The precursors of object-oriented programming can be traced back to the late 1960's: Classes, inheritance and virtual member functions were integral features of Simula67, a programming language that was mainly used for writing event-driven simulations. When Smalltalk first appeared back in 1972, it offered a pure object-oriented programming environment. In fact, Smalltalk defined object-oriented programming. This style of programming was so innovative and revolutionary at the time that it took more than a decade for it to become a standard in the software industry. Undoubtedly, the emergence of C++ in the early '80s provided the most considerable contribution to this revolution.
In 1979, a young engineer at Bell (now AT&T) Labs, Bjarne Stroustrup, started to experiment with extensions to C to make it a better tool for implementing large-scale projects. In those days, an average project consisted of tens of thousands of lines of code (LOC).
NOTE: Today, Microsoft's Windows 2000 (formerly NT 5.0) consists of more than 30 million lines of code (and counting).
When projects leaped over the 100,000 LOC count, the shortcomings of C became noticeably unacceptable. Efficient teamwork is based, among other things, on the capability to decouple development phases of individual teams from one another--something that was difficult to achieve in C.
By adding classes to C, the resultant language -- "C with classes" -- could offer better support for encapsulation and information hiding. A class provides a distinct separation between its internal implementation (the part that is more likely to change) and its external interface. A class object has a determinate state right from its construction, and it bundles together the data and operations that manipulate it.
In 1983, several modifications and extensions had already been made to C with classes. In that year, the name "C++" was coined. Ever since then, the ++ suffix has become a synonym for object-orientation. (Bjarne Stroustrup could have made a fortune only by registering ++ as a trademark) It was also in that year that C++ was first used outside AT&T Labs. The number of users was doubling every few months -- and so was the number of compilers and extensions to the language.
Between 1985 and 1989, C++ underwent a major reform. Protected members, protected inheritance, templates, and a somewhat controversial feature called multiple inheritance were added to the language. It was clear that C++ needed to become standardized.
In 1989, the American National Standards Institution (ANSI) committee for the standardization of C++ was established. The official name of the committee was X3J16, and later it was changed to J16. Generally, standardization committees don't write a standard from scratch; rather, they adopt an existing de facto reference, and use it as their baseline. The ANSI C committee used The C Programming Language by Kernighan and Ritchie as a starting point. Likewise, the ANSI C++ committee used the Annotated C++ Reference Manual (ARM) by Ellis and Stroustrup as its base document. The ARM provided a clear and detailed starting point for the committee's work. The committee's policy was to not rush into establishing a half-baked standard that would become obsolete in a year or two. Instead, the policy was to allow the demands for changes to emerge from the users of the language, the C++ community. Nonetheless, the committee also initiated extensible modifications and changes to the language, such as runtime type information (RTTI) and the new cast notation.
By that time, hundreds of thousands of people were using the language. C++ compilers were available for almost every platform. New C++-based frameworks, such as MFC and OWL, had emerged. The committee had to face enormous pressure from several directions. Some organizations were advocating new features and extensions to the language that were borrowed from other object-oriented languages, while other parties strove to keep it as efficient as possible. On top of this, C++ had to retain its backward compatibility with C, including the support of eight different flavors for integral types, cumbersome pointer syntax, structs, unions, global functions, and many other features that don't exactly go hand in hand with object orientated programming.
C++ standardization was a joint international endeavor in which national standardization bodies from all over the world were intensively involved. This is different from the standardization of C. C standardization was first carried out by ANSI as an American standard and was later adopted, with some modifications (mainly internationalization issues), as an international standard by the International Standardization Organization (ISO). The international venture of C++ guaranteed a worldwide acceptance of the standard, albeit at the price of somewhat more complicated procedures. Thus, the committee's meetings were actually joint meetings of both the ANSI working group and the ISO working group. Officially, the ANSI working group served as an advisor to ISO. Therefore, two votes were taken on every technical issue: an ANSI vote, to decide what the ANSI recommendation was, and a subsequent ISO vote, to actually make the decision. Some important changes were made in order to meet the criteria for ISO approval, including the addition of wchar_t as a built-in type, the templatization of the iostream library, the templatization of class string, and the introduction of the locale library, which encapsulates cultural-dependent differences.
The committee's initiatory task was to produce a draft of the standard, known as the Committee Draft (CD). For that purpose, the committee convened three times a year, one week at a time, in different places of the world. The first CD received several disapproving votes as well as many comments from ISO. The committee resolved these technical issues and addressed the comments in the second CD. The second CD was approved by ISO; however, there were still 5 "nay" votes and accompanying comments. Following the ISO balloting, the CD's were made available to the public. The public review process enabled C++ users from all over the world to comment on the proposed CD and point out contradictions and omissions.
After the approval of the second CD in November 1996, the committee's task was mainly to respond to the 5 "nay" votes and the accompanying comments and turn them into "aye" votes. The resultant document was the Final Draft International Standard, or the FDIS. At the meeting of the standardization committee in November, 1997 at Morristown, New Jersey, the FDIS was unanimously approved. In 1998, after a few minor changes, the FDIS was approved by ISO and became an international standard. In accordance with ISO rules, after it was approved, the Standard entered a freeze period of five years; during this time, the only modifications that are allowed are error fixes. People who find such defects can submit a Defect Report to the committee for consideration.
C++ differs from other object-oriented languages in many ways. For instance, C++ is not a root-based language, nor does it operate on a runtime virtual machine. These differences significantly broaden the domains in which C++ can be used.
The fact that legacy C code can be combined seamlessly with new C++ code is a major advantage. Migration from C to C++ does not force you to throw away good, functional C code. Many commercial frameworks, and even some components of the Standard Library itself, are built upon legacy C code that is wrapped in an object-oriented interface.
Interpreted languages allow easier code porting, albeit at the cost of significant performance overhead. C++, on the other hand, uses the compile and link model it inherited from C. One of the goals of C++ designers has been to keep it as efficient as possible; a compile-and-link model enables very efficient code generation and optimization. Another performance factor is the use of a garbage collector. This feature is handy and prevents some common programming bugs; however, garbage collected languages are disqualifies for time-critical application development, where determinacy is paramount. For that reason, C++ does not have a garbage collector.
In addition to object-oriented programming, C++ supports other useful programming styles, including procedural programming, object-based programming, and generic programming -- making it a multi-paradigm, general-purpose programming language.
Procedural programming is not very popular these days. However, there are some good reasons for C++ to support this style of programming, even today.
Gradual Migration of C Programmers To C++
C programmers who make their first steps in C++ are not forced to throw all their expertise away. Many primitives and fundamental concepts of C++ were inherited from C, including built-in operators and fundamental types, pointers, the notion of dynamic memory allocation, header files, preprocessor, and so on. As a transient phase, C programmers can still remain productive when the shift to C++ is made.
Bilingual Environments
C++ and C code can work together. Under certain conditions, this combination is synergetic and robust.
Machine-Generated Code
Many software tools and generators generate C code as an intermediate stage of application build. For example, SQL queries on most relational databases are translated into C code, which is in turn compiled and linked. There's not much point in forcing these generators to produce C++ code (although some do so) when the generated code is not going to be used by human programmers. Furthermore, many early C++ compilers were not really compilers in the true meaning of the word; they are better described as translators because they translated C++ code into intermediate C code that was later compiled and linked. In fact, any valid C++ programs can be translated directly into pure C code.
This is the most widely used style of programming in C++. The intent of this book is to deliver useful guidelines and rules of thumb for efficient, reliable, reusable, and easy to maintain object-oriented code. But there is no universal consensus as to what OO really is; the definitions vary among schools, languages, and users. There is, however, a consensus about a common denominator -- a combination of encapsulation, information hiding, polymorphism, dynamic binding, and inheritance. Some maintain that advanced object-oriented consists of generic programming support and multiple inheritance. These concepts will be discussed in depth in the chapters that follow.
Generic programming proceeds one step beyond object-oriented programming in pursuing reusability. Two important features of C++, templates and operator overloading, are the basis of generic programming. STL, a collection of generic algorithms and containers, is probably the most impressive manifestation of this paradigm.
This book is aimed at experienced C++ developers who seek a guide for enhancing their design and programming proficiency. It discloses facts and techniques and provides a knowledge base for advanced, Standard-compliant, and efficient use of C++. In addition, the book also explains the underlying mechanism behind the high-level features of the language, and it explains the philosophy behind the design and evolution of C++.
The target audience is intermediate and advanced level C++ developers who want to improve their proficiency by acquiring new programming techniques and design idioms. On top of adding many new features to the language, the standardization committee has deprecated several features and library components. In this book, readers who want to develop long lasting, future-proof C++ software can find a comprehensive list of deprecated features and their recommended alternatives.
Chapter 2, "Standard Briefing: The Latest Addenda to ANSI/ISO C++," presents some of the key terms that are used in the C++ Standard, and which are used extensively in this book. Following this, the recent changes and extensions to C++ are described. Finally, Chapter 2 gives an overview of the deprecated features that are listed in the Standard, and suggests standard-conforming replacements for them.
Chapter 3, "Operator Overloading," explores the benefits as well as the potential problems of operator overloading. It discusses the restrictions that apply to operator overloading and explains how to use conversion operators.
Chapter 4, "Special Member Functions: Default Constructor, Copy Constructor, Destructor, and Assignment Operator," explains the semantics of the special member functions and their role in class design. It also demonstrates several techniques and guidelines for an effective usage of these special member functions.
Chapter 5, "Object-Oriented Programming and Design," provides a brief survey of the various programming styles that are supported by C++, focusing on the principles of object-oriented design and programming.
Chapter 6, "Exception Handling," first describes traditional error handling methods and their disadvantages, and then presents standard exception handling. A brief historical account of the design of exception handling is provided and, finally, exception handling-related performance issues are discussed.
Chapter 7, "Runtime Type Information," discusses the three components of runtime type information (RTTI), namely typeid, dynamic_cast and class type_info. In addition, it explains when the use of RTTI is necessary. Finally, it discusses the performance overhead associated with runtime type information.
Chapter 8, "Namespaces," elucidates the rationale behind the addition of namespaces to the language and the problems that namespaces solve. Then it demonstrates how namespaces are used in practice, and how they interact with other language features.
Chapter 9, "Templates," discusses various aspects of designing and implementing templates, including class templates, function templates, and template issues that are of special concern (such as pointers to members, virtual member functions within a template class, inheritance relations, and explicit instantiations).
Chapter 10, "STL and Generic Programming," is an introduction to the Standard Template Library and to generic programming in general. It discusses the principles of generic programming, focusing on STL as an exemplary framework of generic programming. This chapter also demonstrates the use of STL components: containers, algorithms, iterators, allocators, adapters, binders, and function objects. The most widely used STL components, std::vector and std::string, are explored in detail.
Chapter 11, "Memory Management," explains the memory model of C++. It describes the three types of data storage: static, automatic, and free store. This chapter also delves into the semantics of operators new and delete and their underlying implementation. In addition, it demonstrates the use of advanced memory management techniques and guides you in avoiding common memory-related errors.
Chapter 12, "Optimizing Your Code," is dedicated to code optimization. It provides useful guidelines and tips for writing more efficient code, and it proceeds toward more aggressive optimization techniques for minimizing space and accelerating runtime speed.
Chapter 13, "C Language Compatibility Issues," demonstrates how to migrate from C to C++ and, in particular, how to migrate from procedural programming to object-orientation. It lists the differences between the C subset of C++ and ISO C. Finally, it delves into the underlying representation of C++ objects in memory and their compatibility with C structs.
Chapter 14, "Concluding Remarks and Future Directions," seals this book. It describes the principles and guidelines in the design and evolution of C++ throughout the last two decades, and compares it to the evolution of other, less successful programming languages. Then it lists features that almost made it into the Standard. Next, it discusses possible future extensions, including automated garbage collection, object persistence, and concurrency. Other hypothetical future extensions that are described are dynamically linked libraries, rule-based programming, and extensible member functions.
© Copyright 1999, Macmillan Computer Publishing. All rights reserved.