This article by Richard Grimes, author of the book, Beginning C++ Programming explains the dependencies of a C++ application. A C++ project will produce an executable or library, and this will be built by the linker from object files. The executable or library is dependent upon these object files. An object file will be compiled from a C++ source file (and potentially one or more header files). The object file is dependent upon these C++ source and header files. Understanding dependencies is important because it helps you understand the order to compile the files in your project, and it allows you to make your project builds quicker by only compiling those files that have changed.
(For more resources related to this topic, see here.)
When you include a file within your source file the code within that header file will be accessible to your code. Your include file may contain whole function or class definitions (these will be covered in later chapters) but this will result in a problem: multiple definitions of a function or class. Instead, you can declare a class or function prototype, which indicates how calling code will call the function without actually defining it. Clearly the code will have to be defined elsewhere, and this could be a source file or a library, but the compiler will be happy because it only sees one definition.
A library is code that has already been defined, it has been fully debugged and tested, and therefore users should not need to have access to the source code. The C++ Standard Library is mostly shared through header files, which helps you when you debug your code, but you must resist any temptation to edit these files. Other libraries will be provided as compiled libraries.
There are essentially two types of compiled libraries: static libraries and dynamic link libraries. If you use a static library then the compiler will copy the compiled code that you use from the static library and place it in your executable. If you use a dynamic link (or shared) library then the linker will add information used during runtime (it may be when the executable is loaded, or it may even be delayed until the function is called) to load the shared library into memory and access the function.
Windows uses the extension lib for static libraries and dll for dynamic link libraries. GNU gcc uses the extension a for static libraries and so for shared libraries.
If you use library code in a static or dynamic link library the compiler will need to know that you are calling a function correctly—to make sure your code calls a function with the correct number of parameters and correct types. This is the purpose of a function prototype—it gives the compiler the information it needs to know about calling the function without providing the actual body of the function, the function definition.
In general, the C++ Standard Library will be included into your code through the standard header files. The C Runtime Library (which provides some code for the C++ Standard Library) will be static linked, but if the compiler provides a dynamic linked version you will have a compiler option to use this.
When you include a file into your source file the preprocessor will include the contents of that file (after taking into account any conditional compilation directives) and recursively any files included by that file. As illustrated earlier, this could result in thousands of lines of code. As you develop your code you will often compile the project so that you can test the code. Every time you compile your code the code defined in the header files will also be compiled even though the code in library header files will not have changed. With a large project this can make the compilation take a long time.
To get around this problem compilers often offer an option to pre-compile headers that will not change. Creating and using precompiled headers is compiler specific. For example, with gcc you compile a header as if it is a C++ source file (with the /x switch) and the compiler creates a file with an extension of gch. When gcc compiles source files that use the header it will search for the gch file and if it finds the precompiled header it will use that, otherwise it will use the header file.
In Visual C++ the process is a bit more complicated because you have to specifically tell the compiler to look for a precompiled header when it compiles a source file. The convention in Visual C++ projects is to have a source file called stdafx.cpp which has a single line that includes the file stdafx.h. You put all your stable header file includes in stdafx.h. Next, you create a precompiled header by compiling stdafx.cpp using the /Yc compiler option to specify that stdafx.h contains the stable headers to compile. This will create a pch file (typically, Visual C++ will name it after your project) containing the code compiled up to the point of the inclusion of the stdafx.h header file. Your other source files must include the stdafx.h header file as the first header file, but it may also include other files. When you compile your source files you use the /Yu switch to specify the stable header file (stdafx.h) and the compiler will use the precompiled header pch file instead of the header.
When you examine large projects you will often find precompiled headers are used, and as you can see, it alters the file structure of the project. The example later in this chapter will show how to create and use precompiled headers.
It is important to organize your code into modules to enable you to maintain it effectively. Even if you are writing C-like procedural code (that is, your code involves calls to functions in a linear way) you will also benefit from organizing it into modules. For example, you may have functions that manipulate strings and other functions that access files, so you may decide to put the definition of the string functions in one source file, string.cpp, and the definition of the file functions in another file, file.cpp. So that other modules in the project can use these files you must declare the prototypes of the functions in a header file and include that header in the module that uses the functions.
There is no absolute rule in the language about the relationship between the header files and the source files that contain the definition of the functions. You may have a header file called string.h for the functions in string.cpp and a header file called file.h for the functions in file.cpp. Or you may have just one file called utilities.h that contains the declarations for all the functions in both files. The only rule that you have to abide by is that at compile time the compiler must have access to a declaration of the function in the current source file, either through a header file, or the function definition itself.
The compiler will not look forward in a source file, so if a function calls another function in the same source file that called function must have already been defined before the calling function, or there must be a prototype declaration. This leads to a typical convention of having a header file associated with each source file that contains the prototypes of the functions in the source file, and the source file includes this header. This convention becomes more important when you write classes.
When a project is built with a building tool, checks are performed to see if the output of the build exist and if not, perform the appropriate actions to build it. Common terminology is that the output of a build step is called a target and the inputs of the build step (for example, source files) are the dependencies of that target. Each target’s dependencies are the files used to make them. The dependencies may themselves be a target of a build action and have their own dependencies.
For example, the following picture shows the dependencies in a project:
In this project there are three source files (main.cpp, file1.cpp, file2.cpp) each of these includes the same header utils.h which is precompiled (and hence why there is a fourth source file, utils.cpp, that only contains utils.h). All of the source files depend on utils.pch, which in turn depends upon utils.h. The source file main.cpp has the main function and calls functions in the other two source files (file1.cpp and file2.cpp), and accesses the functions through the associated header files file1.h and file2.h.
On the first compilation the build tool will see that the executable depends on the four object files and so it will look for the rule to build each one. In the case of the three C++ source files this means compiling the cpp files, but since utils.obj is used to support the precompiled header, the build rule will be different to the other files. When the build tool has made these object files it will then link them together along with any library code (not shown here).
Subsequently, if you change file2.cpp and build the project, the build tool will see that only file2.cpp has changed and since only file2.obj depends on file2.cpp all the make tool needs to do is compile file2.cpp and then link the new file2.obj with the existing object files to create the executable. If you change the header file, file2.h, the build tool will see that two files depend on this header file, file2.cpp and main.cpp and so the build tool will compile these two source files and link the new two object files file2.obj and main.obj with the existing object files to form the executable. If, however, the precompiled header source file, util.h, changes it means that all of the source files will have to be compiled.
For a small project, dependencies are easy to manage, and as you have seen, for a single source file project you do not even have to worry about calling the linker because the compiler will do that automatically. As a C++ project gets bigger, managing dependencies gets more complex and this is where development environments like Visual C++ become vital.