From Wikibooks, the open-content textbooks collection

Foreword

This book covers the C++ programming language, its interactions with software design and real life use of the language. Its presented as an introductory to advance course but can also be used as reference book. If you are already familiar with programming in other languages you can skip most of the Getting Started Chapter (it deals with introducing the language and the first steps needed to get you started). You should not skip the Programming Paradigms introduction, since C++ does have some particulars on that topic that should be useful even if you already know an Object Oriented Programming language. The Language Comparisons Section, providing comparisons for some language(s) you may already know, is important for veterans. However if this is your first contact with programming then continue on reading, and take in consideration that the Programming Paradigms section can be hard to digest if you lack some experience, don't despair, the relevant points will be extended when other concepts are introduced, that section is provided to give you a mental framework to help you not only to understand C++, but to let you easily adapt to (and from) other languages that share those concepts.

The book is organized into different parts, but as this is a work that is always evolving, things may be missing or just not where they should be, you are free to become a writer and contribute to fix things up...

Reader Comments

If you have comments about the technical accuracy, content, or organization of this document, please tell us (e.g. by using the "discussion" pages or by email). Be sure to include the section or the part title of the document with your comments and the date of your copy of the book. If you are really convinced of your point, information or correction then become a writer (at Wikibooks) and do it, it can always be rolled back if someone disagrees.

Guide to Writers

Authors/Contributors should register if intending to make non-anonymous contributions to the book (this will give more value and relevance to your opinions and views on the evolution of the work and enable others to talk to you) and try to follow the structure. If you have major ideas or big changes use the discussion area; as a rule just go with the flow.

Conventions 

A set of conventions have been adopted on the creation of this book, please read about them before you contribute any content on the book's talk page.

Authors

The following people are authors to this book

Panic

This work is licensed under the Creative Commons Attribution-Share Alike 3.0 Unported license. In short: you are free to share and to make derivatives of this work under the conditions that you appropriately attribute it, and that you only distribute it under the same, similar or a compatible license. Any of the above conditions can be waived if you get permission from the copyright holder.

Acknowledgment is given for using some contents from other works like Wikipedia, the Wikibooks Java Programming, C Programming and C++ Exercises for beginners, the C/C++ Reference Web Site, and from Wikisource, as from the authors Scott Wheeler, Stephen Ferg and Ivor Horton.

There are many other contributors/editors to the book; a verifiable list of all contributions exist as History Logs at Wikibooks (http://en.wikibooks.org/).

Introducing C++

C++ (pronounced "see plus plus") is a general-purpose, object-oriented, statically typed, free-form, multi-paradigm programming language supporting procedural programming, data abstraction, and generic programming. During the 1990s, C++ became one of the most popular computer programming languages.

History

Bjarne Stroustrup a Computer Scientist, from Bell Labs was the designer and original implementer of C++ (originally named "C with Classes") during the 1980s as an enhancement to the C programming language. Enhancements started with the addition of classes, followed by, among many features, virtual functions, operator overloading, multiple inheritance, templates, and exception handling, these and other features are covered in detail along this book.

The C++ programming language is a standard recognized by the ANSI (The American National Standards Institute), BSI (The British Standards Institute), DIN (The German national standards organization), several other national standards bodies, and was ratified in 1998 by the ISO (The International Standards Organization) as ISO/IEC 14882:1998, consists of two parts: the Core Language and the Standard Library; the latter includes the Standard Template Library and the Standard C Library (ANSI C 89).

Features introduced in C++ include declarations as statements, function-like casts, new/delete, bool, reference types, const, inline functions, default arguments, function overloading, namespaces, classes (including all class-related features such as inheritance, member functions, virtual functions, abstract classes, and constructors), operator overloading, templates, the :: operator, exception handling, run-time type identification, and more type checking in several cases. Comments starting with two slashes ("//") were originally part of BCPL, and was reintroduced in C++. Several features of C++ were later adopted by C, including const, inline, declarations in for loops, and C++-style comments (using the // symbol).

The current version, which is the 2003 version, ISO/IEC 14882:2003 redefines the standard language as a single item. The STL that pre-dated the standardization of C++, and was originally implemented in Ada is now an integral part of the standard and requirement for a compliant implementation of the same. Many other C++ libraries exist which are not part of the Standard, such as Boost. Also, non-Standard libraries written in C can generally be used by C++ programs.

Since 2004, the standards committee (includes Bjarne Stroustrup) has been busy working out the details of a new revision of the standard, that has been temporarily titled C++0x, due publication in the end of 2011. Some implementations already support some of the proposed alterations.

C++ source code example

// 'Hello World!' program 
 
#include <iostream>
 
int main()
{
  std::cout << "Hello World!" << std::endl;
  return 0;
}

Traditionally the first program people write in a new language is called "Hello World." because all it does is print the words Hello World. Hello World Explained offers a detailed explanation of this code; the included source code is to give you an idea of a simple C++ program.

Overview

Before you begin your journey to understand how to write programs using C++, it is important to understand a few key concepts that you may encounter. These concepts are not unique to C++, but are helpful to understanding computer programming in general. Readers who have experience in another programming language may wish to skim through or skip this section entirely.

There are many different kinds of programs in use today. From the operating system you use that makes sure everything works as it should, to the video games and music applications you use for fun, programs can fulfill many different purposes. What all programs (also called software or applications) have in common is that they all are made up of a sequence of instructions written in some form of programming language. These instructions tell a computer what to do, and generally how to do it. Programs can contain anything from instructions to solve math problems or send emails, to how to behave when a video game character is shot in a game. The computer will follow the instructions of a program one instruction at a time from start to finish.

Why learn C++?

Why not? This is the most clarifying approach to the decision to learn anything. Although learning is always good, selecting what you learn is more important as it is how you will prioritize tasks. Another side of this problem is that you will be investing some time in getting a new skill set. You must decide how will this benefit you. Check your objectives and compare similar projects or see what the programming market is in need of. In any case, the more programming languages you know, the better.

If you are approaching the learning process only to add another notch under your belt, that is, willing only to dedicate enough effort to understand its major quirks and learn something about its dark corners then you should be best served in learning first two other languages, this will clarify what makes C++ special in its approach to programming problems. You should select one imperative language, and in this C will probably have a better market value and will have a direct relation to C++ (a good substitute would be ASM) and the second language should be an Object Oriented language like Java for the same reasons, as there is a close relation between the three languages.

If you are willing to dedicate a more than passing interest in C++ then you can even learn C++ as your first language, but dedicate some time understanding the different paradigms and why C++ is a multi-paradigm language, or how some like to call it, a hybrid language.

Learning C is not a requirement for understanding C++, but knowing how to use an imperative language is, C++ will not make it easy for you to understand and distinguish some of this deeper concepts, since that in C++ you are free to implement solutions with a greater range of freedom. Understanding what options to make will become the cornerstone of mastering the language.

You should not learn C++ if you are only interested in applying or learning about Object Oriented Programing since the nomenclature used and some of the approaches C++ takes to the problem will probably increase the difficulty level in learning and mastering those concepts, if you are truly interested in Object Oriented programming, the best language for that is Smalltalk.

As with all languages C++ has a specific scope of application, where it can truly shine, and if we take a quick comparison with the previous mentioned languages, C++ is harder to learn than C and Java but more powerful than both. C++ enables you to abstract from the little things you have to deal with in C or other lower level languages but will grant you a bigger control and responsibility than Java, but it will not provide the default features you can obtain in similar higher level languages. You will have to search and examine several external implementations of these features and freely select those that best serve your purposes or you may even have to implement your own solution.

Where to get a compiler

When you select your compiler you must take in consideration your system OS, your personal preferences and the documentation that you can get on using it.

One of most actualized and compatible compilers is GCC. The next section will show how to get a copy and install it on Windows. You can easily find information on the GCC website on how to do it under another OS. GCC is a decent choice, and can be obtained for free. Many Open Source platforms include a recent GCC version. Version 4.0 or later gives fairly good conformance to the C++ standard. Various IDEs are available to support GCC. For Windows, Microsoft Visual Studio Express is currently available free of charge (but not free as in non-proprietary) with a C++ compiler that can be used from the command line or from the supplied IDE. An IDE, or Integrated Development Environment, is generally a graphical environment which integrates functionality like editing, compiling, linking, and usually a help system etc.).

GCC

The GNU Compiler Collection is a free set of compilers developed by the Free Software Foundation, with Richard Stallman as one of the main architects.

There are many different pre-compiled GCC binaries on the Internet, some popular choices are listed below (with detailed steps for installation).

On Windows

Cygwin:

1. Go to http://www.cygwin.com and click on the "Install Cygwin Now" button in the upper right corner of the page.
2. Click "run" in the window that pops up, and click "next" several times, accepting all the default settings.
3. Choose any of the Download sites ("ftp.easynet.be", etc.) when that window comes up; press "next" and the Cygwin installer should start downloading.
4. When the "Select Packages" window appears, scroll down to the heading "Devel" and click on the "+" by it. In the list of packages that now displays, scroll down and find the "gcc-c++" package; this is the compiler. Click once on the word "Skip", and it should change to some number like "3.4" etc. (the version number), and an "X" will appear next to "gcc-core" and several other required packages that will now be downloaded.
5. Click "next" and the compiler as well as the Cygwin tools should start downloading; this could take a while. While you're waiting, go to http://www.crimsoneditor.com and download that free programmer's editor; it's powerful yet easy to use for beginners.
6. Once the Cygwin downloads are finished and you have clicked "next", etc. to finish the installation, double-click the Cygwin icon on your desktop to begin the Cygwin "command prompt". Your home directory will automatically be set up in the Cygwin folder, which now should be at "C:\cygwin" (the Cygwin folder is in some ways like a small Unix/Linux computer on your Windows machine -- not technically of course, but it may be helpful to think of it that way).
7. Type "g++" at the Cygwin prompt and press "enter"; if "g++: no input files" or something like it appears you have succeeded and now have the gcc C++ compiler on your computer (and congratulations -- you have also just received your first error message!).

MinGW + DevCpp-IDE

1. Go to http://www.bloodshed.net/devcpp.html, choose the version you want (eventually scrolling down), click on the appropriate download link! For the most current version, you will be redirected to http://www.bloodshed.net/dev/devcpp.html
2. Scroll down to read the license and then to the download links. Download a version with Mingw/GCC. It's much easier than to do this assembling yourself. With a very short delay (only some days) you will always get the most current version of mingw packaged with the devcpp IDE. It's absolutely the same as with manual download of the required modules.
3. You get an executable that can be executed at user level under any WinNT version. If you want it to be setup for all users, however, you need admin rights. It will install devcpp and mingw in folders of your wish.
4. Start the IDE and experience your first project!
   You will find something mostly similar to MSVC, including menu and button placement. Of course, many things are somewhat different if you were familiar with the former, but it's as simple as a handfull of clicks to let your first program run.

For DOS

DJGPP:

• Go to Delorie Software and download the GNU C++ compiler and other necessary tools. The site provides a Zip Picker in order to help identify which files you need, which is available from the main page.
• Use unzip32 or other extraction utility to place files into the directory of your choice (ie. C:\DJGPP).
• Set the envionment variables to configure DJGPP for compilation, by either adding lines to autoexec.bat or a custom batch file:

  set PATH=C:\DJGPP\BIN;%PATH%

  set DJGPP=C:\DJGPP\DJGPP.ENV

• If you are running MS-DOS or Windows 3.1, you need to add a few lines to config.sys if they are not already present:

  shell=c:\dos\command.com c:\dos /e:2048 /p

  files=40

  fcbs=40,0

Note: The GNU C++ compiler under DJGPP is named gpp.

For Linux

• For Redhat, get a gcc-c++ RPM, e.g. using Rpmfind and then install (as root) using rpm -ivh gcc-c++-version-release.arch.rpm
• For Fedora Core, install the GCC C++ compiler (as root) by using yum install gcc-c++
• For Mandrake, install the GCC C++ compiler (as root) by using urpmi gcc-c++
• For Debian, install the GCC C++ compiler (as root) by using apt-get install g++
• For Ubuntu, install the GCC C++ compiler by using sudo apt-get install g++
• If you cannot become root, get the tarball from ftp://ftp.gnu.org/ and follow the instructions in it to compile and install in your home directory.

For Mac OS X

Xcode has GCC C++ compiler bundled. It can be invoked from the Terminal in the same way as Linux, but can also be compiled in one of XCode's projects.

What is a Programming Language?

In the most basic terms, a "programming language" is a means of communication between a human being (programmer) and a computer. A programmer uses this means of communication in order to give the computer instructions. These instructions are called "programs".

Like the many languages we use to communicate with each other, there are many languages that a programmer can use to communicate with a computer. Each language has its own set of words and rules, called semantics. If you're going to write a program, you have to follow the semantics of the language you're writing in, or you won't be understood.

Programming languages can basically be divided in to two categories: Low-Level and High-level, next we will introduce you to these concepts and their relevance to C++.

Low-level Languages

There are two general types of low level "languages".

Machine code (also called binary) is the lowest form of a low-level language. Machine code consists of a string of 0s and 1s, which combine to form meaningful instructions that computers can take action on. If you look at a page of binary it becomes apparent why binary is never a practical choice for writing programs; what kind of person would actually be able to remember what a bunch of strings of 1 and 0 mean ?

Assembly language (also called ASM), is just above machine code on the scale from low level to high level. It is a human-readable translation of the machine language instructions the computer executes. For example, instead of referring to processor instructions by their binary representation (0s and 1s), the programmer refers to those instructions using a more memorable (mnemonic) form. These mnemonics are usually short collections of letters that symbolize the action of the respective instruction, such as "ADD" for addition, and "MOV" for moving values from one place to another.

You do not have to understand assembly language to program in C++, but it does help to have an idea of what's going on "behind-the-scenes". Learning about assembly language will also allow you to have more control as a programmer and help you in debugging and understanding code.

High-level Languages

Higher level languages partially solve the problem of abstraction to the hardware (CPU, co-processors, number of registers etc...) by providing portability of code. High-level languages do more with less code, although there is sometimes a loss in performance and less freedom for the programmer. They also attempt to use English language words in a form which can be read and generally interpreted by the average person with little experience in them. A program written in one of these languages is sometimes referred to as "human-readable code". In general, more abstraction makes it easier for a language be learned. No programming language is written in what one might call "plain English" though, (although BASIC comes close.) Because of this, the text of a program is sometimes referred to as "code", or more specifically as "source code." This is discussed in more detail in the Code Section of the book.

Keep in mind that this classification scheme is evolving. C++ is still considered a high-level language, but with the appearance of newer languages (Java, C#, Ruby etc...), C++ is beginning to be grouped with lower level languages like C.

Translating Programming Languages

Since a computer is only capable of understanding machine code, human-readable code must be either interpreted or translated into machine code.

An interpreter is a program (often written in a lower level language) that interprets the instructions of a program one instruction at a time into commands that are to be carried out by the interpreter as it happens. Typically each instruction consists of one line of text or provides some other clear means of telling each instruction apart and the program must be reinterpreted again each time the program is run.

A compiler is a program that translates the instruction of a program one instruction at a time into machine code. The translation into machine code may involve splitting one instruction understood by the compiler into multiple machine instructions. The instructions are only translated once and after that the machine can understand and follow the instructions directly whenever it is instructed to do so. A complete examination is given on the Compiler Section of the book.

The words and statements used to instruct the computer may differ, but no matter what words and statements are used, just about every programming language will include statements that will accomplish the following:

Input

Input is the act of getting information from a device such as a keyboard or mouse, or sometimes another program.

Output

Output is the opposite of input; it gives information to the computer monitor or another device or program.

Math/Algorithm

All computer processors (the brain of the computer), have the ability to perform basic mathematical computation, and every programming language has some way of telling it to do so.

Testing

Testing involves telling the computer to check for a certain condition and to do something when that condition is true or false. Conditionals are one of the most important concepts in programming, and all languages have some method of testing conditions.

Repetition

Perform some action repeatedly, usually with some variation.

An further examination is provided on the Statements Section of the book.

Believe it or not, that's pretty much all there is to it. Every program you've ever used, no matter how complicated, is made up of functions that look more or less like these. Thus, one way to describe programming is the process of breaking a large, complex task up into smaller and smaller subtasks until eventually the subtasks are simple enough to be performed with one of these simple functions.

C++ is mostly compiled rather than interpreted (there are some C++ interpreters), and then "executed" later. As complicated as this may seem, later you will see how easy it really is.

So as we have seen in the Introducing C++ Section, C++ evolved from C by adding some levels of abstraction (so we can correctly state that C++ is of a higher level than C). We will learn the particulars of those differences in the Programming Paradigms Section of the book and for some of you that already know some other languages should look into Programming Languages Comparisons Section.

Programming Paradigms

A programming paradigm is a style or model of programming that affects the way programmers can design, organize and write programs. A multi-paradigm programming language allows programmers to choose from a number of different programming paradigms. C++ is a multi-paradigm programming language.

Procedural Programming

Procedural programming can be defined as a subtype of imperative programming as a programming paradigm based upon the concept of procedure calls, in which statements are structured into procedures (also known as subroutines or functions). Procedure calls are modular and are bound by scope. A procedural program is composed of one or more modules. Each module is composed of one or more subprograms. Modules may consist of procedures, functions, subroutines or methods, depending on the programming language. Procedural programs may possibly have multiple levels or scopes, with subprograms defined inside other subprograms. Each scope can contain names which cannot be seen in outer scopes.

Procedural programming offers many benefits over simple sequential programming since procedural code:

• is easier to read and more maintainable
• is more flexible
• facilitates the practice of good program design
• allows modules to be reused in the form of code libraries.

Object-Oriented Programming

Object-oriented programming can be seen as an extension of procedural programming in which programs are made up of collection of individual units called objects that have a distinct purpose and function with limited or no dependencies on implementation. For example, a car is like an object; it gets you from point A to point B with no need to know what type of engine the car uses or how the engine works. Object-oriented languages usually provide a means of documenting what an object can and cannot do, like instructions for driving a car.

Objects and Classes

An object is composed of members and methods. The members (also called data members, characteristics, attributes, or properties) describe the object. The methods generally describe the actions associated with a particular object. Think of an object as a noun, its members as adjectives describing that noun, and its methods as the verbs that can be performed by or on that noun.

For example, a sports car is an object. Some of its members might be its height, weight, acceleration, and speed. An object's members just hold data about that object. Some of the methods of the sports car could be "drive", "park", "race", etc. The methods really don't mean much unless associated with the sports car, and the same goes for the members.

The blueprint that lets us build our sports car object is called a class. A class doesn't tell us how fast our sports car goes, or what color it is, but it does tell us that our sports car will have a member representing speed and color, and that they will be say, a number and a word, respectively. The class also lays out the methods for us, telling the car how to park and drive, but these methods can't take any action with just the blueprint - they need an object to have an effect.

Encapsulation

Encapsulation, the principle of information hiding (from the user), is the process of hiding the data structures of the class and allowing changes in the data through a public interface where the incoming values are checked for validity, and so not only it permits the hiding of data in an object but also of behavior. This prevents clients of an interface from depending on those parts of the implementation that are likely to change in future, thereby allowing those changes to be made more easily, that is, without changes to clients. In modern programming languages, the principle of information hiding manifests itself in a number of ways, including encapsulation and polymorphism.

Inheritance

This concept describes a relationship between two (or more) types, or classes, of objects in which one is said to be a "subtype" or "child" of the other, as result the "child" object is said to inherit features of the parent, allowing for shared functionality, this lets programmers re-use or reduce code and simplifies the development and maintenance of software.

Inheritance is also commonly held to include subtyping, whereby one type of object is defined to be a more specialized version of another type (see Liskov substitution principle), though non sub-typing inheritance is also possible.

Inheritance is typically expressed by describing classes of objects arranged in an inheritance hierarchy reflecting common behavior.

For example, one might create a variable class "Mammal" with features such as eating, reproducing, etc.; then define a subtype "Cat" that inherits those features without having to explicitly program them, while adding new features like "chasing mice". This allows commonalities among different kinds of objects to be expressed once and reused multiple times.

In C++ we can then have classes which are related to other classes (a class can be defined by means of an older, pre-existing, class ). This leads to a situation in which a new class has all the functionality of the older class, and additionally introduces its own specific functionality. Instead of composition, where a given class contains another class, we mean here derivation, where a given class is another class.

This OOP property will be explained further when we talk about Classes (and Structures) inheritance in the Classes Inheritance Section of the book.

If one wants to use more than one totally orthogonal hierarchy simultaneously, such as allowing "Cat" to inherit from "Cartoon character" and "Pet" as well as "Mammal" we are using multiple inheritance.

Multiple Inheritance

Multiple inheritance is the process by which one class can inherit the properties of two or more classes (variously known as its base classes, or parent classes, or ancestor classes, or super classes).

In some similar language, multiple inheritance is restricted in various ways to keep the language simple, such as by allowing inheritance from only one real class and a number of "interfaces", or by completely disallowing multiple inheritance. C++ places the full power of multiple inheritance in the hands of programmers, but it is needed only rarely, and (as with most techniques) can complicate code if used inappropriately. Because of C++'s approach to multiple inheritance, C++ has no need of separate language facilities for "interfaces"; C++'s classes can do everything that interfaces do in some related languages.

Polymorphism

Polymorphism allows a single name to be reused for several related but different purposes. The purpose of polymorphism is to allow one name to be used for a general class. Depending on the type of data, a specific instance of the general case is executed.

The concept of polymorphism is wider. Polymorphism exists every time we use two functions that have the same name, but differ in the implementation. They may also differ in their interface, e.g., by taking different arguments. In that case the choice of which function to make is via overload resolution, and is performed at compile time, so we refer to static polymorphism.

Dynamic polymorphism will be covered deeply in the Classes Section where we will address its use on redefining the method in the derived class.

Generic Programming

Generic programming or polymorphism is a programming style that emphasizes techniques that allow one value to take on different types as long as certain contracts such as subtypes and signature are kept. In simpler terms generic programming is based in finding the most abstract representations of efficient algorithms. Templates popularized the notion of generics. Templates allow code to be written without consideration of the type with which it will eventually be used. Templates are defined in the Standard Template Library (STL), where generic programming was introduced into C++.

Statically Typed

Typing refers to how a computer language handles its variables. Variables are values that the program uses during execution. These values can change; they are variable, hence their name. Static typing usually results in compiled code that executes more quickly. When the compiler knows the exact types that are in use, it can produce machine code that does the right thing easier. In C++, variables need to be defined before they are used so that compilers know what type they are, and hence is statically typed. Languages that are not statically typed are called dynamically typed.

Static typing usually finds type errors more reliably at compile time, increasing the reliability of compiled programs. Simply put, it means that "A round peg won't fit in a square hole", so the compiler will report it when a type leads to ambiguity or incompatible usage. However, programmers disagree over how common type errors are and what proportion of bugs that are written would be caught by static typing. Static typing advocates believe programs are more reliable when they have been type checked, while dynamic typing advocates point to distributed code that has proved reliable and to small bug databases. The value of static typing, then, presumably increases as the strength of the type system is increased.

A statically typed system constrains the use of powerful language constructs more than it constrains less powerful ones. This makes powerful constructs harder to use, and thus places the burden of choosing the "right tool for the problem" on the shoulders of the programmer, who might otherwise be inclined to use the most powerful tool available. Choosing overly powerful tools may cause additional performance, reliability or correctness problems, because there are theoretical limits on the properties that can be expected from powerful language constructs. For example, indiscriminate use of recursion or global variables may cause well-documented adverse effects.

Static typing allows construction of libraries which are less likely to be accidentally misused by their users. This can be used as an additional mechanism for communicating the intentions of the library developer.

Free-form

Free-form refers to how the programmer crafts the code. Basically, there are no rules on how you choose to write your program, save for the semantic rules of C++. Any C++ program should compile as long as it is legal C++.

The free-form nature of C++ is used (or abused, depending on your point of view) by some programmers in crafting obfuscated C++ (C++ that is purposefully written to be difficult to understand). The use of obfuscation is regarded by some as a security device, ensuring that the source code can is harder to analyzed by the average user.

Language Comparisons

There isn't a perfect language. It all depends on the tools and the objective. The optimal language (in terms of run-time performance) is machine code but machine code (binary) is the least efficient programming language in terms of coder time. The complexity of writing large systems is enormous with high-level languages, and beyond human capabilities with machine code. In the next section C++ will be compared with other closely related languages like C, Java, C# and C++/CLI.

The quote above is shown to indicate that no programming language at present can translate directly concepts or ideas into useful code, there are solutions that will help. We will cover the use of Computer-aided software engineering (CASE) tools that will address part of this problem but its use does require planning and some degree of complexity.

The intention of these sections is not to promote one language above another; each has its applicability. Some are better in specific tasks, some are simpler to learn, others only provide a better level of control to the programmer. This all may depend also on the level of control the programmer has of a given language.

Garbage Collection

In C++ garbage collection is optional rather than required. In the Garbage Collection Section of this book we will cover this issue deeply.

Why doesn't C++ include a finally keyword?

As we will see in the Resource Acquisition Is Initialization (RAII) Section of the book, RAII can be used to provide a better solution for most issues. When finally is used to clean up, it has to be written by the clients of a class each time that class is used (for example, clients of a File class have to do I/O in a try/catch/finally block so that they can guarantee that the File is closed). With RAII, the destructor of the File class can make that guarantee. Now the cleanup code has to be coded only once — in the destructor of File; the users of the class don't need to do anything.

TODO Split this explanation to RAII and only provide the reference

Mixing Languages

TODO Add relevant information

By default, C++ compilers normally "mangle" the names of functions in order to facilitate function overloading, and generic functions. In some cases, you need to gain access to a function that wasn't created in a C++ compiler. For this to occur, you need to declare a function as external:

extern "C" void LibraryFunction();

C 89/99

C was essentially the core language of C++ when Bjarne Stroustrup, decided to create a "better C". Many of the syntax conventions and rules still hold true and so we can even state that C was a subset of C++, most recent C++ compilers will also compile C code taking into consideration the small incompatibilities, since C99 and C++ 2003 are not compatible any more. You can also check more information about the C language on the C Programming Wikibooks ( http://en.wikibooks.org/wiki/C ).

C++ as defined by the ANSI standard in 98 (called C++98 at times) is very nearly, but not quite, a superset of the C language as it was defined by its first ANSI standard in 1989 (known as C89). There are a number of ways in which C++ is not a strict superset, in the sense that not all valid C89 programs are valid C++ programs, but the process of converting C code to valid C++ code is fairly trivial (avoiding reserved words, getting around the stricter C++ type checking with casts, declaring every called function, and so on).

In 1999, C was revised and many new features were added to it. As of 2004, most of these new "C99" features are not there in C++. Some (including Stroustrup himself) have argued that the changes brought about in C99 have a philosophy distinct from what C++98 adds to C89, and hence these C99 changes are directed towards increasing incompatibility between C and C++.

The merging of the languages seems a dead issue as coordinated actions by the C and C++ standards committees leading to a practical result didn't happen and it can be said that the languages started even to diverge.

Some of the differences are:

• C++ supports function overloading (absent in C89, allowed only for some standard library code in C99).
• C++ supports inheritance and polymorphism.
• C++ adds keyword class, but keeps struct from C, with compatible semantics.
• C++ supports access control for class members.
• C++ supports generic programming through the use of templates.
• C++ extends the C89 standard library with its own standard library.
• C++ and C99 offer different complex number facilities.
• C++ has bool and wchar_t as primitive types, while typedefs in C.
• C++ comparison operators return bool, while C returns int.
• C++ supports overloading of operators.
• C++ character constants have type char, while C character constants have type int.
• C++ has additional cast operators (static_cast, dynamic_cast, const_cast and reinterpret_cast).
• C++ adds mutable keyword to address the imperfect match between physical and logical constness.
• C++ extends the type system with references.
• C++ supports member functions, constructors and destructors for user-defined types to establish invariants and to manage resources.
• C++ supports runtime type identification (RTTI), via typeid and dynamic_cast.
• C++ includes exception handling.
• C++ has std::vector as part of its standard library instead of variable-length arrays as in C.
• C++ treats sizeof operator as compile time operation, while C allows it be a runtime operation.
• C++ has new and delete operators, while C uses malloc and free library functions exclusively.
• C++ supports object-oriented programming without extensions.
• C++ does not require use of macros and careful information-hiding and abstraction for code portability.
• C++ supports per-line comments denoted by //. (C99 started official support for this comment system, and most compilers supported this as an extension.)

Reasons to chose one of the languages over the other

It is not uncommon to find someone defending C over C++ or vice versa or complain about some of those languages features. There is no scientific evidence to put most languages above another in general terms, the only reasons that do have some traction is if the language is still very recent and prone to deep changes or as yet unknown bugs, in the case of C or C++ this is not the case both languages are very mature even if both are still evolving, the new features keep an high level of compatibility with old code, making the use of those new constructs a programmer's decision. It is not uncommon to establish rules in a project to limit the use of parts of a language (such as RTTI, exceptions, or virtual-functions in inner loops), due the proficiency of the programmers or the needs of the project, as it is also common for new hardware to support lower level languages first. Due to C being less extensive and lower level than C++, it is easier to check and comply with strict industry guidelines and automate those steps. Another benefit is that it is easier for the programmer to do low level optimizations, even if most C++ compilers can guarantee near perfect optimizations automatically, a human can still do more and C has less complex structures.

Any of the valid reasons to choose a language over another is mostly due to programmers choice that indirectly deals with choosing the best tool for the job and having the resources needed to complete it. It would be hard to validate selecting C++ for a project if the available programmers only knew C or even the reverse, it is somewhat expected for a C++ programmer to produce functional C code, but the mindset and experience needed aren't the same, the same rational is valid for C programmers and ASM, this is due to the close relations that exist in the language structure and historic evolution.

One could argue that using the C subset of C++, in a C++ compiler, is the same as using C but in reality we find that it will generate slightly different results depending on the compiler used.

Java

This is a comparison of the Java programming language with the C++ programming language. C++ and Java share many common traits. You can get a better understanding of Java in the Java Programming WikiBook.

Java was created initially to support network computing on embedded systems. Java was designed to be extremely portable, secure, multi-threaded and distributed, none of which were design goals for C++. The syntax of Java was chosen to be familiar to C programmers, but direct compatibility with C was not maintained. Java also was specifically designed to be simpler than C++ but it keeps evolving above that simplification.

C++	Java
backwards compatible with C	backwards compatibility with previous versions
execution efficiency	developer productivity
trusts the programmer	restrains the programmer's abilities
arbitrary memory access possible	memory access only through objects
concise expression	explicit operation
can arbitrarily override types	type safety
procedural or object-oriented	object-oriented
operator overloading	meaning of operators immutable
powerful capabilities of language	feature-rich, easy to use standard library

Differences between C++ and Java are:

• C++ parsing is somewhat more complicated than with Java; for example, Foo<1>(3); is a sequence of comparisons if Foo is a variable, but it creates an object if Foo is the name of a class template.
• C++ allows namespace level constants, variables, and functions. All such Java declarations must be inside a class or interface.
• const in C++ indicates data to be 'read-only,' and is applied to types. final in java indicates that the variable is not to be reassigned. For basic types such as const int vs final int these are identical, but for complex classes, they are different.
• C++ doesn't support constructor delegation.
• C++ runs on the hardware, Java runs on a virtual machine so with C++ you have greater power at the cost of portability.
• C++, int main() is a function by itself, without a class.
• C++ access specification (public, private) is done with labels and in groups.
• C++ access to class members default to private, in Java it is package access.
• C++ classes declarations end in a semicolon.
• C++ lacks language level support for garbage collection while Java has built-in garbage collection to handle memory deallocation.
• C++ supports goto statements; Java does not, but its labeled break and labeled continue statements provide some structured goto-like functionality. In fact, Java enforces structured control flow, with the goal of code being easier to understand.
• C++ provides some low-level features which Java lacks. In C++, pointers can be used to manipulate specific memory locations, a task necessary for writing low-level operating system components. Similarly, many C++ compilers support inline assembler. In Java, assembly code can still be accessed as libraries, through the Java Native Interface. However, there is significant overhead for each call.
• C++ allows a range of implicit conversions between native types, and also allows the programmer to define implicit conversions involving compound types. However, Java only permits widening conversions between native types to be implicit; any other conversions require explicit cast syntax.
  • A consequence of this is that although loop conditions (if, while and the exit condition in for) in Java and C++ both expect a boolean expression, code such as if(a = 5) will cause a compile error in Java because there is no implicit narrowing conversion from int to boolean. This is handy if the code were a typo for if(a == 5), but the need for an explicit cast can add verbosity when statements such as if (x) are translated from Java to C++.
• For passing parameters to functions, C++ supports both true pass-by-reference and pass-by-value. As in C, the programmer can simulate by-reference parameters with by-value parameters and indirection. In Java, all parameters are passed by value, but object (non-primitive) parameters are reference values, meaning indirection is built-in.
• Generally, Java built-in types are of a specified size and range; whereas C++ types have a variety of possible sizes, ranges and representations, which may even change between different versions of the same compiler, or be configurable via compiler switches.
  • In particular, Java characters are 16-bit Unicode characters, and strings are composed of a sequence of such characters. C++ offers both narrow and wide characters, but the actual size of each is platform dependent, as is the character set used. Strings can be formed from either type.
• The rounding and precision of floating point values and operations in C++ is platform dependent. Java provides a strict floating-point model that guarantees consistent results across platforms, though normally a more lenient mode of operation is used to allow optimal floating-point performance.
• In C++, pointers can be manipulated directly as memory address values. Java does not have pointers—it only has object references and array references, neither of which allow direct access to memory addresses. In C++ one can construct pointers to pointers, while Java references only access objects.
• In C++ pointers can point to functions or member functions (function pointers or functors). The equivalent mechanism in Java uses object or interface references.
• C++ features programmer-defined operator overloading. The only overloaded operators in Java are the "+" and "+=" operators, which concatenate strings as well as performing addition.
• Java features standard API support for reflection and dynamic loading of arbitrary new code.
• Java has generics. C++ has templates.
• Both Java and C++ distinguish between native types (these are also known as "fundamental" or "built-in" types) and user-defined types (these are also known as "compound" types). In Java, native types have value semantics only, and compound types have reference semantics only. In C++ all types have value semantics, but a reference can be created to any object, which will allow the object to be manipulated via reference semantics.
• C++ supports multiple inheritance of arbitrary classes. Java supports multiple inheritance of types, but only single inheritance of implementation. In Java, a class can derive from only one class, but a class can implement multiple interfaces.
• Java explicitly distinguishes between interfaces and classes. In C++ multiple inheritance and pure virtual functions makes it possible to define classes that function just as Java interfaces do.
• Java has both language and standard library support for multi-threading. The synchronized keyword in Java provides simple and secure mutex locks to support multi-threaded applications. While mutex lock mechanisms are available through libraries in C++, the lack of language semantics makes writing thread safe code more difficult and error prone.

Memory management

• Java requires automatic garbage collection. Memory management in C++ is usually done by hand, or through smart pointers. The C++ standard permits garbage collection, but does not require it; garbage collection is rarely used in practice. When permitted to relocate objects, modern garbage collectors can improve overall application space and time efficiency over using explicit deallocation.
• C++ can allocate arbitrary blocks of memory. Java only allocates memory through object instantiation. (Note that in Java, the programmer can simulate allocation of arbitrary memory blocks by creating an array of bytes. Still, Java arrays are objects.)
• Java and C++ use different idioms for resource management. Java relies mainly on garbage collection, while C++ relies mainly on the RAII (Resource Acquisition Is Initialization) idiom. This is reflected in several differences between the two languages:
  • In C++ it is common to allocate objects of compound types as local stack-bound variables which are destructed when they go out of scope. In Java compound types are always allocated on the heap and collected by the garbage collector (except in virtual machines that use escape analysis to convert heap allocations to stack allocations).
  • C++ has destructors, while Java has finalizers. Both are invoked prior to an object's deallocation, but they differ significantly. A C++ object's destructor must be implicitly (in the case of stack-bound variables) or explicitly invoked to deallocate the object. The destructor executes synchronously at the point in the program at which the object is deallocated. Synchronous, coordinated uninitialization and deallocation in C++ thus satisfy the RAII idiom. In Java, object deallocation is implicitly handled by the garbage collector. A Java object's finalizer is invoked asynchronously some time after it has been accessed for the last time and before it is actually deallocated, which may never happen. Very few objects require finalizers; a finalizer is only required by objects that must guarantee some clean up of the object state prior to deallocation—typically releasing resources external to the JVM. In Java safe synchronous deallocation of resources is performed using the try/finally construct.
  • In C++ it is possible to have a dangling pointer – a reference to an object that has been destructed; attempting to use a dangling pointer typically results in program failure. In Java, the garbage collector won't destruct a referenced object.
  • In C++ it is possible to have an object that is allocated, but unreachable. An unreachable object is one that has no reachable references to it. An unreachable object cannot be destructed (deallocated), and results in a memory leak. By contrast, in Java an object will not be deallocated by the garbage collector until it becomes unreachable (by the user program). (Note: weak references are supported, which work with the Java garbage collector to allow for different strengths of reachability.) Garbage collection in Java prevents many memory leaks, but leaks are still possible under some circumstances.

Libraries

• C++ standard library only provides components that are relatively general purpose, such as strings, containers, and I/O streams. Java has a considerably larger standard library. This additional functionality is available for C++ by (often free) third party libraries, but third party libraries do not provide the same ubiquitous cross-platform functionality as standard libraries.
• C++ is mostly backward compatible with C, and C libraries (such as the APIs of most operating systems) are directly accessible from C++. In Java, the richer functionality of the standard library is that it provides cross-platform access to many features typically available in platform-specific libraries. Direct access from Java to native operating system and hardware functions requires the use of the Java Native Interface.

Runtime

• C++ is normally compiled directly to machine code which is then executed directly by the operating system. Java is normally compiled to byte-code which the Java virtual machine (JVM) then either interprets or JIT compiles to machine code and then executes.
• Due to the lack of constraints in the use of some C++ language features (e.g. unchecked array access, raw pointers), programming errors can lead to low-level buffer overflows, page faults, and segmentation faults. The Standard Template Library, however, provides higher-level abstractions (like vector, list and map) to help avoid such errors. In Java, such errors either simply cannot occur or are detected by the JVM and reported to the application in the form of an exception.
• In Java, bounds checking is implicitly performed for all array access operations. In C++, array access operations on native arrays are not bounds-checked, and bounds checking for random-access element access on standard library collections like std::vector and std::deque is optional.

Miscellaneous

• Java and C++ use different techniques for splitting up code in multiple source files. Java uses a package system that dictates the file name and path for all program definitions. In Java, the compiler imports the executable class files. C++ uses a header file source code inclusion system for sharing declarations between source files. (See Comparison of imports and includes.)
• Templates and macros in C++, including those in the standard library, can result in duplication of similar code after compilation. Second, dynamic linking with standard libraries eliminates binding the libraries at compile time.
• C++ compilation features a textual preprocessing phase, while Java does not. Java supports many optimizations that mitigate the need for a preprocessor, but some users add a preprocessing phase to their build process for better support of conditional compilation.
• In Java, arrays are container objects which you can inspect the length of at any time. In both languages, arrays have a fixed size. Further, C++ programmers often refer to an array only by a pointer to its first element, from which they cannot retrieve the array size. However, C++ and Java both provide container classes (std::vector and java.util.ArrayList respectively) which are resizable and store their size.
• Java's division and modulus operators are well defined to truncate to zero. C++ does not specify whether or not these operators truncate to zero or "truncate to -infinity". -3/2 will always be -1 in Java, but a C++ compiler may return either -1 or -2, depending on the platform. C99 defines division in the same fashion as Java. Both languages guarantee that (a/b)*b + (a%b) == a for all a and b (b != 0). The C++ version will sometimes be faster, as it is allowed to pick whichever truncation mode is native to the processor.
• The sizes of integer types is defined in Java (int is 32-bit, long is 64-bit), while in C++ the size of integers and pointers is compiler-dependent. Thus, carefully-written C++ code can take advantage of the 64-bit processor's capabilities while still functioning properly on 32-bit processors. However, C++ programs written without concern for a processor's word size may fail to function properly with some compilers. In contrast, Java's fixed integer sizes mean that programmers need not concern themselves with varying integer sizes, and programs will run exactly the same. This may incur a performance penalty since Java code cannot run using an arbitrary processor's word size.

Performance

Computing performance is a measure of resource consumption when a system of hardware and software performs a piece of computing work such as an algorithm or a transaction. Higher performance is defined to be 'using fewer resources'. Resources of interest include memory, bandwidth, persistent storage and CPU cycles. Because of the high availability of all but the latter on modern desktop and server systems, performance is colloquially taken to mean the least CPU cycles; which often converts directly into the least wall clock time. Comparing the performance of two software languages requires a fixed hardware platform and (often relative) measurements of two or more software subsystems. This section compares the relative computing performance of C++ and Java on common operating systems such as Windows and Linux.

Early versions of Java were significantly outperformed by statically compiled languages such as C++. This is because the program statements of these two closely related Level 6 languages may compile to a few machine instructions with C++, while compiling into several byte codes involving several machine instructions each when interpreted by a Java JVM. For example:

While this may still be the case for embedded systems because of the requirement for a small footprint, advances in just in time (JIT) compiler technology for long-running server and desktop Java processes has closed the performance gap and in some cases given the performance advantage to Java. In effect, Java byte code is compiled into machine instructions at run time, in a similar manner to C++ static compilation, resulting in similar instruction sequences.

C++ is still faster in most operations than Java at the moment, even at low-level and numeric computation. For in-depth information you could check Performance of Java versus C++. It's a bit pro-Java but very detailed.

C#

C# (pronounced "See Sharp") is a multi-purpose computer programming language suitable for all development needs. There is a WikiBook (http://en.wikibooks.org/wiki/C_sharp) that introduces C# language fundamentals and covers a variety of the base class libraries (BCL) provided by the Microsoft .NET Framework.

C# is very similar to Java in that it takes the basic operators and style of C++ but forces programs to be type safe, in that it executes the code in a controlled sandbox called the virtual machine. As such, all code must be encapsulated inside an object, among other things. C# provides many additions to facilitate interaction with Microsoft's Windows, COM, and Visual Basic.

There are several shortcomings to C++ which are resolved in C#. One of the more subtle ones is the use of reference variables as function arguments. When a code maintainer is looking at C++ source code, if a called function is declared in a header somewhere, the immediate code does not provide any indication that an argument to a function is passed as a reference. An argument passed by reference could be changed after calling the function whereas an argument passed by value cannot be changed. A maintainer not be familiar with the function looking for the location of an unexpected value change of a variable would additionally need to examine the header file for the function in order to determine whether or not that function could have changed the value of the variable. C# insists that the ref keyword be placed in the function call (in addition to the function declaration), thereby cluing the maintainer in that the value could be changed by the function.

TODO Should refer MS & MONO and Portable.NET http://getdotgnu.com/pnet, as well as the ECMA and ISO standards for C# and CLI

Managed C++ (C++/CLI)

Managed C++ is a shorthand notation for Managed Extensions for C++, which are part of the .NET framework from Microsoft. This extension of the C++ language was developed to add functionality like automatic garbage collection and heap management, automatic initialization of arrays, and support for multidimensional arrays, simplifying all those details of programming in C++ that would otherwise have to be done by the programmer.

Managed C++ is not compiled to machine code. Rather, it is compiled to Common Intermediate Language, which is an object-oriented machine language and was formerly known as MSIL

D

The D programming language, also known simply as D, was developed in-house by Digital Mars, the language specification and D compiler (Windows and Linux) is freely distributed for free on their web site. Only the compiler front-end is licensed under both the Artistic License and the GNU GPL; sources for the front-end are distributed along with the compiler binaries. The compiler back-end is proprietary.

Digital Mars is a small US software company, also known for producing a C compiler (known over time as Datalight C compiler, then Zorland C and then Zortech C) and a C++ compiler (known as Zortech C++ that is attribute to be the first C++ compiler for Windows, renamed later as Symantec C++ and now Digital Mars C++ (DMC++) and associated to utilities such as an IDE for Windows (supporting the MFC library).

D originated as a re-engineering of C++, but even though it is predominantly influenced by that language, it is also a multi-paradigm language, it is not a variant of it. D has redesigned some C++ features and has been influenced by concepts used in other programming languages, such as Java, C# and Eiffel.

Wikibooks has a D Programming book, in this section we will point out some features were D is distinct from C++:

• D does not support multiple inheritance.

TODO Complete with distinctions from C++ to D

Chapter Summary

1. Introducing C++
2. Programming languages
   1. Programming paradigms - the versatility of C++ as a multi-paradigm language, concepts of Object-Oriented Programming (Objects and Classes, Inheritance, Polymorphism).
3. Comparisons - to other languages, relation to other computer science constructs and idioms.
   1. with C
   2. with Java
   3. with C#
   4. with Managed C++ (C++/CLI)
   5. with D

The Code

The task of programming, while not easy in its execution, is actually fairly simple in its goals. A programmer will envision, or be tasked with, a specific goal. Goals are usually provided in the form of "I want a program that will perform...fill in the blank..." The job of the programmer then is to come up with a "working model" (a model that may consist of one or more algorithms). That "working model" is sort of an idea of how a program will accomplish the goal set out for it. It gives a programmer an idea of what to write in order to turn the idea in to a working program. Once the programmer has an idea of the structure their program will need to take in order to accomplish the goal, they set about actually writing the program itself with all of the proper commands, functions and syntax. The code that they write is what actually implements the program, or causes it to perform the necessary task, and for that reason, it is sometimes called "implementation code".

How the instructions of a program are written out and stored is generally not a concept determined by a programming language. Punch cards used to be in common use, however under most modern operating systems the instructions are commonly saved as plain text files that can be edited with any text editor. These files are the source of the instructions that make up a program and so are sometimes referred to as source files but a more exclusive definition is source code.

When referring to source code or just source, you are considering only the files that contain code, the actual text that makes up the functions (actions) for computer to execute. By referring to source files you are extending the idea to not only the files with the instructions that make up the program but all the raw files resources that together can build the program.

Source code

Source code is the halfway point between human language and machine code. As mentioned before, it can be read by people to an extent, but it can also be parsed (converted) into machine code by a computer. The machine code is the strings of 1's and 0's that the computer can fully understand and act on.

In a small program, you might have as little as a few dozen lines of code at the most, whereas in larger programs, this number might stretch into the thousands or even millions. For this reason, it is sometimes more practical to split large amounts of code across many files. This makes it easier to read, as you can do it bit by bit, and it also reduces compile time of each source file. It takes much less time to compile a lot of small source files than it does to compile a single massive source file.

Managing size is not the only reason to split code, though. Often, especially when a piece of software is being developed by a large team, source code is split. Instead of one massive file, the program is divided into separate files, and each individual file contains the code to perform one particular set of tasks for the overall program. This creates a condition known as Modularity. Modularity is a quality that allows source code to be changed, added to, or removed a piece at a time. This has the advantage of allowing many people to work on separate aspects of the same program, thereby allowing it to move faster and more smoothly. Source code for a large project should always be written with modularity in mind. Even when working with small or medium sized projects, it is good to get in the habit of writing code with ease of editing and use in mind.

C++ source code is case sensitive. This means that it distinguishes between lowercase and capital letters, so that it sees the words "hello," "Hello," and "HeLlO" as being totally different things. This is important to remember and understand, it will be discussed further in the Coding Style Section.

File Organization

Most operating systems require C++ files to be designated by a name followed by a specific extension. The C++ standard doesn't impose any specific rules on how files are named or organized.

The specific conventions for the file organizations has both technical reasons and organizational benefits, very similar to the code style conventions we will examine later. Most of the conventions governing files derive from historical preferences and practices, that are especially related with lower level languages that preceded C++. This is especially true when we take into consideration that C++ was built over the C89 ANSI standard, with compatibility in mind so most practices have remained mostly static, except for the operating systems improved support for files and greater ease of management of file resources.

One of the evolutions when dealing with filenames on the language standard was that the default include files would have no extension. Most implementations still provide the old C style headers that use C's file extension ".h" for the C Standard Library, but C++-specific header filenames that were terminated in the same fashion now have no extension (e.g. iostream.h is now iostream). This change to old C++ headers was simultaneous with the implementation of namespaces, in particular the std namespace.

File Names

TODO Add reference to over .cpp .h, common rules to file naming and code distribution.

Do not reuse a standard header file name

As you will see later, the C++ Standard defines a list of headers. The behavior is undefined if a file with the same name as a standard header is placed in the search path for included source files.

Extensions

The extension serves one purpose: to indicate to the Operating System, the IDE or the compiler what resides within the file. By itself an extension will not serve as a guarantee for the content.

Since the C language sources usually have the extension ".c" and ".h", in the beginning it was common for C++ source files to share the same extensions or use a distinct variation to clearly indicate the C++ code file. Today this is the practice, most C++ implementation files will use the ".cpp" extension and ".h" for the declaration or header files (the last one is still shared across most assembler and C compilers).

There are other common extensions variations, such as, ".cc", ".C", ".cxx", and ".c++" for "implementation" code. For header files, the same extension variations are used, but the first letter of the extension is usually replaced with an "h" as in, ".hh", ".H", ".hxx", "hpp", ".h++" etc...

Header files will be discussed with more detail later in the Preprocessor Section when introducing the #include directive and the standard headers, but in general terms a header file is a special kind of source code file that is included (by the preprocessor) by way of the #include directive, traditionally used at the beginning of a ".cpp" file.

Source Code

C++ programs would be compilable even if using a single file, but any complex project will benefit from being split into several source files in order to be manageable and permit re-usability of the code. The beginning programmer sees this as an extra complication, where the benefits are obscure, especially since most of the first attempts will probably result in problems. This section will cover not only the benefits and best practices but also explain how a standardized method will avoid and reduce complexity.

Why split code into several files?

• Increases organization and better code structure.
• Improves compilation speed.
• Promotes code reuse.

TODO Complete

Source File Types

Some authors will refer to files with a .cpp extension as "source files" and files with the .h extension as "header files". However, both of those qualify as source code. As a convention for this book, all code, whether contained within a .cpp extension (where a programmer would put it), or within a .h extension (for headers), will be called source code. Any time we're talking about a .cpp file, we'll call it an "implementation file", and any time we're referring to a header file, we'll call it a "declaration file". You should check the editor/IDE or alter the configuration to a setup that best suits you and others that will read and use this files.

Declaration vs Definition

In general terms a declaration specifies for the linker, the identifier, type and other aspects of language elements such as variables and functions. It is used to announce the existence of the element to the compiler which require variables to be declared before use.

The definition assigns values to an area of memory that was reserved during the declaration phase. For functions, definitions supply the function body. While a variable or function may be declared many times, it is typically defined once.

An object may be declared many times but may only be defined one time.

This concept will be further explained and with some particulars noted (such as inline) as we introduce other components of C++. Here are some examples, ignore them for now if some of the concepts are beyond you and came back later to see the distinctions.

    int an_integer;                                 // defines an_integer
    extern const int a = 1;                         // defines a
    int function( int b ) { return b+an_integer; }  // defines function and defines b
    struct a_struct { int a; int b; };              // defines a_struct, a_struct::a, and a_struct::b
    struct another_struct {                         // defines another_strct
      int a;                                        // defines nonstatic data member x
      static int b;                                 // declares static data member y
      another_struct(): a(0) { } };                 // defines a constructor of another_struct
    int another_struct::b = 1;                      // defines another_struct::b
    enum { right, left };                           // defines right and left 
    namespace FirstNamespace { int a; }             // defines FirstNamespace  and FirstNamespace::a
    namespace NextNamespace = FirstNamespace ;      // defines NextNamespace 
    another_struct MySruct;                         // defines MySruct
    extern int b;                                   // declares b
    extern const int c;                             // declares c
    int another_function( int );                    // declares another_function
    struct aStruct;                                 // declares aStruct
    typedef int MyInt;                              // declares MyInt
    extern another_struct yet_another_struct;       // declares yet_another_struct
    using NextNamespace::a;                         // declares NextNamespace::a

.cpp

An implementation file includes the specific details, that is the definitions, for what is done by the program. While the header file for the light declared what a light could do, the light's .cpp file defines how the light acts.

We will go into much more detail on class definition later; here is a preview:

#include "light.h"
 
Light::Light () : on(false) {
}
 
void Light::toggle() {
  on = (!on);
}
 
bool Light::isOn() const {
  return on;
}

.h

Header files contain mostly declarations, to be used in the rest of the program. The skeleton of a class is usually provided in a header file, while an accompanying implementation file provides the definitions to put the meat on the bones of it. Header files are not compiled, but rather provided to other parts of the program through the use of #include.

A typical header file looks like the following:

// Inside sample.h
#ifndef SAMPLE_H
#define SAMPLE_H
 
// Contents of the header file are placed here.
 
#endif /* SAMPLE_H */

Since header files are included in other files, problems can occur if they are included more than once. This often results in the use of "header guards" using preprocessors (#ifndef, #define, and #endif). #ifndef checks to see if SAMPLE_H has appeared already, if it has not, the header becomes included and SAMPLE_H is defined. If SAMPLE_H was originally defined, then the file has already been included, and is not included again.

Classes are usually declared inside header files. We will go into much more detail on class declaration later; here is a preview:

// Inside light.h
#ifndef LIGHT_H
#define LIGHT_H
 
// A light which may be on or off.
class Light {
  private:
    bool on;
 
  public:
    Light ();       // Makes a new light.
    void toggle (); // If light is on, turn it off, if off, turn it on
    bool isOn();    // Is the light on?
};
 
#endif /* LIGHT_H - comment indicating which if this goes with */

This header file "light.h" declares that there is going to be a light class, and gives the properties of the light, and the methods provided by it. Other programmers can now include this file by typing #include "light.h" in their implementation files, which allows them to use this new class. Note how these programmers do not include the actual .cpp file that goes with this class that contains the details of how the light actually works. We'll return to this case study after we discuss implementation files.

Object Files

An object file is a temporary file used by the compiler as an intermediate step between the source code and the final executable file.

All other source files that are not or resulted from source code, the support data needed for the build (creation) of the program. The extensions of this files may vary from system to system, since they depend on the IDE/Compiler and necessities of the program, they may include graphic files, or raw data formats.

Object Code

The compiler produces machine code equivalent (object code) of the source code, contain the binary language (machine language) instruction to be used by the computer to do as was instructed in the source code, that can then be linked into the final program. This step ensures that the code is valid and will sequence into an executable program. Most object files have the file extension (.o) with the same restrictions explained above for the (.cpp/.h) files.

Libraries

Libraries are commonly distributed in binary form, using the (.lib) extension and header (.h) that provided the interface for its utilization. Libraries can also be dynamically linked and in that case the extension may depend on the target OS, for instance windows libraries as a rule have the (.dll) extension, this will be covered later on in the book in the libraries section of this book.

Makefiles

It is common for source code to come with a a specific script file named "Makefile" (without a standard extension or a standard interpreter). This type of script files is not covered by the C++ Standard, even though it is in common use.

In some projects, especially if dealing with a high level of external dependencies or specific configurations, like supporting special hardware, there is need to automate a vast number of incompatible compile sequences. This scripts are intended to alleviate the task. Explaining in detail the myriad of variations and of possible choices a programmer may make in using (or not) such a system goes beyond the scope of this book. You should check the documentation of the IDE, make tool or the information available on the source you are attempting to compile.

TODO If someone wants to tackle this problem please change the text and point it to the relevant section, it was on the TODO list, do attempt to cover at least two distinct ones or the most used...

• The Apache Ant wikibook describes how to write and use a "build.xml", one way to automate the build process.
• The "make" wikibook describes how to write and use a "Makefile", another way to automate the build process.
• ... many IDEs have a "build" button ...

Statements

Most programming languages share the concept of a statement. A statement is a command the programmer gives to the computer. It is also referred to as an expression.

Example

cout << "Hi there!"; // a single statement

Each valid C++ statement is terminated by a semicolon (;). Each statement performs an action. That statement and command will be examined in detail later on, for now consider that it has a verb ("cout") and the other details as information (what to print). In this case, the command "cout" means "send to the standard output stream," (in this case we assume the default, the console).

The programmer either enters the statement directly to the computer (by typing it while running a special program), or creates a text file with the command in it (you can use any text editor for that). You could create a file called "hi.txt", put the above command in it, and give the file to the computer.

If one were to write multiple statements, it is recommended that each statement be entered on a separate line and should end with a semicolon (;).

cout << "Hi there!";                   // a statement
cout << "Strange things are afoot..."; // another statement

However, there is no problem writing the code this way:

cout << "Hi there!"; cout << "Strange things are afoot...";

The former code gathers appeal in the developer circles. Writing statements as in the second example only makes your code look more complex and incomprehensible. We will speak of this deeply in the Coding Style Conventions Section of the book.

If you have more than one command in the file, each will be performed in order, top to bottom.

The computer will perform each of these commands sequentially. It's invaluable to be able to "play computer" when programming. Ask yourself, "If I were the computer, what would I do with these statements?" If you're not sure what the answer is, then you are very likely to write incorrect code. Stop and check the manual for the programming language you're using.

In the above case, the computer will look at the first statement, determine that it's a cout statement, look at what needs to be printed, and display that text on the computer screen. It'll look like this:

Hi there!

Note that the quotation marks aren't there. Their purpose in the program is to tell the computer where the text begins and ends, just like in English prose. The computer will then continue to the next statement, perform its command, and the screen will look like this:

Hi there! Strange things are afoot...

When the computer gets to the end of the text file, it stops. There are many different kinds of statements, depending on which programming language is being used. For example, there could be a beep statement that causes the computer to output a beep on its speaker, or a window statement that causes a new window to pop up.

Also, the way statements are written will vary depending on the programming language. These differences are fairly superficial. The set of rules like the first two is called a programming language's syntax. The set of verbs is called its library.

cout << "Hi there!";

Statement Blocks

Also referred to Code Blocks (or in C++-speak, a compound statement), consist on one or more statements or commands that are contained between a pair of curly braces { }. Such a block of statement can be named or be provided a condition for execution. Below is how you'd place a series of statements in a block.

{
  int a = 10;
  int b = 20;
  int result = a + b;
}

Blocks are used primarily in loops, conditionals and functions. Blocks can be nested inside one another, for instance as an if structure inside of a loop inside of a function.

Program Control Flow

As seen above the statements are evaluated in the order as they occur (sequentially). The execution of flow begins at the top most statement and proceed downwards till the last statement is encountered. A statement can be substituted by a statement block. There are special statements that can redirect the execution flow based on a condition, those statements are called branching statements, described in detail in the Control Flow Construct Statements Section of the book.

Coding Style Conventions

The use of a guide or set of convention gives programmers a set of rules for code normalization or coding style that establishes how they should format code, name variables, place comments or any other non language dependent structural decision that is used on the code. This is very important, as you share a project with others. Agreeing to a common set of coding standards and recommendations saves time and effort, by enabling a greater understandings and transparency of the code base, providing a common ground for undocumented structures, making for easy debugging, and increasing code maintainability. These rules can also be referred to as Source Code Style, Code Conventions, Coding Standards or a variation of those.

A list of different approaches can be found on the C++ coding conventions Reference Section. The most commonly used style in C++ programming is ANSI or Allman while much C programming is still done in the Kernighan and Ritchie (K&R) style. You should be warned that this should be one of the first decisions you make on a project and in a democratic environment, a consensus can be very hard to achieve.

Programmers tend to stick to a coding style, they have it automated and any deviation can be very hard to conform with, if you don't have a favorite style try to use the smallest possible variation to a common one or get as broad a view as you can get, so that you can adapt easily to changes or defend your approach. There is software that can help to format or beautify the code, but automation can have its drawbacks. As seen earlier, indentation and the use of white spaces or tabs are completely ignored by the compiler. A coding style should vary depending on the lowest common denominator of the needs to standardize.

Field impacted by the selection of a Code Style are:

• Reusability
  • Self documenting code
  • Internationalization
  • Maintainability
  • Portability
• Optimization
• Build process
• Error avoidance
• Security

Standardization is Important

No matter which particular coding style you pick, once it is selected, it should be kept throughout the same project. Reading code that follows different styles can become very difficult. In the next sections we try to explain why some of the options are common practice without forcing you to adopt a specific style.

25 lines 80 columns

This is a commonly recommended but often inapplicable rule. Many people say it's an outdated rule, that it comes from prehistoric times when terminals could only display 25 lines 80 columns.

This rules signifies that if you are writing code that will go further than 80 columns or 25 lines, it's time to think about splitting the code into functions, as it helps allow you to review code without having to scroll the display. This practice will save you precious time when you have to return to a project you haven't been working on for 6 months.

For example, you may want to split long output statements across multiple lines:

    fprintf(stdout,"The quick brown fox jumps over the lazy dog. "
                   "The quick brown fox jumps over the lazy dog.\n"
                   "The quick brown fox jumps over the lazy dog - %d", 2);

   This recommended practice relates also to the 0 means success convention for functions, that we will cover on the Functions Section of this book.

Whitespace and Indentation

Conventions followed when using whitespace to improve the readability of code is called an indentation style. Every block of code and every definition should follow a consistent indention style. This usually means everything within { and }. However, the same thing goes for one-line code blocks.

Use a fixed number of spaces for indentation. Recommendations vary; 2, 3, 4, 8 are all common numbers. If you use tabs for indention you have to be aware that editors and printers may deal with, and expand, tabs differently. The K&R standard recommends an indentation size of 4 spaces. ^[1]

For example, a program could as well be written using as follows:

// Using an indentation size of 2
if ( a > 5 )  { b=a; a++; }

However, the same code could be made much more readable with proper indentation:

// Using an indentation size of 2
if ( a > 5 )  {
  b = a;
  a++;
}
 
// Using an indentation size of 4
if ( a > 5 )
{
    b = a;
    a++;
}

Placement of braces (curly brackets)

As we have seen early on the Statements Section, compound statement are very important in C++, they also are subject of different coding styles, that recommend different placements of opening and closing braces ({ and }). Some recommend putting the opening brace on the line with the statement, at the end (K&R). Others recommend putting these on a line by itself, but not indented (ANSI C++). GNU recommends putting braces on a line by itself, and indenting them half-way. We recommend picking one brace-placement style and sticking with it.

Examples:

if (a > 5) {
  // This is K&R style
}
 
if (a > 5) 
{
  // This is ANSI C++ style
}
 
if (a > 5) 
  {
    // This is GNU style
  }

Comments

Comments are portions of the code ignored by the compiler which allow the user to make simple notes in the relevant areas of the source code. Comments come either in block form or as single lines.

• Single-line comments (informally, C++ style), start with // and continue until the end of the line. If the last character in a comment line is a \ the comment will continue in the next line.
• Multi-line comments (informally, C style), start with /* and end with */.

We will now describe how a comment can be added to the source code, but not where, how, and when to comment; we will get into that later.

C style Comments

If you use this kind of comment try to use it like this... Commented

/*void EventLoop(); /**/

or for multiple lines

/*
void EventLoop();
void EventLoop();
/**/

this opens you the option to do this... Uncommented

void EventLoop(); /**/

or for multiple lines

void EventLoop();
void EventLoop();
/**/

... by removing only the start of comment and so activating the next one, you did re-activate the commented code, because if you start a comment this way it will be valid until it finds the close of comment */.

int function() /* This is a comment /*
{              
 return 0;  
}              and this is the same comment */
               so this isn't in the comment, and will give an error*/

C++ style Comments

Examples:

// This is a single one line comment

or

if (expression) // This needs a comment
{
  statements;   
}
else
{
  statements;
}

The backslash is a continuation character and will continue the comment to the following line:

// This comment will also comment the following line \
std::cout << "This line will not print" << std::endl;

Using comments to temporarily ignore code

Comments are also sometimes used to enclose code that we temporarily want the compiler to ignore. This can be useful in finding errors in the program. If a program does not give the desired result, it might be possible to track which particular statement contains the error by commenting out code.

Example with C style comments

/* This is a single line comment */

or

/*
   This is a multiple line comment
*/

C and C++ style

Combining multi-line comments (/* */) with c++ comments (//) to comment out multiple lines of code:

Commenting out the code:

/*
void EventLoop();
void EventLoop();
void EventLoop();
void EventLoop();
void EventLoop();
//*/

uncommenting the code chunk

//*
void EventLoop();
void EventLoop();
void EventLoop();
void EventLoop();
void EventLoop();
//*/

This works because a //* is still a c++ comment. And //*/ acts as a c++ comment and a multi-line comment terminator. However this doesn't work if there are any multi-line comments are used for function descriptions.

Note on doing it with preprocessor statements

Another way (considered bad practice) is to selectively enable disable sections of code:

#if(0)   // Change this to 1 to uncomments.
void EventLoop();
#endif

this is considered a bad practice because the code often becomes illegible when several #if's are mixed, if you use them don't forget to add a comment at the #endif saying what #if it correspond

#if (FEATURE_1 == 1)
do_something;
#endif //FEATURE_1 == 1

you can prevent illegibility by using inline functions (often considered better than macros for legibility with no performance cost) containing only 2 sections in #if #else #endif

inline do_test()
  {
    #if (Feature_1 == 1)
      do_something
    #endif  //FEATURE_1 == 1
  }

and call

do_test();

in the program

Naming identifiers

Identifiers are names given to variables, functions, objects, etc. to refer to them in the program. C++ identifiers must start with a letter or an underscore character "_", possibly followed by a series of letters, underscores or digits. None of the C++ keywords can be used as identifiers. Identifiers with successive underscores are reserved for use in the header files or by the compiler for special purpose, e.g. name mangling.

This leaves a lot of freedom in naming, one could use specific prefixes or suffixes, start names with an initial upper or lower case letter, keep all the letters in a single case or, with compound words, use a word separator character like "_" or flip the case of the first letter of each component word.

Hungarian Notation

Hungarian notation, which would now be called Apps Hungarian, was invented by Charles Simonyi, a programmer who worked at Xerox PARC circa 1972-1981, and who later became Chief Architect at Microsoft and has been until recently the preeminent naming convention used on most Microsoft code. It uses prefixes, like "m_" to indicates it is a member variable, the "p" indicates that a pointer and the rest of the name is normally written out with caps on the first letter. We mention this convention because you will very probably find it in use, even more probable if you do any programming in Windows, if you are interested on learning more you can check Wikipedia's entry on this notation.

This notation is considered outdated, since it is highly prone to errors and requires some effort to maintain without no real gains in today's IDEs. Today refactoring is now be everyday task, the IDEs have evolved to provide help with identifier popups and the use of color schemes. All this informational aids removes the need to use this notation.

Leading underscores

In most contexts, leading underscores are better avoided. They are reserved for the compiler or internal variables of a library, and can make your code less portable and more difficult to maintain. Those variables can also be stripped from a library (i.e. the variable isn't accessible anymore, it is hidden from external world) so unless you want to override an internal variable of a library, don't do it.

Reusing existing names

Do not use the names of standard library functions and objects for your identifiers as these names are considered reserved words and programs may become difficult to understand when used in unexpected ways.

Sensible names

Always use good, unabbreviated, correctly-spelled meaningful names.

Prefer the English language (since C++ and most libraries already use English) and avoid short cryptic names. This will make it easier to read and to type a name without having to look it up.

Names indicate purpose

An identifier should indicate the function of the variable/function/etc. that it represents, e.g. foobar is probably not a good name for a variable storing the age of a person.

Identifier names should also be descriptive. n might not be a good name for a global variable representing the number of employees. However, a good medium between long names and lots of typing has to be found. Therefore, this rule can be relaxed for variables that are used in a small scope or context. Many programmers prefer short variables (such as i) as loop iterators.

Capitalization

Conventionally, variable names start with a lower case character. In identifiers which contain more than one natural language words, either underscores or capitalization is used to delimit the words, e.g. num_chars (K&R style) or numChars (Java style). It is recommended that you pick one notation and do not mix them within one project.

Constants

When naming #defines, constant variables, enum constants. and macros put in all uppercase using '_' separators; this makes it very clear that the value is not alterable and in the case of macros, makes it clear that you are using a construct that requires care.

Functions and Member Functions

The name given to functions and member functions should be descriptive and make it clear what it does. Since usually functions and member functions perform actions, the best name choices typically contain a mix of verbs and nouns in them such as CheckForErrors() instead of ErrorCheck() and dump_data_to_file() instead of data_file(). Clear and descriptive names for functions and member functions can sometimes make guessing correctly what functions and member functions do easier, aiding in making code more self documenting. By following this and other naming conventions programs can be read more naturally.

People seem to have very different intuitions when using names containing abbreviations. It's best to settle on one strategy so the names are absolutely predictable. Take for example NetworkABCKey. Notice how the C from ABC and K from key are confused. Some people don't mind this and others just hate it so you'll find different policies in different code so you never know what to call something.

Prefixes and suffixes are sometimes useful:

• Min - to mean the minimum value something can have.
• Max - to mean the maximum value something can have.
• Cnt - the current count of something.
• Count - the current count of something.
• Num - the current number of something.
• Key - key value.
• Hash - hash value.
• Size - the current size of something.
• Len - the current length of something.
• Pos - the current position of something.
• Limit - the current limit of something.
• Is - asking if something is true.
• Not - asking if something is not true.
• Has - asking if something has a specific value, attribute or property.
• Can - asking if something can be done.
• Get - get a value.
• Set - set a value.

Examples

In most contexts, leading underscores are also better avoided. For example, these are valid identifiers:

• i loop value
• numberOfCharacters number of characters
• number_of_chars number of characters
• num_chars number of characters
• get_number_of_characters() get the number of characters
• get_number_of_chars() get the number of characters
• is_character_limit() is this the character limit?
• is_char_limit() is this the character limit?
• character_max() maximum number of a character
• charMax() maximum number of a character
• CharMin() minimum number of a character

These are also valid identifiers but can you tell what they mean?:

• num1
• do_this()
• g()
• hxq

The following are valid identifiers but better avoided:

• _num as it could be used by the compiler/system headers
• num__chars as it could be used by the compiler/system headers
• main as there is potential for confusion
• cout as there is potential for confusion

The following are not valid identifiers:

• if as it is a keyword
• 4nums as it starts with a digit
• number of characters as spaces are not allowed within an identifier

Reduced/Abuse the use of keywords

This can be defended both ways. This can mean less typing but also make the reader and the compiler (depending on the situation) to do extra work, on the other hand if you write more keywords the resulting code will be clearer and reduces errors, or more defined (self documented) but this can lead to adding limitations to the code's evolution. This is a thin line were an equilibrium must be reached in accord to the projects nature. The important fact is to be consistent as with any other rule.

inline

Use inline if the member function is implicitly inlined.

const

Unless you plan on modifying it, you're arguably better off using const data types. The compiler can easily optimize more with this restriction, and you're unlikely to accidentally corrupt the data. Ensure that your methods take const data types unless you absolutely have to modify the parameters. Similarly, when implementing accessors for private member data, you should in most cases return a const. This will ensure that if the object that you're operating on is passed as const, methods that do not affect the data stored in the object still work as they should and can be called. For example, for an object containing a person, a getName() should return a const data type where as walk() might be non-const as it might change some internal data in the Person such as tiredness.

typedef

It is common practice to avoid using this keyword since it can obfuscate code if not properly used or it can cause programmers to accidentally misuse large structures thinking them to be simple types. If used, define a set of rules for the types you rename and be sure to document them.

volatile

This keyword informs the compiler that the variable it is qualifying as volatile (can change at anytime) is excluded from any optimization techniques. Usage of this variable should be reserved for variables that are known to be modified due to an external influence of a program (whether it's hardware update, third party application, or another thread in the application).

Since the volatile keyword impacts performance, you should consider a different design that avoids this situation: most platforms where this keyword is necessary provide an alternative that helps maintain scalable performance.

Note that using volatile was not intended to be used as a threading or synchronization primitive, nor are operations on a volatile variable guaranteed to be atomic.

Pointer declaration

Due to historical reasons some programmers refer to a specific use as:

// C codestyle
int *z;
 
// C++ codestyle
int* z;

Since the second variation is by far preferred by C++ programmers and will help identify a C programmer or legacy code.

One argument against the C++ codestyle version is when chaining declarations of more than one item, like:

// C codestyle
int *ptrA, *ptrB;
 
// C++ codestyle
int* ptrC, ptrD;

As you can see, in n this case, the C codestyle makes it more obvious that ptrA and ptrB are pointers to int, and the C++ codestyle makes it less obvious that ptrD is an int, not a pointer to int.

It is rare to use chains of multiple objects in C++ code with the exception of the basic types and even so it is a not often used and it is extremely rare to see it used in pointers or other complex types, since it will make it harder to for a human to visually parse the code.

In any case most programmers would use a naming convention for pointer variables that would address this downside, in the example above we would see something like:

// C++ codestyle
int* ptrC, D;

Making it easily detectable.

Document your code

There are a number of good reasons to document your code, and a number of aspects of it that can be documented. Documentation provides you with a shortcut for obtaining an overview of the system or for understanding the code that provides a particular feature.

Why?

The purpose of comments is to explain and clarify the source code to anyone examining it (or just as a reminder to yourself). Good commenting conventions are essential to any non-trivial program so that a person reading the code can understand what it is expected to do and to make it easy to follow on the rest of the code. In the next topics some of the most How? and When? rules to use comments will be listed for you.

Documentation of programming is essential when programming not just just in C++, but in any programming language. Many companies have moved away from the idea of "hero programmers" (i.e., one programmer who codes for the entire company) to a concept of groups of programmers working in a team. Many times programmers will only be working on small parts of a larger project. In this particular case, documentation is essential because:

• Other programmers may be tasked to develop your project;
• Your finished project may be submitted to editors to assemble your code into other projects;
• A person other than you may be required to read, understand, and present your code.

Even if you are not programming for a living or for a company, documentation of your code is still essential. Though many programs can be completed in a few hours, more complex programs can take longer time to complete (days, weeks, etc.). In this case, documentation is essential because:

• You may not be able to work on your project in one session;
• It provides a reference to what was changed the last time you programmed;
• It allows you to record why you made the decisions you did, including why you chose not to explore certain solutions;
• It can provide a place to document known limitations and bugs (for the latter a defect tracking system may be the appropriate place for documentation);
• It allows easy searching and referencing within the program (from a non-technical stance);
• It is considered to be good programming practice.

Comments Should Be Written For the Appropriate Audience

When writing code to be read by those who are in the initial stages of learning a new programming language, it can be helpful to include a lot of comments about what the code does. For "production" code, written to be read by professionals, it is considered unhelpful and counterproductive to include comments which say things that are already clear in the code. Some from the Extreme Programming community say that excessive commenting is indicative of code smell -- which is not to say that comments are bad, but that they are often a clue that code would benefit from refactoring. Adding comments as an alternative to writing understandable code is considered poor practice.

What?

What needs to be documented in a program/source code can be divided into what is documented before the specific program execution (that is before "main") and what is executed ("what is in main").

Documentation before program execution:

• Programmer information and license information (if applicable)
• User defined function declarations
• Interfaces
• Context
• Relevant standards/specifications
• Algorithm steps
• How to convert the source code into executable file(s) (perhaps by using make)

Documentation for code inside main:

• Statements, Loops, and Cases
• Public and Private Sectors within Classes
• Algorithms used
• Unusual features of the implementation
• Reasons why other choices have been avoided
• User defined function implementation

If used carelessly comments can make source code hard to read and maintain and may be even unnecessary if the code is self-explanatory -- but remember that what seems self-explanatory today may not seem the same six months or six years from now.

Document Decisions

Comments should document decisions. At every point where you had a choice of what to do place a comment describing which choice you made and why. Archaeologists will find this the most useful information.

Comment Layout

Each part of the project should at least have a single comment layout, and it would be better yet to have the complete project share the same layout if possible.

TODO Add more here.

How?

Documentation can be done within the source code itself through the use of comments (as seen above) in a language understandable to the intended audience. It is good practice to do it in English as the C++ language is itself English based and English being the current lingua franca of international business, science, technology and aviation, you will ensure support for the broadest audience possible.

Comments are useful in documenting portions of an algorithm to be executed, explaining function calls and variable names, or providing reasons as to why a specific choice or method was used. Block comments are used as follows:

/*
get timepunch algorithm - this algorithm gets a time punch for use later
1. user enters their number and selects "in" or "out"
2. time is retrieved from the computer
3. time punch is assigned to user
*/

Alternately, line comments can be used as follows:

GetPunch(user_id, time, punch); //this function gets the time punch

An example of a full program using comments as documentation is:

/*
Chris Seedyk
BORD Technologies
29 December 2006
Test
*/
int main()
{
 cout << "Hello world!" << endl; //predefined cout prints stuff in " " to screen
 return 0;
}

It should be noted that while comments are useful for in-program documentation, it is also a good idea to have an external form of documentation separate from the source code as well, but remember to think first on how the source will be distributed before making references to external information on the code comments.

Commenting code is also no substitute for well-planned and meaningful variable, function, and class names. This is often called "self-documenting code," as it is easy to see from a carefully chosen and descriptive name what the variable, function, or class is meant to do. To illustrate this point, note the relatively equal simplicity with which the following two ways of documenting code, despite the use of comments in the first and their absence in the second, are understood. The first style is often encountered in very old C source by people who understood well what they were doing and had no doubt anyone else might not comprehend it. The second style is more "human-friendly" and while much easier to read is nevertheless not as frequently encountered.

// Returns the area of a triangle cast as an int
int area_ftoi(float a, float b) { return (int) a * b / 2; }
 
int iTriangleArea(float fBase, float fHeight)
{
   return (int) fBase * fHeight / 2;
}

Both functions perform the same task, however the second has such practical names chosen for the function and the variables that its purpose is clear even without comments. As the complexity of the code increases, well-chosen naming schemes increase vastly in importance.

Regardless of what method is preferred, comments in code are helpful, save time (and headaches), and ensure that both the author and others understand the layout and purpose of the program fully.

Automatic Documentation

Various tools are available to help with documenting C++ code; Literate Programming is a whole school of thought on how to approach this, but a very effective tool is Doxygen (also supports several languages), it can even use hand written comments in order to generate more than the bare structure of the code, bringing Javadoc-like documentation comments to C++ and can generate documentation in HTML, PDF and other formats.

Comments Should Tell a Story

Consider your comments a story describing the system. Expect your comments to be extracted by a robot and formed into a manual page. Class comments are one part of the story, method signature comments are another part of the story, method arguments another part, and method implementation yet another part. All these parts should weave together and inform someone else at another point of time just exactly what you did and why.

Do not use comments for flowcharts or pseudocode

You should refrain from using comments to do ASCII art or pseudocode (some programmers attempt to explain their code with an ASCII-art flowchart). If you want to flowchart or otherwise model your design there are tools that will do a better job at it using standardized methods. See for example: UML.

Internal storage of data types

Bits and Bytes

The byte is the smallest individual piece of data that we can access or modify on a computer. The computer only works on bytes or groups of bytes, never on bits. If you want to modify individual bits, you have to use binary operations on the whole byte that tell the computer how to modify individual bits, but the operation is still done on whole bytes. Before getting too far ahead of ourselves, we'll look at the internal representation of a byte.

Here's a look at a byte as the computer stores it.

There is actually quite a lot of information here. A byte (usually) contains 8 bits. A bit can only have a value of 0 or 1. The bit number is used to label each bit in the byte (so that we can tell which bit we are talking about). You may be wondering why the bits are labeled from 7 to 0 instead of 0 to 7 or even 1 to 8. The reason 0 is used is because computers always start counting at 0. Technically, we COULD start counting at 1, but this would go against the counting nature of the computer. It is simply more convenient to use 0 for computers as we shall see. Now as to why we numbered them in descending order. In decimal numbers (normal base 10), we put the more significant digits to the left. Example: 254. The 2 here is more significant than the other digits because it represents hundreds as opposed to tens for the 5 or singles for the 4. The same is done in binary. The more significant digits are put towards the left. Counting in binary and in decimal is done in exactly the same manner, except that in binary, instead of counting from 0 to 9, we only count from 0 to 1. If we want to count higher than 1, then we need a more significant digit to the left. In decimal, when we count beyond 9, we need to add a 1 to the next significant digit. It sometimes may look confusing or different only because we as humans are used to counting with 10 digits. In binary, there are only 2 digits, but counting is done by the exact same principles as counting in decimal.

In decimal, each digit represents multiple of a power of 10. Let's take another look at the decimal number 254.

• The 4 represents four multiples of one ( since 10⁰ = 1).
• Since we're working in decimal (base 10), the 5 represents five multiples of 10 ()
• Finally the 2 represents two multiples of 100 ()

All this is elementary. The key point to recognize is that as we move from right to left in the number, the significance of the digits increases by a multiple of 10. This should be obvious when we look at the following equation:

Do you see any similarities between this and the diagram above? In binary, each digit can only be one of two possibilities (0 or 1), therefore when we work with binary we work in base 2 instead of base 10. So, to convert the binary number 1101 to decimal we can use the following base 10 equation, which you should find very much like the one above:

So, to convert the number we simply add the bit values (2ⁿ) where a 1 shows up. Let's take a look at our example byte again, and try to find its value in decimal.

First off, we see that bit #5 is a 1, so we have 2⁵ = 32 in our total. Next we have bit#3, so we add 2³ = 8. This gives us 40. Then next is bit#2, so 40 + 4 is 44. And finally is bit#0 to give 44 + 1 = 45. So this binary number is 45 in decimal.

As can be seen, it is impossible for different bit combinations to give the same decimal value. Here is a quick example to show the relationship between counting in binary (base 2) and counting in decimal (base 10). The bases that these numbers are in are shown in subscript to the right of the number.

00₂ = 0₁₀

01₂ = 1₁₀

10₂ = 2₁₀

11₂ = 3₁₀

Data and Variables

When programming in C++, we need a way to store data that can be manipulated by our program. Data comes in a variety of formats, so the compiler needs a way to differentiate between the different types. Right now, we'll concentrate on using bytes. The type name for a byte in C++ is 'char'. It's called char because a byte is often used to represent characters. We won't go into that right now. We only want to use it for its numerical representation.

Let's write a program that will print each value that a byte can hold. How do we do that? We could write a loop that goes from 0 to 255. We'll set our byte to 0 and add one to it every time through the loop. As a side note, do you know what would happen if you added 1 to 255? No combination will represent 256 unless we add more bits. If you look at the diagram above, you will see that the next value (if we could have another digit) would be 256. So our byte would look like this.

But this 9^th bit (bit#8) doesn't exist. So where does it go? It actually goes into the carry bit. The carry bit, you say? The processor of the computer has an internal bit used exclusively for carry operations such as this. So if you add 1 to 255 stored in a byte, you'd get 0 with the carry bit set in the CPU. Of course, being a C++ programmer, you never get to use this bit directly. You'll need to learn assembler if you want to do that, but that's a whole other ball game.

In our program, we can start off with a value of 0 and wait until it becomes 0 again before exiting. This will make sure we go through every value a byte can hold.

Inside your main() function, write the following. Don't worry about the loop just yet. We are more concerned with the output right now.

char b=0;
 
do
{
  cout << (int)b << " ";
  b++;
  if ((b&15)==0) cout << endl;
} while(b!=0);

b is our byte and we initialize it to 0. Inside the loop we print its value. We cast it to an int so that a number is printed instead of a character. Don't worry about casting or int's right now. The next line increments the value in our byte b. Then we print a new line (carriage return/endl) after every 16 numbers by using the code "b&15" which uses '&', the bitwise AND operator, and the value 15 (00001111 in binary), to give the value of the lowest 4 bits, which will be 0 every multiple of 16, because 16 is 00010000 in binary, so multiplying this by something will just change the higher bits, and will leave the lower bits alone. We do this so that we can see all 256 values on the screen at once by printing 16 values per line, and so only using 16 lines, assuming 16 values will fit on one line.

If you were to run this program, you would notice something strange. After 127, we got -128! Negative numbers! Where did these come from? Well, it just so happens that the compiler needs to be told if we're using numbers that can be negative or number that can only be positive. These are called signed and unsigned numbers respectively. By default, the compiler assumes we want to use signed numbers unless we explicitly tell it otherwise. To fix our little problem, add "unsigned" in front of the declaration for b so that it reads:

unsigned char b=0;

Problem solved!

Two's Complement

Two's complement is a way to store negative numbers in a pure binary representation. The reason that the two's complement method of storing negative numbers was chosen is because this allows the CPU to use the same add and subtract instructions on both signed and unsigned numbers.

To convert a positive number into it's negative two's complement format, you begin by flipping all the bits in the number (1's become 0's and 0's become 1's) and then add 1. (This also works to turn a negative number back into a positive number Ex: -34 into 34 or vice-versa).

Let's try to convert our number 45.

First, we flip all the bits...

And add 1.

Now if we add up the values for all the one bits, we get... 128+64+16+2+1=211? What happened here? Well, this number actually is 211. It all depends on how you interpret it. If you decide this number is unsigned, then it's value is 211. But if you decide it's signed, then it's value is -45. It is completely up to you how you treat the number.

If and only if you decide to treat it as a signed number, then look at the msb (most significant bit [bit#7]). If it's a 1, then it's a negative number. If it's a 0, then it's positive. In C++, using "unsigned" in front of a type will tell the compiler you want to use this variable as an unsigned number, otherwise it will be treated as signed number.

Now, if you see the msb is set, then you know it's negative. So convert it back to a positive number to find out it's real value using the process just described above.

Let's go through a few examples.

Treat the following number as an unsigned byte. What is it's value in decimal?

Since this is an unsigned number, no special handling is needed. Just add up all the values where there's a 1 bit. 128+64+32+4=228. So this binary number is 228 in decimal.

Now treat the number above as a signed byte. What is its value in decimal?

Since this is now a signed number, we first have to check if the msb is set. Let's look. Yup, bit #7 is set. So we have to do a two's complement conversion to get its value as a positive number (then we'll add the negative sign afterwards).

Ok, so let's flip all the bits...

And add 1. This is a little trickier since a carry propagates to the third bit. For bit#0, we do 1+1 = 10 in binary. So we have a 0 in bit#0. Now we have to add the carry to the second bit (bit#1). 1+1=10. bit#1 is 0 and again we carry a 1 over to the 3^rd bit (bit#2). 0+1 = 1 and we're done the conversion.

Now we add the values where there's a one bit. 16+8+4 = 28. Since we did a conversion, we add the negative sign to give a value of -28. So if we treat 11100100 (base 2) as a signed number, it has a value of -28. If we treat it as an unsigned number, it has a value of 228.

Let's try one last example.

Give the decimal value of the following binary number both as a signed and unsigned number.

First as an unsigned number. So we add the values where there's a 1 bit set. 4+1 = 5. For an unsigned number, it has a value of 5.

Now for a signed number. We check if the msb is set. Nope, bit #7 is 0. So for a signed number, it also has a value of 5.

As you can see, if a signed number doesn't have its msb set, then you treat it exactly like an unsigned number.

Endian

Now that we've seen many ways to use a byte, it is time to look at ways to represent numbers larger than 255. By grouping bytes together, we can represent numbers that are much larger than 255. If we use 2 bytes together, we double the number of bits in our number. In effect, 16 bits allows us to represent numbers up to 65535 (unsigned). And 32 bits allows us to represent numbers above 4 billion. We already saw the type for a byte. It is called a 'char'.

Here are a few basic primitive types:

1. char (1 byte (by definition), max unsigned value: at least 255)

2. short int (at least 16 bits, max unsigned value: at least 65535)

3. long int (at least 32 bits, max unsigned value: at least 4294967295)

4. float (typically 4 bytes, floating point)

5. double (typically 8 bytes, floating point)

For 'short int' and 'long int', you can leave out the 'int' because the compiler will know what type you want. You can also use 'int' by itself and it will default to whatever your compiler is set at for an int. On most recent compilers, int defaults to a 32-bit type.

All of the topics explained above also apply to short int's and long int's. The difference is simply the number of bits used is different and the msb is now bit#15 for a short and bit#31 for a long (assuming a 32-bit long type).

Let's look at a (16-bit) short. You may think that in memory the byte for bits 15 to 8 would be followed by the byte for bits 7 to 0 (because bits 15 to 8 appears first). In other words, byte #0 would be the high byte and byte #1 would be the low byte. This is true for some other systems. For example, the Motorola 68000 series CPUs do work this way. The Amiga and old Macintoshes use the M68000 and they indeed do use this byte ordering.

However, on PCs (with 8088/286/386/486/Pentiums) this is not so. The ordering is reversed so that the low byte comes before the high byte. The byte that represents bits 0 to 7 always comes before all other bytes on PCs. This is called little-endian ordering. The other ordering, such as on the M68000, is called big-endian ordering. This is very important to remember when doing low level byte operations.

For big-endian computers, the basic idea is to keep the higher bits on the left or in front. For little-endian computers, the idea is to keep the low bits in the low byte. There is no inherent advantage to either scheme except perhaps for an oddity. Using a little-endian long int as a smaller type of int is theoretically possible as the low byte(s) is/are always in the same location (first byte). With big-endian the low byte is always located differently depending on the size of the type. For example (in big-endian), the low byte is the 4^th byte in a long int and the 2^nd byte in a short int. So a proper cast must be done and low level tricks become rather dangerous.

To convert from one endianness to the other, you reverse the values of the bytes, putting the highest bytes value in the lowest byte and the lowest bytes value in the highest byte, and swap all the values for the in between bytes, so that if you had a 4 byte little-endian integer 0x0A0B0C0D (the 0x signifies that the value is hexadecimal) then converting it to big-endian would change it to 0x0D0C0B0A.

Bit endianness, where the bit order inside the bytes changes, is rarely used in data storage and only really ever matters in serial communication links, where the hardware deals with it.

Floating point representation

A generic real number with a decimal part can also be expressed in binary format. For instance 110.01 in binary corresponds to:

Exponential notation (also known as scientific notation, or standard form, when used with base 10, as in ) can be also used and the same number expressed as:

When there is only one non-zero digit on the left of the decimal point, the notation is termed normalized.

In computing applications a real number is represented by a sign bit (S) an exponent (e) and a mantissa (M). The exponent field needs to represent both positive and negative exponents. To do this, a bias E is added to the actual exponent in order to get the stored exponent, and the sign bit (S), which indicates whether or not the number is negative, is transformed into either +1 or -1, giving s. A real number is thus represented as:

S, e and M are concatenated one after the other in a 32-bit word to create a single precision floating point number and in a 64-bit doubleword to create a double precision one. For the single float type, 8 bits are used for the exponent and 23 bits for the mantissa, and the exponent offset is E=127. For the double type 11 bits are used for the exponent and 52 for the mantissa, and the exponent offset is E=1023.

There are two types of floating point numbers. Normalized and denormalized. A normalized number will have an exponent e in the range 0<e<2⁸ - 1 (between 00000000 and 11111111, non-inclusive) in a single precision float, and an exponent e in the range 0<e<2¹¹ - 1 (between 00000000000 and 11111111111, non-inclusive) for a double float. Normalized numbers are represented as sign times 1.Mantissa times 2^e-E. Denormalized numbers are numbers where the exponent is 0. They are represented as sign times 0.Mantissa times 2^1-E. Denormalized numbers are used to store the value 0, where the exponent and mantissa are both 0. Floating point numbers can store both +0 and -0, depending on the sign. When the number isn't normalized or denormalized (it's exponent is all 1s) the number will be plus or minus infinity if the mantissa is zero and depending on the sign, or plus or minus NaN (Not a Number) if the mantissa isn't zero and depending on the sign.

For instance the binary representation of the number 5.0 (using float type) is:

0 10000001 01000000000000000000000

The first bit is 0, meaning the number is positive, the exponent is 129-127=2, and the mantissa is 1.01 (note the leading one is not included in the binary representation). 1.01 corresponds to 1.25 in decimal representation. Hence 1.25*4=5.

Floating point numbers are not always exact representations of values. a number like 1010110110001110101001101 couldn't be represented by a single precision floating point number because, disregarding the leading 1 which isn't part of the mantissa, there are 24 bits, and a single precision float can only store 23 numbers in it's mantissa, so the 1 at the end would have to be dropped because it is the least significant bit. Also, there are some value which simply cannot be represented in binary which can be easily represented in decimal, E.g. 0.3 in decimal would be 0.0010011001100110011... or something. A lot of other numbers cannot be exactly represented by a binary floating point number, no matter how many bits it use for it's mantissa, just because it would create a repeating pattern like this.

TODO Add a few comments on different standards... Add some images showing the bit representations like w:IEEE_754 has

Variables

Much like a person has a name that distinguishes him or her from other people, a variable assigns a particular instance of an object type, a name or label by which the instance can be referred to. Depending on its use in the code a variable has also a specific locality in relation to the hardware and based on the structure of the code it also has specific scope where the compiler will recognize it as valid. All these characteristics are defined by a programmer.

Locality (hardware)

Variables have two distinct characteristics: those that are created on the stack (local variables), and those that are accessed via a hard-coded memory address (global variables).

Global variables

Typically a variable is bound to a particular address in computer memory that is automatically assigned to at runtime, with a fixed number of bytes determined by the size of the object type of a variable and any operations performed on the variable effects one or more values stored in that particular memory location.

The only scope that can be defined for a global variable is a namespace, this deals with the visibility of variable not its validity.

All global defined variables will have static lifetime. Only those not defined as const will permit external linkage by default.

Local variables

If the size and location of a variable is unknown beforehand, the location in memory of that variable is stored in another variable instead, and the size of the original variable is determined by the size of the type of the second value storing the memory location of the first. This is called referencing, and the variable holding the other variables memory location is called a pointer.

Scope (code)

Variables also reside in a specific scope. The scope of a variable determines the life-time of a variable. Entrance into a scope begins the life of a variable and leaving scope ends the life of a variable. This becomes important later as the constructors of variables are called when entering scope and the destructors of variables are called when leaving scope. A variable is visible when in scope unless it is hidden by a variable with the same name inside an enclosed scope. A variable can be in global scope, namespace scope, file scope or block scope.

Definition vs. Declaration

There is an important concept, the distinction between the declaration of a variable and its definition. The declaration announces the properties (the type, size, etc.), on the other hand the definition causes storage to be allocated in accordance to the declaration.

Types

Just as there are different types of values (integer, character, etc.), there are different types of variables. A variable can refer to simple values like integers and strings called a primitive type or to a set of values called a composite type that are made up of primitive types and other composite types. Types consist of a set of valid values and a set of valid operations which can be performed on these values. A variable must declare what type it is before it can be used in order to enforce value and operation safety and to know how much space is needed to store a value.

Major functions that type systems provide are:

• Safety - types make it impossible to code some operations which cannot be valid in a certain context. This mechanism effectively catches the majority of common mistakes made by programmers. For example, an expression "Hello, Wikipedia"/1 is invalid because a string literal cannot be divided by an integer in the usual sense. As discussed below, strong typing offers more safety, but it does not necessarily guarantee complete safety (see type-safety for more information).
• Optimization - static type checking might provide useful information to a compiler. For example, if a type says a value is aligned at a multiple of 4, the memory access can be optimized.
• Documentation - using types in languages also improves documentation of code. For example, the declaration of a variable as being of a specific type documents how the variable is used. In fact, many languages allow programmers to define semantic types derived from primitive types; either composed of elements of one or more primitive types, or simply as aliases for names of primitive types.
• Abstraction - types allow programmers to think about programs in higher level, not bothering with low-level implementation. For example, programmers can think of strings as values instead of a mere array of bytes.
• Modularity - types allow programmers to express the interface between two subsystems. This localizes the definitions required for interoperability of the subsystems and prevents inconsistencies when those subsystems communicate.

Table of Data Types

Table of Data Types Footnotes

standard types

C++ has five basic primitive types called standard types, specified by particular keywords, that store a single value.

The type of a variable determines what kind of values it can store:

• bool - a boolean value: true; false
• int - Integer: -5; 10; 100
• char - a character in some encoding, often something like ASCII, ISO-8859-1 ("Latin 1") or ISO-8859-15: 'a', '=', 'G', '2'.
• float - floating-point number: 1.25; -2.35*10^23
• double - double-precision floating-point number: like float but more decimals

The float and double primitive data types are called 'floating point' types and are used to represent real numbers (numbers with decimal places, like 1.435324 and 853.562). Floating point numbers and floating point arithmetic can be very tricky, due to the nature of how a computer calculates floating point numbers.

Declaration

C++ is a statically typed language. Hence, any variable cannot be used without specifying its type. This is why the type figures in the declaration. This way the compiler can protect you from trying to store a value of an incompatible type into a variable, e.g. storing a string in an integer variable. Declaring variables before use also allows spelling errors to be easily detected. Consider a variable used in many statements, but misspelled in one of them. Without declarations, the compiler would silently assume that the misspelled variable actually refers to some other variable. With declarations, an "Undeclared Variable" error would be flagged. Another reason for specifying the type of the variable is so the compiler knows how much space in memory must be allocated for this variable.

The simplest variable declarations look like this (the parts in []s are optional):

[specifier(s)] type variable_name [ = initial_value];

To create an integer variable for example, the syntax is

int sum;

where sum is the name you made up for the variable. This kind of statement is called a declaration. It declares sum as a variable of type int, so that sum can store an integer value. Every variable has to be declared before use and it is common practice to declare variables as close as possible to the moment where they are needed. This is unlike languages, such as C, where all declarations must precede all other statements and expressions.

In general, you will want to make up variable names that indicate what you plan to do with the variable. For example, if you saw these variable declarations:

char firstLetter; 
char lastLetter; 
int hour, minute;

you could probably make a good guess at what values would be stored in them. This example also demonstrates the syntax for declaring multiple variables with the same type in the same statement: hour and minute are both integers (int type). Notice how a comma separates the variable names.

int a = 123;
int b (456);

Those lines also declare variables, but this time the variables are initialized to some value. What this means is that not only is space allocated for the variables but the space is also filled with the given value. The two lines illustrate two different but equivalent ways to initialize a variable. The assignment operator '=' in a declaration has a subtle distinction in that it assigns an initial value instead of assigning a new value. The distinction becomes important especially when the values we are dealing with are not of simple types like integers but more complex objects like the input and output streams provided by the iostream class.

The expression used to initialize a variable need not be constant. So the lines:

int sum;
sum = a + b;

can be combined as:

int sum = a + b;

or:

int sum (a + b);

Declare a floating point variable 'f' with an initial value of 1.5:

float f = 1.5 ;

Floating point constants should always have a '.' (decimal point) somewhere in them. Any number that does not have a decimal point is interpreted as an integer, which then must be converted to a floating point value before it is used.

For example:

double a = 5 / 2;

will not set a to 2.5 because 5 and 2 are integers and integer arithmetic will apply for the division, cutting off the fractional part. A correct way to do this would be:

double a = 5.0 / 2.0;

You can also declare floating point values using scientific notation. The constant .05 in scientific notation would be . The syntax for this is the base, followed by an e, followed by the exponent. For example, to use .05 as a scientific notation constant:

double a = 5e-2;

Below is a program storing two values in integer variables, adding them and displaying the result:

// This program adds two numbers and prints their sum.
#include <iostream.h>
 
int main()
{
  int a = 123;
  int b (456);
  int sum;
 
  sum = a + b;
 
  std::cout << "The sum of " << a << " and " << b << " is " << sum << "\n";
 
  return 0;
}

OR, if you like to save some space, the same above statement can be written as:

// This program adds two numbers and prints their sum, variation 1
#include <iostream>
#include <ostream>
 
using namespace std;
 
int main()
{
  int a = 123, b (456), sum = a + b;
 
  cout << "The sum of " << a << " and " << b << " is " << sum << endl;
 
  return 0;
}

typedef

typedef is a languages keyword, used to give a data type a new name. The intent is to make comprehension of source easier. Most of the time this occurs in old external libraries. The Style Conventions section of this book also mentions this keyword.

typedef int Apples;
typedef int Oranges;
Apples coxes;
Oranges jaffa;

Type Modifiers

There are several modifiers that can be applied to data types to change the range of numbers they can represent.

const

A variable declared with this specifier cannot be changed (as in read only). Either local or class-level variables (scope) may be declared const indicating that you don't intend to change their value after they're initialized. You declare a variable as being constant using the const keyword. Global const variables have static linkage. If you need to use a global constant across multiple files the best option is to use a special header file that can be included across the project.

const unsigned int DAYS_IN_WEEK = 7 ;

declares a positive integer constant, called DAYS_IN_WEEK, with the value 7. Because this value cannot be changed, you must give it a value when you declare it. If you later try to assign another value to a constant variable, the compiler will print an error.

int main(){
  const int i = 10;
 
  i = 3;            // ERROR - we can't change "i"
 
  int &j = i;       // ERROR - we promised not to
                    // change "i" so we can't
                    // create a non-const reference
                    // to it
 
  const int &x = i; // fine - "x" is a const
                    // reference to "i"
 
  return 0;
}

The full meaning of const is more complicated than this; when working through pointers or references, const can be applied to mean that the object pointed (or referred) to will not be changed via that pointer or reference. There may be other names for the object, and it may still be changed using one of those names so long as it was not originally defined as being truly const.

It has an advantage for programmers over #define command because it is understood by the compiler, not just substituted into the program text by the preprocessor, so any error messages can be much more helpful.

With pointer it can get messy...

T const *p;                     // p is a pointer to a const T
T *const p;                     // p is a const pointer to T
T const *const p;               // p is a const pointer to a const T

If the pointer is a local, having a const pointer is useless. The order of T and const can be reversed:

const T *p;

is the same as

T const *p;

volatile

A hint to the compiler that a variable's value can be changed externally; therefore the compiler must avoid aggressive optimization on any code that uses the variable.

Unlike in Java, C++'s volatile specifier does not have any meaning in relation to multi-threading. Standard C++ does not include support for multi-threading (though it is a common extension) and so variables needing to be synchronized between threads need a synchronization mechanisms such as mutexes to be employed, keep in mind that volatile implies only safety in the presence of implicit or unpredictable actions by the same thread (or by a signal handler in the case of a volatile sigatomic_t object). Accesses to mutable volatile variables and fields are viewed as synchronization operations by most compilers and can affect control flow and thus determine whether or not other shared variables are accessed, this implies that in general ordinary memory operations cannot be reordered with respect to a mutable volatile access. This also means that mutable volatile accesses are sequentially consistent. This is not (as yet) part of the standard, it is under discussion and should be avoided until it gets defined.

mutable

This specifier may only be applied to a non-static, non-const member variables. It allows the variable to be modified within const member functions.

mutable is usually used when an object might be logically constant, i.e, no outside observable behavior changes, but not bitwise const, i.e. some internal member might change state.

The canonical example is the proxy pattern. Suppose you have created an image catalog application that shows all images in a long, scrolling list. This list could be modeled as:

class image {
 public:
   // construct an image by loading from disk
   image(const char* const filename); 
 
   // get the image data
   char const * data() const;
 private:
   // The image data
   char* m_data;
}
 
class scrolling_images {
   image const* images[1000];
};

Note that for the image class, bitwise const and logically const is the same: If m_data changes, the public function data() returns different output.

At a given time, most of those images will not be shown, and might never be needed. To avoid having the user wait for a lot of data being loaded which might never be needed, the proxy pattern might be invoked:

class image_proxy {
  public:
   image_proxy( char const * const filename )
      : m_filename( filename ),
        m_image( 0 ) 
   {}
   ~image_proxy() { delete m_image; }
   char const * data() const {
      if ( !m_image ) {
         m_image = new image( m_filename );
      }
      return m_image->data();
   }
  private:
   char const* m_filename;
   mutable image* m_image;
};
 
class scrolling_images {
   image_proxy const* images[1000];
};

Note that the image_proxy does not change observable state when data() is invoked: it is logically constant. However, it is not bitwise constant since m_image changes the first time data() is invoked. This is made possible by declaring m_image mutable. If it had not been declared mutable, the image_proxy::data() would not compile, since m_image is assigned to within a constant function.

short

The short specifier can be applied to the int data type. It can decrease the number of bytes used by the variable, which decreases the range of numbers that the variable can represent. Typically, a short int is half the size of a regular int -- but this will be different depending on the compiler and the system that you use. When you use the short specifier, the int type is implicit. For example:

short a;

is equivalent to:

short int a;

long

The long specifier can be applied to the int and double data types. It can increase the number of bytes used by the variable, which increases the range of numbers that the variable can represent. A long int is typically twice the size of an int, and a long double can represent larger numbers more precisely. When you use long by itself, the int type is implied. For example:

long a;

is equivalent to:

long int a;

The shorter form, with the int implied rather than stated, is more idiomatic (i.e., seems more natural to experienced C++ programmers).

Use the long specifier when you need to store larger numbers in your variables. Be aware, however, that on some compilers and systems the long specifier may not increase the size of a variable. Indeed, most common 32-bit platforms (and one 64-bit platform) use 32 bits for int and also 32 bits for long int.

unsigned

The unsigned specifier makes a variable only represent positive numbers and zero. It can be applied only to the char, short,int and long types. For example, if an int typically holds values from -32768 to 32767, an unsigned int will hold values from 0 to 65535. You can use this specifier when you know that your variable will never need to be negative. For example, if you declared a variable 'myHeight' to hold your height, you could make it unsigned because you know that you would never be negative inches tall.

signed

The signed specifier makes a variable represent both positive and negative numbers. It can be applied only to the char, int and long data types. The signed specifier is applied by default for int and long, so you typically will never use it in your code.

static

Using the static modifier makes a variable have static lifetime and on global variables makes them require internal linkage (variables will not be accessible form code of the same project that resides in other files).

static lifetime

Means that a static variable will needs to be initialized in the file scope and at run time, will exist and maintain changes across until the program's process is closed, the particular order of destruction of static variables is undefined.

The static keyword can also be used on functions, inside functions, on classes, on classes members (data and functions), in structs, unions (but not in a union's member) we will cover each use separately.

Enumerated data types

In programming it is often necessary to deal with data types that describe a fixed set of alternatives. For example, when designing a program to play a card game it is necessary to keep trak of the suit of an individual card.

One method for doing this may be to create unique constants to keep track of the suit. For example one could define

const int Clubs=0;
const int Diamonds=1;
const int Hearts=2;
const int Spades=3;
.
.
.
int current_card_suit=Diamonds;

Unfortunately there are several problems with this method. The most minor problem is that this can be a bit cumbersome to write. A more serious problem is that this data is indistinguishable from integers. It becomes very easy to start using the associated numbers instead of the suits themselves. Such as

int current_card_suit=1;

and worse to make mistakes that may be very difficult to catch such as a typo

current_card_suit=11;

which produces a valid experession in C++, but would be meaningless in representing the card's suit.

One way around these difficulty is to create a new data type specifically designed to keep track of the suit of the card, and resticts you to to only use valid possibilities. We can accomplish this using an enumerated data type using the C++ "enum" keyword. In this case we could create the desired data type with the code

enum card_suit {Clubs,Diamonds,Hearts,Spades};
card_suit first_cards_suit=Diamonds;
card_suit second_cards_suit=Hearts;
card_suit third_cards_suit=0; //Would cause an error, 0 is an "integer" not a "card_suit" 
card_suit forth_cards_suit=first_cards_suit; //OK, they both have the same type.

The line of code creates a new data type "card_suit" that may take on only one of four possible values: "Clubs", "Diamonds", "Hearts", and "Spades". In general the enum command takes the form

enum new_type_name {possible_value_1, possible_value_1, ..., possible_value_n'} Optional_Variable_With_This_Type;

While the second line of code creates a new varible with this data type and initializes it to value to "Diamonds". The other lines create new variables of this new type and show some initializations that are (and are not) possible.

Internally enumerated types are stored as integers, that begin with 0 and increment by 1 for each new possible value for the data type.

enum apples {Fuji,Macintosh,GrannySmith};
enum oranges {Blood,Navel,Persian};
apples pie_filling=Navel; //error can't make an apple pie with oranges.
apples my_fav_apple=Macintosh;
oranges my_fav_orange=Navel; //This has the same internal integer value as my_favorite_apple
if(my_fav_apple==my_fav_orange) //Many compilers will produce an error or warning letting you know your comparing two different quantities.
  std::cout << "You shouldn't compare apples and oranges" << std::endl;

While enumerated types are not integers, they are in some case converted into integers. For example, when we try to send an enumerated type to standard output. For example

enum color {Red, Green, Blue};
color hair=Red;
color eyes=Blue;
color skin=Green;
std::cout << "My hair color is " << hair << std::endl;
std::cout << "My eye color is " << eyes << std::endl;
std::cout << "My skin color is " << skin << std::endl;
if (skin==Green)
  std::cout << "I am seasick!" << std::endl;

Will produce the output

My hair color is 0
My eye color is 2
My skin color is 1
I am seasick!

We could improve this example by introducing an array that holds the names of our enumerated type such as

std::string color_names[3]={"Red", "Green", "Blue"};
enum color {Red, Green, Blue};
color hair=Red;
color eyes=Blue;
color skin=Green;
std::cout << "My hair color is " << color_names[hair] << std::endl;
std::cout << "My eye color is " << color_names[eyes] << std::endl;
std::cout << "My skin color is " << color_names[skin] << std::endl;

In this case hair is automatically converted to an integer when it is index arrays. This technique is intimately tide to the fact that the color Red is internally stored as "0", Green is internally stored as "1", and Blue is internally stored as "2". Be Careful! One may override these default choices for the internal values of the enumerated types.

This is done by simply setting the value in the "enum" such as:

enum color {Red=2, Green=4, Blue=6};

In fact it is not necessary to an integer for every value of an enumerated type. In the case the value, the complier will simply increase the value of the previous possible value by one. Consider the following example,

enum colour {Red=2, Green, Blue=6, Orange};

Here the internal value of "Red" is 2, "Green" is 3, "Blue" is 6 and "Orange is 7. Be careful to keep in mind when using this that the internal values do not need to be unique.

Enumerated types are also automatically converted into integers in arithmetic expressions. Which makes it useful to be able to choose particular integers for the internal representations of an enumerated type.

One may have enumerated for the width and height of a standard computer screen. This may allow a program to do meaningful calculations, while still maintaining the benefits of an enumerated type.

enum screen_width {SMALL=800, MEDIUM=1280};
enum screen_height {SMALL=600, MEDIUM=768};
screen_width MyScreenW=SMALL;
screen_height MyScreenH=SMALL;
std::cout << "The number of pixels on my screen is " << MyScreenW*MyScreenH << std::endl;

It should be noted that the internal values used in an enumerated type are constant, and cannot be changed during the execution of the program.

It is perhaps useful to notice that while the enumerated types can be converted to integers for the purpose arithmetic, they cannot be iterated through.

For example

enum month { JANUARY=1, FEBRUARY, MARCH, APRIL, MAY, JUNE, JULY, AUGUST, SEPTEMBER, OCTOBER, NOVEMBER, DECEMBER};
 
for( month cur_month = JANUARY; cur_month <= DECEMBER; cur_month=cur_month+1)
{
  std::cout << cur_month << std::endl;
}

will fail to compile. The problem is with the for loop. The first two statements in the loop are fine. We may certainly create a new month variable and initialize it. We may also compare two months, where they will be compared as integers. We may not increment the cur_month variable. "cur_month+1" evaluates to an integer which may not be stored into a "month" data type.

In the code above we might try to fix this by replacing the for loop with:

for( int monthcount = JANUARY; monthcount <= DECEMBER; monthcount++)
{
  std::cout << monthcount  << std::endl;
}

and this will work because we can increment the integer "mounthcount".

Derived Types

TODO Complete... ref to: C++ Programming/Operators/Pointers C++ Programming/Operators/Arrays

Scope

As with any other type of language, context (i.e. what is the background of a given action or statement) has a high impact on its validity. The same is true in a programming language. For instance, variables have a finite lifetime when your program executes. The scope of an object or variable is simply that part of a program in which the variable name exists or is visible to the compiler.

We will see that in a program we have various constructs, may they be objects, variables or any other such, and they come into existence from the point were you declare them (before they are declared they are unknown) and then, at some point, they are destroyed (as we will see there are many reasons to be so) and all are destroyed when your program terminates.

As an example, in the following fragment of code, the variable 'i' is in scope only in the lines between the appropriate comments:

{
  int i; /*'i' is now in scope */
  i = 5;
  i = i + 1;
  cout << i;
}/* 'i' is now no longer in scope */

The concept of scope is straightforward unless procedures are included in the mix; then it becomes more difficult to follow:

// Confusing Scope Program
#include <iostream>
 
using namespace std;
 
int i = 5;           /* The first version of the variable 'i' is now in scope */
 
void p(){
  int i = -1;        /* A ''new'' variable, also called 'i' has come into existence. The 'i' declared above is now out of scope,*/
  i = i + 1;
  cout << i << ' ';  /* so this ''local'' 'i' will print out the value 0.*/
}                    /* The newest variable 'i' is now out of scope. The first variable, also called 'i', is in scope again now.*/
 
 
main(){
  cout << i << ' ';  /* The first variable 'i' is still in scope here so a ''5'' will be output here.*/
  char ch;
  int i = 6;         /* A ''new'' variable, also called 'i' has come into existence. The first variable 'i' is now out of scope again,*/
  i = i + 1;
  p();
  cout << i << endl; /* so this line will print out a ''7''.*/
}                    /* End of program: all variables are, of course, now out of scope.*/

The first variable 'i' is put out of scope in two separate sections. Thus the repeated statement cout << i << ' '; means something very different each time it is written. The 'i' referred to is a different location in memory on each occasion. This is what was meant by `context' in the introduction: the context or background in which the statement is placed of is different each time, so the statement does something different in each place.

It is always an error to declare the same variable name twice within the same level of scope, and as a rule should be avoided in general, this was one of the reasons behind the implementation of namespaces.

The default scope is defined as global scope, this is commonly used to defines and use global variables or other global constructs (classes, structure, functions, etc...), this makes them valid and visible to the compiler at all times.

The purpose of Scope

There are important things to note about the above example. That program is only an example program and unusually convoluted; it is useful to demonstrate the idea of scope and little else. So while it illustrates the concept of scope, it fails to illustrate usefully the purpose of scope.

Some variables are required to store information for an entire program, while other variables are short-term variables which are brought into existence momentarily for a single small purpose and then disposed of, by going out of scope. In the following program, an array of numbers is read in and then a procedure is called that computes the average of the numbers in the array. Within the procedure, in order to move through the array and select the elements in the array in turn, a variable 'i' is created for that one purpose. Contrast the two kinds of variable: the array itself, which is in scope throughout the entire program, and the variable 'i' which is in scope only for a small section of the code, to do its own little job.

// Program Average
#include <iostream>
 
using namespace std;
 
float a[20];                      /* a is now in scope.*/
int length;                       /* length is now in scope.*/
 
float average(){
  float result = 0.0;             /* result is now in scope.*/
 
  for(int i = 0; i < length; i++){ /* i is now in scope.*/
   result += a[i];
  }                               /* i is now out of scope.*/
 
  return result/length;
}                                 /* result is now out of scope.*/
 
int main(){
 length = 0;
 while(cin >> a[length++]){
 };
 length--;
 float av = average();            /* av is now in scope.*/
 cout << av << endl;
}                                 /* All variables now out of scope.*/

Scope and control structures

Within a single procedure it is possible to begin a new level of scoping. In fact it occurs every time a left brace `{' is written and it ends where its matching right brace '}' is written. Thus layers of scope to any depth can be built up as can be seen in the following program. The following program has four levels of scoping. The innermost level is said to be enclosed by the levels around it, thus we talk about inner levels and enclosing levels of scoping. Again this program is for illustration purposes only and does nothing of real value.

// Complicated Scope Program
#include <iostream>
 
using namespace std;  /* outermost level of scope starts here */
 
int i;
 
main(){               /* next level of scope starts here */
  int i;
  i = 5;
 
  {                   /* next level of scope starts here */
    int j,i;
    j = 1;
    i = 0;
 
    {                 /* innermost level of scope of this program starts here */
      int k, i;
      i = -1;
      j = 6;
      k = 2;
    }                 /* innermost level of scope of this program ends here */
 
    cout << j << ' ';
  }                   /* next level of scope ends here */
 
  cout << i << endl;
}                     /* next and outermost levels of scope end here */

The output if this program is 6 5. To understand why, we look first at the simpler situation of 'i', and we see why a five is printed for it, and then next at the more complicated situation with 'j' and why a six is printed for it.

A new variable 'i' is created at each new level of scope. Thus only the first assignment to 'i', where 'i' is assigned the value of five, affects this particular variable 'i': the variable that is printed. That variable does not alter its value after that first assignment and so the final statement prints a five. The other assignments to 'i' are irrelevant to the final print statement.

In contrast, the variable 'j' is created only once, hence the assignment where `j' is assigned the value of six, alters this only existing variable called 'j' even though the variable is declared at an enclosing level of scope. If a program has a scope level inside another, and no variable of the right name is declared at this inner level, then the computer `looks outwards' to the next enclosing level of scope. If there is no variable of that name declared there either, then it will look out further and again further outwards until the variable's declaration is found. (Of course, if a declaration for the variable is never found then the compiler indicates an error, stating that the variable is undeclared and so the program is not compiled.)

Scope using other control structures

Above we stated that a new level of scope started with each left brace `{'. However, it is not common to use just a naked left brace: usually the left brace is associated with an if statement or a while statement or some such. We add these to the above program to make the more usual (but still useless) example program following. The program does compile and it prints a 5 when it runs, but does little of value as a program.

// Complicated Scope Program, variation 1
#include <iostream>  
 
using namespace std;  /* outermost level of scope starts here */
 
int i;
 
main(){               /* next level of scope starts here */
  int i;
  i = 5;
 
  while(i != 5) {     /* next level of scope starts here */
    int j,i;
    j = 1;
    i = 0;
    switch (i) {      /* next level of scope starts here */
      int i;
 
      case 1:
        if (i != 4) { /* innermost level of scope of this program starts here */
          int k, i;
          i = -1;
          j = 6;
          k = 2;
        }             /* innermost level of scope of this program ends here */
        break;
 
      case 2: 
        j = 5;
        break;
    }                 /* next level of scope ends here */
 
    cout << j << ' ';
  }                   /* next level of scope ends here */
 
  cout << i << endl;
}                     /* next and outermost levels of scope end here */

We added an extra level of scoping at the switch statement because this statement demands a left brace, and it shows that even at that point a new level of scope is opened (because we were able to declare yet another variable called `i' without getting an error). As you can see, the various control structures (i.e. the while, if and switch statements) all end at their own matching right brace and hence at the same point as the scope ends. It is fair to say that each of these control statements operates over an area of scope. But remember, scope is a concept about variable names and the area in which they are defined, not control structures.

It is not essential that either the if or the while statements have an opening brace following, and program average above shows an example of this with its while statement. In this situation, no new level of scope is begun. It is the left brace that opens the new level of scope, not the if or while statement itself. The switch statement demands a left brace at that point, but notice that the `i' switch variable is at the old level of scope. The new level of scope comes into existence immediately following the left brace as always.

For all practical purposes, the for loop control structure can be seen to have scoping that operates in the same manner. However it is permissible to declare the for loop variable within the for statement itself, as can be seen in line eight of program average above. This is good programming practice as it makes the structure of the program clearer.

With all the programs in this section the statements start further and further to the right as the level of scope deepens. This is referred to as indentation and is a very important feature of program presentation. It makes the scope level explicit at all times and also makes it clear where a statement ends. For example, in the above program the line cout << j << ' '; is within the while loop whereas cout << i << endl; is outside that loop. The loop and the scope level end at the same point: the right brace between these two statements signals the end of both of these.

The scoping of the for control statement in detail

This subsection is probably more confusing than useful, but it is offered for completeness; feel free to skip it.

The for control statement has an unusual scoping, in that the left round bracket also starts its own level of scope. Thus the following program is legal:

// Complicated Scope Program, variation 2
#include <iostream>
 
using namespace std; /* outermost level of scope starts here */
 
int i;
 
main(){              /* next level of scope starts here */
  int i;
  i = 5;
 
  for(               /* next level of scope starts here */
      int i = 1;
      i<10 && cout << i << ' ';
      ++i )  
  {                  /* next level of scope starts here */
    int i = -1;
    cout << i << ' ';
  }                  /* two levels of scope end here*/
 
  cout << i << endl;
}                    /* next and outermost levels of scope end here */

and it gives the output:

 1 -1 2 -1 3 -1 4 -1 5 -1 6 -1 7 -1 8 -1 9 -1 5

This special feature of the for statement is not shared by, for example, the while statement. It is a syntax error to attempt to declare a variable within the while statement, so while(int i < 22)i++; gives a syntax error.

This special scope level enables one to declare a for loop variable within the for loop itself (like the variable 'i' in the above program), instead of having to declare it in the enclosing scope level, creating a neater program. But it is a little peculiar.

The above program does show one interesting feature: in order to check if a new level of scope has been opened, it is only necessary to attempt to declare an existing variable again at that level, as was done with the variable `i' in the above example program, which has a new variable `i' declared at every level possible.

The above program also illustrates one further very important point: there is a saying in computer programming that is it possible to write bad code in any language. The above program is quite unclear in its operation and a good example of bad coding: it is good code to illustrate a point, but bad code to read. All code should be written to make the program as clear as possible as discussed elsewhere under the topic of program style.

Scope and lifetime

The scope of a variable should be contrasted with its lifetime. In the program above called `Confusing Scope Program' the first variable `i' goes out of scope for a time but it remains in existence, thus its lifetime is continuing while it is out of scope. In older programming languages, it is difficult to contrive examples where scope and lifetime are different - in general in these older languages the two are the same, so lifetime equals scope. Not only is it difficult, it is not all that useful in the general case when it does occur. In recently created computer languages like C++ however, the idea of having variables that are out of scope but still alive is heavily exploited and creates the principle distinguishing feature of C++: the C++ class. This is the above program re-written with a class:

// Program Average rewritten using a class
#include <iostream>
 
using namespace std;
 
class StatisticsPackage{
private:
  float aa[20];                     /* aa scope start*/
  int length;                       /* length scope start*/
public:                           
  float average(){
    float result = 0.0;             /* result scope start*/
 
    for(int i = 0; i < length; ++i) /* i scope start*/
      result += aa[i];
 
  return result/length;
  }                                 /* result and i scope end*/
 
  void get_data(){
    length = 0;
    while(cin >> aa[length++]);
    --length;
  }
};                                  /* aa and length scope end*/
 
main(){
  StatisticsPackage sp;             /* aa and length lifetimes start */
  sp.get_data();
  float av = sp.average();          /* av scope start*/
  cout << av << endl;
}                                   /* av scope end*/

In this version of the program, the variables `length' and `aa' are alive after the class `sp' comes into existence. However, their scope has been limited: they are alive but out of scope, storing the information but not being directly available in the main program. Keeping variables out of scope in this manner is very helpful in debugging a program, by narrowing the number of lines in which the variable in question can possibly change its value.

TODO evolve and add references to the further insight and practical usefulness in sections like class space.

Namespace

The keyword as used on the several code examples examined before, this section will give an explanation to the line using namespace std;.

In many programming languages, a namespace is a context for identifiers. C++ can handle multiple namespaces within the language. By using namespace (or the using namespace keyword), one is offered a clean way to aggregate code under a shared label, so as to prevent naming collisions or just to ease recall and use of very specific scopes. There are other "name spaces" besides "namespaces"; this can be confusing.

Name spaces (note the space there), as we will see, go beyond the concept of scope by providing an easy way to differentiate what is being called/used. As we will see, classes are also name spaces, but they are not namespaces.

A namespace is defined with a namespace block.

namespace foo {
  int bar;
}

Within this block, identifiers can be used exactly as they are declared. Outside of this block, the namespace specifier must be prefixed (that is, it must be qualified). For example, outside of namespace foo, bar must be written foo::bar. C++ includes another construct which makes this verbosity unnecessary. By adding the line using namespace foo; to a piece of code, the prefix foo:: is no longer needed.

using namespace std;

This using-directive indicates that any names used but not declared within the program should be sought in the ‘standard (std) namespace’.

To make a single name from a namespace available, the following using-declaration exists:

using foo::bar;

After this declaration, the name bar can be used inside the current namespace instead of the more verbose version foo::bar. Note that programmers often use the terms declaration and directive interchangeably, despite their technically different meanings.

It is good practice to use the narrow second form (using declaration), because the broad first form (using directive) might make more names available than desired. Example:

namespace foo {
  int bar;
  double pi;
}
 
using namespace foo;
 
int* pi;
pi = &bar;  // ambiguity: pi or foo::pi?

In that case the declaration using foo::bar; would have made only foo::bar available, avoiding the clash of pi and foo::pi. This problem (the collision of identically-named variables or functions) is called "namespace pollution" and as a rule should be avoided wherever possible.

using-declarations can appear in a lot of different places. Among them are:

• namespaces (including the default namespace)
• functions

A using-declaration makes the name (or namespace) available in the scope of the declaration. Example:

namespace foo {
  namespace bar {
   double pi;
  }
 
  using bar::pi;
  // bar::pi can be abbreviated as pi
}
 
// here, pi is no longer an abbreviation. Instead, foo::bar::pi must be used.

Namespaces are hierarchical. Within the hypothetical namespace food::fruit, the identifier orange refers to food::fruit::orange if it exists, or if not, then food::orange if that exists. If neither exist, orange refers to an identifier in the default namespace.

Code that is not explicitly declared within a namespace is considered to be in the default namespace.

Another property of namespaces is that they are open. Once a namespace is declared, it can be redeclared (reopened) and namespace members can be added. Example:

namespace foo {
  int bar;
}
 
// ...
 
namespace foo {
  double pi;
}

Namespaces are most often used to avoid naming collisions. Although namespaces are used extensively in recent C++ code, most older code does not use this facility. For example, the entire standard library is defined within namespace std, and in earlier standards of the language, in the default namespace.

For a long namespace name, a shorter alias can be defined (a namespace alias declaration). Example:

namespace ultra_cool_library_for_image_processing_version_1_0 {
  int foo;
}
 
namespace improc1 = ultra_cool_library_for_image_processing_version_1_0;
// from here, the above foo can be accessed as improc1::foo

There exists a special namespace: the unnamed namespace. This namespace is used for names which are private to a particular source file or other namespace:

namespace {
  int some_private_variable;
}
// can use some_private_variable here

In the surrounding scope, members of an unnamed namespace can be accessed without qualifying, i.e. without prefixing with the namespace name and :: (since the namespace doesn't have a name). If the surrounding scope is a namespace, members can be treated and accessed as a member of it. However, if the surrounding scope is a file, members cannot be accessed from any other source file, as there is no way to name the file as a scope. An anonymous namespace declaration is semantically equivalent to the following construct

namespace $$$ {
  // ...
}
using namespace $$$;

where $$$ is a unique identifier manufactured by the compiler.

As you can nest an anonymous namespace in an ordinary namespace, and vice versa, you can also nest two anonymous namespaces.

namespace {
 
  namespace {
    // ok
  }
 
}

Because of space considerations, we cannot actually show the namespace command being used properly: it would require a very large program to show it working usefully. However, we can illustrate the concept itself easily.

// Namespaces Program, an example to illustrate the use of namespaces
#include <iostream>
 
namespace first {
  int first1;
  int x;
}
 
namespace second {
  int second1;
  int x;
}
 
namespace first {
  int first2;
}
 
main(){
  //first1 = 1;
  first::first1 = 1;
  using namespace first;
  first1 = 1;
  x = 1;
  second::x = 1;
  using namespace second;
 
  //x = 1;
  first::x = 1;
  second::x = 1;
  first2 = 1;
 
  //cout << 'X';
  std::cout << 'X';
  using namespace std;
  cout << 'X';
}

We will examine the code moving from the start down to the end of the program, examining fragments of it in turn.

#include <iostream>

This just includes the iostream library so that we can use std::cout to print stuff to the screen.

namespace first {
  int first1;
  int x;
}
 
namespace second {
  int second1;
  int x;
}
 
namespace first {
  int first2;
}

We create a namespace called first and add to it two variables, first1 and x. Then we close it. Then we create a new namespace called second and put two variables in it: second1 and x. Then we re-open the namespace first and add another variable called first2 to it. A namespace can be re-opened in this manner as often as desired to add in extra names.

  main(){
1  //first1 = 1;
2  first::first1 = 1;

The first line of the main program is commented out because it would cause an error. In order to get at a name from the first namespace, we must qualify the variable's name with the name of its namespace before it and two colons; hence the second line of the main program is not a syntax error. The name of the variable is in scope: it just has to be referred to in that particular way before it can be used at this point. This therefore cuts up the list of global names into groups, each group with its own prefixing name.

3  using namespace first;
4  first1 = 1;
5  x = 1;
6  second::x = 1;

The third line of the main program introduces the using namespace command. This commands pulls all the names in the first namespace into scope. They can then be used in the usual way from there on. Hence the fourth and fifth lines of the program compile without error. In particular, the variable x is available now: in order to address the other variable x in the second namespace, we would call it second::x as shown in line six. Thus the two variables called x can be separately referred to, as they are on the fifth and sixth lines.

7  using namespace second;
8  //x = 1;
9  first::x = 1;
10 second::x = 1;

We then pull the declarations in the namespace called second in, again with the using namespace command. The line following is commented out because it is now an error (whereas before it was correct). Since both namespaces have been brought into the global list of names, the variable x is now ambiguous, and needs to be talked about only in the qualified manner illustrated in the ninth and tenth lines.

11 first2 = 1;

The eleventh line of the main program shows that even though first2 was declared in a separate section of the namespace called first, it has the same status as the other variables in namespace first. A namespace can be re-opened as many times as you wish. The usual rules of scoping apply, of course: it is not legal to try to declare the same name twice in the same namespace.

12 //cout << 'X';
13 std::cout << 'X';
14 using namespace std;
15 cout << 'X';
}

There is a namespace defined in the computer in special group of files. Its name is std and all the system-supplied names, such as cout, are declared in that namespace in a number of different files: it is a very large namespace. Note that the #include statement at the very top of the program does not fully bring the namespace in: the names are there but must still be referred to in qualified form. Line twelve has to be commented out because currently the system-supplied names like cout are not available, except in the qualified form std::cout as can be seen in line thirteen. Thus we need a line like the fourteenth line: after that line is written, all the system-supplied names are available, as illustrated in the last line of the program. At this point we have the names of three namespaces incorporated into the program.

As the example program illustrates, the declarations that are needed are brought in as desired, and the unwanted ones are left out, and can be brought in in a controlled manner using the qualified form with the double colons. This gives the greater control of names needed for large programs. In the example above, we used only the names of variables. However, namespaces also control, equally, the names of procedures and classes, as desired.

The Compiler

A compiler is a program that translates a computer program written in one computer language (the source code) into an equivalent program written in the computer's native machine language. This process of translation is called compilation.

Compilation

The compilation output of a compiler is the result from translating or compiling a program. The most important part of the output is saved to a file called an object file. As we have seen before in the The Code Section of the book, it consists of the transformation of source files into object files.

The instructions of this compiled program can then be run (executed) by the computer if the object file is in an executable format. Often, however, there are additional steps that may be required to create an executable program: preprocessing and linking.

Compile Time

Defines the time and operations performed by a compiler (ie, compile-time operations) during a build (creation) of a program (executable or not).

The operations performed at compile time usually include lexical analysis, syntax analysis, various kinds of semantic analysis (eg, type checks, and instantiation of template) and code generation.

The definition of a programming language will specify compile time requirements that source code must meet to be successfully compiled.

Compile time occurs before link time (when the output of one or more compiled files are joined together) and runtime (when a program is executed). In some programming languages it may be necessary for some compilation and linking to occur at runtime. The concept of runtime will be introduced later.

TODO Add run time concept, and mention it here (probably on Debugging)

Lexical analysis

This is alternatively known as scanning or tokenisation. It happens before syntax analysis and converts the code into tokens, which are the parts of the code that the program will actually use. The source code as expressed as characters (arranged on lines) into a sequence of special tokens for each reserved keyword, and tokens for data types and identifiers and values. The lexical analyzer is the part of the compiler which removes whitespace and other non compilable characters from the source code. It uses whitespace to separate different tokens, and ignores the whitespace.

To give a simple illustration of the process:

int main()
{
    std::cout << "hello world" << std::endl;
    return 0;
}

Depending on the lexical rules used it might be tokenized as:

1 = string "int"
2 = string "main"
3 = opening parenthesis
4 = closing parenthesis
5 = opening brace
6 = string "std"
7 = namespace operator
8 = string "cout"
9 = << operator
10 = string ""hello world""
11 = string "endl"
12 = semicolon
13 = string "return"
14 = number 0
15 = closing brace

And so for this program the lexical analyzer might send something like:

1 2 3 4 5 6 7 8 9 10 9 6 11 12 13 14 12 15

To the syntactical analyzer, which is talked about next, to be parsed. It is easier for the syntactical analyzer to apply the rules of the language when it can work with numerical values and can distinguish between language syntax (such as the semicolon) and everything else, and knows what data type each thing has.

Syntax Analysis

This step (also called sometimes syntax checking) ensures that the code is valid and will sequence into an executable program. The syntactical analyzer applies rules to the code, checking to make sure that each opening brace has a corresponding closing brace, and that each declaration has a type, and that the type exists, and that.... syntax analysis is more complicated that lexical analysis =). As an example

int main()
{
    std::cout << "hello world" << std::endl;
    return 0;
}

The syntax analyzer would first look at the string "int", check it against defined keywords, and find that it is a type for integers. The analyzer would then look at the next token as an identifier, and check to make sure that it has used a valid identifier name. It would then look at the next token. Because it is an opening parenthesis it will treat "main" as a function, instead of a declaration of a variable if it found a semicolon or the initialization of an integer variable if it found an equals sign. After the opening parenthesis it would find a closing parenthesis, meaning that the function has 0 parameters. Then it would look at the next token and see it was an opening brace, so it would think that this was the implementation of the function main, instead of a declaration of main if the next token had been a semicolon, even though you can't declare main in c++. It would probably create a counter also to keep track of the level of the statement blocks to make sure the braces were in pairs. After that it would look at the next token, and probably not do anything with it, but then it would see the :: operator, and check that "std" was a valid namespace. Then it would see the next token "cout" as the name of an identifier in the namespace "std", and see that it was a template. The analyzer would see the << operator next, and so would check that the << operator could be used with cout, and also that the next token could be used with the << operator. The same thing would happen with the next token after the ""hello world"" token. Then it would get to the "std" token again, look past it to see the :: operator token and check that the namespace existed again, then check to see if "endl" was in the namespace. Then it would see the semicolon and so it would see that as the end of the statement. Next it would see the keyword "return", and then expect an integer value as the next token because main returns an integer, and it would find 0, which is an integer. Then the next symbol is a semicolon so that is the end of the statement. The next token is a closing brace so that is the end of the function. And there are no more tokens, so if the syntax analyzer didn't find any errors with the code, it would send the tokens to the compiler so that the program could be converted to machine language. This is a simple view of syntax analysis, and real syntax analyzers don't really work this way, but the idea is the same.

Here are some keywords which the syntax analyzer will look for to make sure you aren't using any of these as identifier names, or to know what type you are defining your variables as or what function you are using which is included in the c++ language.

Compile Speed

There are several factors that dictate how fast a compilation proceeds, like:

• Hardware
  • Resources (Slow CPU, low memory and even a slow HDD can have an influence)

• Software
  • The compiler itself, new is always better, but may depend on how portable you want the project to be.
  • The design selected for the program (structure of object dependencies, includes) will also factor in.

Experience tells that most likely if you are suffering from slow compile times, the program you are trying to compile is poorly designed, take the time to structure your own code to minimize re-compilation after changes. Large projects will always compile slower. Use pre-compiled headers and external header guards. We will discuss ways to reduce compile time in the Optimization Section of this book.

ISO C++ (C++98) Keywords

Specific compilers may (in a non-standard compliant mode) also treat some other words as keywords, including cdecl, far, fortran, huge, interrupt, near, pascal, typeof. Old compilers may recognize the overload keyword, an anachronism that has been removed from the language.

The next revision of C++, informally known as C++0x for now, is likely to add some keywords, probably including at least:

• static_assert
• decltype
• nullptr

(These are being considered carefully to minimize breakage to existing code; see http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2105.html for some details.)

Old compilers may not recognize some or all of the following keywords:

C++ Reserved Identifiers

Some "nonstandard" identifiers are reserved for distinct uses, to avoid conflicts on the naming of identifiers by vendors, library creators and users in general.

Reserved identifiers include keywords with two consecutive underscores (__), all that start with an underscore followed by an uppercase letter and some other categories of reserved identifiers carried over from the C library specification.

A list of C reserved identifiers can be found at the Internet Wayback Machine archived page: http://web.archive.org/web/20040209031039/http://oakroadsystems.com/tech/c-predef.htm#ReservedIdentifiers

TODO It would be nice to list those C reserved identifiers, for the moment All Standard C Library Functions have already been listed

Compiler Keywords

A limited set of keywords exists to directly control the compiler's behavior, these keywords are very powerful and must be used with care, they may make a huge difference on the program's compile time and running speed.

In C++ Standard, these keywords are called Specifiers.

auto

The auto keyword used to have a different behavior, but in C++0x it will allow one to omit the type of a variable and let the compiler decide. This is particularly useful for generic programming in which the return type of a function may depend on the type of its arguments. Thus, rather than this:

int x = 42;
std::vector<double> numbers;
numbers.push_back(1.0);
numbers.push_back(2.0);
for(std::vector<double>::iterator i = numbers.begin();
    i != numbers.end(); ++i) {
  cout << *i << " ";
}

we could write this:

auto x = 42; // We can use auto on base types...
std::vector<double> numbers;
numbers.push_back(1.0);
numbers.push_back(2.0);
// But auto is most useful for complicated types.
for(auto i = numbers.begin(); i != numbers.end(); ++i) {
  cout << *i << " ";
}