C++ essentials

Many diploma and university computing courses in Tanzania teach programming in C++, and its ideas — classes, inheritance, careful use of memory, a compiler that checks your work before it runs — appear in every serious language. This chapter is written for those courses. You will rebuild the Form 4 results calculation from Chapter 5 in C++: grades and divisions in their own files, a Student class that refuses bad marks, people of different kinds sharing a common base class, and the standard library doing the sorting and counting.

By the end of this chapter you will be able to:

• install a C++ compiler and turn source code into a program;
• explain how C++‘s types differ from JavaScript’s and Python’s;
• split a program into header and source files;
• use std::vector, std::map, std::sort and std::accumulate;
• write classes with private data, constructors and const member functions;
• use inheritance and virtual functions so different classes share one interface;
• throw and catch exceptions;
• read the compiler’s error messages and fix what they point to.

The files for this chapter go in the cpp folder of your results-system project. This chapter assumes you have worked through Chapter 5, and moves faster than the earlier ones.

6.1 A compiled language

JavaScript and Python are run by another program — the browser, or python — which reads your code line by line as it goes. C++ is different. A compiler reads all of your code first, checks it, and translates it into a program that the computer’s processor runs directly.

 grades.cpp ─┐
 student.cpp ├─►  compiler (g++)  ─►  results.exe  ─►  run it
 main.cpp ───┘     checks types,        a program
                   translates           on its own

This has three consequences you will notice at once:

• Mistakes are found early. Use a number where text is expected, or call a function that does not exist, and the compiler refuses before the program ever runs.
• Programs are fast. There is nothing in between your program and the processor. This is why C++ is used for games, phones, cars, databases — PostgreSQL itself is written in C — and anywhere speed matters.
• There is a step between writing and running. Every change must be compiled again.

6.2 Setting up

On Windows, the easiest free compiler to install is g++ from MSYS2:

1. Download the installer from msys2.org and run it, accepting the defaults.
2. In the MSYS2 UCRT64 window that opens at the end, install the compiler:

pacman -S mingw-w64-ucrt-x86_64-gcc

3. Add the compiler’s folder to your PATH, so any terminal can find it: press Windows, type environment variables, open Edit the system environment variables → Environment Variables, select Path under your user variables, click Edit → New, and add C:\msys64\ucrt64\bin.
4. Close and reopen VS Code, open a terminal, and check:

g++ --version

In VS Code, also install the C/C++ extension from Microsoft. If you cannot install anything — on a shared college computer, for example — free websites such as onlinegdb.com compile and run C++ in the browser. They are fine for practice; keep your own copy of anything important.

The first program

Create hello.cpp:

#include <iostream>
#include <string>

int main() {
    std::string name = "Mwl. Grace Mushi";
    int students = 40;
    std::cout << "Habari, " << name << "! You have " << students << " students.\n";
    std::cout << "7 / 2 = " << 7 / 2 << '\n';
    std::cout << "7.0 / 2 = " << 7.0 / 2 << '\n';
    return 0;
}

Compile it, then run the program the compiler made:

g++ -std=c++17 -Wall -o hello hello.cpp
.\hello

Habari, Mwl. Grace Mushi! You have 40 students.
7 / 2 = 3
7.0 / 2 = 3.5

• #include brings in parts of the standard library: <iostream> for printing, <string> for text.
• Every C++ program starts at the function main. Returning 0 tells the computer it finished successfully.
• std::cout << ... prints. The << arrows send each value to the screen, one after another. '\n' starts a new line.
• std:: means the name comes from the standard library’s namespace, called std.
• In the compile command, -std=c++17 chooses the 2017 version of the language, -Wall turns on warnings about suspicious code, and -o hello names the program. On Linux and macOS, run it with ./hello.

Look at the second and third lines of output. 7 / 2 is 3: when both numbers are whole numbers, C++ does integer division and throws the remainder away. With one decimal number, 7.0 / 2 is 3.5. Averages computed with whole numbers will be wrong unless you plan for this.

6.3 Types

In C++, every variable has a type that is fixed when it is created and checked by the compiler.

const before a type means the value can never change after it is set — the compiler enforces it. Use it whenever you can; it tells the reader, and the compiler, what will stay fixed.

To turn a whole number into a double before dividing, use static_cast<double>(total) / count. You will see exactly that in Student::average later.

6.4 Header and source files

A C++ program is usually split across several files. Each part comes in two halves:

• a header (.h) that declares what exists — the names and types of the functions, without their bodies;
• a source file (.cpp) that defines them — the bodies that do the work.

Other files #include the header, so they know what they can call. Here is the header for grades, grades.h:

// Grades, points and divisions for Form 4 results.
#pragma once

#include <string>
#include <vector>

namespace results {

char gradeFor(int mark);
int pointsFor(char grade);

struct Division {
    int points;        // points of the best seven subjects; 0 if incomplete
    std::string name;  // "I" to "IV", "0", or "INC"
};

Division divisionFor(const std::vector<char>& grades);

}  // namespace results

• #pragma once stops the header being included twice in the same file, which would be an error.
• namespace results { ... } puts these names in a group of their own. Outside, they are written results::gradeFor, so they can never clash with another library’s gradeFor.
• A struct groups values under one name. Division returns two answers — the points and the division’s name — as one value, like the pair Python returned in Chapter 5.
• const std::vector<char>& grades is a reference to a list that the function promises not to change. Without the &, C++ would make a complete copy of the list every time the function is called.

The definitions are in grades.cpp. The divisions follow the rule explained in Chapter 5 — the best seven subjects, with A worth 1 point and F worth 5:

int pointsFor(char grade) {
    switch (grade) {
        case 'A': return 1;
        case 'B': return 2;
        case 'C': return 3;
        case 'D': return 4;
        case 'F': return 5;
    }
    throw std::invalid_argument(std::string("not a grade: ") + grade);
}

switch picks the case matching the value — a neat way to write a chain of if statements that all test the same variable. If no case matches, the function reaches throw, which reports an error instead of returning a made-up number.

Division divisionFor(const std::vector<char>& grades) {
    const std::size_t best = 7;
    if (grades.size() < best) return {0, "INC"};

    std::vector<int> points;
    for (char grade : grades) points.push_back(pointsFor(grade));
    std::sort(points.begin(), points.end());
    int total = std::accumulate(points.begin(), points.begin() + best, 0);

    if (total <= 17) return {total, "I"};
    if (total <= 21) return {total, "II"};
    if (total <= 25) return {total, "III"};
    if (total <= 33) return {total, "IV"};
    return {total, "0"};
}

This is where the standard library earns its place:

• for (char grade : grades) is a range-based for loop: each grade in turn, like Python’s for.
• push_back adds an item to the end of a vector.
• std::sort(points.begin(), points.end()) sorts the whole vector, smallest first. Standard algorithms work on a range, given as a beginning and an end.
• std::accumulate(points.begin(), points.begin() + best, 0) adds up the first seven, starting from 0. Python’s sum(sorted(...)[:7]) became one line here too.
• return {total, "I"} builds a Division from its two values.

6.5 Classes

A class bundles data with the functions that work on it, and decides who may touch the data. Here is the student, in student.h:

class Student : public Person {
public:
    Student(std::string name, std::string admissionNo);

    void setMark(const std::string& subject, int mark);
    double average() const;
    results::Division division() const;
    std::string describe() const override;

private:
    std::string admissionNo_;
    std::map<std::string, int> marks_;  // subject -> mark
};

Ignore : public Person and override for a moment; they are the subject of section 6.6.

• Everything under public: can be used by any code. Everything under private: can only be used by the class’s own functions. The trailing underscore in marks_ is a common way to mark private data.
• The constructor — the function with the same name as the class — runs when a Student is created, and sets it up.
• A function marked const promises not to change the object. average() only reads marks, so it is const; setMark changes them, so it is not.

Why make the marks private? Because then the only way to change a mark is setMark, and setMark checks it:

void Student::setMark(const std::string& subject, int mark) {
    if (mark < 0 || mark > 100) {
        throw std::invalid_argument(name() + ", " + subject + ": " + std::to_string(mark) +
                                    " is not between 0 and 100");
    }
    marks_[subject] = mark;
}

No part of the program, however large it grows, can put a mark of 147 into a student. If someone tries to reach past the check and write s.marks_["Biology"] = 147; directly, the compiler stops them:

error: 'std::map<std::__cxx11::basic_string<char>, int> Student::marks_' is private within this context

This idea — data kept private, and changed only through functions that protect it — is called encapsulation. It is the same reason the database in Chapter 4 has CHECK constraints: put the rule in one place that cannot be avoided.

Student:: in front of the name says that this setMark belongs to the Student class. std::to_string turns a number into text, and marks_[subject] = mark stores the mark in the map under the subject’s name, replacing any mark already there.

The constructor uses a special form, the member initialiser list after the colon, to set up the base class and the data before the body runs:

Student::Student(std::string name, std::string admissionNo)
    : Person(std::move(name)), admissionNo_(std::move(admissionNo)) {}

std::move hands the text over instead of copying it — a small speed-up that you will see in most modern C++.

The average needs care with integer division, from section 6.2:

double Student::average() const {
    if (marks_.empty()) return 0.0;
    int total = 0;
    for (const auto& [subject, mark] : marks_) total += mark;
    return static_cast<double>(total) / marks_.size();
}

const auto& [subject, mark] takes each entry of the map apart into its key and value, like Python’s for subject, mark in marks.items(). auto lets the compiler work out the type itself.

6.6 Inheritance and virtual functions

A school has students and teachers. Both have names; each describes themselves differently. Inheritance lets them share what is common. The base class, in person.h:

class Person {
public:
    explicit Person(std::string name) : name_(std::move(name)) {}
    virtual ~Person() = default;

    const std::string& name() const { return name_; }

    // Every kind of person describes themselves in their own way.
    virtual std::string describe() const = 0;

private:
    std::string name_;
};

And a teacher is a kind of person:

class Teacher : public Person {
public:
    Teacher(std::string name, std::string subject)
        : Person(std::move(name)), subject_(std::move(subject)) {}

    std::string describe() const override {
        return name() + " teaches " + subject_;
    }

private:
    std::string subject_;
};

• class Teacher : public Person means a Teacher is a Person: it has everything a person has — here, name() — plus its own subject. Student is declared the same way.
• virtual marks a function that each derived class may provide its own version of. When the program calls describe() on a Person, C++ runs the version belonging to the real object, whether it is a teacher or a student. This is polymorphism: one call, many forms.
• = 0 makes describe a pure virtual function: Person has no version of its own, so every derived class must write one. A class with a pure virtual function is abstract — you cannot create a plain Person, only teachers and students. Try Person p("Amina"); and the compiler says:

error: cannot declare variable 'p' to be of abstract type 'Person'

• override says “this replaces a virtual function from the base class”. If you misspell the name, or the parameters do not match, the compiler tells you, instead of quietly creating a second, unrelated function.
• virtual ~Person() = default; is a virtual destructor. Any class meant to be used through a base-class pointer needs one, so that deleting the object cleans up the whole of it.
• explicit stops C++ from silently turning a plain string into a Person where you did not ask for one.

The payoff comes in main.cpp, where one list holds both kinds:

    std::vector<std::unique_ptr<Person>> people;
    people.push_back(std::make_unique<Teacher>("Mwl. Grace Mushi", "Biology"));
    people.push_back(std::make_unique<Student>(students.front()));

    std::cout << '\n';
    for (const auto& person : people) std::cout << person->describe() << '\n';

A std::unique_ptr<Person> owns an object of any class derived from Person, and deletes it automatically when the list goes away — no memory is leaked, and you never write delete yourself. -> calls a function through a pointer. The loop does not know or care which kind of person it has; each describes itself correctly.

6.7 Exceptions

When something goes wrong that the current function cannot fix, it throws an exception, as setMark does. Code further up that knows what to do catches it:

    try {
        students.front().setMark("Biology", 147);
    } catch (const std::invalid_argument& error) {
        std::cout << "\nRefused: " << error.what() << '\n';
    }

error.what() is the message given when the exception was thrown. If nothing catches an exception, the program stops — which is better than carrying on with a wrong mark.

6.8 The whole program

main.cpp puts it together: it gives five students their marks, sorts them best first, prints a table, counts the divisions, shows a bad mark being refused, and lets a teacher and a student describe themselves. Sorting uses a lambda — a small function written where it is needed:

    std::sort(students.begin(), students.end(), [](const Student& a, const Student& b) {
        if (rankOf(a) != rankOf(b)) return rankOf(a) < rankOf(b);
        return a.average() > b.average();
    });

The lambda answers one question for std::sort: should a come before b? Fewer points first; if the points are equal, the higher average first. rankOf treats an incomplete result as worse than any division.

Counting uses a map: ++count[d.name]; adds one to the count for that division, and a name not yet in the map starts at zero.

A program made of several files is compiled by naming all the .cpp files. Headers are not listed; they are included by the .cpp files that need them:

g++ -std=c++17 -Wall -Wextra -o results main.cpp grades.cpp student.cpp
.\results

No. Name            Average  Points  Division
1   Neema Kileo        89.2       7  I
2   Amina Juma         79.2       7  I
3   Rehema Mollel      65.8      15  I
4   Baraka Mushi       53.1      20  II
5   Juma Hassan        38.6      26  IV

Divisions: I=3 II=1 IV=1

Refused: Neema Kileo, Biology: 147 is not between 0 and 100

Mwl. Grace Mushi teaches Biology
Neema Kileo (MSS/2023/003), 9 subjects

These five students have the same averages, points and divisions as in the Python report in Chapter 5. Two programs in two languages agreeing is good evidence that both are right.

-Wextra turns on even more warnings than -Wall. This program compiles with none. Aim for that in your own work: a warning is the compiler pointing at something that is probably a mistake.

6.9 Reading compiler errors

The compiler’s messages look alarming at first, but they follow a pattern: file, line, column, then what is wrong. Always fix the first error first — one mistake often causes a cascade of confusing errors after it, which disappear when the first is fixed.

The last one is worth understanding. Compiling has two stages: each .cpp file is compiled on its own, and then the linker joins the pieces into one program. If you forget student.cpp in the command, every file still compiles — the header promised Student::division exists — but the linker cannot find its body.

6.10 The complete files

The cpp folder holds six files. Compile them with the command in section 6.8.

cpp/grades.h

// Grades, points and divisions for Form 4 results.
#pragma once

#include <string>
#include <vector>

namespace results {

char gradeFor(int mark);
int pointsFor(char grade);

struct Division {
    int points;        // points of the best seven subjects; 0 if incomplete
    std::string name;  // "I" to "IV", "0", or "INC"
};

Division divisionFor(const std::vector<char>& grades);

}  // namespace results

cpp/grades.cpp

#include "grades.h"

#include <algorithm>
#include <numeric>
#include <stdexcept>

namespace results {

char gradeFor(int mark) {
    if (mark < 0 || mark > 100) {
        throw std::invalid_argument("a mark must be between 0 and 100");
    }
    if (mark >= 75) return 'A';
    if (mark >= 65) return 'B';
    if (mark >= 45) return 'C';
    if (mark >= 30) return 'D';
    return 'F';
}

int pointsFor(char grade) {
    switch (grade) {
        case 'A': return 1;
        case 'B': return 2;
        case 'C': return 3;
        case 'D': return 4;
        case 'F': return 5;
    }
    throw std::invalid_argument(std::string("not a grade: ") + grade);
}

Division divisionFor(const std::vector<char>& grades) {
    const std::size_t best = 7;
    if (grades.size() < best) return {0, "INC"};

    std::vector<int> points;
    for (char grade : grades) points.push_back(pointsFor(grade));
    std::sort(points.begin(), points.end());
    int total = std::accumulate(points.begin(), points.begin() + best, 0);

    if (total <= 17) return {total, "I"};
    if (total <= 21) return {total, "II"};
    if (total <= 25) return {total, "III"};
    if (total <= 33) return {total, "IV"};
    return {total, "0"};
}

}  // namespace results

cpp/person.h

// People in a school: what every person has, and how each kind differs.
#pragma once

#include <string>
#include <utility>

class Person {
public:
    explicit Person(std::string name) : name_(std::move(name)) {}
    virtual ~Person() = default;

    const std::string& name() const { return name_; }

    // Every kind of person describes themselves in their own way.
    virtual std::string describe() const = 0;

private:
    std::string name_;
};

class Teacher : public Person {
public:
    Teacher(std::string name, std::string subject)
        : Person(std::move(name)), subject_(std::move(subject)) {}

    std::string describe() const override {
        return name() + " teaches " + subject_;
    }

private:
    std::string subject_;
};

cpp/student.h

// A student and their marks for one examination.
#pragma once

#include <map>
#include <string>

#include "grades.h"
#include "person.h"

class Student : public Person {
public:
    Student(std::string name, std::string admissionNo);

    void setMark(const std::string& subject, int mark);
    double average() const;
    results::Division division() const;
    std::string describe() const override;

private:
    std::string admissionNo_;
    std::map<std::string, int> marks_;  // subject -> mark
};

cpp/student.cpp

#include "student.h"

#include <stdexcept>
#include <utility>
#include <vector>

Student::Student(std::string name, std::string admissionNo)
    : Person(std::move(name)), admissionNo_(std::move(admissionNo)) {}

void Student::setMark(const std::string& subject, int mark) {
    if (mark < 0 || mark > 100) {
        throw std::invalid_argument(name() + ", " + subject + ": " + std::to_string(mark) +
                                    " is not between 0 and 100");
    }
    marks_[subject] = mark;
}

double Student::average() const {
    if (marks_.empty()) return 0.0;
    int total = 0;
    for (const auto& [subject, mark] : marks_) total += mark;
    return static_cast<double>(total) / marks_.size();
}

results::Division Student::division() const {
    std::vector<char> grades;
    for (const auto& [subject, mark] : marks_) grades.push_back(results::gradeFor(mark));
    return results::divisionFor(grades);
}

std::string Student::describe() const {
    return name() + " (" + admissionNo_ + "), " + std::to_string(marks_.size()) + " subjects";
}

cpp/main.cpp

// Form 4 results: classes, inheritance and the standard library.
#include <algorithm>
#include <iomanip>
#include <iostream>
#include <map>
#include <memory>
#include <stdexcept>
#include <string>
#include <vector>

#include "person.h"
#include "student.h"

// Lower is better: points of the best seven, with incomplete results last.
int rankOf(const Student& student) {
    results::Division d = student.division();
    return d.name == "INC" ? 100 : d.points;
}

int main() {
    const std::vector<std::string> subjects = {"Civics",  "History",     "Geography", "Kiswahili", "English",
                                               "Biology", "Mathematics", "Chemistry", "Physics"};
    const std::vector<std::vector<int>> marks = {
        {82, 76, 79, 88, 85, 81, 78, 74, 70},
        {58, 49, 61, 66, 61, 47, 52, 44, 40},
        {90, 85, 88, 92, 88, 94, 91, 86, 89},
        {38, 41, 35, 52, 49, 38, 34, 29, 31},
        {68, 70, 63, 75, 72, 59, 66, 58, 61},
    };
    std::vector<Student> students = {
        {"Amina Juma", "MSS/2023/001"},  {"Baraka Mushi", "MSS/2023/002"},
        {"Neema Kileo", "MSS/2023/003"}, {"Juma Hassan", "MSS/2023/004"},
        {"Rehema Mollel", "MSS/2023/005"},
    };

    for (std::size_t i = 0; i < students.size(); ++i) {
        for (std::size_t j = 0; j < subjects.size(); ++j) {
            students[i].setMark(subjects[j], marks[i][j]);
        }
    }

    // Best first: fewest points, then the higher average.
    std::sort(students.begin(), students.end(), [](const Student& a, const Student& b) {
        if (rankOf(a) != rankOf(b)) return rankOf(a) < rankOf(b);
        return a.average() > b.average();
    });

    std::cout << std::left << std::setw(4) << "No." << std::setw(15) << "Name" << std::right
              << std::setw(8) << "Average" << std::setw(8) << "Points" << "  Division\n";

    std::map<std::string, int> count;
    int number = 1;
    for (const Student& s : students) {
        results::Division d = s.division();
        std::cout << std::left << std::setw(4) << number++ << std::setw(15) << s.name() << std::right
                  << std::fixed << std::setprecision(1) << std::setw(8) << s.average()
                  << std::setw(8) << d.points << "  " << d.name << '\n';
        ++count[d.name];
    }

    std::cout << "\nDivisions:";
    for (const auto& [name, n] : count) std::cout << ' ' << name << '=' << n;
    std::cout << '\n';

    // A mistake in the data is refused, not quietly accepted.
    try {
        students.front().setMark("Biology", 147);
    } catch (const std::invalid_argument& error) {
        std::cout << "\nRefused: " << error.what() << '\n';
    }

    // A teacher and a student are both people, so one list can hold both.
    std::vector<std::unique_ptr<Person>> people;
    people.push_back(std::make_unique<Teacher>("Mwl. Grace Mushi", "Biology"));
    people.push_back(std::make_unique<Student>(students.front()));

    std::cout << '\n';
    for (const auto& person : people) std::cout << person->describe() << '\n';

    return 0;
}

std::setw(15) sets the width of the next value printed, and std::left and std::right line it up, as < and > did in Python’s f-strings. std::fixed with std::setprecision(1) prints one decimal place.

Try it yourself

1. Add Elia Kimaro from Chapter 5, who missed Geography: give him every mark except that one. Check that his points and division match the Python report.
2. Add a HeadTeacher class that inherits from Teacher and describes themselves differently. Does people need any change to hold one?
3. Read the class list from class_list.csv instead of the marks written in main. You will need <fstream>, std::getline and a way to split a line at commas.
4. Write a small test program, test_grades.cpp, that uses assert from <cassert> to check the grade and division boundaries tested in Chapter 5.
5. Add a function to Student that returns its best subject. Should it be const?

Key terms

Check your understanding

1. What does a compiler do that Python and the browser do not do before your program runs?
2. Why does 7 / 2 give 3 in C++, and how do you get 3.5?
3. What is the difference between grades.h and grades.cpp?
4. Why is the parameter of divisionFor written const std::vector<char>& grades and not std::vector<char> grades?
5. Why are a student’s marks private, and what would go wrong if they were public?
6. Why is average() a const function, but not setMark?
7. What does = 0 after virtual std::string describe() const mean, and what does it stop you doing?
8. In the loop over people, how does C++ know whether to run Teacher::describe or Student::describe?
9. You compile with g++ -o results main.cpp grades.cpp and get undefined reference to 'Student::division() const'. What is wrong?
10. The compiler reports twelve errors. Which do you fix first, and why?

Answers

1. It reads and checks the whole program before anything runs — types, names, function calls — and translates it into a program the processor runs directly.
2. Both numbers are int, so C++ does integer division and drops the remainder. Make one of them a double: 7.0 / 2, or static_cast<double>(7) / 2.
3. The header declares the functions and types — what exists. The source file defines them — the bodies that do the work. Other files include the header.
4. The & passes a reference, so the vector is not copied each time; const promises the function will not change it.
5. So that the only way to change a mark is setMark, which checks it. If they were public, any code could store a mark of 147 and every result would be wrong.
6. average() only reads the marks, so it promises not to change the object. setMark changes the marks, so it cannot be const.
7. It makes describe pure virtual: Person has no version of its own and every derived class must write one. You cannot create a plain Person object.
8. describe is virtual, so C++ looks at the real type of each object when the program runs and calls that class’s version.
9. student.cpp is missing from the command, so the linker cannot find the bodies of Student’s functions. Add it: g++ -o results main.cpp grades.cpp student.cpp.
10. The first. A single mistake, such as a missing semicolon, often causes many confusing errors after it, which disappear once it is fixed.