C++ Tricks of the Trade: Friend Functions
By Kais Dukes
Introduction
This new "C++ Tricks of the Trade" series of articles will show you how to write efficient yet maintainable C++ code, as well as explaining how to avoid common programming pitfalls.
In this introductory article, Kais discusses how to use friend functions to enhance a class interface.
Friendly Friends
By using the friend keyword, a class can grant access to non-member functions or to another class. These friend functions and friend classes are permitted to access private and protected class members. There are places where friends can lead to more intuitive code, and are often needed to correctly implement operator overloading.
If encountering friend functions for the first time, you might feel slightly uneasy since they seem to violate encapsulation. This feeling may stem from the fact that a friend function is not strictly a member of the class.
By thinking of a friend function as part of the class’s public interface, you can get a better understanding of how friends work. From a design perspective, friends can be treated in a similar way to public member functions. The concept of a class interface can be extended from public members to include friend functions and friend classes.
Put another way:
Friend functions do not break encapsulation; instead they naturally extend the encapsulation barrier.
Friend Functions
Friend functions can be declared anywhere within a class declaration, but it is common practice to list friends at the beginning of the class. The public and protected keywords do not apply to friend functions, as the class has no control over the scope of friends.
To see where a friend function would useful, suppose we are constructing a class to represent a rational number as a pair of integers. If we want to use rational numbers in the same way as intrinsic types like a double or an int, then it would be reasonable to be able to write:
CRational r(2, 3);
cout << r << endl;
In order to do this, we need to stream a rational object to an STL ostream by overloading the binary << operator:
class CRational
{
friend ostream &operator<<(ostream &out, const CRational &r);
public:
CRational(const int nNumerator, const int nDenominator);
~CRational();
// ...
};
The friend function gets full access to the members of CRational, in order to do its job of writing to an ostream:
ostream &operator<<(ostream &out, const CRational &r)
{
// Access private members of CRational to display r.
// ...
return out;
}
This operator overload could be declared as a non-friend function, but that would require using get() and set() members to reach the internals of the rational class. It is considered good practice to declare the operator<< function as a friend, even if get/set members exist. When reading through the CRational class declaration it is clear that the operator<< function should be treated as part of the class’s public interface when marked as a friend.
Another reason for using a friend function is one of efficiency: directly accessing data members saves the overhead of using get/set members, if the compiler has not inlined these.
Commutative Operators
Overloading operator+ allows the rational class to behave more like an intrinsic data type. To add two rational numbers, we can overload the operator as a class member:
const CRational CRational::operator+(const CRational &r)
{
// add *this to r, and return the result by value.
// ...
}
The operator+ function can also be overloaded to handle the addition of a rational, and a non-rational type. If n is an integer, and r is a rational then we could implement r + n as a class member:
const CRational CRational::operator+(const int n)
{
// add *this to n, and return the result by value
// ...
}
Since operator+ is an example of a commutative operator, we would expect r + n to have the same semantics as n + r. In the second case we need to use a non-member function:
class CRational
{
friend const CRational
operator+(const int n, const CRational &r);
// ...
}
This function could be written as a non-friend function, provided that get/set member functions are used. Doing so would turn the function into a plain non-member function, and so it would not be clear how it interacts with the class. By using a friend function, readers of the class declaration see that operator+ forms part of the CRational interface, and is semantically part of the class.
Friends and Virtual Functions
A problem with friends is that class member functions can be virtual while non-member functions can’t. The equivalent of a virtual friend function is useful when deriving from a class that has a friend function.
An example of where the equivalent of a virtual friend would be useful is if serializing an entire class hierarchy. The trick in doing this is to turn the friend function into a lightweight inline proxy function, which forwards the request back to the class. For example, consider the simple base class:
class CPerson
{
friend ostream &operator<< (ostream &out, CPerson &person);
// ...
protected:
virtual ostream &display(ostream &out);
// ...
};
The friend function forwards the display request back to the CPerson class. Note that we pass an instance of CPerson by reference:
inline ostream &operator<< (ostream &out, CPerson &person)
{
return person.display(out);
}
In a class hierarchy using the base CPerson class, each derived class has its own implementation of the display function:
class CBob : public CPerson
{
protected:
virtual ostream &display(ostream &out)
{
out << "Bob";
return out;
}
};
This is just as efficient as using a virtual friend function (if such a thing existed) since a good compiler will replace a call to operator<< with a call to the display function, which each derived class implements.
Friend Classes
As well as choosing a non-member friend function, a class has two other possibilities for its friends. A class can declare a member function of another class as a friend, or declare another class as a friend class.
Friend classes are used in cases where one class is tightly coupled to another class. For example, suppose we have a class CPoint that represents a coordinate, and a class CPointCollection that holds a list of points. Since the collection class may need to manipulate point objects, we could declare CPointCollection as a friend of the CPoint class:
// Forward declaration of friend class.
class CPointCollection;
// Point class.
class CPoint
{
friend CPointCollection;
public:
CPoint(const double x, const double y) :
m_x(x),
m_y(y)
{
}
~CPoint(void)
{
}
// ...
private:
double m_x;
double m_y;
};
Since the collection class is a friend of CPoint, it has access to the internal data of any point object. This is useful when individual elements of the collection need to be manipulated. For example, a set method of the CPointCollection class could set all CPoint elements to a particular value:
class CPointCollection
{
public:
CPointCollection(const int nSize) :
m_vecPoints(nSize)
{
}
~CPointCollection(void);
void set(const double x, const double y);
// ...
private:
vector<CPoint> m_vecPoints;
};
The set member can iterate over the collection and reset each point:
void CPointCollection::set(const double x, const double y)
{
// Get the number of elements in the collection.
const int nElements = m_vecPoints.size();
// Set each element.
for(int i=0; i<nElements; i++)
{
m_vecPoints[i].m_x = x;
m_vecPoints[i].m_y = y;
}
}
One thing worth noting about friend classes is that friendship is not mutual: although CPointCollection can access CPoint, the converse is not true. Friendship is also not something that is passed down a class hierarchy. Derived classes of CPointCollection will not be able to access CPoint. The principle behind this is that friendship is not implicitly granted; each class must explicitly choose its friends.
Conclusion
Friend functions and friend classes allow a class interface to be extended in a natural and efficient way. For some problems, friend functions lead to more intuitive code, while friend classes are needed when two classes are tightly coupled. Non-member friend functions are especially useful for implementing operator overloading, a good example being operator<<.
Friend functions do not break encapsulation, but rather enhance a class interface. Although this is the case, friends should be used judiciously; a member function will often do just as well. As a general rule of thumb, use member functions where possible, and friends where necessary. A well-designed project will not be free of friend functions, but will instead use them in places where non-members are naturally part of a class.
Top quality daily ASP, PHP and .NET articles, tutorials, news, reviews, interviews AND FREE EBOOKS! devArticles is the ultimate online resource for the serious web developer. Visit
http://www.devarticles.com today!
| DISCLAIMER: The content provided in this article is not warranted or guaranteed by Developer Shed, Inc. The content provided is intended for entertainment and/or educational purposes in order to introduce to the reader key ideas, concepts, and/or product reviews. As such it is incumbent upon the reader to employ real-world tactics for security and implementation of best practices. We are not liable for any negative consequences that may result from implementing any information covered in our articles or tutorials. If this is a hardware review, it is not recommended to open and/or modify your hardware. |
More Web Development Articles
More By Developer Shed
 developerWorks - FREE Tools!
|
Learn field-tested SOA principles, methodology, technology and implementation from the global SOA market leader - in a new e-book by an IBM SOA expert. Written by IBM Certified SOA Solution Designer Bobby Woolf, "Exploring IBM SOA Technology & Practice" is the ultimate insider's guide to SOA - a PDF e-book packed cover to cover with IBM's specific advice on how to make your SOA implementation a success.
FREE! Go There Now!
|
|
|
|
Visit IBM developerWorks to download a free trial version of Lotus Quickr 8.0, which enables collaboration by transforming the way everyday business content such as documents, rich media, photos, and video can be shared. Lotus Quickr makes it faster and easier to share content of all types (not just documents) within virtual teams. It is designed to make it easier to collaborate across organizational boundaries, while continuing to work within the context of familiar desktop applications.
FREE! Go There Now!
|
|
|
|
Download a free trial version of IBM Rational Developer for System z, software that can help you deliver core development capabilities; the power of Java Platform, Enterprise Edition (Java EE); and rapid application development support to diverse enterprise application development teams. With comprehensive development tools to help create, deploy and maintain traditional enterprise and composite applications, Rational Developer for System z enables developers with different technical backgrounds to easily participate in important technology projects.
FREE! Go There Now!
|
|
|
|
Visit IBM developerWorks to download a free trial of the Rational Host Access Transformation Services (HATS) Toolkit. The HATS toolkit provides a set of plug-ins for the IBM Rational Software Delivery Platform to help you easily extend your legacy applications. HATS makes your 3270 and 5250 applications available as HTML through the most popular Web browsers, while converting your host screens to a Web look and feel and it also enables you to develop new Web, portal, and rich-client applications.
FREE! Go There Now!
|
|
|
|
Visit IBM developerWorks to download a free trial version of WebSphere Extended Deployment Compute Grid, which lets you schedule, execute, and monitor batch jobs. Because online transaction processing and batch jobs execute simultaneously on the same server resources, you can avoid costly duplication of resources. Compute Grid supports job types of Java transactional batch, compute-intensive and a new type called "native execution", which enables non-Java workloads to run on distributed end points.
FREE! Go There Now!
|
|
|
|
Manage, govern, and share services across your organization by using WebSphere Service Registry and Repository. Follow the hands-on exercises to learn how to navigate the Web interface to publish, find, reuse, and update services.
FREE! Go There Now!
|
|
|
|
Learn from the best! Find out how developers use Rational ClearCase to be more flexible, innovative and deliver higher quality code in the Rational ClearCase Power Users eKit. This complimentary eKit provides a collection of materials, like articles, whitepapers, and demos that can help you become a power user of Rational ClearCase.
FREE! Go There Now!
|
|
|
|
This webcast outlines the best practices that must be instituted to gain the maximum benefit from SOA while maintaining high quality of service. Whether you are deploying new applications or managing and monitoring your existing infrastructure, learn how you can ensure high quality of services with SOA based solutions from IBM. All registrants who attend this live Web Seminar will receive complimentary access to a white paper titled “Maintaining QoS in an SOA Environment”.
FREE! Go There Now!
|
|
|
|
Portfolio Management is about effectively managing portfolio value by aligning portfolio investments with business goals. This complimentary e-kit provides a collection of materials that can help you understand how IBM Rational enables and automates best practices for improved governance and clear visibility into portfolio and project performance across the entire IT project lifecycle.
FREE! Go There Now!
|
|
|
|
Join this webcast to learn how IBM Rational's Functional Testing solution enables you to implement automation your way, at your pace, with your existing staff. In this webcast, you’ll learn how you can eliminate redundancy of manual test scripts, reduce errors, and increase test coverage through test automation. After this presentation you will understand how IBM Rational Functional Testing solution can streamline your manual testing and make test automation easily attainable.
FREE! Go There Now!
|
|
|
|
All FREE IBM® developerWorks Tools!
|
/ 4 
