-1>(m);
template<>
int power<1>(int m){
return m;
template<>
int power<0>(int m){
return 1;
int main(){
int main(){
std::cout << power<10>(2) << std::endl;
Explanation#
In the above code, the power function template exists in three variations.
First in the primary template in line 6. Second and third in the full
specializations for 1 and 0 in lines 11, and 16. The call power<10>(2) triggers
the recursive invocation of the primary template. The recursion ends with
the full specialization for 1. When you study the example carefully, you’ll see
that the full specialization for is not necessary in this case because the full
specialization for 0 is also a valid boundary condition.
When we invoke the power<10>(2) function, the argument in () brackets is

evaluated at runtime and the argument in <> brackets is evaluated at
compile-time. Therefore, we can say that the round brackets are run time
arguments and the angle brackets are compile-time arguments.
Let’s have a look at the assembler code and how they are managing this.
Click here (https://godbolt.org/z/wLujdp) to view the code.
In the lesson, we’ll learn about class templates in detail.
Back Next
- Exercise Class Templates
Mark as Completed
Report an Issue
3/20/22, 7:08 PM Class Templates - Generic Programming Templates in C++
Class Templates
In this lesson, we'll learn about the class templates.
• Syntax
• Explanation
• Instantiation
• Method Templates
• Inheritance
• 3 Solutions:
• Templates: Alias Templates
A class template defines a family of classes.
Syntax#
template < parameter-list >
class-declaration
Explanation#
parameter-list: A non-empty comma-separated list of the template
parameters, each of which is either a non-type parameter, a type parameter,
a template parameter, or a mixture of any of those.
class-declaration: A class declaration. The class name declared becomes a

https://www.educative.io/courses/generic-templates-in-cpp/JEM9l4NrKZJ 1/5
class declaration: A class declaration. The class name declared becomes a
template name.
Instantiation#
The process of substituting the template parameters with the template
arguments is called instantiation .
In contrast to a function template, a class template is not capable of

automatically deriving the template parameters.
Each template
argument must be explicitly specified. This restriction no longer exists
with C++17.
Let’s have a look at the declaration of function and class templates:
Function Template Class Template

Declaration Declaration
template <typename T>
template <typename T, int N>
void xchg(T& x, T&y){
class Array{
...
...
}
};
int a, b;
Array<double, 10> doubleArray;
xchg(a, b);
Array<Account, 1000> accountArray
;
Method Templates#
Method templates are function templates used in a class or class template.
Method templates can be defined inside or outside the class. When you
define the method template outside the class, the syntax is quite
complicated because you have to repeat the class template declaration
and the method template declaration.
Let’s have a look at the declaration of the method template inside the class
and its definition outside the class:
1 template <class T, int N> class Array{
2 public:
3 template <class T2>
4 Array<T, N>& operator = (const Array<T2, N>& a); ...
5 };
6
7 template<class T, int N>
8 template<class T2>
9 Array<T, N>& Array<T, N>::operator = (const Array<T2, N>& a{
10 ...
11 }
The destructor and copy constructor cannot be templates.
Inheritance#
Class and class template can inherit from each other in arbitrary
combinations.
If a class or a class template inherits from a class template, the methods

of the base class or base class template are not automatically available
in the derived class.
2 struct Base{
3 void func(){ ...
4 };
5
6 template <typename T> struct Derived: Base<T>{
7 void func2(){
8 func(); // ERROR
9 }
10 }
There are three ways to make a method from the derived class template
available.
(/learn)
3 Solutions:#
Qualification via this pointer: this->func()
Introducing the name using Base<T>::func
Full qualified access Base<T>::func()
Templates: Alias Templates#

Alias templates, aka template typedefs, allow you to give a name to partially
bound templates. An example of partial specialization from templates is
given below:
1 template <typename T, int Line, int Col> class Matrix{
2 ...
3 };
4
5 template <typename T, int Line>
6 using Square = Matrix<T, Line, Line>;
7
8 template <typename T, int Line>
9 using Vector = Matrix<T, Line, 1>;
10
11 Matrix<int, 5, 3> ma;
12 Square<double, 4> sq;
13 Vector<char, 5> vec;
, ;
To learn more about class templates, click here

(https://en.cppreference.com/w/cpp/language/class_template).
In the next lesson, we’ll learn about the examples of class templates.

Back Next
- Solution - Examples
Completed
Report an Issue
- Examples
In this lesson, we'll learn about the examples of class templates.
• Example 1: Templates in Class
• Explanation
• Example 2: Inheritance in Class Templates
• Explanation
• Example 3: Methods in Class Templates
• Explanation
Example 1: Templates in Class#

1 // templateClassTemplate.cpp
2
4
5 class Account{
6 public:
7 explicit Account(double amount=0.0): balance(amount){}
8
9 void deposit(double amount){
10 balance+= amount;
11 }
12
13 void withdraw(double amount){
14 balance-= amount;
15 }
16
17 double getBalance() const{
18 return balance;
19 }
https://www.educative.io/courses/generic-templates-in-cpp/7DxJ7y5k0Zj 1/7
19 }
20
21 private:
22 double balance;
23 };
24
25 template <typename T, int N>
26 class Array{
27
28 public:
29 Array()= default;
30 int getSize() const;
31
32 private:
33 T elem[N];
34 };
35
37 int Array<T,N>::getSize() const {
38 return N;
39 }
40
41 int main(){
42
44
45 Array<double,10> doubleArray;
46 std::cout << "doubleArray.getSize(): " << doubleArray.getSize() << std::endl;
47
48 Array<Account,1000> accountArray;
49 std::cout << "accountArray.getSize(): " << accountArray.getSize() << std::endl;
50
52 }
Output 1.28s
doubleArray.getSize(): 10
accountArray.getSize(): 1000
Explanation#
We have created two Array class objects, i.e., doubleArray and accountArray
in lines 45 and 48. By calling generic function getSize() in line 37, we can
access the size of different objects.
Example 2: Inheritance in Class

Templates#
1 // templateClassTemplateInheritance.cpp
2
4
6 class Base{
7 public:
8 void func1() const {
9 std::cout << "func1()" << std::endl;
10 }
13 }
16 }
17 };
18
20 class Derived: public Base<T>{
21 public:
22 using Base<T>::func2;
23
24 void callAllBaseFunctions(){
25 this->func1();
26 func2();
27 Base<T>::func3();
28 }
29 };
30
31 int main(){
33
34 Derived<int> derived;
35 derived.callAllBaseFunctions();
36
38 }
Output 2.23s
func1()
func2()
func3()
Explanation#
We have implemented both a Base and a Derived class. Derived is publicly
inherited from Base and may, therefore, use in its method
callAllBaseFunctions in line 24, the methods func1 , func2 , and func3 from
the Base class.
Make the name dependent: The call this->func1 in line 25 is

dependent. The name lookup will consider in this case all base classes.
Introduce the name into the current scope: The expression using
Base<T>::func2 (line 22) introduces func2 into the current scope.
Call the name fully qualified: Calling func3 fully qualified (line 27)
will break a virtual dispatch and may cause new surprises.
We have created a Derived class object named derived . By using this object,
we can access the base class functions by calling the method
callAllBaseFunctions .
Example 3: Methods in Class

Templates#
1 // templateClassTemplateMethods.cpp
2
3 #include <algorithm>
5 #include <vector>
6
8 class Array{
9
10 public:
12
13 template <typename T2>
14 Array<T, N>& operator=(const Array<T2, N>& arr){
15 elem.clear();
16 elem.insert(elem.begin(), arr.elem.begin(), arr.elem.end());
17 return *this;
18 }
19
21
22 std::vector<T> elem;
23 };
24
26 (/learn)
int Array<T, N>::getSize() const {
27 return N;
28 }
29
30 int main(){
31
32 Array<double, 10> doubleArray{};
33 Array<int, 10> intArray{};
34
35 doubleArray= intArray;
36
37 Array<std::string, 10> strArray{};
38 Array<int, 100> bigIntArray{};
39
40 // doubleArray= strArray; // ERROR: cannot convert ‘const std::basic_st
41 // doubleArray= bigIntArray; // ERROR: no match for ‘operator=’ in ‘double
42 }
Succeeded
Explanation#
In the example above, we have initialized two instances of Array class
namely doubleArray and intArray in lines 32 and 33. We’re using the
generic = operator to copy the intArray elements to doubleArray in line 35.
When you look carefully, you see that the generic = is only applicable, when
both arrays have the same length.
In the next lesson, we have a small challenge for you to test your knowledge
of class templates.

Back Next
Class Templates - Exercise
Completed
Report an Issue
- Exercise
Let's do a quick exercise on class templates.
Problem Statement#
Uncomment the final assignment doubleArray = strArray in line 41 and use
the function static_assert in combination with the function
std::is_convertible to catch the erroneous instantiation at compile-time.
3 #include <vector>
4
6 class Array{
7
8 public:
9 Array()= default;
10
13 // write your code here
14 // uncomment line 41 to check if your code runs fine
15
16 elem.clear();
18 return *this;
19 }
20
22
https://www.educative.io/courses/generic-templates-in-cpp/xlrKg7OpKwl 1/3
24 };
25
27 int Array<T, N>::getSize() const {
28 return N;
29 }
30 (/learn)
31 int main(){
32
35
37
40
41 //doubleArray= strArray; // ERROR: cannot convert ‘const std::basic_str
43
44 }
Succeeded
In the next lesson, we’ll look at the solution to this problem.

Back Next
Completed
Report an Issue
- Solution
Let's review the solution of the previous problem in this lesson.
• Solution Review
• Explanation
Solution Review#
1 // templateClassTemplateMethods.cpp
2 #include <type_traits>
5 #include <vector>
6
8 class Array{
9
10 public:
12
14 Array<T,N>& operator=(const Array<T2, N>& arr){
15 static_assert(std::is_convertible<T2, T>::value, "Cannot convert source type to
16 elem.clear();
18 return *this;
19 }
20
22
24 };
25
https://www.educative.io/courses/generic-templates-in-cpp/g7ZOpwpW4j3 1/3
28 return N;
29 }
30
31
32 int main(){
33
36
38
41
42 // doubleArray= strArray; // ERROR: cannot convert ‘const std::basic_st
44
45 }
(/learn)
Succeeded
Explanation#
In the above code, we have created two arrays of int and double types in
lines 34 and 35. We’re copying the data of integer array to double array in
line 37. If we try to copy the string array data to double type in line 42, this
gives us an error because of the type mismatch. To see for yourself
uncomment the line and run to check the error. The function
std::is_convertible (https://en.cppreference.com/w/cpp/types/is_convertible)
in line 15 from the type-traits
(https://www.educative.io/collection/page/10370001/5799373671759872/6380
229176066048) library checks if one type can be converted to the other. Of
course, it is not possible to convert string into double .
Let’s move on to template parameters in the next lesson.

Back Next
- Exercise Template Parameters
Completed
Report an Issue
3/20/22, 5:36 PM Template Parameters - Generic Programming Templates in C++
(/learn)
Template Parameters
Let's familiarize ourselves with template parameters in this lesson.
• Template Parameter
• Types
• Non-Types
Template Parameter#
Every template is parameterized by one or more template parameters,
indicated in the parameter-list of the template.
C++ supports three different kinds of template parameters
1. Type parameter
std::vector<int> vec = {1, 2, 3, 4, 5};
2. Non-type parameter
std::array<int, 5> arr = {1, 2, 3, 4, 5};
3. Template-template parameter
template <typename T, template <typename, typename> class Cont> class Ma
trix{
...
https://www.educative.io/courses/generic-templates-in-cpp/JYv86M3D2EP 1/3
Matrix<int, std::vector> myIntVec;
Types#
A type parameter is a typical case for template arguments.
Type parameters are class types and fundamental types
Non-Types#
Non-types are template parameters which can be evaluated at compile-time.
The following types are possible
Integers and enumerations
Pointers to objects, functions, and attributes of a class
References to objects and functions
std::nullptr_t constant
With C++17, floating-point numbers and strings cannot be used as non-

type parameters.
To learn more about template parameters, click here

(https://en.cppreference.com/w/cpp/language/template_parameters).
In the next lesson, we’ll look at the examples of the three different types of
template parameters.
Back Next
Solution Examples
Mark as Completed
Report an Issue
- Examples
In this lesson, we'll learn about the examples of template parameters.
• Example 1: Type Parameter
• Explanation
• Example 2: Non-Type Template Parameter
• Explanation
• Example 3: Template-Template Parameter
• Explanation
Example 1: Type Parameter#

1 // templateTypeParameter.cpp
2
4 #include <typeinfo>
5
6 class Account{
7 public:
8 explicit Account(double amt): balance(amt){}
9 private:
10 double balance;
11
12 };
13
14 union WithString{
15 std::string s;
16 int i;
17 WithString():s("hello"){}
18 ~WithString(){}
19 };
https://www.educative.io/courses/generic-templates-in-cpp/gk9Eq0rPLxZ 1/7
19 };
20
22 class ClassTemplate{
23 public:
24 ClassTemplate(){
25 std::cout << "typeid(T).name(): " << typeid(T).name() << std::endl;
26 }
27 };
28
29 int main(){
30
32
33 ClassTemplate<int> clTempInt;
34 ClassTemplate<double> clTempDouble;
35 ClassTemplate<std::string> clTempString;
36
37 ClassTemplate<Account> clTempAccount;
38 ClassTemplate<WithString> clTempWithString;
39
41
42 }
Output 1.82s
typeid(T).name(): i
typeid(T).name(): d
typeid(T).name(): NSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE
typeid(T).name(): 7Account
typeid(T).name(): 10WithString
Explanation#
In the above code, we are identifying the type of different data types that we
have passed in the parameter list We can identify the type of variable
have passed in the parameter list. We can identify the type of variable
passed to the function by using the keyword typeid in line 25. If we pass
string or class type object in the parameter list, it will display the type of
parameter passed along with the size of the object.
Example 2: Non-Type Template

Parameter#
1 // array.cpp
2
4 #include <array>
6
7 int main(){
8
10
11 // output the array
12 std::array <int,8> array1{1,2,3,4,5,6,7,8};
13 std::for_each( array1.begin(),array1.end(),[](int v){std::cout << v << " ";});
14
16
17 // calculate the sum of the array by using a global variable
18 int sum = 0;
19 std::for_each(array1.begin(), array1.end(),[&sum](int v) { sum += v; });
20 std::cout << "sum of array{1,2,3,4,5,6,7,8}: " << sum << std::endl;
21
22 // change each array element to the second power
23 std::for_each(array1.begin(), array1.end(),[](int& v) { v=v*v; });
24 std::for_each( array1.begin(),array1.end(),[](int v){std::cout << v << " ";});
26
28 }
Output 1.93s
1 2 3 4 5 6 7 8
sum of array{1,2,3,4,5,6,7,8}: 36
1 4 9 16 25 36 49 64
Explanation#
When you define an std::array in line 12, you have to specify its size. The
size is a non-type template argument, which has to be specified at compile-
time.
Therefore, you can output array1 in line 13 with a lambda-function [] and

the range-based for-loop. By using the summation variable sum in line 19,
you can sum up the elements of the std::array . The lambda-function in line
23 takes its arguments by reference and can, therefore, map each element to
its square. There is nothing really special, but we are dealing with an
std::array .
With C++11 we have the free function templates std::begin and std::end
returning iterators for a C array. A C array is quite comfortable and safe to
use with these function templates because we don’t have to remember its
size.
Example 3: Template-Template
Parameter#
1 // t l t T l t T l t P
https://www.educative.io/courses/generic-templates-in-cpp/gk9Eq0rPLxZ
t 4/7
1 // templateTemplateTemplatesParameter.cpp
2
3 #include <initializer_list>
5 #include <list>
6 #include <vector>
7
8 template <typename T, template <typename, typename> class Cont >
9 class Matrix{
10 public:
11 explicit Matrix(std::initializer_list<T> inList): data(inList){
12 for (auto d: data) std::cout << d << " ";
13 }
14 int getSize() const{
15 return data.size();
16 }
17
18 private:
19 Cont<T, std::allocator<T>> data;
20
21 };
22
23 int main(){
24
26
27 Matrix<int,std::vector> myIntVec{1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
29 std::cout << "myIntVec.getSize(): " << myIntVec.getSize() << std::endl;
30
32
33 Matrix<double,std::vector> myDoubleVec{1.1, 2.2, 3.3, 4.4, 5.5};
35 std::cout << "myDoubleVec.getSize(): " << myDoubleVec.getSize() << std::endl;
36
38
39 Matrix<std::string,std::list> myStringList{"one", "two", "three", "four"};
41 std::cout << "myStringList.getSize(): " << myStringList.getSize() << std::endl;
42
44 }
(/learn)
(/learn)
Output 1.56s
1 2 3 4 5 6 7 8 9 10
myIntVec.getSize(): 10
1.1 2.2 3.3 4.4 5.5
myDoubleVec.getSize(): 5
one two three four
myStringList.getSize(): 4
Explanation#
We have declared a Matrix class which contains a function, i.e., getSize ,
and an explicit constructor that prints all entries of the passed parameter.
Cont in line 8 is a template, which takes two arguments. There’s no need for
us to name the template parameters in the template declaration. We have to
specify them in the instantiation of the template (line 19). The template used
in the template parameter has exactly the signature of the sequence
containers (https://en.cppreference.com/w/cpp/container). The result is, that
we can instantiate a matrix with an std::vector , or an std::list . Of course
std::deque and std::forward_list would also be possible. In the end, you
have a Matrix , which stores its elements in a vector or in a list.
If you want to study more examples for template-template parameter, you

can check container adaptors (https://en.cppreference.com/w/cpp/container).
We’ll be solving a small exercise on template parameters in the next lesson.

Interviewing soon? We ve partnered with Hired so that companies apply to
Back Next
Template Parameters - Exercise
Completed
Report an Issue
(/learn)
- Exercise
Let's solve an exercise in this lesson.
Problem Statement#
Implement four different class templates which accept an integral, a pointer,
a reference, or a nullptr as a non-type parameter. You need to display the
type of data passed as a parameter.
2
3 // Write class templates for given commented code
4
5
6 int main() {
7 // uncomment these lines after implementing codes
8
9 /*
10 AcceptChar<'c'> acceptChar;
11 AcceptFunction< myFunc> acceptFunction;
12 AcceptReference< arr > acceptReference;
13 AcceptNullptr< nullptr > acceptNullptr;
14 */
15 }
https://www.educative.io/courses/generic-templates-in-cpp/B1QQ5241XWo 1/2
In the next lesson, we’ll look at the solution to this exercise.
Back Next
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/B1QQ5241XWo 2/2
- Solution
In this lesson, we'll look at the solution review of the previous challenge.
• Solution Review
• Explanation
Solution Review#
1 #include <cstddef>
4
5 template <char c>
6 class AcceptChar{
7 public:
8 AcceptChar(){
9 std::cout << "AcceptChar: " << typeid(c).name() << std::endl;
10 }
11 };
12
13 template < int(*func)(int) >
14 class AcceptFunction{
15 public:
16 AcceptFunction(){
17 std::cout << "AcceptFunction: " << typeid(func).name() << std::endl;
18 }
19 };
20
21 template < int(&arr)[5] >
22 class AcceptReference{
23 public:
24 AcceptReference(){
25 std::cout << "AcceptReference: " << typeid(arr).name() << std::endl;
26 }
https://www.educative.io/courses/generic-templates-in-cpp/xlKWvlmxVVJ 1/3
27 };
28
29 template < std::nullptr_t N >
30 class AcceptNullptr{
31 public:
32 AcceptNullptr(){
33 std::cout << "AcceptNullpt: " << typeid(N).name() << std::endl;
34 }
35 };
36
37 int myFunc(int){ return 2011; };
38 int arr[5];
39
40 int main(){
41
43
44 AcceptChar<'c'> acceptChar;
45 AcceptFunction< myFunc> acceptFunction;
46 AcceptReference< arr > acceptReference;
47 AcceptNullptr< nullptr > acceptNullptr;
48
50
51 }
52
(/learn)
Output 1.29s
AcceptChar: c
AcceptFunction: PFiiE
AcceptReference: A5_i
AcceptNullpt: Dn
Explanation#
We have created four different class templates which include AcceptChar ,

AcceptFunction , AcceptReference , and AcceptNull in lines 6, 14, 22, and 30.
Each class template accepts a different non-type. To verify all, we have
declared a character variable, a reference to an array, a function, and
nullptr in lines 44 – 47. This matches the order of declaration in the code.
We identify their type using the operator typeid in lines 9, 17, 25, and 33.
Let’s move on to template arguments in the next lesson.

Back Next
- Exercise Template Arguments
Completed
Report an Issue
3/20/22, 5:38 PM Template Arguments - Generic Programming Templates in C++
(/learn)
Template Arguments
In this lesson, we'll learn about template arguments.
• Template Arguments
• Template Arguments (C++17)

• Argument Deduction
• Explicit Template Arguments

• Default Template Arguments
Template Arguments#
Template arguments can, in general, automatically be deduced for function
templates.
The compiler deduces the template arguments for the function
arguments. From the user’s perspective, function templates feel like
functions.
Conversion:
The compiler uses simple conversions for deducing the template

arguments from the function arguments.
The compiler removes const or volatile from the function arguments
and converts C-arrays and functions to pointers.
Template argument deduction for function templates:

https://www.educative.io/courses/generic-templates-in-cpp/m2rWRJGYmVG 1/5
void func(ParamType param);
Two datatypes were deduced:
ParamType
ParamType can be
Reference or pointer
Universal reference( && )

Value (copy)
1. The parameter type is a reference or a pointer
void func(T& param);
// void func(T* param);
func(expr);
T ignores reference or pointer
Pattern matching on expr for T& or T
2. The parameter type is a universal reference ( && )
void func(T&& param);
func(expr);
expr is an lvalue: T and ParamType become lvalue references

expr is an rvalue: T is deduced such as the ParamType is a reference
(case 1)
3. Parameter type is a value (copy)
void func(T param);

void func(T param);
func(expr);
1. expr is a reference: the reference (pointer) of the argument is ignored
2. expr is const or volatile : const or volatile is ignored
Template Arguments (C++17)#

The constructor can deduce its template arguments from its function
arguments.
Template Argument deduction for a constructor is available since C++17, but

for function templates since C++98.
std::pair<int, double> myPair(2011, 1.23);
std::pair myPair(2011, 1.23);
Many of the make_ functions such as std::make_pair are not necessary any
more:
auto myPair = std::make_pair(2011, 1.23);
Argument Deduction#
The types of function arguments have to be exact, otherwise, no conversion
takes place.
bool isSmaller(T fir, T sec){
return fir < sec;
isSmaller(1, 5LL); // ERROR int != long long int
Providing a second template parameter makes this

example work.
g p p p
template <typename T, typename U>
bool isSmaller(T fir, U sec){
return fir < sec;
isSmaller(1, 5LL); // OK
Explicit Template Arguments#

Unlike in line 5 in the previous example, sometimes the template argument
types need to be explicitly specified. This is necessary in the following cases:
Explicit Template Arguments
if the template argument cannot be deduced from the function

argument.
if a specific instance of a function template is needed.
template <typename R, typename T, typename U>
R add(T fir, U sec){
return fir * sec;
add<long long int>(1000000, 1000000LL);
Missing template arguments are automatically derived from the

function arguments.
Default Template Arguments#

The default for template parameters can be specified for class templates and
function templates. If a template parameter has a default parameter, all
subsequent template parameters also need a default argument.
template <typename T, typename Pred = std::less<T>>
bool isSmaller(T fir, T sec, Pred pred = Pred()){
return pred(fir, sec);
To learn more about template arguments, click here

(https://en.cppreference.com/w/cpp/language/template_argument_deduction
).
In the next lesson, we’ll take a look at the examples of template arguments.
Back Next
Mark as Completed
Report an Issue
(/learn)
- Examples
Let's check out the examples of template arguments.
• Example 1: Deduction of Template Arguments
• Explanation
• Example 2: Template Default Arguments
• Explanation
• Example 3: Function Template Argument Deduction by Reference
• Explanation
• Example 4: Function Template Argument Deduction by Universal

Reference
• Explanation
• Example 5: Function Template Argument Deduction by Value
• Explanation
Example 1: Deduction of Template

Arguments#
1 // templateArgumentDeduction.cpp
2
4
6 bool isSmaller(T fir, T sec){
7 return fir < sec;
https://www.educative.io/courses/generic-templates-in-cpp/m20Mg8J806A 1/10
7 return fir < sec;
8 }
9
10 template <typename T, typename U>
11 bool isSmaller2(T fir, U sec){
12 return fir < sec;
13 }
14
15 template <typename R, typename T, typename U>
16 R add(T fir, U sec){
17 return fir + sec;
18 }
19
20 int main(){
21
22 std::cout << std::boolalpha << std::endl;
23
24 std::cout << "isSmaller(1,2): " << isSmaller(1,2) << std::endl;
25 // std::cout << "isSmaller(1,5LL): " << isSmaller(1,5LL) << std::endl; // ERROR
26
27 std::cout << "isSmaller<int>(1,5LL): " << isSmaller<int>(1,5LL) << std::endl;
28 std::cout << "isSmaller<double>(1,5LL): " << isSmaller<double>(1,5LL) << std::endl
29
31
32 std::cout << "isSmaller2(1,5LL): " << isSmaller2(1,5LL) << std::endl;
33
35
36 std::cout << "add<long long int>(1000000,1000000): " << add<long long int>(1000000
37 std::cout << "add<double,double>(1000000,1000000): " << add<double,double>(1000000
38 std::cout << "add<double,double,float>(1000000,1000000): " << add<double,double,fl
39
41 }
Output 1.65s
isSmaller(1,2): true
isSmaller<int>(1,5LL): true
isSmaller<double>(1,5LL): true
isSmaller2(1 5LL): true

isSmaller2(1,5LL): true
add<long long int>(1000000,1000000): 2000000
add<double,double>(1000000,1000000): 2e+06
Explanation#
In the above example, we have defined 3 function templates
isSmaller takes two arguments which have the same type and returns
true if the first element is less than the second element (line 6).
Invoking the function with arguments of different types would give a
compile-time error (line 25).
isSmaller2 takes two arguments which can have a different type. The
function returns true if the first element is less than the second
element (line 11).
add takes two arguments which can have different types (line 16). The
return type must be specified because it cannot be deduced from the
function arguments.
Example 2: Template Default

Arguments#
1 // templateDefaultArgument.cpp
2
3 #include <functional>
5 #include <string>
6
7 class Account{
8 public:
9 explicit Account(double b): balance(b){}
10 double getBalance() const {
11 return balance;
12 }
13 private:
14 double balance;
15 };
16
17 template <typename T, typename Pred= std::less<T> >
18 bool isSmaller(T fir, T sec, Pred pred= Pred() ){
19 return pred(fir,sec);
20 }
21
22 int main(){
23
25
27 std::cout << "isSmaller(2.14,3.14): " << isSmaller(2.14,3.14) << std::endl;
28 std::cout << "isSmaller(std::string(abc),std::string(def)): " << isSmaller(std::st
29
30 bool resAcc= isSmaller(Account(100.0),Account(200.0),[](const Account& fir, const
31 std::cout << "isSmaller(Account(100.0),Account(200.0)): " << resAcc << std::endl;
32
33 bool acc= isSmaller(std::string("3.14"),std::string("2.14"),[](const std::string&
34 std::cout << "isSmaller(std::string(3.14),std::string(2.14)): " << acc << std::end
35
37 }
Output 2.65s
isSmaller(2.14,3.14): true
isSmaller(std::string(abc),std::string(def)): true
isSmaller(Account(100.0),Account(200.0)): true
isSmaller(std::string(3.14),std::string(2.14)): false
Explanation#
In the first example, we have passed only the built-in data types. In this
example, we have used the built-in types int , double , std::string , and an
Account class in lines 26 – 28. The function template isSmaller is
parametrized by a second template parameter, which defines the
comparison criterion. The default for the comparison is the predefined
function object std::less . A function object is a class for which the call
operator ( operator () ) is overloaded. This means that instances of function
objects behave similarly as a function. The Account class doesn’t support the
< operator. Thanks to the second template parameter, a lambda expression
like in lines 30 and 33 can be used. This means Account can be compared by
their balance and strings by their number. stod converts a string to a
double.
Since C++17, the constructor of a class template can deduce its arguments.
Study the first example of Class template argument deduction
(https://en.cppreference.com/w/cpp/language/class_template_argument_ded
uction) for a deeper understanding.
Example 3: Function Template

Argument Deduction by Reference#
1 // functionTemplateArgumentDeductionReference.cpp
2
4 void func(T& param){}
5
7 void constFunc(const T& param){}
8
9 int main(){
10
11 int x = 2011;
12 const int cx = x;
13 const int& rx = x;
14
15 func(x);
16 func(cx);
17 func(rx);
18
19 constFunc(x);
20 constFunc(cx);
21 constFunc(rx);
22 }
Succeeded
Explanation#
In the above example, we have created two functions func and constFunc in
lines 4 and 7. Both of these functions accept parameters by reference (19 –
21).
For better understanding, click here (https://cppinsights.io/lnk?

code=dGVtcGxhdGUgPHR5cGVuYW1lIFQ+CnZvaWQgZnVuYyhUJiBwYXJhbSl
7fQoKdGVtcGxhdGUgPHR5cGVuYW1lIFQ+CnZvaWQgY29uc3RGdW5jKGNvb
nN0IFQmIHBhcmFtKXt9CgppbnQgbWFpbigpewogIAogIGludCB4ID0gMjAxM
TsKICBjb25zdCBpbnQgY3ggPSB4OwogIGNvbnN0IGludCYgcnggPSB4OwogIAo
gIGZ1bmMoeCk7CiAgZnVuYyhjeCk7CiAgZnVuYyhyeCk7CiAgIAogIGNvbnN0R
nVuYyh4KTsKICBjb25zdEZ1bmMoY3gpOwogIGNvbnN0RnVuYyhyeCk7IAp9
&insightsOptions=cpp17&std=cpp17&rev=1.0) to analyze the process using
C++ Insight.

Argument Deduction by Universal
Reference#
1 // functionTemplateArgumentDeductionUniversalReference.cpp
2
4 void funcUniversal(T&& param){}
5
6
7 int main(){
8
9 int x = 2011;
12
13 funcUniversal(x);
14 funcUniversal(cx);
15 funcUniversal(rx);
16 funcUniversal(2014);
17 }
Succeeded
Explanation#
In the above code, we have defined a function funcUniversal in line 4 which
accepts its parameters with a universal reference.
For better understanding click here (https://cppinsights.io/lnk?

code=dGVtcGxhdGUgPHR5cGVuYW1lIFQ+CnZvaWQgZnVuY1VuaXZlcnNhbC
hUJiYgcGFyYW0pe30KCgppbnQgbWFpbigpewogIAogIGludCB4ID0gMjAxMTs
KICBjb25zdCBpbnQgY3ggPSB4OwogIGNvbnN0IGludCYgcnggPSB4OwogIAogI
GZ1bmNVbml2ZXJzYWwoeCk7CiAgZnVuY1VuaXZlcnNhbChjeCk7CiAgZnVuY
1VuaXZlcnNhbChyeCk7CiAgZnVuY1VuaXZlcnNhbCgyMDE0KTsgCn0=&insigh
tsOptions=cpp17&std=cpp17&rev=1.0) to analyze the process using C++
Insight.

Argument Deduction by Value#
1 // functionTemplateArgumentDeductionValue.cpp
2
4 void funcValue(T param){}
5
6 int main(){
7
8 int x = 2011;
9 const int cx = x;
11
12 funcValue(x);
13 funcValue(cx);
14 funcValue(rx);
15 }
Succeeded
Explanation#
In the above example, we have implemented a function funcValue in line 4
which takes its parameter by value.
For better understanding click here (https://cppinsights.io/lnk?

code=dGVtcGxhdGUgPHR5cGVuYW1lIFQ+CnZvaWQgZnVuY1ZhbHVlKFQgcG
FyYW0pe30KCmludCBtYWluKCl7CiAgCiAgaW50IHggPSAyMDExOwogIGNvb
nN0IGludCBjeCA9IHg7CiAgY29uc3QgaW50JiByeCA9IHg7CiAgCiAgZnVuY1Zh
bHVlKHgpOwogIGZ1bmNWYWx1ZShjeCk7CiAgZnVuY1ZhbHVlKHJ4KTsgCn0
=&insightsOptions=cpp17&std=cpp17&rev=1.0) to analyze the process using
C++ Insight.
Let’s solve an exercise in the next lesson.

Back Next
Template Arguments - Exercise
Completed
Report an Issue
- Exercise
In this lesson, we'll solve an exercise on template arguments.
Problem Statement#
The class Matrix holds its values in the container Cont .
Cont should have a default argument std::vector .
Instantiate myIntVec and myDoubleVec without specifying the container

explicitly.
3 #include <list>
4 #include <vector>
5
6 template <typename T, template <typename, typename> class Cont >
7 class Matrix{
8 public:
11 }
14 }
15
16 private:
18
19 };
https://www.educative.io/courses/generic-templates-in-cpp/NE5gyo1VZ3m 1/3
20
21 int main(){
22
24
25 // Define myIntVec and myDoubleVec with specifying containers explicitly
26 // Call getSize() function on it to check the result
27 (/learn)
31
33
34 }
Output 1.57s
one two three four
We’ll look at the solution of this problem in the next lesson.

Back Next
Completed
Report an Issue
- Solution
Let's review the solution of the exercise in this lesson.
• Solution Review
• Explanation
Solution Review#
1 // templateClassTemplateMethods3.cpp
2
5 #include <list>
6 #include <vector>
7
8 template <typename T, template <typename, typename> class Cont = std::vector>
9 class Matrix{
10 public:
13 }
16 }
17
18 private:
20
21 };
22
23 int main(){
24
26
https://www.educative.io/courses/generic-templates-in-cpp/3j28jxWmYGQ 1/3
27 Matrix<int> myIntVec{1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
29 std::cout << "myIntVec.getSize(): " << myIntVec.getSize() << std::endl;
30
(/learn)
32
33 Matrix<double> myDoubleVec{1.1, 2.2, 3.3, 4.4, 5.5};
35 std::cout << "myDoubleVec.getSize(): " << myDoubleVec.getSize() << std::endl;
36
38
42
44 }
Output 4.07s
1 2 3 4 5 6 7 8 9 10
myIntVec.getSize(): 10
1.1 2.2 3.3 4.4 5.5
myDoubleVec.getSize(): 5
one two three four
Explanation#
The class template Matrix uses, by default, an std::vector for holding it’s
elements (line 8). Thanks to the default container, a Matrix can be
instantiated (lines 27 and 33) without explicitly specifying a container.
Let’s move on to specialization in the next lesson.

Back Next
- Exercise Specialization
Completed
Report an Issue
3/20/22, 5:41 PM Specialization - Generic Programming Templates in C++
Specialization
Let's learn about template specialization in this lesson.
• Specialization
• Primary Template
• Partial Specialization
• Rules for Partial Specializations:

• Rules for Right Specialization:
• Full Specialization
Specialization#
Template specialization addresses the need to have different code for
different template argument types. Templates define the behavior of
families of classes and functions.
Often it is necessary that special types, non-types, or templates as

arguments are treated as special.
You can fully specialize templates; class templates can even be partially
specialized.
The methods and attributes of specialization don’t have to be identical.
General or Primary templates can coexist with partially or fully
specialized templates.
https://www.educative.io/courses/generic-templates-in-cpp/g758AwL0jQj 1/5
The compiler prefers fully specialized to partially specialized templates

and partially specialized templates to primary templates.
Primary Template#
The primary template has to be declared before the partially or fully
specialized templates.
If the primary template is not needed, just a declaration will suffice.
template <typename T, int Line, int Column> class Matrix;
class Matrix<T, 3, 3>{};
template <>class Matrix<int, 3, 3>{};
Partial Specialization#
The partial specialization of a template is only supported for class templates
and it has template arguments and template parameters.
1 template <typename T, int Line, int Column> class Matrix{};
2
4 class Matrix<T, 3, 3>{};
5
6 template <int Line, int Column>
7 class Matrix<double, Line, Column>{};
8
9 Matrix<int, 3, 3> m1; // class Matrix<T, 3, 3>
10 Matrix<double, 10, 10> m2; // class Matrix<double, Line, Column>
11 Matrix<std::string, 4, 3> m3; // class Matrix<T, Line, Column>
(/learn)
Rules for Partial Specializations:#

1. The compiler uses a partial specialization if the parameters of the class
are a subset of the template arguments.
2. The template parameters which are not specified must be given as

template arguments, for example, lines 3 and 4 in the code snippet.
3. The number and sequence of the template arguments must match with
the number and sequence of the template parameters of the primary
template.
4. If you use defaults for template parameters, you don’t have to provide
the template arguments. Only the primary template accepts defaults.
Rules for Right Specialization:#

Three rules for getting the right specialization:
1. If the compiler finds only one specialization, it uses that specialization.
2. If the compiler finds more than one specialization, it uses the most
specialized one.
If the compiler can’t find the most specialized one, it
throws an error.
3. If the compiler finds no specialization, it simply uses the primary

template.
Template A is more specialized than template B if:
B can accept all arguments that A can accept.
B can accept arguments that A cannot accept.
To learn more about partial specialization, click here

(https://en.cppreference.com/w/cpp/language/partial_specialization).
Full Specialization#
For a fully specialized template, you have to provide all template arguments
for the template parameters. The number of template parameters is reduced
to an empty list.
1 template <typename T> struct Type{
2 std::string getName() const {
3 return "unknown";
4 };
5 };
6 template <>
7 struct Type<Account>{
9 return "Account";
10 };
If you define the methods of a class template outside of the class, you have to
specify the template arguments in angle brackets after the name of the class.
Define the method of a fully specialized class template outside the class body
without the empty template parameter list: template <> .
template <typename T, int Line, int Column>
struct Matrix;
template <>
struct Matrix<int, 3, 3>{
int numberOfElements() const;
};
// template <>
int Matrix<int, 3, 3>::numberOfElements() const {
return 3 * 3;
};
To learn more about full specialization, click here

(https://en.cppreference.com/w/cpp/language/template_specialization).
In the next lesson, we’ll look at a few examples of template specialization.
Back Next
Mark as Completed
Report an Issue
- Examples
In this lesson, we'll look at a few examples of template specialization.
• Example 1: Template Specialization
• Explanation
• Example 2: Template Specialization External
• Explanation
• Example 3: Template Specialization Full
• Explanation
• Example 4: Template Specialization Type Traits

• Explanation
• Example 5: Template Types
• Explanation
Example 1: Template Specialization#

1 // TemplateSpecialization.cpp
2
4
5 class Account{
6 public:
7 explicit Account(double b): balance(b){}
8 double getBalance() const {
9 return balance;
10 }
11 private:
https://www.educative.io/courses/generic-templates-in-cpp/mE0mkoNgogA 1/10
12 double balance;
13 };
14
15 template <typename T, int Line, int Column>
16 class Matrix{
17 std::string getName() const { return "Primary Template"; }
18 };
19
21 class Matrix<T,3,3>{
22 std::string name{"Partial Specialization"};
23 };
24
25 template <>
26 class Matrix<int,3,3>{};
27
28
29 template<typename T>
32 }
33
34 template <>
35 bool isSmaller<Account>(Account fir, Account sec){
36 return fir.getBalance() < sec.getBalance();
37 }
38
39 int main(){
40
42
43 Matrix<double,3,4> primaryM;
44 Matrix<double,3,3> partialM;
45 Matrix<int,3,3> fullM;
46
48 std::cout << "isSmaller(Account(100.0),Account(200.0)): "<< isSmaller(Account(100.
49
51
52 }
53
Output 1.67s
isSmaller(Account(100.0),Account(200.0)): true
Explanation#
In the above example, we’re modifying the codes that we have used in the
previous lesson
754565951488/#primary-template).
Primary template is called when we use values other than

Matrix<data_type, 3, 3> (line 43).
Partial specialization is called when we instantiate Matrix<data_type,

3, 3> where data_type is not int (line 44).
Full specialization is called when we explicitly use int as a data type:

Matrix<int, 3, 3> (line 45)
Full specialization of the function template isSmaller is only applicable

for Account objects. This allows it to compare two Account objects
based on their balance (line 48).
Example 2: Template Specialization

External#
1 //TemplateSpecializationExternal.cpp
2
4
5
6 template <typename T=std::string, int Line=10, int Column=Line>
7 class Matrix{
8 public:
9 int numberOfElements() const;
10 };
10 };
11
12 template <typename T, int Line, int Column>
13 int Matrix<T,Line,Column>::numberOfElements() const {
14 return Line * Column;
15 }
16
17
19 class Matrix<T,3,3>{
20 public:
22 };
23
25 int Matrix<T,3,3>::numberOfElements() const {
26 return 3*3;
27 }
28
29
30 template <>
31 class Matrix<int,3,3>{
32 public:
34 };
35
36 int Matrix<int,3,3>::numberOfElements() const {
37 return 3*3;
38 }
39
40 int main(){
41
43
44 Matrix<double,10,5> matBigDouble;
45 std::cout << "matBigDouble.numberOfElements(): " << matBigDouble.numberOfElements(
46
47 // Matrix matString; // ERROR
48 Matrix<> matString;
49 std::cout << "matString.numberOfElements(): " << matString.numberOfElements() << s
50
51 Matrix<float> matFloat;
52 std::cout << "matFloat.numberOfElements(): " << matFloat.numberOfElements() << std
53
54 Matrix<bool,20> matBool;
55 std::cout << "matBool.numberOfElements(): " << matBool.numberOfElements() << std::
56
57 Matrix <double,3,3> matSmallDouble;
58 std::cout << "matSmallDouble.numberOfElements(): " << matSmallDouble.numberOfEleme
59
60 Matrix <int,3,3> matInt;
61 std::cout << "matInt.numberOfElements(): " << matInt.numberOfElements() << std::en
62
64
65 }
66
Output 1.38s
matBigDouble.numberOfElements(): 50
matString.numberOfElements(): 100
matFloat.numberOfElements(): 100
matBool.numberOfElements(): 400
matSmallDouble.numberOfElements(): 9
matInt.numberOfElements(): 9
Explanation#
In the above example, we have set the default value of line to 10 (line 6)
and used the value for line as the default for column . The method
numberOfElements returns the product of both numbers as a result. If we call
the Matrix with arguments, then these passed arguments override the
default. For float and string , it returns the 100 as no arguments are passed
and the default arguments are used (lines 48 and 51).

Full#
1 // templateSpecializationFull.cpp
2
4 #include <string>
5
7 T min(T fir, T sec){
8 return (fir < sec) ? fir : sec;
9 }
10
11 template <>
12 bool min<bool>(bool fir, bool sec){
13 return fir & sec;
14 }
15
16 int main(){
17
19
20 std::cout << "min(3.5, 4.5): " << min(3.5, 4.5) << std::endl;
21 std::cout << "min<double>(3.5, 4.5): " << min<double>(3.5, 4.5) << std::endl;
22
23 std::cout << "min(true, false): " << min(true, false) << std::endl;
24 std::cout << "min<bool>(true, false): " << min<bool>(true, false) << std::endl;
25
27
28 }
29
Output 1.52s
min(3.5, 4.5): 3.5
min<double>(3.5, 4.5): 3.5
min(true, false): false
min<bool>(true, false): false
Explanation#
In the above example, we have defined a full specialization for bool . The
primary and the full specialization are implicitly invoked in the lines (20 and
23) and explicitly invoked in the lines (21 and 24).

Type Traits#
1 // templateSpecializationTypeTraits.cpp
2
5
6 using namespace std;
7
9 void getPrimaryTypeCategory(){
10
11 cout << boolalpha << endl;
12
13 cout << "is_void<T>::value: " << is_void<T>::value << endl;
14 cout << "is_integral<T>::value: " << is_integral<T>::value << endl;
15 cout << "is_floating_point<T>::value: " << is_floating_point<T>::value << endl;
16 cout << "is_array<T>::value: " << is_array<T>::value << endl;
17 cout << "is_pointer<T>::value: " << is_pointer<T>::value << endl;
18 cout << "is_reference<T>::value: " << is_reference<T>::value << endl;
19 cout << "is_member_object_pointer<T>::value: " << is_member_object_pointer<T>::val
20 cout << "is_member_function_pointer<T>::value: " << is_member_function_pointer<T>:
21 cout << "is_enum<T>::value: " << is_enum<T>::value << endl;
22 cout << "is_union<T>::value: " << is_union<T>::value << endl;
23 cout << "is_class<T>::value: " << is_class<T>::value << endl;
24 cout << "is_function<T>::value: " << is_function<T>::value << endl;
25 cout << "is_lvalue_reference<T>::value: " << is_lvalue_reference<T>::value << endl
26 cout << "is_rvalue_reference<T>::value: " << is_rvalue_reference<T>::value << endl
27
28 cout << endl;
29
30 }
31
Output 3.93s
is_void<T>::value: true
is_integral<T>::value: false
is_floating_point<T>::value: false
is_array<T>::value: false
is_pointer<T>::value: false
is_reference<T>::value: false
is_member_object_pointer<T>::value: false
is_member_function_pointer<T>::value: false
Explanation#
We have used the type_traits
229176066048) library which detects at compile-time to which primary type
category void (line 13) belongs to. The primary type categories are complete
and exclusive. This means each type belongs exactly to one primary type
category. For example, void returns true for the type-trait std::is_void
and false
(/learn)for all the other type categories.
Example 5: Template Types#

1 // templateTypes.cpp
2
4 #include <string>
5
7 struct Type{
9 return "unknown";
10 }
11 };
12
13 int main(){
13 int main(){
14
16
17 Type<float> tFloat;
18 std::cout << "tFloat.getName(): " << tFloat.getName() << std::endl;
19
21
22 }
23
Output 1.39s
tFloat.getName(): unknown
Explanation#
In the above example, the method getName returns unknown for any type
passed in the argument of type function (line 8). If we specialize the class
template for further types, we will implement a type deduction system at
compile-time. We’ll look at it in the coming exercise.
We’ll solve an exercise in the next lesson.


Back Next
Specialization - Exercise
Completed
Report an Issue
(/learn)
- Exercise
Let's solve a template specialization problem in this lesson.
Problem Statement#
The class template Type in the code below returns to each type the name
unknown .
Use the class template Type as a starting point to write a type

introspection system with the help of partial and full specialization.
You need to write code for int , double , an arbitrary class named
Account , pointer , const , and string .
2 #include <string>
3
4 // Implement with full and partial specialization
6
8 struct Type{
10 return "unknown";
11 }
12 };
13
14 int main(){
15
https://www.educative.io/courses/generic-templates-in-cpp/qZq27KK1WgG 1/3
17
18 // call objects for each defined templetes here
19
20 // An example of float is given below which returns "unknown"
23
25
26 }

Back Next
Mark as Completed
Report an Issue
- Solution
Let's look at the solution review for the exercise we presented in the last lesson.
• Solution Review
• Explanation
Solution Review#
1 // Template Types
2
4 #include <string>
5
6 class Account{};
7
9 struct Type{
11 return "unknown";
12 }
13 };
14
16 struct Type<T*>{
18 return "pointer";
19 }
20 };
21
23 struct Type<const T>{
25 return "const";
26 }
https://www.educative.io/courses/generic-templates-in-cpp/x1mQqPOmE3q 1/4
27 };
28
29
30 template<>
31 struct Type<int>{
33 return "int";
34 }
35 };
36
37 template<>
38 struct Type<double>{
40 return "double";
41 }
42 };
43
44 template<>
45 struct Type<std::string>{
47 return "std::string";
48 }
49 };
50
51 template<>
52 struct Type<Account>{
54 return "Account";
55 }
56 };
57
58 int main(){
59
61
64
65 Type<const float> tConstFloat;
66 std::cout << "tConstFloat.getName(): " << tConstFloat.getName() << std::endl;
67
68 Type<float*> tFloatPointer;
69 std::cout << "tFloatPointer.getName(): " << tFloatPointer.getName() << std::endl;
70
71 Type<double> tDouble;
72 std::cout << "tDouble.getName(): " << tDouble.getName() << std::endl;
73
74 Type<std::string> tString;
75 std::cout << "tString.getName(): " << tString.getName() << std::endl;
76
77 Type<int> tInt;
78 std::cout << "tInt.getName(): " << tInt.getName() << std::endl;
78 std::cout << tInt.getName(): << tInt.getName() << std::endl;
79
80 Type<Account> tAccount;
81 std::cout << "tAccount.getName(): " << tAccount.getName() << std::endl;
82
84
85 }
86
Output 1.31s
tFloat.getName(): unknown
tConstFloat.getName(): const
tFloatPointer.getName(): pointer
tDouble.getName(): double
tString.getName(): std::string
tInt.getName(): int
tAccount.getName(): Account
(/learn)
Explanation#
In the above code, we have separately defined
the partial and full
specialization of the class template Type . The partial and full specializations
accept an int , double , Account , string , const , and Pointer . On calling
each type, the relative type is returned. We have not defined the full
specialization for float , so when instantiating the class template for float ,
it gives an unknown in response.
We have covered the basics of templates in this chapter. In the next chapter,
we’ll learn about the details of templates. Let’s start with template
instantiation in the next lesson.

Back Next
- Exercise Template Instantiation
Completed
Report an Issue
3/20/22, 5:53 PM Template Instantiation - Generic Programming Templates in C++
(/learn)
Template Instantiation
In this lesson, we'll learn about template instantiation.
• Template Instantiation
• Implicit
• Explicit
• Lazy Evaluation
Template Instantiation#
Templates can be implicitly and explicitly instantiated. Implicit instantiation
means automatically and explicit means manually.
Implicit#
std::vector<int> vec{};
bool isSmaller<double>(fir, sec);
bool isSmaller(fir, sec);
Explicit#
template class std::vector<int>;
template bool std::vector<double>::empty() const;
template bool isSmaller<double>(double, double);
template bool isSmaller(double, double);
https://www.educative.io/courses/generic-templates-in-cpp/B8A2jn0D4LJ 1/3
Lazy Evaluation#
When a class is instantiated, only the method declarations are available.
The definition of a method is only instantiated when it is used.
It is not necessary that all methods of class templates are valid for the
template arguments. You can only use the methods, which are valid for a
given instantiation.
In the next lesson, we’ll look at an example of template instantiation.

Back Next
- Solution - Example
Mark as Completed
Report an Issue
3/20/22, 8:07 PM - Example - Generic Programming Templates in C++
- Example
Let's have a look at an example of template instantiation.
• Example: Template Instantiation
• Explanation
Example: Template Instantiation#

1 // templateInstantiation.cpp
2
4 #include <vector>
5
7 class Array{
8
9 public:
11
13 return N;
14 }
15
17 };
18
22 }
23
24 template class std::vector<int>;
25 template bool std::vector<double>::empty() const;
26
https://www.educative.io/courses/generic-templates-in-cpp/YQzNvWngvNM 1/4
27 template class Array<int, 20>;
28 template int Array<double, 5>::getSize() const;
29
30 template bool isSmaller(double, double);
31 template bool isSmaller<int>(int, int);
32
33 int main(){
34
36
37 std::cout << std::boolalpha << "implicit" << std::endl;
38
40
41 std::vector<int> vec{};
42 std::cout << "vec.size(): " << vec.size() << std::endl;
43
44 Array<int, 10> arr;
45 std::cout << "arr.getSize(): " << arr.getSize() << std::endl;
46
48
49 std::cout << "isSmaller(5, 10): " << isSmaller(5,10) << std::endl;
50
51 std::cout << "isSmaller<double>(5.5, 6.5): " << isSmaller<double>(5.5, 6.5) << std
52
54
55 }
56
Output 1.89s
implicit
vec.size(): 0
arr.getSize(): 10
isSmaller(5, 10): true
isSmaller<double>(5.5, 6.5): true
(/learn)
Explanation#
In the above example, we have implemented a template class Array which
includes a function getSize() that returns the size of the element N passed
into the constructor. We have also defined a template function isSmaller
bool and its return type is declared explicitly which returns true if the first
passed argument is less than the second argument.
Lines 24 – 31 contain explicit template instantiation. The main program

contains implicit template instantiation. Line 24 is an explicit instantiation
for int and line 25 is an explicit instantiation of the method getSize for
double . The lines 27 and 28 are quite similar for Array . The compiler can
automatically deduce the template argument for the function argument in
line 30.
In the next lesson, we’ll solve a small exercise on template instantiation.

Back Next
Template Instantiation - Exercise
Completed
Report an Issue
(/learn)
- Exercise
Problem Statement#
Define a class template with at least one method. This method should not be
valid for all possible template arguments. Instantiate the class template for
an invalid template argument.
What happens, when you
don’t instantiate?
implicitly instantiate?
or explicitly instantiate the method?
2
3 // Define a class template here with one method
4
5 int main() {
6 // call these methods here
7 }
In the next lesson, we’ll look at the solution for this exercise.
https://www.educative.io/courses/generic-templates-in-cpp/7AgkJ127Xyy 1/2

Back Next
- Example - Solution
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/7AgkJ127Xyy 2/2
- Solution
Let's have a look at the solution review of the last exercise.
• Solution Review
• Explanation
Solution Review#
1 // templateInstantiationInvalid.cpp
2
3 (/learn)
#include <iostream>
4 #include <vector>
5
6 template <int Nom, int Denom>
7 class Rational{
8 public:
9 int getFloor(){
10 return Nom / Denom;
11 }
12 };
13
14 // template int Rational<5, 0>::getFloor();
15
16 int main(){
17
19
20 Rational<5, 3> rat1;
21 std::cout << "rat1.getFloor(): " << rat1.getFloor() << std::endl;
22
23
24 Rational<5, 0> rat2;
25 // std::cout << "rat2.getFloor(): " << rat2.getFloor() << std::endl;
26
https://www.educative.io/courses/generic-templates-in-cpp/JQRnWgvjQgD 1/3
28
29 }
30
Explanation#
In the above code, we have called the getFloor function for 5 and 3 in line 9,
and it returns 1. To invoke the function for an invalid call, we can give the
arguments 5 and 0 which gives an error.
In the next lesson, we’ll discuss variadic templates.

Back Next
- Exercise Variadic Templates
Mark as Completed
Report an Issue
3/20/22, 5:55 PM Variadic Templates - Generic Programming Templates in C++
Variadic Templates
Let's learn about variadic templates in detail in this lesson.
• Variadic Templates
• Parameter Pack
Variadic Templates#
A variadic template is a template that can has an arbitrary number of
parameters.
template <typename ... Args>
void variadicTemplate(Args ... args){ . . . . }
Parameter Pack#
A template parameter pack is a template parameter that accepts zero or
more template arguments (non-types, types, or templates). A function
parameter pack is a function parameter that accepts zero or more function
arguments.
(/learn)
By using the ellipse (…), Args- or args becomes a parameter pack.
Args is a template parameter pack; args is a function parameter pack.
Parameter packs can only be packed and unpacked.

If the ellipsis is left from Args , the parameter pack will be packed and if
https://www.educative.io/courses/generic-templates-in-cpp/B6qkMM2LjKx 1/3
the ellipse is right from Args , the parameter pack will be unpacked.
The compiler can automatically deduce the template arguments in case

of a function template.
For example, the following classes/functions in STL extensively use variadic

templates.
Variadic Templates are often used in the Standard Template
Library:
sizeof -Operator, std::tuple , std::thread , std::make_unique , std::lock
The usage of parameter packs obeys a typical pattern for class templates.
Perform an operation on the first element of the parameter pack and

recursively invoke the operation on the remaining elements.
The recursion ends after a finite number of steps.
The boundary condition is typically a fully specialized template.
1 template<>
2 struct Mult<>{ .... }
3 template<int i, int ... tail >
4 struct Mult<i, tail ...>{ ....
To learn more about variadic templates, click here

(https://en.cppreference.com/w/cpp/language/parameter_pack).
In the next lesson, we’ll study examples of variadic templates.

d i &
https://www.educative.io/courses/generic-templates-in-cpp/B6qkMM2LjKx di & i M h 20222/3
Back Next
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/B6qkMM2LjKx 3/3
(/learn)
- Examples
Let's look at the examples of variadic templates.
• Example 1: Variadic Template
• Explanation
• Example 2: Template Perfect Forwarding
• Explanation
Example 1: Variadic Template#

1 // templateVariadicTemplates.cpp
2
4
5 template <typename... Args>
6 int printSize(Args... args){
7 return sizeof ...(args);
8 }
9
10 template<int ...>
11 struct Mult;
12
13 template<>
14 struct Mult<>{
15 static const int value= 1;
16 };
17
18 template<int i, int ... tail>
19 struct Mult<i, tail ...>{
20 static const int value= i * Mult<tail ...>::value;
21 };
22
23 int main(){
https://www.educative.io/courses/generic-templates-in-cpp/JEPE51ykBOg 1/6
23 int main(){
24
26
27 std::cout << "printSize(): " << printSize() << std::endl;
28 std::cout << "printSize(template,2011,true): " << printSize("template",2011,true)
29 std::cout << "printSize(1, 2.5, 4, 5, 10): " << printSize(1, 2.5, 4, 5, 10) << st
30
Explanation#
In the above example, we have used printSize function, which prints the
number of elements (of any type) passed as arguments. It detects the
number of elements on compile-time using the sizeof operator, and in case
of an empty argument list, it returns 0.
There is a struct defined as Mult which takes arguments of integer type

and return their product. If there is no argument passed, then it returns 1
which is the neutral element for multiplication. The result is stored in the
value in the fully specialized template in lines 13 – 16. The partial
specialization in lines 18 – 21 starts the recursion, which ends with the
aforementioned fully specialization for 0. The primary template in line 10 is
never used and must, therefore, never be defined.
To visualize the template instantiation for the above-mentioned example

click here (https://cppinsights.io/lnk?
code=I2luY2x1ZGUgPGlvc3RyZWFtPgoKdGVtcGxhdGUgPHR5cGVuYW1lLi4uI
EFyZ3M+CmludCBwcmludFNpemUoQXJncy4uLiBhcmdzKXsKICByZXR1cm4g
c2l6ZW9mIC4uLihhcmdzKTsKfQoKdGVtcGxhdGU8aW50IC4uLj4Kc3RydWN0
IE11bHQ7Cgp0ZW1wbGF0ZTw+CnN0cnVjdCBNdWx0PD57CiAgc3RhdGljIGN
vbnN0IGludCB2YWx1ZT0gMTsKfTsKCnRlbXBsYXRlPGludCBpLCBpbnQgLi4u
IHRhaWw+CnN0cnVjdCBNdWx0PGksIHRhaWwgLi4uPnsKICBzdGF0aWMgY
29uc3QgaW50IHZhbHVlPSBpICogTXVsdDx0YWlsIC4uLj46OnZhbHVlOwp9O
woKaW50IG1haW4oKXsKCiAgc3RkOjpjb3V0IDw8IHN0ZDo6ZW5kbDsKCiAgc
3RkOjpjb3V0IDw8ICJwcmludFNpemUoKTogIiA8PCBwcmludFNpemUoKSA8P
CBzdGQ6OmVuZGw7CiAgc3RkOjpjb3V0IDw8ICJwcmludFNpemUodGVtcGxh
dGUsMjAxMSx0cnVlKTogIiA8PCBwcmludFNpemUoInRlbXBsYXRlIiwyMDEx
LHRydWUpIDw8IHN0ZDo6ZW5kbDsKICBzdGQ6OmNvdXQgPDwgInByaW50
U2l6ZSgxLCAyLjUsIDQsIDUsIDEwKTogIiA8PCAgcHJpbnRTaXplKDEsIDIuNSw
gNCwgNSwgMTApIDw8IHN0ZDo6ZW5kbDsKCiAgc3RkOjpjb3V0IDw8IHN0ZD
o6ZW5kbDsKCiAgc3RkOjpjb3V0IDw8ICJNdWx0PDEwPjo6dmFsdWU6ICIgPD
wgTXVsdDwxMD46OnZhbHVlIDw8IHN0ZDo6ZW5kbDsKICBzdGQ6OmNvdX
QgPDwgIk11bHQ8MTAsMTAsMTA+Ojp2YWx1ZTogIiA8PCBNdWx0PDEwLDE
wLDEwPjo6dmFsdWUgPDwgc3RkOjplbmRsOwogIHN0ZDo6Y291dCA8PCAiT
XVsdDwxLDIsMyw0LDU+Ojp2YWx1ZTogIiA8PCBNdWx0PDEsMiwzLDQsNT4
6OnZhbHVlIDw8IHN0ZDo6ZW5kbDsKCiAgc3RkOjpjb3V0IDw8IHN0ZDo6ZW
5kbDsKCn0K&insightsOptions=cpp17&std=cpp17&rev=1.0).
Example 2: Template Perfect

Forwarding#
// templatePerfectForwarding.cpp
#include <iostream>
#include <utility>
template<typename T, typename ... Args>
T createT(Args&& ... args){
return T(std::forward<Args>(args) ...);
struct MyStruct{
MyStruct(int&, double&, double&&){}
friend std::ostream& operator<< (std::ostream& out, const MyStruct&){
out << "MyStruct" << std::endl;
return out;
}
};
int main(){
std::cout << std::endl;
double myDouble= createT<double>();
std::cout << "myDouble: " << myDouble << std::endl;
int myInt= createT<int>(1);
std::cout << "myInt: " << myInt << std::endl;
std::string myString= createT<std::string>("My String");
std::cout << "myString: " << myString << std::endl;
MyStruct myStruct= createT<MyStruct>(myInt, myDouble, 3.14);
std::cout << "myStruct: " << myStruct << std::endl;
Explanation#
In the above example, we have created a createT function which invokes
the constructor T with the arguments args . If there is no value passed, it
invokes the default constructor. The magic of the factory function createT is
g y
that it can invoke each constructor. Thanks to perfect forwarding, each
value can be used such as an lvalue or an rvalue; thanks to parameter packs,

any number of arguments can be used. In the case of MyStruct , a
constructor that requires three arguments is used.
The pattern of the function template createT is exactly the pattern, factory
functions such as std::make_unique , std::make_shared , std::make_pair , or
std::make_tuple use.
In the next lesson, we’ll solve exercises related to variadic templates.

Back Next
Variadic Templates - Exercises
Mark as Completed
Report an Issue
3/20/22, 5:56 PM - Exercises - Generic Programming Templates in C++
(/learn)
- Exercises
Let's solve a couple of exercises to test your understanding of variadic templates.
• Try It Out!
• Problem Statement 2
Try It Out!#
The function createT in the Example 2
046497624064#example-2-template-perfect-forwarding)
is the perfect factory
function because it can invoke each constructor. Try to find a constructor
call, which cannot be used in the function template createT .
2 #include <utility>
3
4 template<typename T, typename ... Args>
5 T createT(Args&& ... args){
6 return T(std::forward<Args>(args) ...);
7 }
8
9 // Try to break the code
10
11 struct MyStruct{
12 MyStruct(int&, double&, double&&){}
13 friend std::ostream& operator<< (std::ostream& out, const MyStruct&){
14 out << "MyStruct" << std::endl;
15 return out;
16 }
17 };
https://www.educative.io/courses/generic-templates-in-cpp/JQKyvqxRnnJ 1/3
17 };
18
19 int main(){
20
21 // Call it here after breaking the code
22
24 }
Problem Statement 2#
A typical example of a variadic template is the C function printf . Try to
implement it with variadic templates.
2
3 void myPrintf(const char* format) {
4 std::cout << format;
5 }
6
7 template<typename T, typename... Args>
8 void myPrintf(const char* format, T value, Args... args){
9 // Implement the function here
10 }
11
12 int main(){
13 // the given line should print "Hello world! 2011"
14 myPrintf("% world% %\n","Hello",'!',2011);
15 }
16
In the next lesson, we’ll look at the solution of problem statement 2.

Back Next
Mark as Completed
Report an Issue
(/learn)
- Solution
In this lesson, we'll look at the solution review of the problem statement 2 form last exercise.
• Solution Review of Problem Statement 2
• Explanation
Solution Review of Problem

Statement 2#
1 // printf.cpp
2
4
5 void myPrintf(const char* format)
6 {
7 std::cout << format;
8 }
9
10 template<typename T, typename... Args>
11 void myPrintf(const char* format, T value, Args... args)
12 {
13 ····for·(·;·*format·!=·'\0';·format++·)·{
14 if ( *format == '%' ) {
15 std::cout << value;
16 myPrintf(format+1, args...);
17 return;
18 }
19 std::cout << *format;
20 }
21 }
22
23 int main(){
24 P i tf("% ld% %\ " "H ll " '!' 2011)
https://www.educative.io/courses/generic-templates-in-cpp/qAK1PwqXKGp 1/3
24 myPrintf("% world% %\n","Hello",'!',2011);
25 }
26
Output 1.58s
Hello world! 2011
Explanation#
The printf function uses variadic templates to implement a type-safe
printf variant. The format string is specified and thanks to function
template argument deduction, the compiler deduces the types.
The function templates loops (line 13) as long as the format symbol is not
equal to \0 . When the format symbol is not equal to \0 , two control flows
are possible. First, if the format starts with % (line 14), the first argument
value is displayed, and myPrintf is invoked once more but with a new
format symbol and an argument less (line 16). Second, if the format string
does not start with % , the format symbol is just displayed (line 19). The
function myPrintf (line 5) is the boundary condition for the recursive calls.
Let’s study fold expressions in the next lesson.

Back Next
- Exercises Fold Expressions
Mark as Completed
Report an Issue
3/20/22, 5:59 PM Fold Expressions - Generic Programming Templates in C++
Fold Expressions
In this lesson, we'll study fold expressions.
• Fold Expressions (C++17)
• Two variations
Fold Expressions (C++17)#

Fold expressions is a nice syntax to evaluate binary operators at compile-
time. Fold expressions reduce parameter packs on binary operators.
C++11 provides support for parameter packs:
1 bool all_14(){
2 return true;
3 }
4
5 template<typename T, typename ...Ts>
6 bool all_14(T t, Ts ... ts){
7 return t && all_14(ts...);
8 }
C++17 provides support for fold expressions:
1 template<typename ... Args>
2 bool all_17(Args ... args){
3 return ( ... && args);
4 }
https://www.educative.io/courses/generic-templates-in-cpp/3YXgz8j22OR 1/5
5
6 bool val == all_14(true, true, true, false)
7 == all_17(true, true, true, false
8 == ((true && true)&& true)&& false
9 == false;
Two variations#
The fold expression either has or does not have an initial value
The parameter pack will be processed from left or right
C++17 supports the following 32 operators in fold expressions:
+ - * / % ^ & | = < > << >> += -= *= /= %= ^= &= |= <<= >>= ==

!= <= >= && || , .* ->*
Operators with their default values:
Operator Symbol Defualt Value
Logical AND && true
Logical OR || false
Comma operator , void()
For binary operators that have no default value, you have to provide an
initial value. For binary operators that have a default value, you can specify
an initial value.
If the ellipsis stands left of the parameter pack, the parameter pack will be
processed from the left. The same holds for right. This is also true if you
(/learn)
provide an initial value.
The following table shows the four variations and their Haskell pendants.
The C++17 standard requires that fold expressions with initial value use the
same binary operator op .
C++ templates support Haskells fold* variants at compile-time, i.e., foldl ,

foldl1 , foldr and foldr1
Fold Expressions Haskell Description
Processes from left

… op pack fold1 op list
with operator op
Processes from right

pack op … foldr1 op list
with operator op
Processes from left

init op … op pack foldl op init list with operator op and
initial value init
Processes from right

pack op … op init foldr op init list with operator op and
initial value init
The C++ and Haskell variations differ in two points. The C++ version uses the
default value as the initial value while the Haskell version uses the first
element as the initial value. The C++ version processes the parameter pack at
compile-time and the Haskell version, at run time.
The small code snippet shows once more the algorithm all – this time, we
use true as the initial value.
use true as the initial value.
template<typename... Args>
bool all(Args... args){
return (true && ... && args);
To learn more about fold expressions, click here

(https://en.cppreference.com/w/cpp/language/fold).
In the next lesson, we’ll have a look at the examples of fold expressions.

Back Next
Mark as Completed
Report an Issue
- Examples
In this lesson, we'll look at an examples of the fold expressions.
• Example 1: Fold Expression
• Explanation
• Example 2: String Concatenation
• Explanation
Example 1: Fold Expression#

1 // foldExpression.cpp
2
4
5 template<typename... Args>
6 bool all(Args... args) { return (... && args); }
7
9 bool any(Args... args) { return (... || args); }
10
12 bool none(Args... args) { return not(... || args); }
13
14 (/learn)
15 int main(){
16
18
19 std::cout << std::boolalpha;
20
21 std::cout << "all(true): " << all(true) << std::endl;
22 std::cout << "any(true): " << any(true) << std::endl;
23 std::cout << "none(true): " << none(true) << std::endl;
https://www.educative.io/courses/generic-templates-in-cpp/R8zjBEkQMWE 1/4
24
26
27 std::cout << "all(true, true, true, false): " << all(true, true, true, false) << s
28 std::cout << "any(true, true, true, false): " << any(true, true, true, false) << s
29 std::cout << "none(true, true, true, false): " << none(true, true, true, false) <<
30
32
33 std::cout << "all(false, false, false, false): " << all(false, false, false, false
34 std::cout << "any(false, false, false, false): " << any(false, false, false, false
35 std::cout << "none(false, false, false, false): " << none(false, false, false, fal
36
38 }
Output 1.38s
all(true, true, true, false): false
any(true, true, true, false): true
none(true, true, true, false): false
all(false, false, false, false): false
any(false, false, false, false): false
none(false, false, false, false): true
Explanation#
In the above example, we have three predicates.
all function returns true only if all the values passed to it are true ,
else false because we’re using && as an operator.
any function returns true if any passed value is true , else false
because we’re using || as an operator.
none function returns true only if all the passed parameters are false
because we’re using || operator with not and it will invert the result.
Example 2: String Concatenation#

2 #include <string>
3
4 template<typename ...Args>
5 auto addLeft(Args ... args){
6 return (std::string("0") + ... + args); // (((std::string("0")+"1")+"2")+"3")
7 }
8
10 auto addRight(Args ... args){
11 return (args + ... + std::string("0")); // ("1"+("2"+("3" + std::string("0"))))
12 }
13
14 int main(){
15
16 std::cout << addLeft("1", "2", "3") << std::endl; // 0123
17 std::cout << addRight("1", "2", "3") << std::endl; // 1230
18 }
19
Output 1.63s
0123
1230
Explanation#
Explanation#
The above-mentioned example shows the difference between left and right
fold. We had to start with a std::string(“0”) and not “0” because “0” +
“1” gives an error. String concatenation requires at least one string.

Back Next
Fold Expressions - Exercise
Completed
Report an Issue
- Examples
• Explanation
• Explanation

(/learn)
4
7
10
13
14
15 int main(){
16
18
20
24
26
30
32
34 std::cout << "any(false false false false): " << any(false false false false
Output 1.37s
all(true): true
any(true): true
none(true): false
Explanation#
In the above example, we have three predicates.

2 #include <string>
3
7 }
8
12 }
13
14 int main(){
15
18 }
19
Output 2.34s
0123
1230
Explanation#
We’ll solve an exercise in the next lesson

We ll solve an exercise in the next lesson.

Back Next
Mark as Completed
Report an Issue
- Examples
• Explanation
• Explanation

1 // foldExpression.cpp
2
4
7
10
13
14
15 int main(){
16
18
20
24
26
30
32
34 std::cout << "any(false, false, false, false): " << any(false, false, false, false
35 std::cout << "none(false, false, false, false): " << none(false, false, false, fal
36
38 }
Output 1.37s
all(true): true
any(true): true
none(true): false
Explanation#
In the (/learn)
above example, we have three predicates.

2 #include <string>
3
7 }
8
12 }
13
14 int main(){
15
18 }
19
Output 2.34s
0123
1230
Explanation#

Back Next
Mark as Completed
Report an Issue
(/learn)
- Exercise
In this lesson, we'll solve an exercise.
Problem Statement#
Write two sum functions which use the fold expressions to add all the values
passed as arguments:
One function without an initial value.

One function with an initial value.
2
3 // implement a function sum without an intital value here
4
5 // implement a function sum with an intital value here
6
7 int main(){
8 // call your functions here
9 }
In the next lesson, we’ll look at the solution for this exercise.
https://www.educative.io/courses/generic-templates-in-cpp/N0o2olBQ2vD 1/2

Back Next
Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/N0o2olBQ2vD 2/2
- Solution
Let's have a look at the solution review for the last exercise.
• Solution Review
• Explanation
Solution Review#
1 // sum.cpp
2
4
5 template<typename ...Args> // Without an initial value
6 auto sum(Args&& ... args){
7 return ( ... + args);
8 }
9
10 template<typename T, typename ...Args> // with an initial value
11 auto sumWithStart(T&& t, Args&& ... args){
12 return ( t + ... + args);
13 }
14
15 int main(){
16 std::cout << sum(1, 2, 3) << std::endl;
17 std::cout << sumWithStart(20, 1, 2, 3) << std::endl;
18 }
https://www.educative.io/courses/generic-templates-in-cpp/RMKoMBMq4gL 1/3
Output 1.7s
26
(/learn)
Explanation#
In the above code, we have declared two functions, i.e., sum and
sumWithStart.
sum function takes a list of arguments without any initial value and
returns their sum.
sumWithStart function takes a list of arguments and uses the first

element of the list as the initial value and adds all the successive
elements to this initial value.
Let’s move on to friends in the next lesson.

Back Next
- Exercise Friends
Mark as Completed
Report an Issue
3/20/22, 6:05 PM Friends - Generic Programming Templates in C++
Friends
Let's discuss friends in this lesson.
• Friends
• General Friends
• Special Friends
• Friend to Types
Friends#
Friends of a class template have access to all members of the class template.
A class or a class template can have a friendship to class or class templates,

function or function templates, and types.
Rules:
1. The declaration of friends can be made at an arbitrary place in the class

declaration.
2. The access rights in the class have no influence.
3. Friendship will not be inherited.

4. Friendship is not transitive.
A friend has unrestricted access to the members of the class.
https://www.educative.io/courses/generic-templates-in-cpp/JQ3NDVv9xxg 1/4
General Friends#
A class or a class template can grant friendship to each instance of a class
template or a function template.
1 template <typename T> int myFriendFunction(T);
2
3 template <typename T> class MyFriend;
4
5 (/learn)
6 class GrantingFriendshipAsClassTemplate{
7 template <typename U> friend int myFriendFunction(U);
8 template <typename U> friend class MyFriend;
9 ...
When a class template grants friendship to a template, the typename of

the class template should be different from the typename of the
template. If both use the same name, the friendship is only granted for
the same types.
Special Friends#
A special friendship is a friendship that depends on the type of the template
parameter.
1 template <typename T> int myFriendFunction(T);
2 template <typename T> class MyFriend;
3
6 friend int myFriendFunction<>(double);
7 friend class MyFriend<int>
8 friend class MyFriend<T>;
If the name of the template parameter is identical to the name of the

template parameter granting the friendship, the friendship will be between
instances of the same type.
Friend to Types#
A class template can grant its friendship to a type parameter.
class Array{
friend T;
...
};
Array<Account> myAccount;
To know more about friends, click here

(https://en.cppreference.com/w/cpp/language/friend).

Back Next
Mark as Completed
Report an Issue
- Examples
In this lesson, we'll look at a few examples of using templates with friends.
• Example 1: Class Template General Friendship
• Explanation
• Example 2: Class Template Special Friendship
• Explanation
• Example 3: Class Template Type Friendship
• Explanation
Example 1: Class Template General

Friendship#
1 // templateClassTemplateGeneralFriendship.cpp
2
4
5 template <typename T> void myFriendFunction(T);
6 template <typename U> class MyFriend;
7
8
9 class GrantingFriendshipAsClass{
10
11 template <typename U> friend void myFriendFunction(U);
13
14 private:
15 std::string secret{"My secret from GrantingFriendshipAsClass."};
16
https://www.educative.io/courses/generic-templates-in-cpp/7nj8Kn01Nr8 1/7
16
17 };
18
21
22 template <typename U> friend void myFriendFunction(U);
24
25 private:
26 std::string secret{"My secret from GrantingFriendshipAsClassTemplate."};
27
28 };
29
31 void myFriendFunction(T){
32 GrantingFriendshipAsClass myFriend;
33 std::cout << myFriend.secret << std::endl;
34
35 GrantingFriendshipAsClassTemplate<double> myFriend1;
36 std::cout << myFriend1.secret << std::endl;
37 }
38
40 class MyFriend{
41 public:
42 MyFriend(){
45
46 GrantingFriendshipAsClassTemplate<T> myFriend1;
48 }
49 };
50
51 int main(){
52
54
55 int a{2011};
56 myFriendFunction(a);
57
58 MyFriend<double> myFriend;
59
61
62 }
63
Output 1.58s
My secret from GrantingFriendshipAsClass.
My secret from GrantingFriendshipAsClassTemplate.
Explanation#
In the above example, we have created a function myFriendFunction and a
class MyFriend. We have defined two classes: GrantingFriendshipAsClass
and GrantingFriendshipAsClassTemplate . As the name mentioned as well, we
are using one class with template and one without a template. The class
MyFriend and the function myFriendFunction have access to the private
members of the other classes by using a friend keyword. We have defined a
private variable secret which is of a string type and can be called with the
object of myFriendFunction and MyFriend.
Example 2: Class Template Special

Friendship#
1 // templateClassTemplateSpecialFriendship.cpp
2
4
5 template <typename T> void myFriendFunction(T);
6 template <typename U> class MyFriend;
7
8
9 class GrantingFriendshipAsClass{
10
11 friend void myFriendFunction<>(int);
12 friend class MyFriend<int>;
13
14 private:
15 std::string secret{"My secret from GrantingFriendshipAsClass."};
16
17 };
18
21
22 friend void myFriendFunction<>(int);
23 friend class MyFriend<int>;
24 friend class MyFriend<T>;
25
26 private:
27 std::string secret{"My secret from GrantingFriendshipAsClassTemplate."};
28
29 };
30
32 void myFriendFunction(T){
35
36 GrantingFriendshipAsClassTemplate<T> myFriend1;
38 }
39
41 class MyFriend{
42 public:
43 MyFriend(){
46
47 GrantingFriendshipAsClassTemplate<int> myFriendInt;
48 std::cout << myFriendInt.secret << std::endl;
49
50 GrantingFriendshipAsClassTemplate<T> myFriendT;
51 std::cout << myFriendT.secret << std::endl;
52 }
53 };
54
55 int main(){
56
58
59 int a{2011};
60 myFriendFunction(a);
61
61
62 MyFriend<int> myFriend;
63
65
66 }
67
Output 3.02s
Explanation#
This example is similar to example 1 with a small change; we have explicitly
stated the type of class template to int . Now, the class template is called for
int and also for any other type mentioned in the typename portion.
Example 3: Class Template Type

Friendship#
1 // templateClassTemplateTypeFriendship.cpp
2
4
6 class Bank{
7 std::string secret{"Import secret from the bank."};
8 friend T;
9 };
10
11 class Account{
12 public:
13 Account(){
14 Bank<Account> bank;
15 std::cout << bank.secret << std::endl;
16 }
17 };
18
19 int main(){
20
22
23 Account acc;
24
26 (/learn)
27 }
28
Output 1.28s
Import secret from the bank.
Explanation#
In the above code, we have created an Account class which contains the
Bank class object. We can access the Bank class member secret with the
help of friend . Now, the value stored in the secret is accessible in the
Account class.

Back Next
Friends - Exercise
Mark as Completed
Report an Issue
- Exercise
Problem Statement#
Make the public member elem a private member of Array . How does that
affect the implementation of the assignment operator?
3 #include <vector>
4
6 class Array{
7
8 public:
9 Array()= default;
10
14 elem.clear();
16 return *this;
17 }
18
20 // you need to make the `elem` private and then run it
22 };
23
24 template <typename T int N>
https://www.educative.io/courses/generic-templates-in-cpp/RL2W0DrZkLK 1/3
26 return N;
(/learn)
27 }
28
29
30 int main(){
31
32 // uncomment these line after the above implementation
33
34 /*
37
38 doubleArray = intArray;
39
40 */
41 }
Succeeded
Let’s move on to the solution review of this exercise in the next lesson.

Back Next
Mark as Completed
Report an Issue
- Solution
Let's have a look at the solution review of the last exercise in this lesson.
• Solution Review
• Explanation
Solution Review#
1 //templateClassTemplateMethods2.cpp
2
5 #include <vector>
6
8 (/learn)
class Array{
9
10 public:
12
13 template <typename T2, int M> friend class Array;
14
18 elem.clear();
20 return *this;
21 }
22
24
25 private:
26
https://www.educative.io/courses/generic-templates-in-cpp/gknPmvqxGYY 1/3
28 };
29
32 return N;
33 }
34
35 int main(){
36
39
40 doubleArray = intArray;
41
42 }
Succeeded
Explanation#
In the above code, we have created an Array class in which we have defined
a method = which copies the entries of integer array to double array.
getSize function returns the size of the array passed in the argument.
Template classes with different template arguments are considered of
different types, which is why Array<int, 10> and Array<double, 10> aren’t
able to access each other. And that’s why the Array template class has to be
declared a friend of itself.
To access the private elem of arr in line 19, class Array is a friend of the
class Array in line 13.
In the next lesson, we’ll study dependent names.

In the next lesson, we ll study dependent names.

Back Next
- Exercise Dependent Names
Mark as Completed
Report an Issue
3/20/22, 6:10 PM Dependent Names - Generic Programming Templates in C++
Dependent Names
In this lesson, we'll study dependent names.
• Dependent Names
• Two-phase name lookup

• The Dependent Name is a Type typename
• The Dependent Name is a Template .template
Dependent Names#
A dependent name is essentially a name that depends on a template
parameter. A dependent name can be a type, a non-type, or a template-
template parameter.
If you use a dependent name in a template declaration or template

definition, the compiler has no idea, whether this name refers to a type, a
non-type, or a template parameter. In this case, the compiler assumes that
the dependent name refers to a non-type, which may be wrong.
Let’s have a look at the example of dependent names:
2 struct X : B<T> // "B<T>" is dependent on T
3 {
4 typename T::A* pa; // "T::A" is dependent on T
5 void f(B<T>* pb) {
6 static int i = B<T>::i; // "B<T>::i" is dependent on T
https://www.educative.io/courses/generic-templates-in-cpp/g7nlq1oqEq9 1/4
7 pb->j++; // "pb->j" is dependent on T
8 }
9 };
T is the template parameter. The names B<T> , T::A , B<T> , B<T>::i , and pb-
>j are dependent.
Two-phase name lookup#

Dependent names are resolved during template instantiation.
Non-dependent names are resolved during template definition.
A from a template parameter T is dependent, qualified name

T::A can be a
Type
Non-type
Template
The compiler assumes by default that T::A is a non-type.
The compiler has to be convinced that T::A is a type or a template.
The Dependent Name is a Type typename #

1 template <typename T> void test(){
2 std::vector<T>::const_iterator* p1; // ERROR
3 typename std::vector<T>::const_iterator* p2; //OK
4 }
Without typename like in line 3, the expression in line 2 would be interpreted

as multiplication.
(/learn)
The Dependent Name is a Template

.template #
2 struct S{
3 template <typename U> void func(){}
4 }
5
7 void func2(){
8 S<T> s;
9 s.func<T>(); // ERROR
10 s.template func<T>(); // OK
11 }
Compare lines 9 and 10. When the compiler reads the name s.func (line 9),
it decides to interpret it as non-type. This means, the < sign stands in this
case for the comparison operator but not opening square bracket of the
template argument of the generic method func . To help the parser, you have
to specify that s.func is a template like in line 10: s.template func .
To learn more about dependent names, click here

(https://en.cppreference.com/w/cpp/language/parameter_pack).
In the next lesson, we’ll look at an example of dependent names.

Back Next
- Solution - Example
Mark as Completed
Report an Issue
- Example
Let's have a look at an example of dependent names in this lesson.
• Example: Template Lookup
• Explanation
Example: Template Lookup#

1 // templateLookup.cpp
2
4
5 void g(double) { std::cout << "g(double)\n"; }
6
7 template<class T>
8 struct S {
9 void f() const {
10 g(1); // non-dependent
11 }
12 };
13
14 void g(int) { std::cout << "g(int)\n"; }
15
16 int main(){
17 g(1); // calls g(int)
18
19 S<int> s;
20 s.f(); // calls g(double)
21 }
(/learn)
22
https://www.educative.io/courses/generic-templates-in-cpp/myoYOYGNBnR 1/3
Output 1.69s
g(int)
g(double)
Explanation#
If we access the defined functions g with double or int type object, they
work fine. We have created the struct object S of int type in line 19.
When we try to access the g function then it follows the same order and
calls, the g with a double type parameter is defined first. The call to g() on
line 17 calls the g(int) version and the call to g() through the call to f()
on line 20 calls g(double) .
In this chapter, we have learned about the details of templates. In the next
chapter, we’ll familiarize ourselves with the techniques used in templates.
Let’s start with Automatic Return type in the next lesson.

Back Next
Dependent Names Automatic Return Type
Mark as Completed
Report an Issue
3/20/22, 6:11 PM Automatic Return Type - Generic Programming Templates in C++
Automatic Return Type

In this lesson, we'll look at the technique that deduces return type automatically.
• Automatic Return Type
• Automatic Return Type: C++14
Automatic Return Type#

A function template is automatically able to deduce their return type.
template <typename T1, typename T2>
auto add(T1 fir, T2 sec) -> decltype( fir + sec ) {
return fir + sec;
The automatic return type deduction is typically used for function

templates but can also be applied to non-template functions.
Rules:
auto : introduces the syntax for the delayed return type
auto : auto type deduction is based on the function template argument

deduction. Function template argument deduction (decays
(https://en.cppreference.com/w/cpp/types/decay)). So it means auto
https://www.educative.io/courses/generic-templates-in-cpp/g7DnLgVkjBk 1/4
does not return the exact type but a decayed type such as for template
argument deduction
(https://www.educative.io/collection/page/10370001/5799373671759872/
6076820204027904)
decltype : declares the return type
The alternative function syntax is obligatory
The C++11 syntax for automatically deducing the return type breaks
the crucial principle of software development: DRY. DRY stands for
Don’t Repeat Yourself
(https://en.wikipedia.org/wiki/Don%27t_repeat_yourself).
(/learn)
Automatic Return Type: C++14#

A function template is automatically able to deduce their return type.
auto add(T1 fir, T2 sec){
return fir + sec;
Rules
auto: introduces the syntax for the delayed return type
decltype : declares the return type
The alternative function syntax is obligatory.
With the expression decltype(auto) , auto uses the same rules to determine
the type as decltype . This means, in particular, no decay
(https://en.cppreference.com/w/cpp/types/decay) takes place.
Both declarations are identical.
decltype(expr) v= expr;
decltype(auto) v= expr;
The syntax also applies for the automatic return type of a function
template.
decltype(auto) add(T1 fir, T2 sec){
return fir + sec;
}
When a function template has more than one return statements, all
return statements must have the same type.
In the next lesson, we’ll study an example of automatic return type

deduction.

Back Next
- Example - Example
Mark as Completed
Report an Issue
- Example
In this lesson, we'll look at an example of the automatic return type.
• Example: Automatic Template Return Type
• Explanation
Example: Automatic Template Return

Type#
1 // templateAutomaticReturnType.cpp
2
5
6 template<typename T1, typename T2>
7 auto add(T1 first, T2 second) -> decltype(first + second){
8 return first + second;
9 }
10
11 int main(){
12
14
15 std::cout << "add(1, 1)= " << add(1,1) << std::endl;
16 std::cout << "typeid(add(1, 1)).name()= " << typeid(add(1, 1)).name() << std::endl
17
19
20 std::cout << "add(1, 2.1)= " << add(1,2.1) << std::endl;
21 std::cout << "typeid(add(1, 2.1)).name()= " << typeid(add(1, 2.1)).name() << std::
22
24
https://www.educative.io/courses/generic-templates-in-cpp/NEwww5qlQ0m 1/3
24
25 (/learn)
std::cout << "add(1000LL, 5)= " << add(1000LL,5) << std::endl;
26 std::cout << "typeid(add(1000LL, 5)).name()= " << typeid(add(1000LL, 5)).name() <<
27
29 }
Output 1.25s
add(1, 1)= 2
typeid(add(1, 1)).name()= i
add(1, 2.1)= 3.1
typeid(add(1, 2.1)).name()= d
add(1000LL, 5)= 1005
typeid(add(1000LL, 5)).name()= x
Explanation#
The example has a function add which takes two arguments and returns
their sum. The return type of the function is deduced by the compiler by
applying the decltype operator on the sum of the arguments. The
expression typeid(add(1, 2.1)).name() such as in line 21 returns a string
representation of the type of result.

you instead of you applying to them See how
Back Next
Automatic Return Type - Exercise
Mark as Completed
Report an Issue
- Exercise
In this lesson, we'll solve an exercise related to the automatic return type.
Problem Statement#
The two algorithms gcdConditional and gcdCommon , in the given code, use the
type-traits library.
Study both algorithms.

Both algorithms determine their return type at compile-time.
Why is it not possible to use auto or decltype(auto) to automatically deduce

the return type?
2 #include <string>
5
6 template<typename T1, typename T2,
7 typename R = typename std::conditional <(sizeof(T1) < sizeof(T2)), T1, T2>:
8 R gcdConditional(T1 a, T2 b){
9 static_assert(std::is_integral<T1>::value, "T1 should be an integral type!");
11 if( b == 0 ){ return a; }
12 else{
13 return gcdConditional(b, a % b);
14 }
15 }
https://www.educative.io/courses/generic-templates-in-cpp/B8zMvKN4KJX 1/3
}
16
17
19 typename R = typename std::common_type<T1, T2>::type>
20 R gcdCommon(T1 a, T2 b){
23 if( b == 0 ){ return a; }
24 else{
25 return gcdCommon(b, a % b);
26 }
27 }
28
29 int main(){
30
31
32 (/learn)
33
34 std::cout << "gcdConditional(100, 10LL) = " << gcdConditional(100, 10LL) << std::e
35 std::cout << "gcdCommon(100, 10LL) = " << gcdCommon(100, 10LL) << std::endl;
36
37 std::conditional <(sizeof(int) < sizeof(long long)), int, long long>::type gcd1 =
38 auto gcd2 = gcdCommon(100, 10LL);
39
41
42 std::cout << "typeid(gcd1).name() = " << typeid(gcd1).name() << std::endl;
44
46
47 }
48
We’ll look at the solution review of this exercise in the next lesson.

Back Next
Mark as Completed
Report an Issue
- Solution
In this lesson, we'll look at the solution review of the last exercise.
• Solution Review
• Explanation
Solution Review#
1 // gcdVariation.cpp
2
6
8 typename R = typename std::conditional <(sizeof(T1) < sizeof(T2)), T1, T2>:
9 R gcdConditional(T1 a, T2 b){
12 if( b == 0 ){ return a; }
13 else{
14 return gcdConditional(b, a % b);
15 }
16 }
17
18
20 typename R = typename std::common_type<T1, T2>::type>
21 R gcdCommon(T1 a, T2 b){
24 if( b == 0 ){ return a; }
25 else{
26 return gcdCommon(b, a % b);
https://www.educative.io/courses/generic-templates-in-cpp/7nDBNLNglzB 1/3
27 }
28 }
29
30 int main(){
31
32
34
35 std::cout << "gcdConditional(100, 10LL) = " << gcdConditional(100, 10LL) << std::e
36 std::cout << "gcdCommon(100, 10LL) = " << gcdCommon(100, 10LL) << std::endl;
37
38 std::conditional <(sizeof(int) < sizeof(long long)), int, long long>::type gcd1 =
39 auto gcd2 = gcdCommon(100, 10LL);
40
42
45
47
48 }
49
(/learn)
Output 1.4s
gcdConditional(100, 10LL) = 10
gcdCommon(100, 10LL) = 10
typeid(gcd1).name() = i
typeid(gcd2).name() = x
Explanation#
In the above code, we have defined an automatic return type R which
returns data based on the data type passed in the function gcdCommon and
gcdConditional .
Automatic return type deduction can’t be used in both algorithms, because

both algorithms solve their job by recursion (lines 24 and 26). The critical
observation is that this recursion swaps their arguments. For example,
gcdCommon(a, b) invokes gcdCommon(b, a % b) . This recursion, therefore,
creates a function with different return types. Having a function with
different return types is not valid.
Let’s move on to template metaprogramming in the next lesson.

Back Next
- Exercise Template Metaprogramming
Mark as Completed
Report an Issue
3/20/22, 6:14 PM Template Metaprogramming - Generic Programming Templates in C++
Template Metaprogramming
In this lesson, we'll learn about template metaprogramming.
• Template Metaprogramming
• How this all started:

• Calculating at Compile-Time
• Type Manipulations
• Explanation
• Metadata and Metafunctions
• Functions vs Meta Functions

• Pure Functional Sublanguage
Template Metaprogramming#
(/learn)
How this all started:#

1994 Erwin Unruh discovered template metaprogramming by accident.
His program failed to compile but calculated the first 30 prime numbers
at compile-time.
To prove his point, he used the error messages to display the first 30
prime numbers.
Let’s have a look at the screenshot of the error:
https://www.educative.io/courses/generic-templates-in-cpp/7n0N88x2oEA 1/8
We have highlighted the important parts in red. We hope you can see the
pattern. The program calculates at compile-time the first 30 prime numbers.
This means template instantiation can be used to do math at compile-time. It
gets even better. Template metaprogramming is Turing-complete
(https://simple.wikipedia.org/wiki/Turing_complete) and can, therefore, be
used to solve any computational problem. Of course, Turing-completeness
holds only in theory for template metaprogramming because the recursion
depth (at least 1024 with C++11) and the length of the names which are
generated during template instantiation provide some limitations.
Calculating at Compile-Time#
The Factorial program is the Hello World of template metaprogramming.
1 template <int N>
2 struct Factorial{
3 static int const value= N * Factorial<N-1>::value;
4 };
5
6 template <> struct Factorial<1>{
8 };
9
10 std::cout << Factorial<5>::value << std::endl;
11 std::cout << 120 << std::endl;
The call Factorial<5>::value in line 10 causes the instantiation of the

primary or general template in line 3. During this instantiation,
Factorial<4>::value will be instantiated. This recursion will end if the fully
specialized class template Factorial<1> (line 6) kicks in as the boundary
condition. Maybe, you like it more pictorial.
The following picture shows this process.
Assembler Instructions
From the assemblers point of view, the Factorial<5>::value boils down to

the constant 0x78 , which is 120.
mov 0x78, %esi
mov 0x601060, %edi
...
mov 0x78, %esi
mov 0x601060, %edi
...
Type Manipulations#
Manipulating types at compile-time is typically for template
metaprogramming
metaprogramming.
2 struct RemoveConst{
3 typedef T type;
4 };
5
7 struct RemoveConst<const T>{
8 typedef T type;
9 };
10
11 int main(){
12 std::is_same<int, RemoveConst<int>::type>::value; // true
13 std::is_same<int, RemoveConst<const int>::type>::value; // true
14 }
Explanation#
In the code, we have defined the class template removeConst in two versions.
We have implemented removeConst the way std::remove_const is probably
implemented in the type-traits
229176066048) library.
std::is_same from the type-traits library helps us to decide at compile-time

if both types are the same. In case of removeConst<int> , the first or general
class template kicks in; in case of removeConst<const int> , the partial
specialization for const T applies. The key observation is that both class
templates return the underlying type in lines 3 and 8, therefore, the
constness is removed.
This kind of technique, which is heavily used in the type-traits library, is a

compile-time if on types.
To jump into more details of type traits click here

229176066048).
Metadata and Metafunctions#

At compile-time, we speak about metadata and metafunctions instead of
data and functions.
Metadata: Types and integral types that are used in metafunctions.
Metafunctions: Functions that are executed at compile-time. Class

templates are used to implement metafunctions.
Return their value by ::value .
template <>
struct Factorial<1>{
static int const value = 1;
};
Return their type by ::type .
struct RemoveConst<const T>{
typedef T type;
};
Functions vs Meta Functions#

From the conceptual view, it helps a lot to compare functions and
metafunctions.
Characteristics Functions Metafunctions
Call power(2,10) Power<2,10>::value
Characteristics Functions Metafunctions
Execution Time Runtime Compile-time
Arguments Function arguments Template arguments
Arguments and re- Types, non-types and

Arbitrary values
turn value templates
Implementation Callable Class template
Data Mutable Immutable
Modification Data can be modified New data are created
State Has state Has no state
What does the table above mean for a concrete function and a concrete
metafunction?
Function Metafunction
int power(int m, int n){
template<int m, int n>
int r = 1;
struct Power{
for(int k=1; k<=n; ++k static int const value = m * Power<m,
){
n-1>::value;
r *= m;
};
}

return r;
template<int m>
}
struct Power<m, 0>{

static int const value =1;

};
Function arguments go into round ( ) braces and template arguments

go into sharp <> braces.
int main(){
std::cout << power(2, 10) << std::endl; // 1024
std::cout << Power<2, 10>::value << std::endl; // 1024
Pure Functional Sublanguage#

Template metaprogramming is
an embedded pure functional language in the imperative language
C++.
Turing-complete. Turing-complete means, that all can be calculated
what is calculatable.
an intellectual playground for C++ experts.
the foundation for many boost (https://www.boost.org/) libraries.
The template recursion depth is limited.
C++03: 17
C++11: 1024
In the next lesson, we’ll look at a few examples of template

metaprogramming.

Back Next
Mark as Completed
Report an Issue
- Examples
We'll look at a few examples of template metaprogramming in this lesson.
• Example 1: Template Prime Number
• Explanation
• Example 2: Template Type Manipulation
• Explanation
• Example 3: Template Power
• Explanation
Example 1: Template Prime Number #

2 // Prime number computation by Erwin Unruh
3
4 template <int i> struct D { D(void*); operator int(); };
5
6 template <int p, int i> struct is_prime {
7 enum { prim = (p==2) || (p%i) && is_prime<(i>2?p:0), i-1> :: prim };
8 };
9
10 template <int i> struct Prime_print {
11 Prime_print<i-1> a;
12 enum { prim = is_prime<i, i-1>::prim };
13 void f() { D<i> d = prim ? 1 : 0; a.f();}
14 };
15
16 template<> struct is_prime<0,0> { enum {prim=1}; };
17 template<> struct is_prime<0,1> { enum {prim=1}; };
18
19 template<> struct Prime_print<1> {
20 enum {prim=0};
https://www.educative.io/courses/generic-templates-in-cpp/RLRgj3MoWpR 1/6
{p };
21 void f() { D<1> d = prim ? 1 : 0; };
22 };
23
24 #ifndef LAST
25 #define LAST 18
26 #endif
27
28 int main() {
29 Prime_print<LAST> a;
30 a.f();
31 }
32
Explanation #
This is the original prime number program by Erwin Unruh, which was the
starting point of template metaprogramming. Current compilers will not
produce the same output as the ancient compiler, which Erwin Unruh used
more than 20 years ago.
Example 2: Template Type

Manipulation #
1 // templateTypeManipulation.cpp
2
5
8 typedef T type;
9 };
10
13 typedef T type;
14 };
15
16
17 int main(){
18
20
21 std::cout << "std::is_same<int, RemoveConst<int>::type>::value: " << std::is_sam
22 std::cout << "std::is_same<int, RemoveConst<const int>::type>::value: " << std::
23
25
26 }
27
Output 2.48s
std::is_same<int, RemoveConst<int>::type>::value: true
std::is_same<int, RemoveConst<const int>::type>::value: true
Explanation #
The code uses the function std::is_same from the type-traits library.
std::is_same compares the type passed and returns at compile time if they
are the same. Thanks to the type-traits function, we can verify the
RemoveConst class template from the previous subsection
934744182784#type-manipulations).
Example 3: Template Power #

1 // templatePower.cpp
2
4
5 i t (i t i t ){
6 int r = 1;
7 for(int k=1; k<=n; ++k) r*= m;
8 return r;
9 }
10
11 template<int m, int n>
12 struct Power{
13 static const int value = Power<m,n-1>::value * m;
14 };
15
16 template<int m>
17 struct Power<m,0>{
18 static const int value = 1;
19 };
20
21 template<int n>
22 int power2(const int& m){
23 return power2<n-1>(m) * m;
24 }
25
26 template<>
27 int power2<1>(const int& m){
28 return m;
29 }
30
31 template<>
32 int power2<0>(const int&){
33 return 1;
34 }
35
36 int main(){
37
39
40 std::cout << "power(2,10): " << power(2,10) << std::endl;
41 std::cout << "power2<10>(2): " << power2<10>(2) << std::endl;
42 std::cout << "Power<2,10>::value: " << Power<2,10>::value << std::endl;
43
45 }
46
Output 1.36s
power(2,10): 1024
power2<10>(2): 1024
(/learn)
Power<2,10>::value: 1024
Explanation #
The program calculates 210 in three different variants.
power is a function in line 5
Power is a class template in line 12
power2 is a function template in line 22
The key question is: When is the function executed?
power runs at runtime
Power runs at compile-time
power2 runs at runtime and at compile-time too

the template argument is evaluated at compile-time
the function argument is evaluated at runtime

Back Next
Template Metaprogramming - Exercise
Mark as Completed
Report an Issue
- Exercise
Problem Statement#
Implement the metafunction to calculate the factorial of a number.
2
3 template <int N>
(/learn)
4 struct Factorial{
5 // Implement the functionality to run the code in the main
6 };
7
8 int main(){
9 // Uncomment the given code after implementing the Factorial function
10
11 // std::cout << Factorial<5>::value << std::endl;
12 }
13
In the next lesson, we’ll look at the solution of this exercise.
https://www.educative.io/courses/generic-templates-in-cpp/39QDv3WJAQr 1/2

Back Next
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/39QDv3WJAQr 2/2
- Solution
Let's have a look at a solution review for the last exercise.
• Solution Review
• Explanation
Solution Review#
1 // templateFactorial.cpp
2
4
5 template <int N>
6 struct Factorial{
7 static int const value = N * Factorial<N-1>::value;
8 };
9
10 template <>
11 struct Factorial<1>{
13 };
14
15 int main(){
16 std::cout << Factorial<10>::value << std::endl;
17 }
18
https://www.educative.io/courses/generic-templates-in-cpp/YQM4OWvwRg9 1/3
Output 1.39s
3628800
(/learn)
Explanation#
The Factorial function calls itself recursively by storing the result of each
iteration in value . For the base case, the result is multiplied by 1 in line 12
and the answer is stored in value.
To be precise, what looks like a recursion, like in the case of Factorial ,

is not a recursion. Each invocation of Factorial<N> triggers a new
instantiation Factorial<N-1> in line 7.
In the next lesson, we’ll discuss type traits.

Back Next
- Exercise Type-Traits Overview
Mark as Completed
Mark as Completed
Report an Issue
3/20/22, 6:17 PM Type-Traits Overview - Generic Programming Templates in C++
Type-Traits Overview
In this lesson, we'll study the type traits library and its goals along with type-checks. This
section could only provide an overview of the many functions of the type-traits library.
• Type-Traits Library
• Type-Traits: Goals
• Optimization
• Correctness
• Type Checks
Type-Traits Library #
Type-traits enable type checks, type comparisons, and type modifications at
compile-time.
Below are some applications of template metaprogramming:
Programming at compile-time
Programming with types and values
Compiler translates the templates and transforms it in C++ source code
We need to add a type_traits library in the header to enable all the

functions present in the library.
#include <type_traits>
https://www.educative.io/courses/generic-templates-in-cpp/m7Ny81pnZ5p 1/6
Type-Traits: Goals #
If you look carefully, you’ll see that type-traits have a significant
optimization potential. In the first step, type-traits help to analyze the code
at compile-time and in the second step, to optimize the code based on that
analysis. How is that possible? Depending on the type of variable, a faster
variant of an algorithm will be chosen.
Optimization #
Code that optimizes itself. Depending on the type of a variable another
code will be chosen.
Optimized version of std::copy , std::fill , or std::equal is used so
that algorithms can work on memory blocks.
The optimized version of operations happens on all the elements in a
container in one step and not on each element individually.
Correctness #
Type checks will be performed at compile-time.
Type information, together with static_assert , defines the
requirements for the code.
With the concepts
(https://www.educative.io/collection/page/10370001/5799373671759872/
5589499255980032) in C++20, the correctness aspect of the type-traits
becomes less important.
Type Checks #
C++ has 14 primary type categories. They are complete and orthogonal. This
th t ht i b
https://www.educative.io/courses/generic-templates-in-cpp/m7Ny81pnZ5p f tl t t Th h k 2/6
means, that each type is a member of exactly one type category. The check
for the type categories is independent of the type qualifiers const or
volatile .
Let’s have a look at these categories syntactically:
1 template <class T> struct is_void;
2 template <class T> struct is_integral;
3 template <class T> struct is_floating_point;
4 template <class T> struct is_array;
5 template <class T> struct is_pointer;
6 template <class T> struct is_reference;
7 template <class T> struct is_member_object_pointer;
8 template <class T> struct is_member_function_pointer;
9 template <class T> struct is_enum;
10 template <class T> struct is_union;
11 template <class T> struct is_class;
12 template <class T> struct is_function;
13 template <class T> struct is_lvalue_reference;
14 template <class T> struct is_rvalue_reference;
We can divide the type-traits into smaller sets for simplicity.
Primary type category ( ::value )
std::is_pointer<T>,
std::is_integral<T>,
std::is_floating_point<T>
Composed type category ( ::value )
std::is_arithmetic<T>,
std::is_object<T>
Type comparisons ( ::value )
std::is_same<T,U>,
std::is_base_of<Base,Derived>,
std::is convertible<From To>
std::is_convertible<From,To>
Type transformation ( ::type )
std::add_const<T>,
std::remove_reference<T>
std::make_signed<T>,
std::make_unsigned<T>
Others ( ::type )
std::enable_if<bool,T>
std::conditional<bool,T,F>
std::common_type<T1, T2, T3, ... >
(/learn)
The above-mentioned functions, from the type-traits, give only a rough idea
of their power. To learn more about type checks, click here
(https://en.cppreference.com/w/cpp/header/type_traits). The above-
mentioned functions are available on the given link with more detail.
1 // removeConst.cpp
2
4 #include <string>
6
7 namespace rgr{
8
9 template<class T, class U>
10 struct is_same : std::false_type {};
11
12 template<class T>
13 struct is_same<T, T> : std::true_type {};
14
15 template< class T >
16 struct remove_const{
17 typedef T type;
18 };
19
20 template< class T >
21 struct remove_const<const T> {
22 typedef T type;
23 };
23 };
24 }
25
26
27 int main(){
28
30
31 std::cout << std::is_same<int,std::remove_const<const int>::type>::value << std::e
32 std::cout << rgr::is_same<int,rgr::remove_const<const int>::type>::value << std::e
33
34 typedef rgr::remove_const<double>::type myDouble;
35 std::cout << rgr::is_same<double,myDouble>::value << std::endl;
36
37 typedef rgr::remove_const<const std::string>::type myString;
38 std::cout << rgr::is_same<std::string,myString>::value << std::endl;
39
40 typedef rgr::remove_const<std::add_const<int>::type>::type myInt;
41 std::cout << rgr::is_same<int,myInt>::value << std::endl;
42
44
45 }
46
We have implemented is_same and remove_const in the namespace rgr .

This corresponds to the type-traits library. For simplicity reason, we use the
static constants std::false_type and std::true_type (lines 10 and 13).
Thanks to the base class std::false_type , the class template has a member
value. Respectively for std::true_type . The key observation of the class
template is_same is to distinguish the general template (lines 9 and 10) from
the partially specialized template (line 12 and 13). The compiler will use the
partially specialized template if both template arguments have the same
type. The partially specialized template, as opposed to the general template,
has only one type parameter.
Our reasoning for the class template remove_const is similar. The general
template returns, via its member type , exactly the same type; the partially
specialized template returns the new type after removing the const
property (line 22) The compiler will choose the partially specialized
property (line 22). The compiler will choose the partially specialized
template if its template argument is const .
The rest is quickly explained. In lines 31 and 32, we used the functions of the
type-traits library and our own versions. We declared a typedef mydouble
(line 34), a type myString (line 37), and a type myInt (line 40). All types are
non-constant.
In the next lesson, we’ll study type-traits correctness and their optimization
with the help of an example.

Back Next
- Solution Type-Traits (Correctness and Optimiza…
Mark as Completed
Report an Issue
3/20/22, 6:19 PM Type-Traits (Correctness and Optimization) - Generic Programming Templates in C++
Type-Traits (Correctness and

Optimization)
In this lesson, we'll study type-traits correctness and their optimization using a gcd (greatest
common divisor) algorithm along with fill and equal (type-trait features).
• Correctness
• gcd - The First

• gcd - The Second
• gcd - The Third

• The Smaller Type
• The Common Type
• gcd - The Fourth

• Type-Traits: Performance
• Type-Traits fill
• Type-Traits std::equal
Correctness#
The reason we use type-traits is correctness and optimization. Let’s start
with correctness. The idea is to implement generic gcd algorithms, step by
step, to make it more type-safe with the help of the type-traits library.
gcd The First#

https://www.educative.io/courses/generic-templates-in-cpp/qA1OYzn7nn0 1/16
gcd - The First#
Our starting point is the euclid algorithm

(https://en.wikipedia.org/wiki/Greatest_common_divisor) to calculate the
greatest common divisor of two numbers.
It’s quite easy to implement the algorithm as a function template and feed it
with various arguments. Let’s start!
2
4 T gcd(T a, T b){
5 if( b == 0 ) return a;
6 else return gcd(b, a % b);
7 }
8
9 int main(){
10 std::cout << gcd(100, 10) << std::endl; // 10
13 std::cout << gcd(3.5, 4.0) << std::endl; // ERROR
14 std::cout << gcd("100", "10") << std::endl; // ERROR
15 std::cout << gcd(100, 10L) << std::endl; // ERROR
16 }
Output 1.35s
main.cpp: In function 'int main()':
main.cpp:15:28: error: no matching function for call to 'gcd(int, long int)'
15 | std::cout << gcd(100, 10L) << std::endl; // ERROR
| ^
main.cpp:4:3: note: candidate: 'template<class T> T gcd(T, T)'
4 | T gcd(T a, T b){
| ^~~
main.cpp:4:3: note: template argument deduction/substitution failed:
main.cpp:15:28: note: deduced conflicting types for parameter 'T' ('int'

main.cpp:15:28: note: deduced conflicting types for parameter T ( int
and 'long int')
15 | std::cout << gcd(100, 10L) << std::endl; // ERROR
| ^
main.cpp: In instantiation of 'T gcd(T, T) [with T = double]':
main.cpp:13:28: required from here
main.cpp:6:24: error: invalid operands of types 'double' and 'double' to

binary 'operator%'
6 | else return gcd(b, a % b);
| ~~^~~
main.cpp: In instantiation of 'T gcd(T, T) [with T = const char*]':
main.cpp:6:24: error: invalid operands of types 'const char*' and 'const

char*' to binary 'operator%'
The function template has two serious issues.
First, it is too generic. The function template accepts doubles (line 13)
and C strings (line 14). But it makes no sense to determine the greatest
common divisor of both data types. The modulo operation ( % ) for the
double and the C string values fails in line 6. But that’s not the only
issue.
Second, gcds depend on one type parameter, T . This shows the function
template signature gcd(T a, T b)) . a and b have to be of the same
type T . There is no conversion for type parameters. Therefore, the
instantiation of gcd with an int type and a long type (line 15) fails.
gcd - The Second#

We can ignore the rest of the examples below where both arguments have to
be positive numbers. The static_assert operator and the predicate
std::is_integral<T>::value will help us to check at compile-time whether T
is an integral type. A predicate always returns a boolean value.
3
3
5 T gcd(T a, T b){
6 static_assert(std::is_integral<T>::value, "T should be integral type!");
7 if( b == 0 ) return a;
8 else return gcd(b, a % b);
9 }
10 int main(){
11 std::cout << gcd(3.5, 4.0) << std::endl;
12 std::cout << gcd("100", "10") << std::endl;
13 }
Output 1.26s
main.cpp:6:38: error: static assertion failed: T should be integral type!
6 | static_assert(std::is_integral<T>::value, "T should be integral

type!");
| ^~~~~

binary 'operator%'
| ~~^~~
main.cpp:6:38: error: static assertion failed: T should be integral type!
6 | static_assert(std::is_integral<T>::value, "T should be integral

type!");
| ^~~~~

| ~~^~~
Great. We have solved the first issue of the gcd algorithm. The compilation
will not fail by accident because the modulo operator is not defined for a
double value and a C string The compilation fails because the assertion in
double value and a C string. The compilation fails because the assertion in
line 6 will not hold true. The subtle difference is that we now get an exact
error message and not a cryptic output of a failed template instantiation as
in the first example.
But what about the second issue. The gcd algorithm should accept
arguments of a different type.
gcd - The Third#

That’s no big deal. But wait, what should the type of result be?
3
5 ??? gcd(T1 a, T2 b){
6 static_assert(std::is_integral<T1>::value, "T1 should be integral!");
8 if( b == 0 )
9 return a;
10 else
11 return gcd(b, a % b);
12 }
13
14 int main(){
15 std::cout << gcd(100, 10L) << std::endl;
16 }
The three questions marks in line 5 show the core of the issue. Should the
first type or the second type be the return type of the algorithm? Or should
the algorithm derive a new type from the two arguments? The type-traits
library comes to the rescue. We will present two variations.
The Smaller Type#

A good choice for the return type is to use the smaller of both types.
Therefore, we need a ternary operator at compile-time. Thanks to the type-
traits library, we have it. The ternary function std::conditional operates on
types and not on values. That’s because we apply the function at compile-
time. So, we have to feed std::conditional with the right constant
expression and we are done. std::conditional<(sizeof(T1) < sizeof(T2)),
T1, T2>::type will return, at compile-time, T1 if T1 is smaller than T2 ; it
will return T2 if T2 is not smaller than T1 .
Let’s apply the logic.
5 typename std::conditional<(sizeof(T1)<sizeof(T2)),T1,T2>::type gcd(T1 a, T2 b){
8 if( b == 0 )
9 return a;
10 else
12 }
13
14 int main(){
15 std::cout << gcd(100,10LL) << std::endl;
16 auto res= gcd(100,10LL);
17 std::conditional<(sizeof(long long)<sizeof(long)), long long, long>::type res2=gcd
18 std::cout << typeid(res).name() << std::endl; // i
19 std::cout << typeid(res2).name() << std::endl; // l
21 }
Output 1.55s
10
The critical line of the program is in line 5 with the return type of the gcd
algorithm. Of course, the algorithm can also deal with template arguments
of the same type. What about line 15? We used the number 100 of type int
and the number 10 of type long long int . The result for the greatest
common divisor is 10. Line 17 is extremely ugly. We have to repeat the
expression std::conditional <(sizeof(100) < sizeof(10LL)), long long,
long>::type to determine the right type of the variable res2 . Automatic type
deduction with auto comes to my rescue (line 16). The typeid operator in
line 18 and 19 shows that the result type of the arguments of type int and
long long int is int ; that the result type of the types long long int and
long int is long int .
The Common Type#

Now to the second variation. Often it is not necessary to determine the
smaller type at compile-time but to determine the type to which all types can
implicitly be converted to. std::common_type can handle an arbitrary
number of template arguments. To say it more formally. std::common_type is
a variadic template
034837340160).
2 #include <type traits>
4
6 typename std::common_type<T1, T2>::type gcd(T1 a, T2 b){
9 if( b == 0 ){
10 return a;
11 }
12 else{
14 }
15 }
16 int main(){
17 std::cout << typeid(gcd(100, 10)).name() << std::endl; // i
18 std::cout << typeid(gcd(100, 10L)).name() << std::endl; // l
19 std::cout << typeid(gcd(100, 10LL)).name() << std::endl; // x
20 }
Output 1.45s
The only difference to the last implementation is that std::common_type in

line 6 determines the return type. We ignored the results of the gcd
algorithm in this example because we’re more interested in the types of
results. With the argument types int and int we get int ; with the
argument types int and long int we get long int , and with int and long
long int we get long long int .
gcd - The Fourth#

But that’s not all. std::enable_if from the type-traits library provides a very
interesting variation. What the previous implementations have in common
is that they will check in the function body if the arguments are of integral
types or not. The key observation is that the compiler always tries to
instantiate the function templates but sometimes fails. We know the result.
If the expression std::is_integral returns false , the instantiation will fail.
That is not the best way. It would be better if the function template is only
available for the valid types. Therefore, we put the check of the function
template from the template body to the template signature.
3
5 typename std::enable_if<std::is_integral<T1>::value,T1 >::type= 0,
6 typename std::enable_if<std::is_integral<T2>::value,T2 >::type= 0,
7 typename R = typename std::conditional<(sizeof(T1) < sizeof(T2)),T1,T2>::type>
8 R gcd(T1 a, T2 b){
9 if( b == 0 ){
10 return a;
11 }
12 else{
14 }
15 }
16
17 int main(){
18 std::cout << "gcd(100, 10)= " << gcd(100, 10) << std::endl;
19 std::cout << "gcd(100, 33)= " << gcd(100, 33) << std::endl;
20 std::cout << "gcd(3.5, 4.0)= " << gcd(3.5, 4.0) << std::endl;
21 }
Output 1.99s
main.cpp:20:49: error: no matching function for call to 'gcd(double, double)'
20 | std::cout << "gcd(3.5, 4.0)= " << gcd(3.5, 4.0) << std::endl;
| ^
main.cpp:8:3: note: candidate: 'template<class T1, class T2, typename

std::enable_if<std::is_integral<_Tp>::value, T1>::type <anonymous>, typename
std::enable_if<std::is_integral<T2>::value, T2>::type <anonymous>, class R> R
gcd(T1, T2)'
8 | R gcd(T1 a, T2 b){
| ^~~
main.cpp:8:3: note: template argument deduction/substitution failed:
main.cpp:5:67: error: no type named 'type' in 'struct std::enable_if<false,

double>'
5 | typename std::enable_if<std::is_integral<T1>::value,T1 >::type= 0,
| ^
Lines 5 and 6 are the key lines of the new program. The expression
std::is_integral determines whether the type parameter T1 and T2 are
integrals. If T1 and T2 are not integrals, and therefore they return false ,
we will not get a template instantiation. This is the decisive observation.
If std::enable_if returns true as the first parameter, std::enable_if will

have a public member typedef type. This type is used in lines 5 and 6. If
std::enable_if returns false as first parameter, std::enable_if will have
no member type . Therefore, lines 5 and 6 are not valid. This is not an error
but a common technique in C++: SFINAE
(https://en.cppreference.com/w/cpp/language/sfinae). SFINAE stands for
Substitution Failure Is Not An Error. Only the template for exactly this type
will not be instantiated and the compiler tries to instantiate the template in
another way.
Type-Traits: Performance#
The idea is quite straightforward and is used in current implementations of

the Standard Template Library (STL). If the elements of a container are
simple enough, the algorithm of the STL like std::copy , std::fill , or
std::equal will directly be applied on the memory area. Instead of using
std::copy to copy the elements one by one, all is done in one step. Internally,
C functions like memcmp , memset , memcpy , or memmove are used. The small
difference between memcpy and memmove is that memmove can deal with
overlapping memory areas.
The implementations of the algorithm std::copy , std::fill , or std::equal

use a simple strategy. std::copy is like a wrapper. This wrapper checks if the
element is simple enough. If so, the wrapper will delegate the work to the
optimized copy function. If not, the general copy algorithm will be used. This
one copies each element after one another. To make the right decision, the
functions of the type-traits library will be used if the elements are simple
enough.
The graphic shows this strategy once more:
Type-Traits: Optimization
Type-Traits fill #
std::fill assigns each element, in the range, a value. The listing shows a
simple implementation which is based on the GCC implementation.
1 // fill.cpp
2
3 #include <cstring>
4 #include <chrono>
7
8 namespace my{
9 (/learn)
10 template <typename I, typename T, bool b>
11 void fill_impl(I first, I last, const T& val, const std::integral_constant<bool, b
12 while(first != last){
13 *first = val;
14 ++first;
15 }
15 }
16 }
17
19 void fill_impl(T* first, T* last, const T& val, const std::true_type&){
20 std::memset(first, val, last-first);
21 }
22
23 template <class I, class T>
24 inline void fill(I first, I last, const T& val){
25 // typedef std::integral_constant<bool,std::has_trivial_copy_assign<T>::value &&
26 typedef std::integral_constant<bool,std::is_trivially_copy_assignable<T>::value
27 fill_impl(first, last, val, boolType());
28 }
29 }
30
31 const int arraySize = 100000000;
32 char charArray1[arraySize]= {0,};
34
35 int main(){
36
38
39 auto begin= std::chrono::system_clock::now();
40 my::fill(charArray1, charArray1 + arraySize,1);
41 auto last= std::chrono::system_clock::now() - begin;
42 std::cout << "charArray1: " << std::chrono::duration<double>(last).count() << " s
43
44 begin= std::chrono::system_clock::now();
45 my::fill(charArray2, charArray2 + arraySize, static_cast<char>(1));
46 last= std::chrono::system_clock::now() - begin;
48
50
51 }
Output 2.35s
charArray1: 0.494381 seconds
my::fill make in line 27 the decision which implementation of

my::fill_impl is applied. To use the optimized variant, the elements should
have a compiler generated copy assignment operator
std::is_trivially_copy_assignable<T> and should be 1 byte large: sizeof(T)
== 1 . The function std::is_trivially_copy_assignable is part of the type-
traits library. The first call my::fill(charArray1, charArray1 + arraySize,1);
has the last parameter 1, which is an int that has a size of 4 bytes. That is
why, the test on line 26 evaluates to false.
Our GCC calls the function std::is_trivially_copy_assignable instead of

std::has_trivial_copy_assign . If we request with the keyword default from
the compiler the copy assignment operator, the operator will be trivial.
Type-Traits std::equal #
The following code snippet shows a part of the implementation of
std::equal in the GCC:
1 template<typename _II1, typename _II2>
2 inline bool __equal_aux(_II1 __first1, _II1 __last1, _II2 __first2){
3 typedef typename iterator_traits<_II1>::value_type _ValueType1;
4 typedef typename iterator_traits<_II2>::value_type _ValueType2;
5 const bool __simple = ((__is_integer<_ValueType1>::__value
6 || __is_pointer<_ValueType1>::__value )
7 && __is_pointer<_II1>::__value
8 && __is_pointer<_II2>::__value
9 &&__are_same<_ValueType1, _ValueType2>::__value
10 );
11 return std::__equal<__simple>::equal(__first1, __last1, __first2);
12 }
We have a different perception of __simple . To use the optimized variant of

std::equal , the container elements have to fulfill some assurances. The
elements of the container have to be of the same type (line 9) and have to be
elements of the container have to be of the same type (line 9) and have to be
an integral or a pointer (lines 5 and 6). In addition, the iterators have to be

pointers (lines 7 and 8).
If you would like to learn about the type-traits library in more detail, the
chapter on Type Traits (https://www.educative.io/courses/cpp-standard-
library-including-cpp-14-and-cpp-17/7nA7550oor8) in our course C++
Standard Library including C++ 14 & C++ 17
(https://www.educative.io/courses/cpp-standard-library-including-cpp-14-
and-cpp-17) is a great place to start.
In the next lesson, we’ll look at a couple of examples of type traits.

Back Next
Type-Traits Overview - Examples
Mark as Completed
Report an Issue
- Examples
Let's have a look at a couple of examples of type-traits.
• Example 1: Template fill
• Explanation
• Example 2: Template Type Manipulation
• Explanation
Example 1: Template fill#

1 // templatefill.cpp
2
3 #include <cstring>
4 #include <chrono>
7
8 namespace my{
9
10 template <typename I, typename T, bool b>
11 void fill_impl(I first, I last, const T& val, const std::integral_constant<bool, b
12 while(first != last){
13 *first = val;
14 ++first;
15 }
16 }
17
19 void fill_impl(T* first, T* last, const T& val, const std::true_type&){
20 std::memset(first, val, last-first);
21 }
22
23 template <class I class T>
https://www.educative.io/courses/generic-templates-in-cpp/mE9MYGjx5R0 1/5
23 template <class I, class T>
24 inline void fill(I first, I last, const T& val){
25 typedef std::integral_constant<bool,std::is_trivially_copy_assignable<T> ::value
26 fill_impl(first, last, val, boolType());
27 }
28 }
29
30 const int arraySize = 100000000;
33
34 int main(){
35
37
38 auto begin= std::chrono::system_clock::now();
39 my::fill(charArray1, charArray1 + arraySize, 1);
40 auto last= std::chrono::system_clock::now() - begin;
42
43 begin= std::chrono::system_clock::now();
44 my::fill(charArray2, charArray2 + arraySize, static_cast<char>(1));
45 last= std::chrono::system_clock::now() - begin;
47
49
50 }
51
Output 2.15s
Explanation#
In line 26, my::fill make the decision as to which implementation of
i li d T th
https://www.educative.io/courses/generic-templates-in-cpp/mE9MYGjx5R0 ti i d i t th l t h ld 2/5
my::fill_impl is applied. To use the optimized variant, the elements should

have a compiler generated copy assignment operator
std::is_trivially_copy_assignable<T> and should be 1 byte large: sizeof(T)
== 1 . The function std::is_trivially_copy_assignable is part of the type-
traits library.
If the expression boolType() in line 26 is true, the optimized version of

my::fill_impl in the lines 17 - 20 will be used. This variant fills in opposite
of the generic variant my::fill_impl (line 10-16) the entire memory area -
consisting of 100 million entries - with the value 1. sizeof(char) is 1.
What’s about the performance of the program? We compiled the program

(/learn)
with full optimization. The execution of the optimized variant is about 3
times faster on windows; about 20 times faster on Linux.
Example 2: Template Type

Manipulation#
1 // templateTypeManipulation.cpp
2
5
8 typedef T type;
9 };
10
13 typedef T type;
14 };
15
16
17 int main(){
18
20
21 std::cout << "std::is_same<int, RemoveConst<int>::type>::value: " << std::is_sam
22 td t " td i i t R
https://www.educative.io/courses/generic-templates-in-cpp/mE9MYGjx5R0
C t t i t t l " td3/5
22 std::cout << "std::is_same<int, RemoveConst<const int>::type>::value: " << std::
23
25
26 }
27
Output 2.01s
std::is_same<int, RemoveConst<int>::type>::value: true
std::is_same<int, RemoveConst<const int>::type>::value: true
Explanation#
The code above uses the function std::is_same from the type-traits
library. std::is_same compares the type passed in the function and the type
given in the function defined by us, and it returns true only when both
types are the same.
In the next lesson, we’ll solve exercise on type-traits.

Back Next
Type-Traits (Correctness and Optimiza… - Exercise
Mark as Completed
Report an Issue
- Exercise
Problem Statement#
Get used to the many functions of the type traits library.
The objective of this exercise is to get familiar with the type-traits library.
What we want to achieve here is to modify an int type by doing a, b, c and
d.
Modify an int type at compile-time.

(/learn)
Add const to it.
Remove const from it.

Compare it with a const int .
3
4 int main() {
5 // Write different functions of type traits here
6 std::cout << "";
7 }
https://www.educative.io/courses/generic-templates-in-cpp/3j44PVqNJwn 1/2
Let’s move on to the solution to this exercise in the next lesson.

Back Next
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/3j44PVqNJwn 2/2
- Solution
Let's have a look at the solution to the previous exercise in this lesson.
• Solution Review
• Explanation
Solution Review#
1 // templateTypeTraitsModifications.cpp
2
5
6 int main(){
7
9
10 std::cout << "std::is_const<std::add_const<int>::type>::value: " << std::is_const
11 std::cout << "std::is_const<std::remove_const<const int>::type>::value: " << std::
12
14 typedef std::add_const<int>::type myConstInt;
15 std::cout << "std::is_const<myConstInt>::value: " << std::is_const<myConstInt>::va
16 typedef const int myConstInt2;
17 std::cout << "std::is_same<myConstInt, myConstInt2>::value: " << std::is_same<myCo
18
20
21 int fir= 1;
22 int& refFir1= fir;
23 using refToIntType= typename std::add_lvalue_reference<int>::type;
24 refToIntType refFir2= fir;
25
26 std::cout << "(fir, refFi1r, refFir2): " << "(" << fir << ", " << refFir1 << ", "
https://www.educative.io/courses/generic-templates-in-cpp/xV5Rm5YB9Zl 1/3
27
28 fir= 2;
29
30 std::cout << "(fir, refFir1, refFir2): " << "(" << fir << ", " << refFir1 << ", "
31
33
34 }
35
Output 1.33s
std::is_const<std::add_const<int>::type>::value: true
std::is_const<std::remove_const<const int>::type>::value: false
std::is_const<myConstInt>::value: true
std::is_same<myConstInt, myConstInt2>::value: true
(/learn)
(fir, refFi1r, refFir2): (1, 1, 1)
(fir, refFir1, refFir2): (2, 2, 2)
Explanation#
Lines 10 and 11 determine at compile-time if the manipulated type is const .
myConstInt is a type alias for an const int . Line 17 shows, with the help of
the function std::is_same , that myConstInt and myConstInt2 are the same
types. An lvalue such as fir can only be bound to an lvalue reference . Line
24 shows this in a complicated way. Line 26 proves that refFir1 and
refFire2 are references.
In the next lesson, we’ll discuss constexpr.

Back Next
- Exercise constexpr
Mark as Completed
Report an Issue
3/20/22, 6:22 PM constexpr - Generic Programming Templates in C++
constexpr
In this lesson, we'll study constexpr.
• Constant Expressions
• constexpr - Variables and Objects

• Variables
• User-Defined Types
• Example
• Functions
• Examples
• Functions with C++14
• constexpr Functions in C++14
• Example
• Template Metaprogramming vs constexpr Functions
Constant Expressions#
You can define, with the keyword constexpr , an expression that can be
evaluated at compile-time. constexpr can be used for variables, functions,
and user-defined types. An expression that is evaluated at compile-time has
a lot of advantages.
A constant expression
can be evaluated at compile-time.
gives the compiler deep insight into the code

https://www.educative.io/courses/generic-templates-in-cpp/B6WpXmVMY0N 1/8
gives the compiler deep insight into the code.
is implicitly thread-safe.
constexpr - Variables and Objects#

If you declare a variable as constexpr , the compiler will evaluate them at
compile-time. This holds not only true for built-in types but also for
instantiations of user-defined types. There are a few serious restrictions for
objects in order to evaluate them at compile-time.
To make life easier for us, we will call the C types like bool , char , int , and
double primitive data types. We will call the remaining data types as user-
defined data types. These are for example, std::string , types from the C++
library and non-primitive data types. Non-primitive data types typically hold
primitive data types.
Variables#
By using the keyword constexpr , the variable becomes a constant
expression.
constexpr double pi= 3.14;
Therefore, we can use the variable in contexts that require a constant

expression. For example, if we want to define the size of an array. This has to
be done at compile-time.
For the declaration of a constexpr variable, you have to keep a few rules in
mind.
The variable:
is implicitly const.
has to be initialized.
requires a constant expression for initialization.
3/20/22, 6:22 PM q p constexpr - Generic Programming Templates in C++
The above rules make sense because if we evaluate a variable at compile-

time, the variable can only depend on values that can be evaluated at
compile time.
The objects are created by the invocation of the constructor and the
constructor has a few special rules as well.
User-Defined Types#
A constexpr constructor
1. can only be invoked with constant expressions.

2. cannot use exception handling.
3. has to be declared as default or delete or the function body must be

empty (C++11).
The constexpr user-defined type
cannot have virtual base classes.

requires that each base object and each non-static member has to be
initialized in the initialization list of the constructor or directly in the
class body. Consequently, it holds that each used constructor (e.g of a
base class) has to be a constexpr constructor and that the applied
initializers have to be constant expressions.
Example#
struct MyDouble{
double myVal;
constexpr MyDouble(double v): myVal(v){}
constexpr double getVal(){return myVal;}
};
The constructor has to be empty and a constant expression.
The user-defined type can have methods which are constant

expressions and cannot to be virtual .
Instances of MyDouble can be instantiated at compile-time.
Functions#
constexpr functions are functions that have the potential to run at compile-
time. With constexpr functions, you can perform a lot of calculations at
compile-time. Therefore, the result of the calculation is available at runtime
and stored as a constant in the ROM available. In addition, constexpr
functions are implicitly inline.
A constexr function can be invoked with a non- constexpr value. In this

case, the function runs at runtime. A constexpr function is executed at
compile-time when it is used in an expression which is evaluated at compile-
time. Some examples would be when using a static_assert or the definition
of a C-array. A constexpr function is also executed at compile-time, when the
result is requested at compile-time, for example: constexpr auto res =
constexprFunction() .
For constexpr functions there are a few restrictions:
The function
has to be non-virtual.
has to have arguments and a return value of a literal type

(https://en.cppreference.com/w/cpp/named_req/LiteralType). Literal
types are the types of constexpr variables.
can only have one return statement.
must return a value.

will be executed at compile-time if invoked within a constant

expression.
can only have a function body consisting of a return statement.
must have a constant return value
is implicitly inline.
Examples#
constexpr int fac(int n){
return n > 0 ? n * fac(n-1): 1;
constexpr int gcd(int a, int b){
return (b==0) ? a : gcd(b, a % b);
Functions with C++14#

The syntax of constexpr functions was massively improved with the change
from C++11 to C++14. In C++11, you had to keep in mind which feature you
can use in a constexpr functions. With C++14, you only have to keep in mind
which feature you can’t use in a constexpr function.
constexpr Functions in C++14#

can have variables that have to be initialized by a constant expression.
cannot have static or thread_local data.
can have conditional jump instructions or loop instructions.

can have more than one instruction.
Example#
1 constexpr auto gcd(int a, int b){
2 while (b != 0){
3 auto t= b;
4 b= a % b;
5 a= t;
6 (/learn)
}
7 return a;
8 }
Template Metaprogramming vs
constexpr Functions#
Template Constant
Characteristics
Metaprogramming Expressions
Compile-time and
Execution time Compile-time
runtime
Arguments Types and values Values
Programming
Functional Imperative
paradigm
Modification No Yes
Control structure Recursion Conditions and Loops
Template Constant
Characteristics
Metaprogramming Expressions
Conditional Template Conditional

execution specialization statement
There are a few remarks about the above-mentioned table.
A template metaprogram runs at compile-time, but a constexpr

functions can run at compile-time or runtime.
Arguments of a template (template metaprogram) can be types and

values. To be more specific, a template can take types,
std::vector<int> , values, std::array<int, 5> , and even templates
std::stack<int, std::vector<int>> . constexpr functions are just
functions which have the potential to run at compile time. Therefore,
they can only accept values.
To learn more about constexpr , click here

(https://en.cppreference.com/w/cpp/language/constexpr).
In the next lesson, we’ll look at a couple of examples of constexpr .

Back Next
Mark as Completed
Report an Issue
- Examples
In this lesson, we'll get into examples of constexpr.
• Example 1: constexpr using C++ 11
• Explanation
• Example 2: constexpr function in C++ 14
• Explanation
Example 1: constexpr using C++ 11#

1 // constExpression.cpp
2
4
5 constexpr int square(int x) { return x * x; }
6 constexpr int squareToSquare(int x){ return square(square(x));}
7
8 int main() {
9
11
12 static_assert(square(10) == 100, "you calculated it wrong");
13 static_assert(squareToSquare(10) == 10000 , "you calculated it wrong");
14
15 std::cout<< "square(10)= " << square(10) << std::endl;
16 std::cout<< "squareToSquare(10)= " << squareToSquare(10) << std::endl;
17 constexpr int constExpr= square(10);
18
19 int arrayClassic[100];
20 int arrayNewWithConstExpression[constExpr];
21 int arrayNewWithConstExpressioFunction[square(10)];
22
https://www.educative.io/courses/generic-templates-in-cpp/B6A6X18AY6Q 1/5
24
25 }
26
Output 1.54s
square(10)= 100
squareToSquare(10)= 10000
Explanation#
In the example above, we have implemented two constexpr functions:
constexpr int square(int x) and constexpr int squareToSquare(int x) .
As you can see, both the functions follow the conventions for constexpr
functions definition in C++11.
The assertion in lines 12 and 13 succeed since 10 is a literal type.

Making a constexpr variable will allow the code compilation to pass the
assertions.
In line 17, we have initialized a constexpr variable constExpr using the

sqaure function.
In lines 19-21, we have initialized three arrays
1. by using a constant 100
2. by using a constexpr variable constExpr

3. by calling the function square(10) . Notice that the input argument

for this function call is constant.
Example 2: constexpr function in C++

14#
1 // constExpressionCpp14.cpp
2
4
5 constexpr int gcd(int a, int b){
6 while (b != 0){
7 auto t= b;
8 b= a % b;
9 a= t;
10 }
11 return a;
12 }
13
14 int main(){
15
17
18 constexpr auto res= gcd(100, 10);
19 std::cout << "gcd(100, 10) " << res << std::endl;
20
21 auto val= 100;
22 auto res2= gcd(val, 10);
23 std::cout << "gcd(val, 10) " << res2 << std::endl;
24
25 }
26
(/learn)
Output 1.7s
gcd(100, 10) 10
gcd(val, 10) 10
Explanation#
Line 18 calculates the result res at compile-time, and line 22 res2 at
runtime.
The difference between ordinary functions and constexpr functions in

C++14 is minimal. Therefore, it’s quite easy to implement the gcd
algorithm in C++14 as a constexpr function.
We have defined res as constexpr variable and its type is

automatically determined by auto .

Back Next
constexpr - Exercise
Mark as Completed
Mark as Completed
Report an Issue
- Exercise
• Try It Out!
Try It Out!#
Use my MyDouble in your program.
How can you be sure that instances of MyDouble are created at compile
time?
What happens, if the instances of my MyDouble are used in non-
constexpr contexts?
(/learn)
2
3 struct MyDouble{
4 double myVal;
5 constexpr MyDouble(double v): myVal(v){}
6 constexpr double getVal(){
7 return myVal;
8 }
9 };
10
11 int main() {
12 // try to call non-constexpr myDouble instance
13 }
https://www.educative.io/courses/generic-templates-in-cpp/R8KXJAAPoZq 1/2
In the next lesson, we’ll study constexpr if .

Back Next
- Examples constexpr if
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/R8KXJAAPoZq 2/2
3/20/22, 6:24 PM constexpr if - Generic Programming Templates in C++
(/learn)
constexpr if
Let's study constexpr if in this lesson.
• constexpr if (C++17)
constexpr if (C++17)#
constexpr if enables us to compile source code
conditionally.
if constexpr (cond) statement1;
else statement2;
The expression cond has to be a constant expression.
The unused code has to be valid.

Thanks to constexpr if , functions can have different return types.
The following code snippet shows a function, which returns an int or a

double .
auto getAnswer(T t){
static_assert(std::is_arithmetic_v<T>); // arithmetic
if constexpr (std::is_integral_v<T>) // integral
return 42;
else
return 42.0; // floating point
https://www.educative.io/courses/generic-templates-in-cpp/qV77XBZBAGR 1/2
3/20/22, 6:24 PM constexpr if - Generic Programming Templates in C++
constexpr if is a replacement for tag dispatch and SFINAE

(https://en.wikipedia.org/wiki/Substitution_failure_is_not_an_error). SFINAE
stands for Substitution Failure Is Not An Error.
To study further about tag dispatching, click here

792309575680).
In the next lesson, we’ll look at an example of constexpr if .

Back Next
- Exercise - Example
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/qV77XBZBAGR 2/2
- Example
Let's have a look at an example of constexpr if.
• Example: constexpr if
• Explanation
Example: constexpr if #
1 // constexprIf.cpp
2
5
6 // SFINAE
7
8 template <typename T, std::enable_if_t<std::is_arithmetic<T>{}>* = nullptr>
9 auto get_value_SFINAE(T) {
10 std::cout << "get_Value_SFINAE(5)" << std::endl;
11 }
12
13 template <typename T, std::enable_if_t<!std::is_arithmetic<T>{}>* = nullptr>
14 auto get_value_SFINAE(T) {
15 std::cout << "get_Value_SFINAE(five)" << std::endl;
16
17 }
18
19 // Tag dispatch
20
22 auto get_value_TAG_DISPATCH(T, std::true_type) {
23 std::cout << "get_Value_TAG_DISPATCH(5)" << std::endl;
24 }
25
https://www.educative.io/courses/generic-templates-in-cpp/7XPOJYAKrgQ 1/4
27 auto get_value_TAG_DISPATCH(T, std::false_type) {
28 std::cout << "get_Value_TAG_DISPATCH(five)" << std::endl;
29 }
30
32 auto get_value_TAG_DISPATCH(T t) {
33 return get_value_TAG_DISPATCH(t, std::is_arithmetic<T>{});
34 }
35
36 // constexpr if
37
39 auto get_value_CONSTEXPR_IF(T) {
40 if constexpr (std::is_arithmetic_v<T>) {
41 std::cout << "get_Value_CONSTEXPR_IF(5)" << std::endl;
42 }
43 else {
44 std::cout << "get_Value_CONSTEXPR_IF(five)" << std::endl;
45 }
46 }
47
48 int main(){
49
51
52 get_value_SFINAE(5);
53 get_value_SFINAE("five");
54
56
57 get_value_TAG_DISPATCH(5);
58 get_value_TAG_DISPATCH("five");
59
61
62 get_value_CONSTEXPR_IF(5);
63 get_value_CONSTEXPR_IF("five");
64
66 (/learn)

67 }
68
Output 1.83s
get_Value_SFINAE(5)
get_Value_SFINAE(five)
get_Value_TAG_DISPATCH(5)
get_Value_TAG_DISPATCH(five)
get_Value_CONSTEXPR_IF(5)
get_Value_CONSTEXPR_IF(five)
Explanation#
We have created get_value functions which use SFINAE, TAG_DISPATCH,
and CONSTEXPR_IF. These functions use the std::is_arithmetic function
from the type-traits library. std::is_arithmetic returns true only if
std::is_integral or std::is_floating_point is true for a given type. All the
calls from main verify that the passed argument falls in their required
category.
get_value_SFINAE uses the function std::enable_if from the type-traits

library. std::enable_if is only true, if the given type is arithmetic.
get_value_TAG_DISPATCH in line 33 dispatches on the result of

std::is_arithmetic .
get_value_CONSTEXR_IF creates another branch of the if -statement

depending on the result of the expression std::is_arithmetic_v<T>
which is a shorthand for std::is_arithmetic<T>::value .
We have learned about the techniques which we used in templates. In the

next chapter, we’ll see different design techniques used in C++ Templates.

Back Next
constexpr if Static Versus Dynamic Polymorphism
Mark as Completed
Report an Issue
3/20/22, 6:26 PM Static Versus Dynamic Polymorphism - Generic Programming Templates in C++
(/learn)
Static Versus Dynamic Polymorphism

Let's dive deep into polymorphism in this lesson.
• Polymorphism
Polymorphism#
Polymorphism means that an object can have different behaviors.
Dynamic Polymorphism
Polymorphism happens at runtime.

A key feature of object-orientation.
Based on interfaces and virtual methods.
Needs one indirection such as a reference or a pointer in C++.
Static Polymorphism
Polymorphism happens at compile-time.

Is not bound to interfaces or derivation hierarchies => Duck Typing
(https://en.wikipedia.org/wiki/Duck_typing)
No indirection such as pointers or references required.
Static polymorphism is typically faster than dynamic
polymorphism.
https://www.educative.io/courses/generic-templates-in-cpp/xlGZgoMJoNE 1/2
3/20/22, 6:26 PM Static Versus Dynamic Polymorphism - Generic Programming Templates in C++

Back Next
- Example - Examples
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/xlGZgoMJoNE 2/2
- Examples
Let's have a look at a couple of examples of polymorphism.
• Example 1: Dispatch with Dynamic Polymorphism
• Explanation
• Example 2: Dispatch with Static Polymorphism
• Explanation
Example 1: Dispatch with Dynamic

Polymorphism#
1 // dispatchDynamicPolymorphism.cpp
2
3 #include <chrono>
5
6 auto start = std::chrono::steady_clock::now();
7
8 void writeElapsedTime(){
9 auto now = std::chrono::steady_clock::now();
10 std::chrono::duration<double> diff = now - start;
11
12 std::cerr << diff.count() << " sec. elapsed: ";
13 }
14
15 struct MessageSeverity{
16 virtual void writeMessage() const {
17 std::cerr << "unexpected" << std::endl;
18 }
19 };
20
https://www.educative.io/courses/generic-templates-in-cpp/m78xylop9wA 1/7
20
21 struct MessageInformation: MessageSeverity{
22 void writeMessage() const override {
23 std::cerr << "information" << std::endl;
24 }
25 };
26
27 struct MessageWarning: MessageSeverity{
28 void writeMessage() const override {
29 std::cerr << "warning" << std::endl;
30 }
31 };
32
33 struct MessageFatal: MessageSeverity{};
34
35 void writeMessageReference(const MessageSeverity& messServer){
36
37 writeElapsedTime();
38 messServer.writeMessage();
39
40 }
41
42 void writeMessagePointer(const MessageSeverity* messServer){
43
45 messServer->writeMessage();
46
47 }
48
49 int main(){
50
52
53 MessageInformation messInfo;
54 MessageWarning messWarn;
55 MessageFatal messFatal;
56
57 MessageSeverity& messRef1 = messInfo;
58 MessageSeverity& messRef2 = messWarn;
59 MessageSeverity& messRef3 = messFatal;
60
61 writeMessageReference(messRef1);
64
65 std::cerr << std::endl;
66
67 MessageSeverity* messPoin1 = new MessageInformation;
68 MessageSeverity* messPoin2 = new MessageWarning;
69 MessageSeverity* messPoin3 = new MessageFatal;
70
71 writeMessagePointer(messPoin1);
74
76
77 }
78
Output 1.65s
3.683e-05 sec. elapsed: information
7.225e-05 sec. elapsed: warning
7.8333e-05 sec. elapsed: unexpected
Note: std::cerr of the class std::ostream represents the standard error

stream. This is not a runtime error.
Explanation#
The structs in lines 15, 21, and 27 know, what they should display if used.
The key idea is that the static type MessageSeverity differs from the dynamic
type such as MessageInformation (line 61); therefore, the late binding will kick
in and the writeMessage methods in lines 71, 72, and 73 are of the dynamic
types. Dynamic polymorphism requires a kind of indirection. We can use
yp y p y p q
references (57-59) or pointers (67-69).
From a performance perspective, we can do better and make the dispatch at

compile time.
Example 2: Dispatch with Static

Polymorphism#
1 // DispatchStaticPolymorphism.cpp
2
3 #include <chrono>
5
7
11
13 }
14
15 template <typename ConcreteMessage>
16 struct MessageSeverity{
17 void writeMessage(){
18 static_cast<ConcreteMessage*>(this)->writeMessageImplementation();
19 }
20 void writeMessageImplementation() const {
21 std::cerr << "unexpected" << std::endl;
22 }
23 };
24
25 struct MessageInformation: MessageSeverity<MessageInformation>{
27 std::cerr << "information" << std::endl;
28 }
29 };
30
31 struct MessageWarning:
32 MessageSeverity<MessageWarning>{
34 std::cerr << "warning" << std::endl;
35 }
36 };
37
38 struct MessageFatal:
39 MessageSeverity<MessageFatal>{};
40
42 void writeMessage(T& messServer){
43
45 messServer.writeMessage();
46
47 }
48
49 int main(){
50
52
53 MessageInformation messInfo;
54 writeMessage(messInfo);
55
56 MessageWarning messWarn;
57 writeMessage(messWarn);
58
59 MessageFatal messFatal;
60 writeMessage(messFatal);
61
63
64 }
65
(/learn)
Output 1.28s
Note: std::cerr of the class std::ostream represents the standard error

stream. This is not a runtime error.
Explanation#
In this case, all concrete structs in lines 25, 31, and 38 are derived from the
base class MessageSeverity . The method writeMessage serves as an interface
that dispatches to the concrete implementations
writeMessageImplementation . To make that happen, the object will be
upcasted to the ConcreteMessage :
static_cast<ConcreteMessage*>(this)->writeMessageImplementation();
This is the dispatch at compile time; therefore, this technique is called static
polymorphism.
To be honest, it took me a bit of time to get used to it but applying static

polymorphism like that on line 42 is actually quite easy.

Back Next
Static Versus Dynamic Polymorphism - Exercise
Mark as Completed
Report an Issue
- Exercise
Problem Statement#
Polymorphism can be simulated in various ways. Implement the
polymorphism with if , with switch , and with a dispatch table
(https://en.wikipedia.org/wiki/Dispatch_table) and compare the three
implementations using static and dynamic polymorphism.
2
(/learn)
4 // After writing your code uncomment the lines in main
5
6 int main() {
7
9 /*
10 writeMessage(MessageSeverity::information);
11 writeMessage(MessageSeverity::warning);
12 writeMessage(MessageSeverity::fatal);
13 */
15
16 }
https://www.educative.io/courses/generic-templates-in-cpp/qV4lXrrgQVy 1/2
Let’s look at different solutions for this exercise in the next lesson.

Back Next
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/qV4lXrrgQVy 2/2
- Solution
In this lesson, we'll look at different solution reivews for the last exercise.
• Solution 1: Using the if Statement
• Explanation
• Solution 2: Using the switch statement
• Explanation
• Solution 3: Using a Dispatch Table
• Explanation
Solution 1: Using the if Statement#

1 // dispatchIf.cpp
2
3 #include <chrono>
5
6 enum class MessageSeverity{
7 information,
8 warning,
9 fatal,
10 };
11
13
17
19 }
https://www.educative.io/courses/generic-templates-in-cpp/qV2qOOk2D5G 1/8
19 }
20
21 void writeInformation(){ std::cerr << "information" << std::endl; }
22 void writeWarning(){ std::cerr << "warning" << std::endl; }
23 void writeUnexpected(){ std::cerr << "unexpected" << std::endl; }
24
25 void writeMessage(MessageSeverity messServer){
26
28
29 if (MessageSeverity::information == messServer){
30 writeInformation();
31 }
32 else if (MessageSeverity::warning == messServer){
33 writeWarning();
34 }
35 else{
36 writeUnexpected();
37 }
38
39 }
40
41 int main(){
42
44
48
50
51 }
52
Output 2.51s
Explanation#
Note: std::cerr of the class std::ostream represents the standard

error stream. This is not a runtime error.
The function writeMessage in line 25 displays the elapsed time in seconds in

line 27 since the start of the program and a log message. It uses an
enumeration in line 6 for the message severity. We used the start time in line
12 and the current time in line 15 to calculate the elapsed time. As the name
suggests, the std::steady_clock cannot be adjusted; therefore, it is the right
choice for this measurement. The key part of the program is the part of the
function writeMessage in line 25, in which we made the decision which
message should be displayed. In this case, we used if-else statements.
Solution 2: Using the switch

statement#
1 // dispatchSwitch
2
3 #include <chrono>
5
7 information,
8 warning,
9 fatal,
10 };
11
13
17
17
19 }
20
24
25 void writeMessage(MessageSeverity messSever){
26
28
29 switch(messSever){
30 case MessageSeverity::information:
31 writeInformation();
32 break;
33 case MessageSeverity::warning:
34 writeWarning();
35 break;
36 default:
37 writeUnexpected();
38 break;
39 }
40
41 }
42
43 int main(){
44
46
50
52
53 }
54
Output 1.67s
3 7315e 05 sec elapsed: information

Explanation#

The following program is quite similar to the previous one. Only the
implementation of the function writeMessage changed. The function
writeMessage in line 25 displays the elapsed time in seconds (line 27) since
the start of the program and a log message. It uses an enumeration (line 6)
for the message severity. We used the start time (line 12) and the current
time (line 15) to calculate the elapsed time. As the name suggested, the
std::steady_clock cannot be adjusted; therefore, it is the right choice for
this measurement. The key part of the program is the part of the function
writeMessage (line 25), in which we made the decision which message
should be displayed. In this case, we used the switch statements.
To be honest, I had to look up the syntax for the switch statements to make
it right.
Solution 3: Using a Dispatch Table#

1 // dispatchHasttables
2
3 #include <chrono>
6 #include <unordered_map>
7
9 information,
10 warning,
11 fatal,
12 };
13
15
19
21 }
22
26
27 std::unordered_map<MessageSeverity, std::function<void()>> mess2Func{
28 {MessageSeverity::information, writeInformation},
29 {MessageSeverity::warning, writeWarning},
30 {MessageSeverity::fatal, writeUnexpected}
31 };
32
33 void writeMessage(MessageSeverity messServer){
34
36
37 mess2Func[messServer]();
38
39 }
40
41 int main(){
42
44
48
50
51 }
52
Output 1.95s
(/learn)
Explanation#

With the if-else or the switch statement, we used enumerator for

dispatching to the right case. The key to our dispatch table behaves in a
similar way.
Dynamic or static polymorphism is totally different. Instead of an

enumerator or a key for dispatching to the right action, we used objects
which decide autonomously at runtime (dynamic polymorphism) or
compile-time (static polymorphism) what should be done.
Let’s move on to CRTP in the next lesson.

Back Next
- Exercise CRTP
Mark as Completed
Report an Issue
3/20/22, 6:30 PM CRTP - Generic Programming Templates in C++
CRTP
Let's learn about CRTP in this lesson.
• CRTP
• Typical use-case
• Mixins
• Static Polymorphism
CRTP#
The acronym CRTP
(https://en.wikipedia.org/wiki/Curiously_recurring_template_pattern) stands
for the C++ idiom Curiously Recurring Template Pattern and is a technique
in C++ in which a Derived class derives from a class template Base . The key
is that Base has Derived as a template argument.
Let’s have a look at an example:
template<class T>
class Base{
...
};
class Derived: public Base<Derived>{
...
};
https://www.educative.io/courses/generic-templates-in-cpp/gkY9jRmxOm6 1/4
CRTP enables static polymorphism.
Typical use-case#
There are two typical use-cases for CRTP: Mixins and static polymorphism.
Mixins#
Mixins (https://en.wikipedia.org/wiki/Mixin) is a popular concept in the
design of classes to mix in new code. Therefore, it’s an often-used technique
in Python to change the behavior of a class by using multiple inheritances.
In contrast to C++, in Python, it is legal to have more than one definition of a
method in a class hierarchy. Python simply uses the method that is first in
the Method Resolution Order (https://docs.python.org/3/glossary.html#term-
method-resolution-order) (MRO).
You can implement mixins in C++ by using CRTP. A prominent example is the
class std::enable_shared_from_this . By using this class, you can create
objects that return an std::shared_ptr to themselves. We have to derive
your class MySharedClass public from std::enable_shared_from_this . Now,
our class MySharedClass has a method shared_from_this .
An additional
(/learn) typical use-case for mixins is a class that you want to extend
with the capability that their instances support the comparison for equality
and inequality.
Static Polymorphism#
Static polymorphism is quite similar to dynamic polymorphism. But
contrary to dynamic polymorphism with virtual methods, the dispatch of the
https://www.educative.io/courses/generic-templates-in-cpp/gkY9jRmxOm6 f 2/4
method calls will take place at compile-time. Now, we are at the center of the
CRTP idiom.
class ShareMe: public std::enable_shared_from_this<ShareMe>{
std::shared_ptr<ShareMe> getShared(){
return shared_from_this();
}
};
std::enable_shared_from_this creates a shared _ptr for an object.
std::enable_shared_from_this: base class of the object.
shared_from_this: returns the shared object
To learn more about CRTP, click here

(https://en.wikipedia.org/wiki/Curiously_recurring_template_pattern).
In the next lesson, we’ll look at a couple of examples of CRTP.

Back Next
Mark as Completed
Report an Issue
- Examples
In this lesson, we'll see a couple of examples of CRTP.
• Example 1: Mixins with CRTP
• Explanation
• Example 2: Static Polymorphism with CRTP
• Explanation
Example 1: Mixins with CRTP#

1 // templateCRTPRelational.cpp
2
4 #include <string>
5
6 template<class Derived>
7 class Relational{};
8
9 // Relational Operators
10
11 template <class Derived>
12 bool operator > (Relational<Derived> const& op1, Relational<Derived> const & op2){
13 Derived const& d1 = static_cast<Derived const&>(op1);
15 return d2 < d1;
16 }
17
19 bool operator == (Relational<Derived> const& op1, Relational<Derived> const & op2){
22 return !(d1 < d2) && !(d2 < d1);
23 }
https://www.educative.io/courses/generic-templates-in-cpp/BnGoJ6LLAVn 1/7
23 }
24
26 bool operator != (Relational<Derived> const& op1, Relational<Derived> const & op2){
29 return (d1 < d2) || (d2 < d1);
30 }
31
33 bool operator <= (Relational<Derived> const& op1, Relational<Derived> const & op2){
36 return (d1 < d2) || (d1 == d2);
37 }
38
40 bool operator >= (Relational<Derived> const& op1, Relational<Derived> const & op2){
43 return (d1 > d2) || (d1 == d2);
44 }
45
46 // Apple
47
48 class Apple:public Relational<Apple>{
49 public:
50 explicit Apple(int s): size{s}{};
51 friend bool operator < (Apple const& a1, Apple const& a2){
52 return a1.size < a2.size;
53 }
54 private:
55 int size;
56 };
57
58 // Man
59
60 class Man:public Relational<Man>{
61 public:
62 explicit Man(const std::string& n): name{n}{}
63 friend bool operator < (Man const& m1, Man const& m2){
64 return m1.name < m2.name;
65 }
66 private:
67 std::string name;
68 };
69
70 int main(){
71
73
74 Apple apple1{5};
75 Apple apple2{10};
76 std::cout << "apple1 < apple2: " << (apple1 < apple2) << std::endl;
77 std::cout << "apple1 > apple2: " << (apple1 > apple2) << std::endl;
78 std::cout << "apple1 == apple2: " << (apple1 == apple2) << std::endl;
79 std::cout << "apple1 != apple2: " << (apple1 != apple2) << std::endl;
80 std::cout << "apple1 <= apple2: " << (apple1 <= apple2) << std::endl;
81 std::cout << "apple1 >= apple2: " << (apple1 >= apple2) << std::endl;
82
84
85 Man man1{"grimm"};
86 Man man2{"jaud"};
87 std::cout << "man1 < man2: " << (man1 < man2) << std::endl;
88 std::cout << "man1 > man2: " << (man1 > man2) << std::endl;
89 std::cout << "man1 == man2: " << (man1 == man2) << std::endl;
90 std::cout << "man1 != man2: " << (man1 != man2) << std::endl;
91 std::cout << "man1 <= man2: " << (man1 <= man2) << std::endl;
92 std::cout << "man1 >= man2: " << (man1 >= man2) << std::endl;
93
95
96 }
97
Output 1.35s
apple1 < apple2: true
apple1 > apple2: false
apple1 == apple2: false
apple1 != apple2: true
apple1 <= apple2: true
apple1 >= apple2: false
man1 < man2: true
Explanation#
We have implemented, for the classes Apple and Man , the smaller operator
separately (lines 51-52 and 63-64). The classes Man and Apple are publicly
derived (line 48 and 60) from the class Relational<Man> and
Ralational<Apple> . We have implemented for classes of the kind Relational
the greater than operator > (lines 11 – 16), the equality operator == (lines 18
– 23), the not equal operator != (lines 25 – 30), the less than or equal
operator <= (line 32 – 37) and the greater than or equal operator >= (lines
39 – 44). The less than or equal ( <= ) and greater than or equal ( >= )
operators used the equality operator == (line 36 and 43). All these operators
convert their operands: Derived const&: Derived const& d1 =
static_cast<Derived const&>(op1) .
In the main program, we have compared Apple and Man classes for all the
above-mentioned operators.
Example 2: Static Polymorphism with

CRTP#
1 // templateCRTP.cpp
2
4
5 template <typename Derived>
6 struct Base{
7 void interface(){
8 static_cast<Derived*>(this)->implementation();
9 }
10
11 void implementation(){
12 std::cout << "Implementation Base" << std::endl;
13 }
14 };
15
16 struct Derived1: Base<Derived1>{
18 std::cout << "Implementation Derived1" << std::endl;
18 std::cout << Implementation Derived1 << std::endl;
19 }
20 };
21
25 }
26 };
27
28 struct Derived3: Base<Derived3>{};
29
31 void execute(T& base){
32 base.interface();
33 }
34
35 int main(){
36
38
39 Derived1 d1;
40 execute(d1);
41
42 Derived2 d2;
43 execute(d2);
44
45 Derived3 d3;
46 execute(d3);
47
49
50 }
51
(/learn)
Output 1.29s
Implementation Derived1
Implementation Base
Explanation#
We have used static polymorphism in the function template execute (lines
30-33). We invoked the method base.interface on each base argument. The
method Base::interface , in lines (7-9), is the key point of the CRTP idiom.
The methods dispatch to the implementation of the derived class:
static_cast<Derived*>(this)->implementation() . That is possible because the
method will be instantiated when called. At this point in time, the derived
classes Derived1 , Derived2 , and Derived3 are fully defined. Therefore, the
method Base::interface can use the details of its derived classes. Especially
interesting is the method Base::implementation (lines 11-13). This method
plays the role of a default implementation for the static polymorphism for
the class Derived3 (line 28).
We’ll solve a few exercises around CRTP in the next lesson.

Back Next
CRTP - Exercises
Mark as Completed
Report an Issue
- Exercises
Let's solve a few exercises around CRTP in this lesson.
• Case 1:
• Case 2:
• Case 3:
Extend the given piece of code with a
Person class. A Person should have a
first name
last name
You need to create two objects of the Person class and compare these objects
using relational operators.
2 #include <string>
3
https://www.educative.io/courses/generic-templates-in-cpp/qA6KLqVWWJy 1/8
6
8
13 return d2 < d1;
14 }
15
20 return !(d1 < d2) && !(d2 < d1);
21 }
22
27 return (d1 < d2) || (d2 < d1);
28 }
29
34 return (d1 < d2) || (d1 == d2);
35 }
36
41 return (d1 > d2) || (d1 == d2);
42 }
43
44 // Implement a person class here
45
46 // Person
47
48 // Apple
49
50 class Apple:public Relational<Apple>{
51 public:
55 }
56 private:
57 int size;
57 int size;
58 };
59
60 // Man
61
62 class Man:public Relational<Man>{
63 public:
67 }
68 private:
70 };
71
72 int main(){
73
75
76 // Call Person class object here checks them for all relational operaotrs
77
79
80 }
81
Output 2.63s
In example 2
384649461760#example-1-static-polymorphism-with-crtp), how can you
prevent a derived class which has the wrong template parameter: Derived4:
p
3/20/22, 6:33 PM
g p p
- Exercises - Generic Programming Templates in C++
Base<Derived3>
1 // templateCRTP.cpp
2
4
6 struct Base{
7 void interface(){
9 }
12 }
13 };
14
18 }
19 };
20
24 }
25 };
26
28
32 }
33
34 // Write the struct here
35
36 int main(){
37
39
40 // call the function here
41
43
44 }
45
Output 2.39s
The functionality of the program in example 1
384649461760#example-1-static-polymorphism-with-crtp) of the previous
lesson can be implemented in various ways. Implement each variant.
Case 1:#
Object-oriented with dynamic polymorphism.
2
3 // Implement the functionality here
4
5 int main() {
6 // call the function here
7 return 0;
8 }
Case 2:#
Just a function template.
2
3 // write your function template here
4
5 int main() {
6 // Call the function here
7 }
Case 3:#
To solve this exercise with Concepts
499255980032), you have to peek into the last chapter of this course. Solving
this exercise with Concepts
499255980032) is also a valid and elegant way.
Concepts with C++20:
2
3 // Implement it using C++20 concept
4
5 int main() {
6 // your code goes here
7 }
Implement the class ShareMe and use it.
(/learn)
Objects of the class ShareMe should return an std::shared_ptr to
themselves.
2
3 // Implement the class here
4
5 int main() {
6 // After implementing the shareMe class uncomment the code given below
7
8 /*
10
11 // share the same ShareMe object
12 std::shared_ptr<ShareMe> shareMe(new ShareMe);
13 std::shared_ptr<ShareMe> shareMe1= shareMe->getShared();
14
15 // both resources have the same address
16 std::cout << "Address of resource of shareMe "<< (void*)shareMe.get() << " " << st
17 std::cout << "Address of resource of shareMe1 "<< (void*)shareMe1.get() << " " <<
18
19 // the use_count is 2
20 std::cout << "shareMe.use_count(): "<< shareMe.use_count() << std::endl;
22 */
23 }
In the next lesson, we’ll look at solutions to these exercises.

(https://join hired com/educative/?
Back Next
- Examples - Solutions
Mark as Completed
Report an Issue
3/20/22, 6:36 PM - Solutions - Generic Programming Templates in C++
- Solutions
Let's look at the solutions of the exercises, we had in the last lesson.
• Solution of Problem Statement 1
• Explanation
• Solution of Problem Statement 2
• Explanation
• Solution to Problem Statement 3 (Case 1)
• Explanation

• Explanation
• Explanation
• Solution to Problem Statement 4

• Explanation
Solution of Problem Statement 1#

1 // templateCRTPRelational.cpp
2
4 #include <string>
5 #include <utility>
6
https://www.educative.io/courses/generic-templates-in-cpp/RLZmo9AyRYY 1/14
9
11
16 return d2 < d1;
17 }
18
23 return !(d1 < d2) && !(d2 < d1);
24 }
25
30 return (d1 < d2) || (d2 < d1);
31 }
32
37 return (d1 < d2) || (d1 == d2);
38 }
39
44 return (d1 > d2) || (d1 == d2);
45 }
46
47 // Apple
48
49 class Apple: public Relational<Apple>{
50 public:
54 }
55 private:
56 int size;
57 };
58
59 // Man
60
61 class Man: public Relational<Man>{
62 public:
66 }
67 private:
69 };
70
71 // class Person
72 class Person: public Relational<Person>{
73 public:
74 Person(std::string fst, std::string sec): first(fst), last(sec){}
75 friend bool operator < (Person const& p1, Person const& p2){
76 return std::make_pair(p1.first, p1.last) < std::make_pair(p2.first, p2.last)
77 }
78 private:
79 std::string first;
80 std::string last;
81 };
82
83
84 int main(){
85
87
88 Apple apple1{5};
89 Apple apple2{10};
90 std::cout << "apple1 < apple2: " << (apple1 < apple2) << std::endl;
91 std::cout << "apple1 > apple2: " << (apple1 > apple2) << std::endl;
92 std::cout << "apple1 == apple2: " << (apple1 == apple2) << std::endl;
93 std::cout << "apple1 != apple2: " << (apple1 != apple2) << std::endl;
94 std::cout << "apple1 <= apple2: " << (apple1 <= apple2) << std::endl;
95 std::cout << "apple1 >= apple2: " << (apple1 >= apple2) << std::endl;
96
98
99 Man man1{"grimm"};
100 Man man2{"jaud"};
101 std::cout << "man1 < man2: " << (man1 < man2) << std::endl;
102 std::cout << "man1 > man2: " << (man1 > man2) << std::endl;
103 std::cout << "man1 == man2: " << (man1 == man2) << std::endl;
104 std::cout << "man1 != man2: " << (man1 != man2) << std::endl;
105 std::cout << "man1 <= man2: " << (man1 <= man2) << std::endl;
106 std::cout << "man1 >= man2: " << (man1 >= man2) << std::endl;
107
109
110 Person rainer{"Rainer", "Grimm"};
111 Person marius{"Marius" "Grimm"};
111 Person marius{ Marius , Grimm };
112 std::cout << "rainer < marius: " << (rainer < marius) << std::endl;
113 std::cout << "rainer > marius: " << (rainer > marius) << std::endl;
114 std::cout << "rainer == marius: " << (rainer == marius) << std::endl;
115 std::cout << "rainer != marius: " << (rainer != marius) << std::endl;
116 std::cout << "rainer <= marius: " << (rainer <= marius) << std::endl;
117 std::cout << "rainer >= marius: " << (rainer >= marius) << std::endl;
118
120 }
Output 1.61s
apple1 < apple2: true
apple1 > apple2: false
apple1 == apple2: false
apple1 != apple2: true
apple1 <= apple2: true
apple1 >= apple2: false
man1 < man2: true
Explanation#
In the above code, we have defined the Person class which contains the
string variables, i.e., first and last and a < operator to compare two
people’s lengths in lines 75 – 77. The class Person is publicly derived (line 72)
from the class Relational<Person> . We have implemented for classes of the
kind Relational the greater than operator > (lines 12 - 17), the equality
operator == (lines 19 - 24), the not equal operator != (lines 26 - 31), the less
than or equal operator <= (line 33 - 38) and the greater than or equal
operator >= (lines 40 - 45). The less than or equal and greater than or equal
operators used the equality operator (line 37 and 44). All these operators
convert their operands: Derived const&: Derived const& d1 =
3/20/22, 6:36 PM
p - Solutions - Generic Programming Templates in C++
static_cast<Derived const&>(op1) .
In the main program, we have compared Apple , Man , and Person classes for
all the above-mentioned operators.
Solution of Problem Statement 2#

1 // templateCRTPCheck.cpp
2
4
6 struct Base{
7 void interface(){
9 }
12 }
13 private:
14 Base(){};
15 friend Derived;
16 };
17
21 }
22 };
23
27 }
28 };
29
31
32 // struct Derived4: Base<Derived3>{};
33
37 }
38
39 int main(){
40
42
43 Derived1 d1;
44 execute(d1);
45
46 Derived2 d2;
47 execute(d2);
48
49 Derived3 d3;
50 execute(d3);
51
52 // Derived4 d4;
53
55
56 }
Output 2.02s
Implementation Base
Explanation#
We have used static polymorphism in the function template execute (lines
34 – 37). We invoked the method base.interface on each base argument.
The method Base::interface , in lines (7 – 9), is the key point of the CRTP
idiom. The method dispatches to the implementation of the derived class:
static_cast<Derived*>(this)->implementation() . That is possible because the
method will be instantiated when called. At this point in time, the derived
classes Derived1 , Derived2 , and Derived3 are fully defined. Therefore, the
method Base::interface can use the details of its derived classes. Especially
interesting is the method Base::implementation (lines 10 – 12). This method

plays the role of a default implementation for the static polymorphism for
the class Derived3 (line 30).
The constructor of the derived class has to call the constructor of the base
class. The constructor in the base class is private. Only type T can invoke the
constructor of the base class. So, if the derived class is different from the
class T , the code doesn’t compile.
Solution to Problem Statement 3

(Case 1)#
1 // dispatchPolymorphism.cpp
2
4
5 struct Base{
6 virtual void interface(){
8 }
9 };
10
11 struct Derived1: Base{
14 }
15 };
16
20 }
21 };
22
23 struct Derived3: Base{};
24
25 void execute(Base& base){
27 }
28
29 int main(){
30
32
33 Derived1 d1;
34 Base& b1 = d1;
35 execute(b1);
36
37 Derived2 d2;
38 Base& b2 = d2;
39 execute(b2);
40
41 Derived3 d3;
42 Base& b3 = d3;
43 execute(b3);
44
46
47 }
48
Output 9.15s
Implementation Base
Explanation#
We have used dynamic polymorphism in the function template execute
(lines 25 – 27). This function only accepts a reference to the class object
passed. Now, we make objects of the derived classes and store references to
them (lines 34, 38 and 42). The static type of b1 , b2 , and b3 in lines 34, 38,
42 is Base and the dynamic type is Dervived1 , Derived2 , or Derived3
3/20/22, 6:36 PM
y yp- Solutions - Generic Programming
, ,
Templates in C++
respectively.

(Case 2)#
1 // dispatchGeneric.cpp
2
4
5 struct Base{
8 }
9 };
10
14 }
15 };
16
20 }
21 };
22
24
26 void execute(T& t){
27 t.interface();
28 }
29
30 int main(){
31
33
34 Derived1 d1;
35 execute(d1);
36
37 Derived2 d2;
38 execute(d2);
39
40 Derived3 d3;
41 execute(d3);
42
42
44
45 }
46
Output 1.66s
Implementation Base
Explanation#
We have used static polymorphism in the function template execute (line 25
– 27). The method execute accepts a reference. Now, we can make objects of
the derived classes and pass them to execute methods (line 34 – 41).

(Case 3)#
1 // dispatchConcepts.cpp
2
4
6 concept bool Interface(){
7 return requires(T a){
8 { a.interface } -> void;
9 };
10 }
10 }
11
12 struct Base{
15 }
16 };
17
21 }
22 };
23
27 }
28 };
29
31
32 template <typename Interface>
33 void execute(Interface& inter){
34 inter.interface();
35 }
36
37 int main(){
38
40
41 Derived1 d1;
42 execute(d1);
43
44 Derived2 d2;
45 execute(d2);
46
47 Derived3 d3;
48 execute(d3);
49
51
52 }
53
Output 1.46s
Implementation Base
Explanation#
This implementation is based on concepts
499255980032) which will be part of C++20. In the concrete case, concepts
mean that execute can only be invoked with types, which supports a
function with the name interface returning void .
Solution to Problem Statement 4#

1 // templatesCRTPShareMe.cpp
2
4 #include <memory>
5
6 class ShareMe: public std::enable_shared_from_this<ShareMe>{
7 public:
8 std::shared_ptr<ShareMe> getShared(){
9 return shared_from_this();
10 }
11 };
12
13 int main(){
14
16
17 (/learn)
// share the same ShareMe object
18 std::shared_ptr<ShareMe> shareMe(new ShareMe);
_p ( )
19 std::shared_ptr<ShareMe> shareMe1= shareMe->getShared();
20
21 // both resources have the same address
22 std::cout << "Address of resource of shareMe "<< (void*)shareMe.get() << " " << st
23 std::cout << "Address of resource of shareMe1 "<< (void*)shareMe1.get() << " " <<
24
25 // the use_count is 2
26 std::cout << "shareMe.use_count(): "<< shareMe.use_count() << std::endl;
28
29 }
Output 1.62s
Address of resource of shareMe 0xf38c30
Address of resource of shareMe1 0xf38c30
shareMe.use_count(): 2
Explanation#
With the class std::enable_shared_from_this
(https://en.cppreference.com/w/cpp/memory/enable_shared_from_this), we
can create objects which return a std::shared_ptr on itself. For that, we
have to derive the class public from std::enable_shared_from_this . The
smart pointer shareMe (line 18) is copied by shareMe1 (line 19).
The call
shareMe->getShared() in line 19 creates a new smart pointer. getShared()
(line 8) internally uses the function shared_from_this . In lines 22 and 23,
shareMe.get() and shareMe.get() returns a pointer to the resource. In line
26, the shareMe.use_count() returns the value of the reference counter.
In the next lesson, we’ll learn about expression templates.

Back Next
- Exercises Expression Templates
Mark as Completed
Report an Issue
3/20/22, 6:37 PM Expression Templates - Generic Programming Templates in C++
Expression Templates
In this lesson, we'll study expression templates in detail.
• Expression Templates
• Lazy Evaluation
• A First Naive Approach
• The Issue
• The Solution
• The Idea
• The Entire Magic
Expression Templates#
Expression templates are “structures representing a computation at compile-
time, which are evaluated only as needed to produce efficient code for the
entire computation.”
Now we are at the center of lazy evaluation.
Lazy Evaluation#
The story about lazy evaluation in C++ is quite short. That will change in
C++20, with the ranges library from Eric Niebler. Lazy evaluation is the
default in Haskell. Lazy evaluation means that an expression is only
evaluated when needed
https://www.educative.io/courses/generic-templates-in-cpp/RMgr0XJ67LK 1/8
evaluated when needed.
This strategy has two benefits.
Lazy evaluation helps you to save time and memory.

You can define an algorithm on infinite data structures. Of course, you
can only ask for a finite number of values at runtime.
Advantages:
Creates a domain-specific language (DSL)
Avoidance of temporary data structures
Disadvantages:
Longer compile-times
Advanced programming technique (template metaprogramming)
What problem do expression templates solve? Thanks to expression

templates, we can get rid of superfluous temporary objects in expressions.
What do we mean by superfluous temporary objects? To demonstrate that,
let’s look at the implementation of the class MyVector below.
A First Naive Approach#

MyVector is a simple wrapper for an std::vector<T> . The wrapper class has
two constructors (line 12 and 15), we have a size function which returns its
size (line 18 - 20), and the reading index operator (line 23 - 25) and writing
index access (line 27 - 29).
Given below is the naive vector implementation:
1 // vectorArithmeticOperatorOverloading.cpp
2
4 #include <vector>
5
7 class MyVector{
8 std::vector<T> cont;
9
10 public:
11 // MyVector with initial size
12 MyVector(const std::size_t n) : cont(n){}
13
14 // MyVector with initial size and value
15 MyVector(const std::size_t n, const double initialValue) : cont(n, initialValue){}
16
17 // size of underlying container
18 std::size_t size() const{
19 return cont.size();
20 }
21
22 // index operators
23 T operator[](const std::size_t i) const{
24 return cont[i];
25 }
26
27 T& operator[](const std::size_t i){
28 return cont[i];
29 }
30
31 };
32
33 // function template for the + operator
35 MyVector<T> operator+ (const MyVector<T>& a, const MyVector<T>& b){
36 MyVector<T> result(a.size());
37 for (std::size_t s= 0; s <= a.size(); ++s){
38 result[s]= a[s]+b[s];
39 }
40 return result;
41 }
42
43 // function template for the * operator
45 MyVector<T> operator* (const MyVector<T>& a, const MyVector<T>& b){
48 result[s]= a[s]*b[s];
49 }
50 return result;
51 }
52
53 // function template for << operator
55 std::ostream& operator<<(std::ostream& os, const MyVector<T>& cont){
;
57 for (int i=0; i<cont.size(); ++i) {
58 os << cont[i] << ' ';
59 }
60 os << std::endl;
61 return os;
62 }
63
64 int main(){
65
66 MyVector<double> x(10,5.4);
67 MyVector<double> y(10,10.3);
68
69 MyVector<double> result(10);
70
71 result= x+x + y*y;
72
73 std::cout << result << std::endl;
74
75 }
Output 1.5s
116.89 116.89 116.89 116.89 116.89 116.89 116.89 116.89 116.89 116.89
Thanks to the overloaded + operator (line 34 - 41), the overloaded *

operator (line 44 - 51), and the overloaded output operator (line 54 - 62), the
objects x , y and result feel like numbers.
Why is this implementation naive? The answer is in the expression result =

x+x + y*y . In order to evaluate the expression, three temporary objects are
needed to hold the result of each arithmetic subexpression.
The Issue#
result = x+x + y*y;
Arithmetic expressions create many temporary objects
The Solution#
The Idea#
The overloaded operators return proxy objects
The final assignment result[i] = x[i] + x[i] + y[i] * y[i] triggers

the calculation
No temporary objects are necessary.
The Entire Magic#

result[i] = x[i] + x[i] + y[i] * y[i];
Thanks to the compiler explorer (https://godbolt.org/z/TXfimX) you can see

that no temporary objects are created. The expression is lazily evaluated in
place.
(/learn)
A sharp view at the previous screenshot helps to unwrap the magic. Let’s
have a look at the sequence of calls:
In the next lesson, we’ll look at some examples of expression templates.

Back Next
- Solutions - Examples
Mark as Completed
Report an Issue
- Examples
Let's have a look at some examples of expression templates.
• Example 1: Vector Arithmetic Based on Object-Orientation
• Explanation
• Example 2: Vector Arithmetic Based on Expression Templates
• Explanation
Example 1: Vector Arithmetic Based

on Object-Orientation#
1 // vectorArithmeticOperatorOverloading.cpp
2
4 #include <vector>
5
7 class MyVector{
9
10 public:
12 explicit MyVector(const std::size_t n) : cont(n){}
13
16
20 }
https://www.educative.io/courses/generic-templates-in-cpp/RLy98qkEyrq 1/7
20 }
21
24 return cont[i];
25 }
26
28 return cont[i];
29 }
30
31 };
32
39 }
40 return result;
41 }
42
49 }
50 return result;
51 }
52
56 os << std::endl;
57 for (std::size_t i=0; i<cont.size(); ++i) {
58 os << cont[i] << ' ';
59 }
60 return os;
61 }
62
63 int main(){
64
67
69
71
73
74 }
75
Output 2.13s
116.89 116.89 116.89 116.89 116.89 116.89 116.89 116.89 116.89 116.89
Explanation#
MyVector is a simple wrapper for an std::vector<T> . The wrapper class has
two constructors (line 12 and 15), we have a size function which returns its
size (line 18 - 20), and the reading index operator (line 23 - 25) and writing
index access (line 27 - 29).
Thanks to the overloaded + operator (line 34 - 41), the overloaded *

operator (line 44 - 51) and the overloaded output operator (line 54 - 61), the
objects x , y , and result feel like numbers on lines 70 and 72.
Example 2: Vector Arithmetic Based

on Expression Templates#
1 // vectorArithmeticExpressionTemplates.cpp
2
3 #include <cassert>
5 #include <vector>
6
7 template<typename T, typename Cont= std::vector<T> >
8 class MyVector{
8 class MyVector{
9 Cont cont;
10
11 public:
14
17
18 // Constructor for underlying container
19 MyVector(const Cont& other) : cont(other){}
20
21 // assignment operator for MyVector of different type
22 template<typename T2, typename R2>
23 MyVector& operator=(const MyVector<T2, R2>& other){
24 assert(size() == other.size());
25 for (std::size_t i = 0; i < cont.size(); ++i) cont[i] = other[i];
26 return *this;
27 }
28
32 }
33
36 return cont[i];
37 }
38
40 return cont[i];
41 }
42
43 // returns the underlying data
44 const Cont& data() const{
45 return cont;
46 }
47
48 Cont& data(){
49 return cont;
50 }
51 };
52
53 // MyVector + MyVector
54 template<typename T, typename Op1 , typename Op2>
55 class MyVectorAdd{
56 const Op1& op1;
57 const Op2& op2;
58
59 public:
60 MyVectorAdd(const Op1& a, const Op2& b): op1(a), op2(b){}
61
63 return op1[i] + op2[i];
64 }
65
67 return op1.size();
68 }
69 };
70
71 // elementwise MyVector * MyVector
72 template< typename T, typename Op1 , typename Op2 >
73 class MyVectorMul {
74 const Op1& op1;
75 const Op2& op2;
76
77 public:
78 MyVectorMul(const Op1& a, const Op2& b ): op1(a), op2(b){}
79
81 return op1[i] * op2[i];
82 }
83
86 }
87 };
88
90 template<typename T, typename R1, typename R2>
91 MyVector<T, MyVectorAdd<T, R1, R2> >
92 operator+ (const MyVector<T, R1>& a, const MyVector<T, R2>& b){
93 return MyVector<T, MyVectorAdd<T, R1, R2> >(MyVectorAdd<T, R1, R2 >(a.data(), b.da
94 }
95
98 MyVector<T, MyVectorMul< T, R1, R2> >
99 operator* (const MyVector<T, R1>& a, const MyVector<T, R2>& b){
100 return MyVector<T, MyVectorMul<T, R1, R2> >(MyVectorMul<T, R1, R2 >(a.data(), b.da
101 }
102
108 os << cont[i] << ' ';
109 }
110 os << std::endl;
111 return os;
111 return os;
112 }
113
114 int main(){
115
118
120
122
124
125 }
Output 1.56s
(/learn)
116.89 116.89 116.89 116.89 116.89 116.89 116.89 116.89 116.89 116.89
Explanation#
The key difference between the first naive implementation and this
implementation with expression templates is that the overloaded + and *
operators return in case of the expression tree proxy objects. These proxies
represent the expression tree (lines 91 and 98). The expression tree is only
created but not evaluated. Lazy, of course. The assignment operator (lines 22
- 27) triggers the evaluation of the expression tree that needs no temporary
objects.
In the next lesson, we’ll solve a few exercises for expression templates.

Back Next
Expression Templates - Exercises
Mark as Completed
Report an Issue
- Exercises
Let's solve a few exercises in this lesson regarding expression templates.
Compare both programs containing + and * operators and study, in
particular, the implementation based on expression templates.
Extend the example to expression templates by adding support

subtraction and the division operators for the MyVector class.
2 #include <vector>
3
5 class MyVector{
7
8 public:
11
14
18 }
https://www.educative.io/courses/generic-templates-in-cpp/YV3wMxYKppp 1/6
18 }
19
22 return cont[i];
23 }
24
26 return cont[i];
27 }
28
29 };
30
37 }
38 return result;
39 }
40
47 }
48 return result;
49 }
50
56 os << cont[i] << ' ';
57 }
58 os << std::endl;
59 return os;
60 }
61
62 // Implement subtraction and division operators here
63
64 int main(){
65 // call these subtraction and division operators here
66 }
67
The solution of the previous example
161208496128#example-2-vector-arithmetic-based-on-expression-templates)
Vector Arithmetic Expression Templates is too laborious. In particular, the
classes MyVectorAdd , MyVectorSub , MyVectorMul , and MyVectorDiv have a lot
in common. Try to simplify the program.
3 #include <vector>
4
6 class MyVector{
7 Cont cont;
8
9 public:
12
15
17 MyVector(const Cont& other) : cont(other){}
18
24 return *this;
25 }
26
30 }
31
32 // i d t
34 return cont[i];
35 }
36
38 return cont[i];
39 }
40
43 return cont;
44 }
45
46 Cont& data(){
47 return cont;
48 }
49 };
50
54 const Op1& op1;
55 const Op2& op2;
56
57 public:
59
62 }
63
66 }
67 };
68
72 const Op1& op1;
73 const Op2& op2;
74
75 public:
76 MyVectorMul(const Op1& a, const Op2& b ): op1(a),·op2(b){}
77
78 ··T·operator[](const·std::size_t·i)·const{·
79 ····return·op1[i]·*·op2[i];·
80 ··}
81
82 ··std::size_t·size()·const{·
p ();
84 }
85 };
86
92 }
93
99 }
100
101 // function template for < operator
106 os << cont[i] << ' ';
107 }
(/learn)
109 return os;
110 }
111
112 int main(){
113 /*
116
118
120
122 */
123 }
In the next lesson, we’ll look at the solutions to these exercises.

Back Next
Mark as Completed
Report an Issue
- Solutions
Let's have a look at the solutions to the exercises we had in the last lesson.
• Problem Statement 1: Solution
• Explanation
• Problem Statement 2: Solution
• Explanation
Problem Statement 1: Solution#

1 // vectorArithmeticExpressionTemplates.cpp
2
5 #include <vector>
6
8 class MyVector{
9 Cont cont;
10
11 public:
14
17
19 explicit MyVector(const Cont& other) : cont(other){}
20
23 MyVector& operator (const MyVector<T2 R2>& other){
https://www.educative.io/courses/generic-templates-in-cpp/Bn2ooJP7RMY 1/9
26 return *this;
27 }
28
32 }
33
36 return cont[i];
37 }
38
40 return cont[i];
41 }
42
45 return cont;
46 }
47
48 Cont& data(){
49 return cont;
50 }
51 };
52
56 const Op1& op1;
57 const Op2& op2;
58
59 public:
61
64 }
65
68 }
69 };
70
71 // MyVector - MyVector
72 template<typename T, typename Op1, typename Op2>
73 class MyVectorSub {
74 const Op1& op1;
75 const Op2& op2;
76
77 public:
78 MyVectorSub(const Op1& a, const Op2& b) : op1(a), op2(b) {}
79
80 T operator[](const std::size_t i) const {
81 return op1[i] - op2[i];
82 }
83
84 std::size_t size() const {
86 }
87 };
88
92 const Op1& op1;
93 const Op2& op2;
94
95 public:
96 MyVectorMul(const Op1& a, const Op2& b ): op1(a), op2(b){}
97
99 return op1[i] * op2[i];
100 }
101
104 }
105 };
106
107 // elementwise MyVector / MyVector
108 template< typename T, typename Op1, typename Op2 >
109 class MyVectorDiv {
110 const Op1& op1;
111 const Op2& op2;
112
113 public:
114 MyVectorDiv(const Op1& a, const Op2& b) : op1(a), op2(b) {}
115
116 T operator[](const std::size_t i) const {
117 return op1[i] / op2[i];
118 }
119
120 std::size_t size() const {
122 }
123 };
124
126 template<typename T typename R1 typename R2>
130 }
131
132 // function template for the - operator
134 MyVector<T, MyVectorSub<T, R1, R2> >
135 operator- (const MyVector<T, R1>& a, const MyVector<T, R2>& b) {
136 return MyVector<T, MyVectorSub<T, R1, R2> >(MyVectorSub<T, R1, R2 >(a.data(), b.da
137 }
138
144 }
145
146 // function template for the / operator
148 MyVector<T, MyVectorDiv< T, R1, R2> >
149 operator/ (const MyVector<T, R1>& a, const MyVector<T, R2>& b) {
150 return MyVector<T, MyVectorDiv<T, R1, R2> >(MyVectorDiv<T, R1, R2 >(a.data(), b.da
151 }
152
153 // function template for < operator
158 os << cont[i] << ' ';
159 }
161 return os;
162 }
163
164 int main(){
165
168
170
171 result = x + x + y * y - x + x + y / y;
172
174
175 }
176
Output 1.54s
117.89 117.89 117.89 117.89 117.89 117.89 117.89 117.89 117.89 117.89
Explanation#
We have implemented the overloaded - operator (line 133 - 137), the
overloaded / operator (line 147 - 151), and the overloaded output operator
(line 154 - 162). Now, the objects x , y , and result feel like numbers.
Problem Statement 2: Solution#

4 #include <vector>
5
7 class MyVector{
8 Cont cont;
9
10 public:
13
16
18 explicit MyVector(const Cont& other) : cont(other){}
19
25 return *this;
26 }
27
31 }
32
35 return cont[i];
36 }
37
39 return cont[i];
40 }
41
44 return cont;
45 }
46
47 Cont& data(){
48 return cont;
49 }
50 };
51
52 template<template<typename> class Oper, typename T, typename Op1 , typename Op2>
53 class MyVectorCalc{
54 const Op1& op1;
55 const Op2& op2;
56 Oper<T> oper;
57
58 public:
59 MyVectorCalc(const Op1& a, const Op2& b): op1(a), op2(b) {}
60
62
63 return oper(op1[i], op2[i]);
64 }
65
68 }
69 };
70
73 MyVector<T, MyVectorCalc<std::plus, T, R1, R2> >
75 return MyVector<T, MyVectorCalc<std::plus, T, R1, R2> >(MyVectorCalc<std::plus, T,
76 }
77
78 // function template for the - operator
80 MyVector<T, MyVectorCalc<std::minus, T, R1, R2> >
81 operator- (const MyVector<T, R1>& a, const MyVector<T, R2>& b) {
82 return MyVector<T, MyVectorCalc<std::minus, T, R1, R2> >(MyVectorCalc<std::minus,
83 }
84
87 MyVector<T, MyVectorCalc<std::multiplies, T, R1, R2> >
89 return MyVector<T, MyVectorCalc<std::multiplies, T, R1, R2> >(MyVectorCalc<std::mu
90 }
91
92 // function template for the / operator
94 MyVector<T, MyVectorCalc<std::divides, T, R1, R2> >
95 operator/ (const MyVector<T, R1>& a, const MyVector<T, R2>& b) {
96 return MyVector<T, MyVectorCalc<std::divides, T, R1, R2> >(MyVectorCalc<std::divid
97 }
98
104 os << cont[i] << ' ';
105 }
107 return os;
108 }
109
110 int main(){
111
114
116
117 result = x + x + y * y - x + x + y / y;
118
120
121 }
122
Output 2.52s
117.89 117.89 117.89 117.89 117.89 117.89 117.89 117.89 117.89 117.89
(/learn)
Explanation#
In contrast to the previous example
161208496128#example-2-vector-arithmetic-based-on-expression-templates),
in which each arithmetic operation such as addition ( MyVectorAdd ),
subtraction ( MyVectorSub ), multiplication ( MyVectorMult ), or division
( MyVectorDiv ) is represented in a type, this improved version uses a generic
binary operator ( MyVectorCalc ) in lines 52 – 69. This generic binary operator
requires the concrete binary operator such as std::plus to become the
concrete binary operator. The predefined function objects
(https://en.cppreference.com/w/cpp/utility/functional) std::plus ,
std::minus , std::multiplies , and std::divides are part of the standard
template library. You can think of them as a lambda-function representing
the requested operation.
In the next lesson, we’ll study policy and traits in idioms and patterns.

Interviewing soon? We ve partnered with Hired so that companies apply to
Back Next
- Exercises Idioms and Patterns: Policy and Traits
Mark as Completed
Report an Issue
3/20/22, 6:41 PM Idioms and Patterns: Policy and Traits - Generic Programming Templates in C++
Idioms and Patterns: Policy and Traits

In this lesson, we will study about policy and traits in idioms and patterns.
• Policy and Traits
• Policy
• Traits
Policy and Traits#

Policy#
A Policy is a generic function or class with adaptable behavior.
Policy parameters have typically default values.
This adaptable behavior is expressed in several type parameters, the so-

called policy parameters. Due to different policy parameters, the concrete
generic function or class behaves differently.
Typical examples for policies are the containers of the Standard Template
Library such as std::vector , or std::unordered_map :
std::vector has a default policy for allocating memory, which is based

on the type of the element: std::allocator<T>
https://www.educative.io/courses/generic-templates-in-cpp/YMNvvpWGrAK 1/3
std::unordered_map has a default policy for generating the hash value

( std::hash<Key> ), comparing two keys ( std::equal_to<Key> ), and
allocating memory ( std::allocator<std::pair<const Key, T> ). The hash
function and the comparison function are based on the key.
template<class T, class Allocator = std::allocator<T>>
class vector;
template<
class Key,
class T,
class Hash = std::hash<Key>,
class KeyEqual = std::equal_to<Key>,
class Allocator = std::allocator<std::pair<const Key, T>>>
class unordered_map;
Traits#
(/learn)
Traits are class templates, which provide characteristics of a generic type.
template< class T >
struct is_integral;
template<T>
struct iterator_traits<T*> {
using difference_type = std::ptrdiff_t;
using value_type = T;
using pointer = T*;
using reference = T&;
using iterator_category = std::random_access_iterator_tag;
};
Traits can extract one or more characteristics of a class template.
The function std::is_integral<T> from the type-traits library

determines, if T is an integral type.
determines, if T is an integral type.
In the next lesson, we’ll look at a few examples of policy and traits in idioms
and patterns.

Back Next
- Solutions - Examples
Mark as Completed
Report an Issue
- Examples
Let's have a look at the examples of policy and traits in this lesson.
• Example 1: Templates Policy
• Explanation
• Example 2: Templates Traits
• Explanation
Example 1: Templates Policy#

1 // PolicytemplatesPolicy.cpp
2
5
6 struct MyInt{
7 explicit MyInt(int v):val(v){}
8 int val;
9 };
10
11 struct MyHash{
12 std::size_t operator()(MyInt m) const {
13 std::hash<int> hashVal;
14 return hashVal(m.val);
15 }
16 };
17
18 struct MyEqual{
19 bool operator () (const MyInt& fir, const MyInt& sec) const {
20 return fir.val == sec.val;
21 }
22 };
23
https://www.educative.io/courses/generic-templates-in-cpp/YM0MpgJ8r69 1/5
23
24 std::ostream& operator << (std::ostream& strm, const MyInt& myIn){
25 strm << "MyInt(" << myIn.val << ")";
26 return strm;
27 }
28
29 int main(){
30
32
33 typedef std::unordered_map<MyInt, int, MyHash, MyEqual> MyIntMap;
34
35 std::cout << "MyIntMap: ";
36 MyIntMap myMap{{MyInt(-2), -2}, {MyInt(-1), -1}, {MyInt(0), 0}, {MyInt(1), 1}};
37
38 for(auto m : myMap) std::cout << '{' << m.first << ", " << m.second << "}";
39
40 std::cout << "\n\n";
41
42 }
43
Output 1.45s
MyIntMap: {MyInt(1), 1}{MyInt(0), 0}{MyInt(-1), -1}{MyInt(-2), -2}
Explanation#
The example uses the user-defined type MyInt as key for an
std:unordered_map . To use MyInt as a key, MyInt must support the hash
value and must be equal comparable. The classes MyHash (lines 11 – 16) and
MyEqual (lines 18 – 22) provides the functionality for MyInt by delegating the
job to the underlying int . The typedef in line 33 brings all together. In line
36 i d k i
https://www.educative.io/courses/generic-templates-in-cpp/YM0MpgJ8r69 Th l 2/5
36, MyInt is used as a key in a std::unordered_map . The class
std::unoredered_map in line 33 is an example of the policy class. The

adaptable behavior is, in this case, the hash
function and the equal function.
Both policy parameters have default but can also be specified.
Example 2: Templates Traits#

1 // TemplatesTraits.cpp
2
5
7
10
12
27
28 cout << endl;
29
30 }
31
32 int main(){
33 getPrimaryTypeCategory<void>();
34 }
35
(/learn)
Output 1.5s
Explanation#
In the example above, we have defined the function getPrimaryTypeCategory
in line 9 which takes a type T and determines which type category T
belongs to. This example uses all 14 primary type categories of the type-traits
library.
In the next lesson, we’ll solve a few exercises to understand more on policy
and traits in idioms and patterns.

Back Next
Idioms and Patterns: Policy and Traits - Exercises
Mark as Completed
Report an Issue
- Exercises
Let's solve a few exercises on policy and traits in this lesson.
Extend the example
113329324032#example-1-templates-policy) below in such a way such that
the type MyInt can be used as a key to an std::map .
1 // PolicytemplatesPolicy.cpp
2
5
6 struct MyInt{
8 int val;
9 };
10
11 struct MyHash{
15 }
16 };
17
https://www.educative.io/courses/generic-templates-in-cpp/Y5RrPKjOyvA 1/5
17
18 struct MyEqual{
21 }
22 };
23
26 return strm;
27 }
28
29 // write your functions here
30
31 int main(){
32
33 // uncomment these line after implementing these functionalities
34
35 /*
37
38 typedef std::unordered_map<MyInt, int, MyHash, MyEqual> MyUnorderedMap;
39
40 std::cout << "MyUnorderedMap: ";
41 MyUnorderedMap myMap{{MyInt(-2), -2}, {MyInt(-1), -1}, {MyInt(0), 0}, {MyInt(1),
42
44
46
47 typedef std::map<MyInt, int, MySmaller> MyOrderedMap;
48
49 std::cout << "MyOrderedMap: ";
50 MyOrderedMap myMap2{{MyInt(-2), -2}, {MyInt(-1), -1}, {MyInt(0), 0}, {MyInt(1),
51
52 for(auto m : myMap2) std::cout << '{' << m.first << ", " << m.second << "}";
53
54 std::cout << "\n\n";
55 */
56 }
Define a std::map<std::string, int> , in which the keys are sorted in a

decreasing way.
2 #include <map>
3
4 int main() {
5
6 // your code goes here
7
8 }
The function template templateTraits.cpp
113329324032#example-2-templates-traits) shows for which primary type
category each type belongs to. Each type is in exactly one primary type
category. Extend the program so that you get a type for each type category.
1 // TemplatesTraits.cpp
2
5
7
10
12 (/learn)
27
28 cout << endl;
29
30 }
31
32 int main(){
33 // check for all types for calling type-traits fucntions
34
35 }
We’ll look at the solutions of these exercises in the next lesson.

Back Next
Mark as Completed
Mark as Completed
Report an Issue
- Solutions
Let's look at the solution reviews of previously solved exercises.
• Explanation
• Explanation
• Explanation

Statement 1#
1 // templatePolicyMap.cpp
2
4 #include <map>
6
7 struct MyInt{
9 int val;
10 };
11
12 struct MyHash{
16 }
https://www.educative.io/courses/generic-templates-in-cpp/3j9Rp85JBV4 1/7
16 }
17 };
18
19 struct MyEqual{
22 }
23 };
24
25 struct MySmaller{
27 return fir.val < sec.val;
28 }
29
30 };
31
34 return strm;
35 }
36
37 int main(){
38
40
41 typedef std::unordered_map<MyInt, int, MyHash, MyEqual> MyUnorderedMap;
42
43 std::cout << "MyUnorderedMap: ";
44 MyUnorderedMap myMap{{MyInt(-2), -2}, {MyInt(-1), -1}, {MyInt(0), 0}, {MyInt(1),
45
47
49
50 typedef std::map<MyInt, int, MySmaller> MyOrderedMap;
51
52 std::cout << "MyOrderedMap: ";
53 MyOrderedMap myMap2{{MyInt(-2), -2}, {MyInt(-1), -1}, {MyInt(0), 0}, {MyInt(1),
54
55 for(auto m : myMap2) std::cout << '{' << m.first << ", " << m.second << "}";
56
57 std::cout << "\n\n";
58
59 }
60
Output 2.11s
MyUnorderedMap: {MyInt(1), 1}{MyInt(0), 0}{MyInt(-1), -1}{MyInt(-2), -2}
MyOrderedMap: {MyInt(-2), -2}{MyInt(-1), -1}{MyInt(0), 0}{MyInt(1), 1}
Explanation#
In order to use MyInt as a key in an std::map , MyInt has to support an
ordering. The class MySmaller supports ordering on MyInt . The typedef in
line 50 allows us to use MyInt as a key (line 53) in MyOrderedMap .

Statement 2#
1 // templatePolicyDecreasing.cpp
2
4 #include <map>
5
6
7 int main(){
8
10
11 std::map<std::string, int, std::greater<std::string>> myDecreaseMap{{"Grimm", 1}
12
13 std::cout << "myDecreaseMap: ";
14 for(auto m : myDecreaseMap) std::cout << '{' << m.first << ", " << m.second << "
15
16 std::cout << "\n\n";
17
18 }
19
Output 1.41s
myDecreaseMap: {Schmitt, 3}{Meyer, -2}{Huber, 2}{Grimm, 1}
Explanation#
In the above example, we have stored in line 11 the keys in increasing order
in the std::map . In line 14, we have used for loop to print the element
stored in the myDecreaseMap .

Statement 3#
1 // TemplateTraitsPrimary.cpp
2
5
7
10
12
_ g_p _ g_p ;
27
28 cout << endl;
29
30 }
31
32 int main(){
33
34 getPrimaryTypeCategory<void>();
35 getPrimaryTypeCategory<short>();
36 getPrimaryTypeCategory<double>();
37 getPrimaryTypeCategory<int []>();
38 getPrimaryTypeCategory<int*>();
39 getPrimaryTypeCategory<int&>();
40 struct A{
41 int a;
42 int f(double){return 2011;}
43 };
44 getPrimaryTypeCategory<int A::*>();
45 getPrimaryTypeCategory<int (A::*)(double)>();
46 enum E{
47 e= 1,
48 };
49 getPrimaryTypeCategory<E>();
50 union U{
51 int u;
52 };
53 getPrimaryTypeCategory<U>();
54 getPrimaryTypeCategory<string>();
55 getPrimaryTypeCategory<int * (double)>();
56 getPrimaryTypeCategory<int&>();
57 getPrimaryTypeCategory<int&&>();
58
59 }
60
(/learn)
Output 3.24s
Explanation#
In the example mentioned above, we have defined the function
getPrimaryTypeCategory in line 9 which takes the built-in type of C++ and
returns true in front of the type we have defined in lines (13 – 26). We’re
checking for each type by passing the type in the function.
Let’s move on to tag dispatching in idioms and patterns in the next lesson.

Back Next
- Exercises Idioms and Patterns: Tag Dispatching

Exercises Idioms and Patterns: Tag Dispatching
Mark as Completed
Report an Issue
3/20/22, 6:46 PM Idioms and Patterns: Tag Dispatching - Generic Programming Templates in C++
Idioms and Patterns: Tag Dispatching

In this lesson, we'll learn about tag dispatching in idioms and patterns.
• Tag Dispatching
Tag Dispatching#
Tag Dispatching enables us to choose a function based on type
characteristics.
The decision takes place at compile-time.

Traits make the decision based on specific properties of an argument.
(/learn)
The main benefit is performance.
The iterator categories from the Standard Template Library are a typical
use-case:
struct input_iterator_tag{};
struct output_iterator_tag{};
struct forward_iterator_tag: public input_iterator_tag{};
struct bidirectional_iterator_tag: public forward_iterator_tag{};
struct random_access_iterator_tag: public bidirectional_iterator_tag{};
In the concrete case, std::random_access_iterator is a refinement of

std::bidirectional_iterator , std::birectional_iterator is a refinement of
std::forward_iterator , and std::forward_iterator is a refinement of
std::input iterator
https://www.educative.io/courses/generic-templates-in-cpp/3jmLBpA223M 1/2
3/20/22, 6:46 PM Idioms and Patterns: Tag Dispatching - Generic Programming Templates in C++
std::input_iterator .
In the next lesson, we’ll look at an example of tag dispatching in idioms and
patterns.

Back Next
- Solutions - Example
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/3jmLBpA223M 2/2
- Example
Let's have a look at an example of tag dispatching.
• Example: Templates Tag Dispatching
• Explanation
Example: Templates Tag Dispatching#

1 // TemplatesTagDispatching.cpp
2
3 #include <iterator>
4 #include <forward_list>
5 #include <list>
6 #include <vector>
8
9 template <typename InputIterator, typename Distance>
10 void advance_impl(InputIterator& i, Distance n, std::input_iterator_tag) {
11 std::cout << "InputIterator used" << std::endl;
12 while (n--) ++i;
13 }
14
15 template <typename BidirectionalIterator, typename Distance>
16 void advance_impl(BidirectionalIterator& i, Distance n, std::bidirectional_iterator_
17 std::cout << "BidirectionalIterator used" << std::endl;
18 if (n >= 0)
19 while (n--) ++i;
20 else
21 while (n++) --i;
22 }
23
24 template <typename RandomAccessIterator, typename Distance>
25 void advance_impl(RandomAccessIterator& i, Distance n, std::random_access_iterator_t
26 std::cout << "RandomAccessIterator used" << std::endl;
https://www.educative.io/courses/generic-templates-in-cpp/JERJp9zLLlo 1/3
27 i += n;
28 }
29
31 void advance_(InputIterator& i, Distance n) {
32 typename std::iterator_traits<InputIterator>::iterator_category category;
33 advance_impl(i, n, category);
34 }
35
36 int main(){
37
38 }
39
(/learn)
Succeeded
Explanation#
The expression std::iterator_traits::iterator_category category in line 32
determines the iterator category at compile-time. Based on the iterator
category, the most specific variant of the function template advance_impl(i,
n, category) is used in line 33. Each container returns an iterator of the
iterator category which corresponds to its structure.
In the next lesson, we’ll solve an exercise for tag dispatching in idioms and
patterns.

// /
https://www.educative.io/courses/generic-templates-in-cpp/JERJp9zLLlo / 2/3
Back Next
Idioms and Patterns: Tag Dispatching - Exercise
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/JERJp9zLLlo 3/3
- Exercise
Let's solve the exercise on tag dispatching in this lesson.
Problem Statement#
Extend the example
223903313920) template for Tag Dispatching given below. The following
program shows a typical example of tag dispatching on iterators. Call the
advance_ algorithm in the main function with different containers such as
std:.vector , std::list , and ::forward_list to put the iterator 5 positions
further.
1 // TemplatesTagDispatching.cpp
2
5 #include <list>
6 #include <vector>
8
12 while (n--) ++i;
13 }
14
16 id d i l(Bidi ti lIt
https://www.educative.io/courses/generic-templates-in-cpp/qVjxlNmKJkp
t & i Di t td bidi ti l it t 1/3
18 if (n >= 0)
19 while (n--) ++i;
20 else
21 while (n++) --i;
22 }
23
(/learn)
27 i += n;
28 }
29
34 }
35
36 int main(){
37 // call the above functions to move them 5 position further
38 }
39

https://www.educative.io/courses/generic-templates-in-cpp/qVjxlNmKJkp 2/3
Back Next
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/qVjxlNmKJkp 3/3
- Solution
Let's have a look at the solution to the last exercise in this lesson.
• Solution Review
• Explanation
• Performance
Solution Review#
1 // templatesTagDispatchingImplementation.cpp
2
5 #include <list>
6 #include <vector>
8
12 while (n--) ++i;
13 }
14
18 if (n >= 0)
19 while (n--) ++i;
20 else
21 while (n++) --i;
22 }
23
(
https://www.educative.io/courses/generic-templates-in-cpp/qZzDY8XBMPk 1/4
27 i += n;
28 }
29
34 }
35
36 int main(){
37
39
40 std::vector<int> myVec{0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
41 auto myVecIt = myVec.begin();
42 std::cout << "*myVecIt: " << *myVecIt << std::endl;
43 advance_(myVecIt, 5);
44 std::cout << "*myVecIt: " << *myVecIt << std::endl;
45
47
48 std::list<int> myList{0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
49 auto myListIt = myList.begin();
50 std::cout << "*myListIt: " << *myListIt << std::endl;
51 advance_(myListIt, 5);
52 std::cout << "*myListIt: " << *myListIt << std::endl;
53
55
56 std::forward_list<int> myForwardList{0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
57 auto myForwardListIt = myForwardList.begin();
58 std::cout << "*myForwardListIt: " << *myForwardListIt << std::endl;
59 advance_(myForwardListIt, 5);
60 std::cout << "*myForwardListIt: " << *myForwardListIt << std::endl;
61
63
64 }
65
Output 1.56s
*myVecIt: 0
RandomAccessIterator used
*myVecIt: 5
*myListIt: 0
BidirectionalIterator used
*myListIt: 5
Explanation#
(/learn)
The expression std::iterator_traits::iterator_category category in line 32

determines the iterator category at compile-time. Based on the iterator
category, the most specific variant of the function template advance_impl(i,
n, category) is used in line 33. Each container returns an iterator of the
iterator category which corresponds to its structure. Therefore, line 41 gives
a random-access iterator, line 49 gives a bidirectional iterator, and line 57
gives a forward iterator, which is also an input iterator.
Performance#
From the performance point of view, this distinction makes a lot of sense
because a random-access iterator can be more quickly incremented than a
bidirectional iterator, and a bidirectional iterator can be more quickly
incremented than an input iterator. From the users perspective, we invoked
std::advance(it, 5) in lines 43, 51, and 59 to get the fastest version of the
function template which our container satisfies.
Let’s move on to type erasure idioms and patterns in the next lesson.

Back Next
- Exercise Idioms and Patterns: Type Erasure
Mark as Completed
Report an Issue
3/20/22, 6:49 PM Idioms and Patterns: Type Erasure - Generic Programming Templates in C++
Idioms and Patterns: Type Erasure

In this lesson, we'll learn about type erasure in detail.
• Type Erasure
• Typical Use Case

• Case 1: Void Pointers
• Case 2: Object-Orientation
• Case 3: std::function
Type Erasure#
Type Erasure enables you to use various concrete types through a single
generic interface.
Type erasure is duck typing applied in C++
“When I see a bird that walks like a duck and swims like
a duck and quacks
like a duck, I call that bird a duck.” (James Whitcomb Rileys
(https://en.wikipedia.org/wiki/James_Whitcomb_Riley))
Of course, you’ve already used type erasure in C++ or C. The C-ish way of
type erasure is a void pointer; the C++ish way of type erasure is object-
orientation.
https://www.educative.io/courses/generic-templates-in-cpp/gk3xZy0yOYD 1/5
Typical Use Case#

Type erasure can be performed with void pointers, object-orientation, or
templates.
Case 1: Void Pointers#

void qsort(void *ptr, std::size_t count, std::size_t size, cmp);
Before the qsort() , consider using:
int cmp(const void *a, const void *b);
The comparison function cmp should return a
negative integer: the first argument is less than the second
zero: both arguments are equal
positive integer: the first argument is greater than the second
Thanks to the void pointer, std::qsort is generally applicable but also quite
error-prone.
Maybe you want to sort a std::vector<int> , but you used a comparator for
C-strings. The compiler cannot catch this error because the type of
information was removed and we end up with undefined behavior.
In C++ we can do better.
Case 2: Object-Orientation#
Having a class hierarchy and using the BaseClass pointer instead of concrete
types is one way to enable type erasure
std::vector<const BaseClass*> vec;
The vec object has a pointer to a constant BaseClasses .
Case 3: std::function #
std::function as a polymorphic function wrapper is a nice example of type
erasure in C++ . std::function can accept everything, which behaves like a
function. To be more precise, this can be any callable such as a function, a
function object, a function object created by std::bind , or just a lambda
function.
Let’s have a look at the difference between the three types listed above in the
following table:
Each Type-safe Easy to Common

Technique
Datatype Interface implment Base Class
void* Yes No Yes No
Object-
Orientatio No Yes Yes Yes
n
(/learn)
Templates Yes Yes No No
1 class Object {
2 public:
3 ...
4 struct Concept {
5 virtual ~Concept() {}
6 };
7
8 template< typename T >
9 struct Model : Concept {
10 ...
11 };
11 };
12
13 std::shared_ptr<const Concept> object;
14 };
The small code snippet shows the structure of type erasure with templates.
The components are:
Object : Wrapper for a concrete type
object : Pointer to the concept
Concept : Generic interface
Model : Concrete class
Don’t be irritated by the names Object , object , Concept , and Model . They
are typically used for type erasure in the literature, so we stick to them. We
will see a more explained version of this in the example in the next lesson.
In the next lesson, we’ll look at a couple of examples of type erasure.

Back Next
Mark as Completed
Report an Issue
- Examples
In this lesson, we'll look at a couple of examples of type erasure.
• Example 1: Type Erasure Using Object-Oriented Programming
• Explanation
• Pros:
• Cons:
• Example 2: Type Erasure with Templates
• Explanation
Example 1: Type Erasure Using

Object-Oriented Programming#
1 // typeErasureOO.cpp
2
4 #include <string>
5 #include <vector>
6
7 struct BaseClass{
8 virtual std::string getName() const = 0;
9 };
10
11 struct Bar: BaseClass{
12 std::string getName() const override {
13 return "Bar";
14 }
15 };
16
https://www.educative.io/courses/generic-templates-in-cpp/mEZB4qG1Ox9 1/6
16
17 struct Foo: BaseClass{
18 std::string getName() const override{
19 return "Foo";
20 }
21 };
22
23 void printName(std::vector<BaseClass*> vec){
24 for (auto v: vec) std::cout << v->getName() << std::endl;
25 }
26
27
28 int main(){
29
31
32 Foo foo;
33 Bar bar;
34
35 std::vector<BaseClass*> vec{&foo, &bar};
36
37 printName(vec);
38
40 }
Output 3.67s
Foo
Bar
Explanation#
The key point is that you can use instances of Foo or Bar instead of an
instance for BaseClass . std::vector<BaseClass*> (line 35) has a pointer to
BaseClass . Well, actually it has two pointers to BaseClass (derived) objects
BaseClass . Well, actually it has two pointers to BaseClass (derived) objects
and it is a vector of such pointers.

BaseClass is an abstract base class, which
is used in line 23. Foo and Bar (lines 11 and 17) are the concrete classes.
What are the pros and cons of this implementation with object-orientation?
Pros:#
Type-safe
Easy to implement
Cons:#
Virtual dispatch
Intrusive because the derived class must know about its base class
Example 2: Type Erasure with

Templates#
1 // TypeErasure.cpp
2
4 #include <memory>
5 #include <string>
6 #include <vector>
7
8 class Object {
9
10 public:
12 explicit Object(T&& obj): object(std::make_shared<Model<T>>(std::forward<T>(obj)
13
15 return object->getName();
16 }
17
18 struct Concept {
19 virtual ~Concept() {}
20 virtual std::string getName() const = 0;
21 };
21 };
22
23 template< typename T >
24 struct Model : Concept {
25 explicit Model(const T& t) : object(t) {}
26 std::string getName() const override {
27 return object.getName();
28 }
29 private:
30 T object;
31 };
32
33 std::shared_ptr<const Concept> object;
34 };
35
36
37 void printName(std::vector<Object> vec){
38 for (auto v: vec) std::cout << v.getName() << std::endl;
39 }
40
41 struct Bar{
43 return "Bar";
44 }
45 };
46
47 struct Foo{
49 return "Foo";
50 }
51 };
52
53 int main(){
54
56
57 std::vector<Object> vec{Object(Foo()), Object(Bar())};
58
59 printName(vec);
60
62 }
Output 1.64s
Output 1.64s
Foo
(/learn)
Bar
Explanation#
First of all, the std::vector uses instances (line 57) of type Object and not
pointers like in the previous object-oriented example. These instances can be
created with arbitrary types because it has a generic constructor (line 12).
The Object has the getName method (line 14) which is directly forwarded to
the getName of the object. object is of type std::shared_ptr<const Concept> .
The emphasis should not lay on the empty but on the virtual destructor in
line 19. When the std::shared_ptr<const Concept> goes out of scope, the
destructor is called. The static type of object is std::shared_ptr<const
Concept> but the dynamic type std::shared_ptr<Model<T>> . The virtual
destructor guarantees that the correct destructor is called. This means, in
particular, the destructor of the dynamic type. The getName method of
Concept is pure virtual (line 18), therefore, due to virtual dispatch, the
getName method of Model (line 24) is used. In the end, the getName methods
of Bar and Foo (lines 42 and 48) are applied in the printName function (line
37).
In the next lesson, we’ll solve an exercise on type erasure.

utm_source educative&utm_medium
3/20/22, 6:50 PM
partner&utm_campaign March_2022
- Examples - Generic Programming Templates in C++
Back Next
Idioms and Patterns: Type Erasure - Exercise
Mark as Completed
Report an Issue
- Exercise
In this lesson, we'll solve an exercise on type erasure.
Problem Statement#
Compare the implementations of type erasure:
Implement a type deduction system with the help of type erasure which
returns a string representation corresponding to each data type. You need to
write the function for int , long long , char , void , and float .
3 #include <memory>
4 #include <string>
6 #include <vector>
7
8 // Implement the object struct
9
10 int main() {
11 // uncomment these lines after implementing object struct
12
13 /*
(/learn)
14 printType(Object::Model<int>{});
15 printType(Object::Model<double>{});
16 printType(Object::Model<void>{});
17 printType(Object::Model<Test>{});
18 printType(Object::Model<Object>{});
19 printType(Object::Model<long long>{});
(
https://www.educative.io/courses/generic-templates-in-cpp/q2W30m1wg1D {}) 1/2
20 printType(Object::Model<std::string>{});
21 printType(Object::Model<std::vector<int>>{});
22 */
23 }

Back Next
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/q2W30m1wg1D 2/2
- Solution
Let's have a look at the solution of the last exercise.
• Solution Review
• Explanation
Solution Review#
1 // TypeErasureTypeInfo.cpp
2
4 #include <memory>
5 #include <string>
7 #include <vector>
8
9 struct Object{
10 std::string getTypeName() const { return _inner->getTypeName(); }
11 struct Concept{
12 using ptr = std::unique_ptr<Concept>;
13 virtual std::string getTypeName() const = 0;
14 };
16 struct Model : Concept{
17 std::string getTypeName() const override { return typeid(T).name(); }
18 };
19 private:
20 typename Concept::ptr _inner;
21 };
22
23 template <>
24 struct Object::Model<long long> : Object::Concept{
25 std::string getTypeName() const override { return "long long";}
26 };
https://www.educative.io/courses/generic-templates-in-cpp/qVpxBkEAVL2 1/4
27
28 template <>
29 struct Object::Model<std::string> : Object::Concept{
30 std::string getTypeName() const override { return "std::string<char>";}
31 };
32
33 template <>
34 struct Object::Model<std::vector<int>> : Object::Concept{
35 std::string getTypeName() const override { return "std::vector<int>";}
36 };
37
38
40 void printType(T&& t){
41 std::cout << t.getTypeName() << std::endl;
42 }
43
44 class Test{};
45
46 int main(){
47
48 printType(Object::Model<int>{});
49 printType(Object::Model<double>{});
50 printType(Object::Model<void>{});
51 printType(Object::Model<Test>{});
52 printType(Object::Model<Object>{});
53 printType(Object::Model<long long>{});
54 printType(Object::Model<std::string>{});
55 printType(Object::Model<std::vector<int>>{});
56
57 }
Output 1.4s
d
(/learn)
v
4Test
6Object
long long
std::string<char>
std::vector<int>
Explanation#
The function template printType returns a string representation for each
type. In the general case, the primary template Model (lines 15 – 18) is used.
The implementation uses the typeid operator to get the name. Full
specializations are also available for the types long long (lines 23 – 26),
std::string (lines 28 – 31), and int (lines 33 – 36). The full specializations
provide personalized string representations.
In this chapter, we have learned about design techniques of C++ templates.

In the next chapter, we’ll look at the future concepts in C++20.

Back Next
- Exercise Overview
Mark as Completed
Report an Issue
3/20/22, 6:52 PM Overview - Generic Programming Templates in C++
Overview
In this lesson, we'll look at an overview of what concepts are needed in C++20.
• Too Specific Versus Too Generic
• Too Specific
• Narrowing Conversion
• Integral Promotion
• Too Generic
• Concepts to Our Rescue
• Advantages
Until C++20, we have two diametral ways to think about functions or user-
defined types (classes). Functions or classes can be defined on specific types
or on generic types. In the second case, we call them to function or class
templates.
There exist two extremes while talking about functions in C++. Let’s take a
closer look at two function implementations:
Too Specific Versus Too Generic#
Too Specific Too Generic

https://www.educative.io/courses/generic-templates-in-cpp/xoLVB57lMrq 1/8
#include <iostream>
#include <iostream>
void needInt(int i){ // 1
template<typenameT>
std::cout<< i << std::endl;
T gcd(T a, T b){
}
if( b == 0 ){

return a;
int main(){
}else{
double d{1.234};
return gcd(b, a % b);
needInt(d);
}
}
}

int main(){

std::cout<< gcd(100, 10) << st

d::endl;

std::cout<< gcd(3.5, 4.0) << st

d::endl;
Too Specific#
It’s quite a job to define for each specific type a function or a class. To avoid
that burden, type conversion comes often to our rescue but it is also part of
the problem.
Narrowing Conversion#
We have a function needInt(int a) in line 3 which we can be invoked with a
double . Now narrowing conversion takes place.
2
3 void needInt(int i){
4 std::cout<< i << std::endl;
5 }
5 }
6
7 int main(){
8 double d{1.234};
9 needInt(d);
10 }
Output 1.52s
We assume that this is not the behavior we wanted. We started with a

double and ended with an int .
But conversion also works the other way around.
Integral Promotion#
We have a user-defined type MyHouse . An instance of MyHouse can be
constructed in two ways. When invoked without any argument in line 5, its
attribute family is set to an empty string. This means the house is still empty.
To quickly check if the house is empty or full, we implemented a conversion
operator to bool in line 8.
2 #include <string>
3
4 struct MyHouse{
5 MyHouse() = default;
6 MyHouse(const std::string& fam): family(fam){}
7
8 operator bool(){ return !family.empty(); }
9
10 std::string family = "";
10 std::string family = ;
11 };
12
13 void needInt(int i){
14 std::cout << "int: " << i << std::endl;
15 }
16
17 int main(){
18
20
21 MyHouse firstHouse;
22 if (!firstHouse){
23 std::cout << "The firstHouse is still empty." << std::endl;
24 };
25
26 MyHouse secondHouse("grimm");
27 if (secondHouse){
28 std::cout << "Family grimm lives in secondHouse." << std::endl;
29 }
30
32
33 needInt(firstHouse);
34 needInt(secondHouse);
35
37
38 }
Output 2.46s
The firstHouse is still empty.
Family grimm lives in secondHouse.
int: 0
int: 1
Now, instances of MyHouse can be used, when an int is required. Strange!
Because of the overloaded operator bool in line 8, instances of MyHouse can

be used as an int and can, therefore, be used in arithmetic expressions:
auto res = MyHouse() + 5 . This was not our intention! But we have noted it
just for completeness. With C++11, we should declare those conversion
operators as explicit. Therefore, implicit conversions will not take place.
Too Generic#
Generic functions or classes can be invoked with arbitrary values. If the
values do not satisfy the requirements of the function or class, it is not a
problem because you will get a compile-time error.
2
4 T gcd(T a, T b){
5 if( b == 0 ){ return a; }
6 else{
8 }
9 }
10
11 int main(){
12
14
15 std::cout << gcd(100, 10) << std::endl;
16
17 std::cout << gcd(3.5, 4.0)<< std::endl;
18 std::cout << gcd("100", "10") << std::endl;
19
21
22 }
Output 1.97s

binary 'operator%'
7 | return gcd(b, a % b);

| ~~^~~

What is the problem with this error message?
Of course, it is quite long and quite difficult to understand but our crucial
concern is a different one. The compilation fails because neither double nor
the C-string supports the % operator. This means the error is due to the
failed template instantiation for double and C-string. This is too late and,
therefore, really bad. No template instantiation for type double or C-strings
(/learn)
should even be possible. The requirements for the arguments should be part
of the function declaration and not a side-effect of an erroneous template
instantiation.
Now concepts come to our rescue.
Concepts to Our Rescue#

With concepts, we get something in between. We can define functions or
classes which act on semantic categories. Meaning, the arguments of
functions or classes are neither too specific nor too generic but named sets
of requirements such as Integral .
Advantages#
Express the template parameter requirements as part of the interface
Support the overloading of functions and the specialization of class

templates
Generate drastically improved error messages by comparing the

requirements of the template parameter with the template arguments
Use them as placeholders for generic programming
Empowers you to define your concepts
Can be used for all kinds of templates
In the next lesson, we’ll learn about the history of C++ and talk about future
concepts.

Back Next
- Solution History
Mark as Completed
Report an Issue
3/20/22, 6:53 PM History - Generic Programming Templates in C++
History
In this lesson, we'll learn about the history of C++ and talk about future concepts.
• The Inspiration from Haskell
• Classical Concepts
• Concepts
• The Advantages of Concepts
The Inspiration from Haskell#

Type classes are interfaces for similar types. If a type is a member of a type
class, it has to have specific properties. Type classes play a similar role in
generic programming as interfaces play in object-oriented programming.
Here you can see a part of Haskell’s type classes hierarchy.
https://www.educative.io/courses/generic-templates-in-cpp/7DA9qmDpQZB 1/4
Haskell Type Class Hierarchy
Haskell’s type classes build a hierarchy. The type class Ord is a subclass of
the type class Eq . Therefore, instances of the type class Ord have to be
members of the type class Eq and in addition support the comparison
operators.
Classical Concepts#
The key idea of generic programming with templates is, to define functions
and classes that can be used with different types. But it will often happen
that you instantiate a template with the wrong type. The result may be a
cryptic error message that is many pages long It is sad but true: templates in
cryptic error message that is many pages long. It is sad but true: templates in
C++ are known for this. Therefore, classical concepts were planned as one of
the great features of C++11. They should allow you to specify constraints for
templates that can be verified by the compiler. Due to their complexity, they
were removed in July 2009 from the standard: “The C++0x concept design
evolved into a monster of complexity.” (Bjarne Stroustrup)
Concepts#
With C++20, we will get concepts. Although concepts are in the first
implementations simplified classical concepts, they have a lot to offer.
The Advantages of Concepts#

(/learn)
Concepts
Empower the programmer to express their requirements as part of the

interface directly.
Support the overloading of functions and the specialization of class
templates based on the requirements of the template parameters.
Produce drastically improved error messages by comparing the

requirements of the template parameter with the applied template
arguments.
Can be used as placeholders for generic programming.
Empower you to define your own concepts.
We will get the benefits without additional compile-time or runtime time

costs. Concepts are similar to Haskell’s type classes. Concepts describe
semantic categories and not syntactic restrictions. For types of the standard
library, we get library concepts
(https://en.cppreference.com/w/cpp/language/constraints) such as
DefaultConstructible , MoveConstructible , CopyConstructible ,
MoveAssignable , CopyAssignable , or Destructible . For the containers, we get
MoveAssignable , CopyAssignable , or Destructible . For the containers, we get
concepts such as ReversibleContainer , AllocatorAwareContainer ,
SequenceContainer , ContiguousContainer , AssociativeContainer , or

UnorderedAssociativeContainer . You can read more about concepts and their
constraints here: cppreference.com
(https://en.cppreference.com/w/cpp/named_req).
In the next lesson, we’ll learn about the details of functions and classes with
reference to concepts.

Back Next
Overview Functions and Classes
Mark as Completed
Report an Issue
3/20/22, 6:54 PM Functions and Classes - Generic Programming Templates in C++
Functions and Classes

Let's learn about coming C++ concepts in detail in this lesson.
• Functions
• Classes
• Methods of a Class
• More Requirements
• Overloading
• Specialization
Concepts are part of the template declaration.
Functions#
Using the concept Sortable .
Implicit The implicit version from the left is

syntactic sugar to the explicit
template<Sortable Cont>
version:
void sort(Cont& container){
...
Explicit
}
The container has to be Sortable .

https://www.educative.io/courses/generic-templates-in-cpp/RLP6yZOO1YK 1/5
template<typename Cont>
requires Sortable<Cont>()
void sort(Cont& container){
...
Sortable has to be a constant expression that is a predicate. That means that

the expression has to be evaluable at compile-time and has to return a
boolean.
If you invoke the sort algorithm with a container lst that is not sortable,
you will get a unique error message from the compiler.
Usage:
std::list<int> lst = {1998, 2014, 2003, 2011};
sort(lst); // ERROR: lst is no random-access container with <
You can use concepts for all kind of templates.
Classes#
We can define a class template MyVector that will only accept objects as
template arguments:
template<Object T>
class MyVector{};
MyVector<int> v1; // OK
MyVector<int&> v2 // ERROR: int& does not satisfy the constraint Object
Now the compiler complains that the reference (int&) is not an object
Now, the compiler complains that the reference (int&) is not an object.
MyVector can be further adjusted.
A reference is not an object.
Methods of a Class#
template<Object T>
class MyVector{
...
requires Copyable<T>()
void push_back(const T& e);
...
};
Now the method push_back from MyVector requires that the template
argument has to be copy-able. The concepts have to be placed before the
method declaration.
More Requirements#
A template can have more than one requirement for its template
parameters.
template <SequenceContainer S,
EqualityComparable<value_type<S>> T>
Iterator_type<S> find(S&& seq, const T& val){
...
The function template find has two requirements. On one hand, the
container has to store its elements in a linear arrangement
( SequenceContainer ), on the other hand, the elements of the container have
to be equally comparable: EqualityComparable<value_type<S>> .
Overloading#
Concepts support the overloading of functions.
template<InputIterator I>
void advance(I& iter, int n){...}
template<BidirectionalIterator I>
template<RandomAccessIterator I>
std::list<int> lst{1,2,3,4,5,6,7,8,9};
std::list<int>:: iterator i = lst.begin();
std::advance(i, 2); // BidirectionalIterator
(/learn)
The function template std::advance puts its iterator n positions further.
Depending if the iterator is a forward, a bidirectional, or a random access
iterator, different function templates will be used. In case of a std::list , the
BidirectionalIterator will be chosen.
Concepts also support the specialization of class templates.
Specialization#
template<typename T>
class MyVector{};
template<Object T>
class MyVector{};
MyVector<int> v1; // Object T
MyVector<int&> v2; // typename T
For MyVector<int&> v2 , the compiler uses the general template in the first
line; on the contrary, the compiler uses for MyVector<int> the specialization
template<Object T> class MyVector{} .
MyVector<int&> goes to the unconstrained template parameter.
MyVector<int> goes to the constrained template parameter.
In the next lesson, we’ll study the placeholder syntax.

Back Next
History Placeholder Syntax
Mark as Completed
Report an Issue
3/20/22, 6:54 PM Placeholder Syntax - Generic Programming Templates in C++
Placeholder Syntax
Let's learn about placeholder syntax in this lesson.
• Placeholder Syntax: auto
• Inconsistency in C++14
• Constrained and Unconstrained Placeholders
• Example:
With auto , C++11 has unconstrained placeholders. We can use concepts in

C++20 as constrained placeholders. Decisive quantum leap does not look so
thrilling at the first glimpse. C++ templates will become easy to use C++
features.
According to our definition, C++98 is not a consistent language. By

consistent, we mean that you have to apply a few rules to derived C++ syntax
from it. C++11 is something in between. For example, we have consistent
rules like initializing all with curly braces (see { } - Initialization
(https://www.modernescpp.com/index.php/initialization)). Of course, even
C++14 has a lot of features where we miss a consistent principle. One of the
favorites is the generalized lambda function.
Placeholder Syntax: auto #
https://www.educative.io/courses/generic-templates-in-cpp/gxvvA60vz4l 1/5
auto genLambdaFunction= [](auto a, auto b) {
return a < b;
};
template <typename T, typename T2> // 3
auto genFunction(T a, T2 b){ // 4
return a < b;
Inconsistency in C++14#
By using the placeholder auto for the parameter a and b , the generalized
lambda function becomes - in a magic way - a function template. We know,
genLambdaFunction is a function object that has an overloaded call operator
which accepts two type parameters. The genFunction is also a function
template. But wouldn’t it be nice to define a function template by just using
auto in a function definition? This would be consistent but is not possible.
Hence, we have to use a lot more syntax (line 3 and 4). That syntax is often
too difficult for a lot of C++ programmer.
Exactly that inconsistency will be removed with the placeholder syntax.

Therefore, we have a new simple principle and C++ will become - according
to my definition - a lot easier to use.
Generic Lambdas introduced a new way to define templates.
Constrained and Unconstrained

Placeholders#
We will get unconstrained and constrained placeholders. auto is an

unconstrained placeholder because a with auto defined variable can be of
any type. A concept is a constrained placeholder because it can only be used
to define a variable that satisfies the concept .
General Rule: Constrained Concepts can be used where unconstrained

templates ( auto ) are usable.
Let’s define and use a simple concept before we dig into the details.
Example:#
1 // conceptsPlaceholder.cpp
2
5 #include <vector>
6
8 concept bool Integral(){
9 return std::is_integral<T>::value;
10 }
11
12 Integral getIntegral(auto val){
13 return val;
14 }
15
16 int main(){
17
19
20 std::vector<int> myVec{1, 2, 3, 4, 5};
21 for (Integral& i: myVec) std::cout << i << " ";
23
24 Integral b= true;
25 std::cout << b << std::endl;
26
27 Integral integ= getIntegral(10);
28 std::cout << integ << std::endl;
28 std::cout << integ << std::endl;
29 (/learn)

30 auto integ1= getIntegral(10);
31 std::cout << integ1 << std::endl;
32
34
35 }
Output 2.34s
1 2 3 4 5
true
10
10
We have defined in line 8 the concept Integral . The concept Integral will
evaluate to true if the predicate std::is_integral<T>::value returns true
for T . std::is_integral<T> is a function of the type-traits
229176066048) library. The functions of the type-traits library enable,
amongst other things, that we can check types at compile-time. Hence, we
iterate over Integral 's in the range-based for-loop in line 21 and the
variable b in line 24 has to be Integral . Our usage of concepts goes on in
lines 27 and 30. We required in line 27 that the return type of getIntegral
(line 12) has to fulfill the concept Integral . We’re not so strict in line 30.
Here we’re fine with an unconstrained placeholder.
In the next lesson, we’ll discuss the predefined concepts in C++20.

Back Next
Functions and Classes Predefined Concepts
Mark as Completed
Report an Issue
3/20/22, 6:56 PM Predefined Concepts - Generic Programming Templates in C++
Predefined Concepts
Let's dive deep into predefined concepts of C++20 in this lesson.
• Predefined Concept
• Concepts Definition: Variable Concepts

• Concepts Definition: Function Concepts
• Concepts TS
• Concepts Draft
• The Concept Equal
Before moving on to predefined concepts, let’s get to know about Syntactic

Sugar.
Syntactic Sugar: This is from Wikipedia

(https://en.wikipedia.org/wiki/Syntactic_sugar): In computer science,
syntactic sugar is syntax within a programming language that is
designed to make things easier to read or to express. It makes the
language sweeter for human use: things can be expressed more clearly,
more concisely, or in an alternative style that some may prefer.
Predefined Concept#
We should use the predefined concepts. cppreference.com
(https://en cppreference com/w/cpp/named req) gives a great introduction to
https://www.educative.io/courses/generic-templates-in-cpp/YQEzLDAKWVM 1/7
(https://en.cppreference.com/w/cpp/named_req) gives a great introduction to
predefined concepts. Here are a few of them:
Core language concepts Object concepts
Same Destructible
DerivedFrom Constructible
ConvertibleTo DefaultConstructible
Common MoveConstructible
Integral Copy Constructible
Signed Integral Movable

Unsigned Integral Copyable
Assignable Semi-regular
Swappable Regular
Comparison concepts Callable concepts
Boolean Callable
EqualityComparable RegularCallable
StrictTotallyOrdered Predicate
Relation
StrictWeakOrder
There are two ways to define concepts: variable concepts and function
concepts. If we use a variable template for our concept, it’s called a variable
concept; in the second case a function concept.
Concepts Definition: Variable

Concepts#
concept bool Integral =
std::is_integral<T>::value;
We have defined the concept Integral by using a variable template.

Variable templates are new with C++14 and declare a family of variables.
The concept Integral will evaluate to true if the predicate
std::is_integral<T>::value returns true for T . std::is_integral<T> is a
function of the type-traits
229176066048) library. The functions of the type-traits library enable,
amongst other things, that we can check types at compile time.
Concepts Definition: Function

Concepts#
The original syntax of Concepts Technical Specification (Concepts TS) was a
bit adjusted to the proposed Draft C++20 Standard. Here is the original
syntax from the Concepts TS, which is used in this course.
Concepts TS#
concept bool Equal(){
return requires(T a, T b) {
{ a == b } -> bool;
{ a != b } -> bool;
};
Integral is a variable concept and Equal is a function concept Both return

Integral is a variable concept and Equal is a function concept. Both return
a boolean.
The type parameter T fulfills the variable concept Integral if

std::is_integral<T>::value returns true.
The type parameter T fulfills the function concept Equal if there are
overloaded operators == and != for T that returns a boolean.
Concepts Draft#
The proposed syntax for C++20 is even more concise.
concept Equal =
requires(T a, T b) {
{ a == b } -> bool;
{ a != b } -> bool;
};
T fulfills the function concept if == and != are overloaded and return

a boolean.
The Concept Equal#

1 // conceptsDefintionEqual.cpp
2
4
6 concept bool Equal(){
7 return requires(T a, T b) {
8 { a == b } -> bool;
9 { a != b } -> bool;
10 };
11 }
12
13 bool areEqual(Equal a, Equal b){
14 return a == b;
15 }
16
17
{
18 struct WithoutEqual{
19 bool operator==(const WithoutEqual& other) = delete;
20 };
21
22 struct WithoutUnequal{
23 bool operator!=(const WithoutUnequal& other) = delete;
24 };
25
26
27
28 int main(){
29
31
32 std::cout << "areEqual(1, 5): " << areEqual(1, 5) << std::endl;
33
34 bool res = areEqual(WithoutEqual(), WithoutEqual());
35
36 bool res2 = areEqual(WithoutUnequal(), WithoutUnequal());
37
39
40 }
41
(/learn)
Output 2.01s
We used the concept Equal in the (generic) function areEqual (line 13 to 15)
and that’s not so exciting.
What’s more interesting is if we use the class WithoutEqual and

WithoutUnequal . We set for both the == or respectively the != operator to
delete . The compiler complains immediately that both types do not fulfill
the concept
the concept.
Let’s have a look at the screenshot of the error taken from the machine:
In the next lesson, we will study the predefined concepts Equal and Ord in
detail.

Back Next
Placeholder Syntax Define your Concepts: Equal and Ord
Mark as Completed
Report an Issue
3/20/22, 6:57 PM Define your Concepts: Equal and Ord - Generic Programming Templates in C++
Define your Concepts: Equal and Ord

In this lesson, we'll define the concepts Equal and Ord for C++.
• Eq versus Equal
• Haskell’s Type Class Ord

• The Concept Equal and the Concept Ord
Eq versus Equal #
The Type Class Eq The Concept Equal

(Haskell) (C++)
class Eq a where
(==) :: a -> a -> Bool
(/=) :: a -> a -> Bool
return requires(T a, T b){

{ a == b } -> bool;

{ a != b } -> bool;

};
(/learn)
Let’s have a closer look at Haskell’s type class Eq . Eq requires from its
instances, that
they have equal == and unequal /= operation that returns a Bool.

https://www.educative.io/courses/generic-templates-in-cpp/qVZG1wrXBp3 1/7
both take two arguments ( a -> a ) of the same type.
If you compare Haskell’s type class with C++'s concept, you see the similarity.
Of course, the instances are the concrete types such as int .
Haskell Type Class
Now we have two questions in mind if we look at Haskell’s type hierarchy

above. How is the definition of the type class Ord in Haskell and can we
model the inheritance relation in C++?
Haskell’s Type Class Ord #

class Eq a => Ord a where
compare :: a -> a -> Ordering
(<) :: a -> a -> Bool
(<=) :: a -> a -> Bool
(>) :: a -> a -> Bool
(>=) :: a -> a -> Bool
max :: a -> a -> a
Each type supporting Ord must support Eq .
The most interesting point about the typeclass Ord is the first line of its
definition. An instance of the typeclass Ord has to be already an instance of
the typeclass Eq . Ordering is an enumeration having the values EQ , LT , and
GT .
The Concept Equal and the Concept

Ord #
Let’s define the corresponding concepts in C++.
The concept Equal The Concept Ord
concept bool Ord(){
{ a == b } -> bool;
requires Equal<T>();
{ a != b } -> bool;
{ a <= b } -> bool;
};
{ a < b } -> bool;
}
{ a > b } -> bool;

{ a >= b } -> bool;

};
To make the job a little bit easier, we ignored the requirements compare and
max from Haskell’s type class in the concept Ord . The key point about the
concept is that the line requires Equal<T>() . Here we required that the type
parameter T has to fulfill the requirement Equal . If we use more
requirements such as in the definition of the concept Equal , each
requirement from top to bottom will be checked. That will be done in a
short-circuiting evaluation. So, the first requirement returning false will
end the process.
1 // conceptsDefintionOrd.cpp
3 #include <unordered_set>
4
6 concept bool Equal(){
7 return requires(T a, T b){
8 { a == b } -> bool;
9 { a != b } -> bool;
10 };
11 }
12
14 concept bool Ord(){
15 return requires(T a, T b){
16 requires Equal<T>();
17 { a <= b } -> bool;
18 { a < b } -> bool;
19 { a > b } -> bool;
20 { a >= b } -> bool;
21 };
22 }
23
24 bool areEqual(Equal a, Equal b){
25 return a == b;
26 }
27
28 Ord getSmaller(Ord a, Ord b){
29 return (a < b) ? a : b;
30 }
31
32 int main(){
33
35
36 std::cout << "areEqual(1, 5): " << areEqual(1, 5) << std::endl;
37
38 std::cout << "getSmaller(1, 5): " << getSmaller(1, 5) << std::endl;
39
40 std::unordered_set<int> firSet{1, 2, 3, 4, 5};
41 std::unordered_set<int> secSet{5, 4, 3, 2, 1};
42
43 std::cout << "areEqual(firSet, secSet): " << areEqual(firSet, secSet) << std::endl
44
45 auto smallerSet= getSmaller(firSet, secSet);
46
48
49 }
Output 1.76s
main.cpp:45:45: error: cannot call function 'auto getSmaller(auto:2, auto:2)

[with auto:2 = std::unordered_set<int>]'
45 | auto smallerSet= getSmaller(firSet, secSet);
| ^
main.cpp:28:5: note: constraints not satisfied
28 | Ord getSmaller(Ord a, Ord b){
| ^~~~~~~~~~
main.cpp:14:14: note: within 'template<class T> concept bool Ord() [with T =

std::unordered_set<int>]'
_ ]
14 | concept bool Ord(){
| ^~~
main.cpp:14:14: note: with 'std::unordered_set<int> a'
main.cpp:14:14: note: with 'std::unordered_set<int> b'
main.cpp:14:14: note: the required expression '(a <= b)' would be ill-formed
main.cpp:14:14: note: the required expression '(a < b)' would be ill-formed
main.cpp:14:14: note: the required expression '(a > b)' would be ill-formed
main.cpp:14:14: note: the required expression '(a >= b)' would be ill-formed
Equality and inequality are defined for the data types int and
std::unordered_set .
What would happen, when we uncomment line 45 and compare firSet and
secSet . To remind you, the type of both variables is std::unordered_set .
This says very explicitly that they don’t support an ordering.
Let’s check what happens when we run this code:
Of course, the compilation would fail.
In the next lesson, we’ll discuss other predefined concepts: Regular and
SemiRegular.

Back Next
Predefined Concepts Define your Concept: Regular and Se…
Mark as Completed
Report an Issue
3/20/22, 6:58 PM Define your Concept: Regular and SemiRegular - Generic Programming Templates in C++
Define your Concept: Regular and

SemiRegular
In this lesson, we'll gain an understanding of the important predefined concepts: Regular
and SemiRegular.
• References are not Regular
The first question we have to answer is quite obvious. What is a Regular or a

SemiRegular type? Our answer is based on the proposal p0898
(http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2018/p0898r3.pdf). We
assume you may have already guessed it that Regular and SemiRegular are
concepts, which are defined by other concepts. Given is the list of all
concepts.
Regular SemiRegular
DefaultConstructible Regular without
CopyConstructible, EqualityComparable
CopyAssignable
MoveConstructible,
MoveAssignable
Destructible
Swappable
https://www.educative.io/courses/generic-templates-in-cpp/B8j9KY0g2ON 1/7
EqualityComparable
The term Regular goes back to the father of the Standard Template Library
Alexander Stepanov (https://en.wikipedia.org/wiki/Alexander_Stepanov). He
introduced the term in his book Fundamentals of Generic Programming.
Here is a short excerpt (http://stepanovpapers.com/DeSt98.pdf). It’s quite
easy to remember the eight concepts used to define a regular type. First,
there is the well-known rule of six
(https://www.modernescpp.com/index.php/c-core-guidelines-constructors-
assignments-and-desctructors):
Default constructor: X()

Copy constructor: X(const X&)
Copy assignment: operator=(const X&)
Move constructor: X(X&&)

Move assignment: operator=(X&&)
Destructor: ~X()
Second, add the Swappable and EqualityComparable concepts to it. There is a

more informal way to say that a type T is regular: T behaves like an int .
To get SemiRegular, we have to subtract EqualityComparable from Regular.
References are not Regular#

Thanks to the type-traits
229176066048) library the following program checks at compile-time if int&
is a SemiRegular type.
Let’s have a look at the example:
1 #i l d i t
3
4 int main(){
5
7
8 std::cout << "std::is_default_constructible<int&>::value: " << std::is_default_c
9 std::cout << "std::is_copy_constructible<int&>::value: " << std::is_copy_constru
10 std::cout << "std::is_copy_assignable<int&>::value: " << std::is_copy_assignable
11 std::cout << "std::is_move_constructible<int&>::value: " << std::is_move_constru
12 std::cout << "std::is_move_assignable<int&>::value: " << std::is_move_assignable
13 std::cout << "std::is_destructible<int&>::value: " << std::is_destructible<int&>
15 //std::cout << "std::is_swappable<int&>::value: " << std::is_swappable<int&>::va
16
18 }
Output 1.28s
std::is_default_constructible<int&>::value: false
std::is_copy_constructible<int&>::value: true
std::is_copy_assignable<int&>::value: true
std::is_move_constructible<int&>::value: true
std::is_move_assignable<int&>::value: true
std::is_destructible<int&>::value: true
First of all, the function std::is_swappable requires C++17 that’s why we

have commented it out otherwise it will give an error. We see that the
reference such as int& is not default-constructible. The output shows that a
reference is not SemiRegular and, therefore, not Regular. To check, if a type
is Regular at compile-time, we need a function isEqualityComparable which
is not part of the type-traits library. Let’s define it.
1 #include <experimental/type_traits>
3
5 using equal_comparable_t = decltype(std::declval<T&>() == std::declval<T&>());
6
8 struct isEqualityComparable:
9 std::experimental::is_detected<equal_comparable_t, T>{};
10
11 struct EqualityComparable { };
12 bool operator == (EqualityComparable const&, EqualityComparable const&) { return tru
13
14 struct NotEqualityComparable { };
15
16 int main() {
17
19
20 std::cout << "isEqualityComparable<EqualityComparable>::value: " <<
21 isEqualityComparable<EqualityComparable>::value << std::endl;
22
23 std::cout << "isEqualityComparable<NotEqualityComparable>::value: " <<
24 isEqualityComparable<NotEqualityComparable>::value << std::endl;
25
27
28 }
Output 1.47s
isEqualityComparable<EqualityComparable>::value: true
isEqualityComparable<NotEqualityComparable>::value: false
The new feature is in the experimental namespace in line 1. Line 9 is a

crucial one. It detects at compile-time if the expression in line 5 is valid for
the type T . The type-trait isEqualityComparable works for an
EqualityComparable (line 11) and a NotEqualityComparable (line 14) type.
Only EqualityCompable returns true because we overloaded the Equal-
Comparison operator.
Now, we have all the ingredients to define Regular and SemiRegular. Here
are our new type-traits.
1 #include <experimental/type_traits>
3
5 using equal_comparable_t = decltype(std::declval<T&>() == std::declval<T&>());
6
(/learn)
8 struct isEqualityComparable:
9 std::experimental::is_detected<equal_comparable_t, T>
10 {};
11
13 struct isSemiRegular: std::integral_constant<bool,
14 std::is_default_constructible<T>::value &&
15 std::is_copy_constructible<T>::value &&
16 std::is_copy_assignable<T>::value &&
17 std::is_move_constructible<T>::value &&
18 std::is_move_assignable<T>::value &&
19 std::is_destructible<T>::value &&
20 std::is_swappable<T>::value >{};
21
23 struct isRegular: std::integral_constant<bool,
24 isSemiRegular<T>::value &&
25 isEqualityComparable<T>::value >{};
26
27
28 int main(){
29
31
32 std::cout << "isSemiRegular<int>::value: " << isSemiRegular<int>::value << std::
33 std::cout << "isRegular<int>::value: " << isRegular<int>::value << std::endl;
33 std::cout << isRegular<int>::value: << isRegular<int>::value << std::endl;
34
36
37 std::cout << "isSemiRegular<int&>::value: " << isSemiRegular<int&>::value << std
38 std::cout << "isRegular<int&>::value: " << isRegular<int&>::value << std::endl;
39
41
42 }
Output 1.31s
isSemiRegular<int>::value: true
isRegular<int>::value: true
isSemiRegular<int&>::value: false
isRegular<int&>::value: false
The usage of the new type-traits isSemiRegular and isRegular makes the
main program quite readable.
In this chapter, we have learned about the future concept of C++20. Let’s
move on to the next lesson for the conclusion of this course.

Back Next
Define your Concepts: Equal and Ord Conclusion
Mark as Completed
Report an Issue
3/20/22, 6:58 PM Conclusion - Generic Programming Templates in C++
(/learn)
Conclusion
Author's concluding words regarding the course.
This course includes all the information that you should know about
templates as a professional C++ developer. I know, sometimes the details are
quite sophisticated and, therefore, the course was not so easy to follow. This
holds, particularly true for the design with templates. But the reward is
outstanding.
Templates are one of the outstanding features of C++. They become more
and more important with each new C++ standard. The reason is quite
simple; templates provide abstraction without an abstraction performance
penalty.
With C++20, we get concepts which will lift templates to the next level. With
concepts, template parameters support semantic categories such as
Integral , provide easy to read error messages and can be implemented as
easily as functions.
You see, the future of C++ speaks templates.

https://www.educative.io/courses/generic-templates-in-cpp/q2njEvOVE33 1/2
3/20/22, 6:58 PM Conclusion - Generic Programming Templates in C++
Back Next
Define your Concept: Regular and Se… About the Author
Mark as Completed
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/q2njEvOVE33 2/2
3/20/22, 6:59 PM About the Author - Generic Programming Templates in C++
(/learn)
About the Author

• About Me
About Me#
I’ve worked as a software architect, team lead and instructor for about 20
years. In my spare time, I enjoy writing articles on topics such as C++,
Python, and Haskell and also enjoy speaking at conferences. In 2016, I have
decided to work for myself. I organize and lead seminars about modern C++
and Python.

https://www.educative.io/courses/generic-templates-in-cpp/YVXQpMEAoAp 1/2
3/20/22, 6:59 PM About the Author - Generic Programming Templates in C++
Did you complete this lesson?
YES!
Back
Conclusion
Report an Issue
https://www.educative.io/courses/generic-templates-in-cpp/YVXQpMEAoAp 2/2
(/learn)
- Examples
Let's look at the examples of variadic templates.
• Example 1: Variadic Template
• Explanation
• Example 2: Template Perfect Forwarding
• Explanation
Example 1: Variadic Template#

1 // templateVariadicTemplates.cpp
2
4
5 template <typename... Args>
6 int printSize(Args... args){
7 return sizeof ...(args);
8 }
9
10 template<int ...>
11 struct Mult;
12
13 template<>
14 struct Mult<>{
15 static const int value= 1;
16 };
17
18 template<int i, int ... tail>
19 struct Mult<i, tail ...>{
20 static const int value= i * Mult<tail ...>::value;
21 };
22
23 int main(){
23 int main(){
24
26
27 std::cout << "printSize(): " << printSize() << std::endl;
28 std::cout << "printSize(template,2011,true): " << printSize("template",2011,true)
29 std::cout << "printSize(1, 2.5, 4, 5, 10): " << printSize(1, 2.5, 4, 5, 10) << st
30
Explanation#
In the above example, we have used printSize function, which prints the
number of elements (of any type) passed as arguments. It detects the
number of elements on compile-time using the sizeof operator, and in case
of an empty argument list, it returns 0.
There is a struct defined as Mult which takes arguments of integer type

and return their product. If there is no argument passed, then it returns 1
which is the neutral element for multiplication. The result is stored in the
value in the fully specialized template in lines 13 – 16. The partial
specialization in lines 18 – 21 starts the recursion, which ends with the
aforementioned fully specialization for 0. The primary template in line 10 is
never used and must, therefore, never be defined.
To visualize the template instantiation for the above-mentioned example

click here (https://cppinsights.io/lnk?
code=I2luY2x1ZGUgPGlvc3RyZWFtPgoKdGVtcGxhdGUgPHR5cGVuYW1lLi4uI
EFyZ3M+CmludCBwcmludFNpemUoQXJncy4uLiBhcmdzKXsKICByZXR1cm4g
c2l6ZW9mIC4uLihhcmdzKTsKfQoKdGVtcGxhdGU8aW50IC4uLj4Kc3RydWN0
IE11bHQ7Cgp0ZW1wbGF0ZTw+CnN0cnVjdCBNdWx0PD57CiAgc3RhdGljIGN
vbnN0IGludCB2YWx1ZT0gMTsKfTsKCnRlbXBsYXRlPGludCBpLCBpbnQgLi4u
IHRhaWw+CnN0cnVjdCBNdWx0PGksIHRhaWwgLi4uPnsKICBzdGF0aWMgY
29uc3QgaW50IHZhbHVlPSBpICogTXVsdDx0YWlsIC4uLj46OnZhbHVlOwp9O
woKaW50IG1haW4oKXsKCiAgc3RkOjpjb3V0IDw8IHN0ZDo6ZW5kbDsKCiAgc
3RkOjpjb3V0IDw8ICJwcmludFNpemUoKTogIiA8PCBwcmludFNpemUoKSA8P
CBzdGQ6OmVuZGw7CiAgc3RkOjpjb3V0IDw8ICJwcmludFNpemUodGVtcGxh
dGUsMjAxMSx0cnVlKTogIiA8PCBwcmludFNpemUoInRlbXBsYXRlIiwyMDEx
LHRydWUpIDw8IHN0ZDo6ZW5kbDsKICBzdGQ6OmNvdXQgPDwgInByaW50
U2l6ZSgxLCAyLjUsIDQsIDUsIDEwKTogIiA8PCAgcHJpbnRTaXplKDEsIDIuNSw
gNCwgNSwgMTApIDw8IHN0ZDo6ZW5kbDsKCiAgc3RkOjpjb3V0IDw8IHN0ZD
o6ZW5kbDsKCiAgc3RkOjpjb3V0IDw8ICJNdWx0PDEwPjo6dmFsdWU6ICIgPD
wgTXVsdDwxMD46OnZhbHVlIDw8IHN0ZDo6ZW5kbDsKICBzdGQ6OmNvdX
QgPDwgIk11bHQ8MTAsMTAsMTA+Ojp2YWx1ZTogIiA8PCBNdWx0PDEwLDE
wLDEwPjo6dmFsdWUgPDwgc3RkOjplbmRsOwogIHN0ZDo6Y291dCA8PCAiT
XVsdDwxLDIsMyw0LDU+Ojp2YWx1ZTogIiA8PCBNdWx0PDEsMiwzLDQsNT4
6OnZhbHVlIDw8IHN0ZDo6ZW5kbDsKCiAgc3RkOjpjb3V0IDw8IHN0ZDo6ZW
5kbDsKCn0K&insightsOptions=cpp17&std=cpp17&rev=1.0).
Example 2: Template Perfect

Forwarding#
// templatePerfectForwarding.cpp
#include <iostream>
#include <utility>
template<typename T, typename ... Args>
T createT(Args&& ... args){
return T(std::forward<Args>(args) ...);
struct MyStruct{
MyStruct(int&, double&, double&&){}
friend std::ostream& operator<< (std::ostream& out, const MyStruct&){
out << "MyStruct" << std::endl;
return out;
}
};
int main(){
double myDouble= createT<double>();
std::cout << "myDouble: " << myDouble << std::endl;
int myInt= createT<int>(1);
std::cout << "myInt: " << myInt << std::endl;
std::string myString= createT<std::string>("My String");
std::cout << "myString: " << myString << std::endl;
MyStruct myStruct= createT<MyStruct>(myInt, myDouble, 3.14);
std::cout << "myStruct: " << myStruct << std::endl;
Explanation#
In the above example, we have created a createT function which invokes
the constructor T with the arguments args . If there is no value passed, it
invokes the default constructor. The magic of the factory function createT is
g y
that it can invoke each constructor. Thanks to perfect forwarding, each
value can be used such as an lvalue or an rvalue; thanks to parameter packs,

any number of arguments can be used. In the case of MyStruct , a
constructor that requires three arguments is used.
The pattern of the function template createT is exactly the pattern, factory
functions such as std::make_unique , std::make_shared , std::make_pair , or
std::make_tuple use.
In the next lesson, we’ll solve exercises related to variadic templates.

Back Next
Variadic Templates - Exercises
Mark as Completed
Report an Issue

Generic Programming

Uploaded by

Copyright:

Available Formats

Generic Programming

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Generic Programming

Uploaded by

Copyright:

Available Formats

3/20/22, 5:28 PM Function Templates - Generic Programming Templates in C++

We'll cover the following

• Passing Arguments in Function Templates

The keyword class or typename declares the parameters.

Passing Arguments in Function Templates#

Automatically creates an instance of the function template.

The following rules hold:

1. Templates do not support an automatic type conversion.

2. If a free function is better or equally as good as a function template that

To learn more about function templates, click here

In the next lesson, we’ll look at the examples of function templates.

Who Should Take this Course? - Examples

We'll cover the following

• Example 1: Templates in Functions

Example 1: Templates in Functions#

(t, f): (1, 0)

(t, f): (0, 1)

(int2011, int2014): (2011, 2014)

(int2011, int2014): (2014, 2011)

(first, second): (first, second)

(first, second): (second, first)

Example 2: Overloading Function

After: 10, 5.5

After: 20, 10, 5

In the next lesson, we’ll solve an exercise related to function templates.

Interviewing soon? We've partnered with Hired so that companies apply to

Function Templates - Exercise

We'll cover the following

We'll cover the following

• Solution 1: Using Function Arguments

Solution 1: Using Function

Solution 2: Using Template

The critical point is that the calculation is done at compile-time.

Solution 3: Using Template

When we invoke the power<10>(2) function, the argument in () brackets is

In the lesson, we’ll learn about class templates in detail.

- Exercise Class Templates

• Solution 1: Using Function Arguments

Solution 1: Using Function

Solution 2: Using Template

template<int m, int n>

static int const value = m * Power<m,n-1>::value;

static int const value = 1;

std::cout << Power<2,10>::value << std::endl;

The critical point is that the calculation is done at compile-time.

Solution 3: Using Template

int power(int m){

int power<1>(int m){

int power<0>(int m){

std::cout << power<10>(2) << std::endl;

When we invoke the power<10>(2) function, the argument in () brackets is

In the lesson, we’ll learn about class templates in detail.

- Exercise Class Templates

We'll cover the following

• Templates: Alias Templates

A class template defines a family of classes.

class-declaration: A class declaration. The class name declared becomes a

In contrast to a function template, a class template is not capable of

Let’s have a look at the declaration of function and class templates:

Function Template Class Template