SOLID Design: Dependency Inversion Principle in C++

1. SRP – Single Responsibility Principle
2. OCP – Open/Closed Principle
3. LSP – Liskov Substitution Principle
4. ISP – Interface Segregation Principle
5. DIP – Dependency Inversion Principle

• If you stumbled here directly, then I would suggest you go through What is design pattern? first, even if it is trivial. I believe it will encourage you to explore more on this topic.
• All of this code you encounter in this series of articles are compiled using C++20(though I have used Modern C++ features up to C++17 in most cases). So if you don’t have access to the latest compiler you can use https://wandbox.org/ which has preinstalled boost library as well.

Intent

1. High-level modules should not depend on low-level modules. Both should depend on abstractions.
2. Abstractions should not depend on details. Details should depend on abstractions.

• Above lines might seem cryptic at first but don’t stick here keep going. You will get it by example.

What are the High-level & Low-level modules?

Motivation: Violating Dependency Inversion Principle

enum class Relationship { parent, child, sibling };

struct Person {
    string      m_name;
};

struct Relationships {      // Low-level <<<<<<<<<<<<-------------------------
    vector<tuple<Person, Relationship, Person>>     m_relations;

    void add_parent_and_child(const Person &parent, const Person &child) {
        m_relations.push_back({parent, Relationship::parent, child});
        m_relations.push_back({child, Relationship::child, parent});
    }
};

struct Research {           // High-level  <<<<<<<<<<<<------------------------
    Research(const Relationships &relationships) {
        for (auto &&[first, rel, second] : relationships.m_relations) {// Need C++17 here
            if (first.m_name == "John" && rel == Relationship::parent)
                cout << "John has a child called " << second.m_name << endl;
        }
    }
};

int main() {
    Person parent{"John"};
    Person child1{"Chris"};
    Person child2{"Matt"};

    Relationships relationships;
    relationships.add_parent_and_child(parent, child1);
    relationships.add_parent_and_child(parent, child2);

    Research _(relationships);

    return EXIT_SUCCESS;
}

• When later on the container of Relationships changes from vector to set or any other container, you need to change in many places which isn’t a very good design. Even if just the name of data member i.e. Relationships::m_relations changes, you will find yourself breaking other parts of code.
• As you can see Low-level module i.e. Relationships directly depend on High-level module i.e. Research which is essentially a violation of DIP.

Solution: Example of Dependency Inversion Principle in C++

• Rather we should create an abstraction and bind Low-level & High-level module to that abstraction. Consider the following fix:

struct RelationshipBrowser {
    virtual vector<Person> find_all_children_of(const string &name) = 0;
};

struct Relationships : RelationshipBrowser {     // Low-level <<<<<<<<<<<<<<<------------------------
    vector<tuple<Person, Relationship, Person>>     m_relations;

    void add_parent_and_child(const Person &parent, const Person &child) {
        m_relations.push_back({parent, Relationship::parent, child});
        m_relations.push_back({child, Relationship::child, parent});
    }

    vector<Person> find_all_children_of(const string &name) {
        vector<Person> result;
        for (auto &&[first, rel, second] : m_relations) {
            if (first.name == name && rel == Relationship::parent) {
                result.push_back(second);
            }
        }
        return result;
    }
};

struct Research {                                // High-level <<<<<<<<<<<<<<<----------------------
    Research(RelationshipBrowser &browser) {
        for (auto &child : browser.find_all_children_of("John")) {
            cout << "John has a child called " << child.name << endl;
        }
    }
    //  Research(const Relationships& relationships)
    //  {
    //    auto& relations = relationships.relations;
    //    for (auto&& [first, rel, second] : relations)
    //    {
    //      if (first.name == "John" && rel == Relationship::parent)
    //      {
    //        cout << "John has a child called " << second.name << endl;
    //      }
    //    }
    //  }
};

• Now no matter, the name of container or container itself changes in Low-level module, High-level module or other parts of code which followed DIP will be in-tact.
• The Dependency Inversion Principle (DIP) suggest that the most flexible systems are those in which source code dependencies refer only to abstractions, not to concretions.
• This is the reason why most experienced dev uses STL or library functions along with generic containers. Even using an auto keyword at appropriate places may help in creating generic behaviour with less fragile code.
• There are many ways you can implement DIP, as long as C++ concerns most people use static polymorphism(i.e. CRTP unless they need dynamic one), template specialization, Adapter Design Pattern, type-erasure, etc.

Yardstick to Craft Dependency Inversion Principle(DIP) Friendly Software in C++

• If you find enforcing DIP difficult then just design abstraction first & implement your high-level module on the bases of abstraction. Without having any knowledge of the low-level module or its implementation. Because of this process DIP is also known as Coding To Interface.
• Keep in mind that all Low-level-modules/subclasses adhere to the Liskov Substitution Principle. This is because the Low-level-modules/subclasses will be used via the abstract interface, not the concrete classes interface.

Benefits

Reusability

• Effectively, the DIP reduces coupling between different pieces of code. Thus we get reusable code.

Maintainability

• It is also important to mention that changing already implemented modules is risky. By depending on abstraction & not on concrete implementation, we can reduce that risk by not having to change high-level modules in our project.
• Finally, DIP when applied correctly gives us flexibility and stability at the level of the entire architecture of our application. Our application will be able to evolve more securely and become stable & robust.

Conclusion

Do not use the concrete object directly unless you have a strong reason to do so. Use abstraction instead.