isk defragmenters, antiviruses, backup utilities, and file compression tools are a few examples of applications that operate on directories. Alas, standard C++ doesn’t have a library for manipulating directories. Consequently, programmers resort to third-party libraries and platform-dependent hacks for mundane operations such as listing files in a directory, creating new directories, and deleting directory files. These workarounds compromise code portability and force developers to learn different libraries every time. Thankfully, you can use the quasi-standard and libraries to manipulate directories in a portable and platform-neutral fashion.

How can you operate on directories using a portable and standardized library?

Use the and libraries to list directory files and perform other directory-related operations.

Reading a Directory’s Contents
Before you can view the files in a directory you need to open it. declares the following functions for opening, reading, closing, and rewinding a directory.

To open a directory, use opendir():

DIR * opendir(const char * pathname);

This function returns a pointer to a DIR data structure that represents a directory. A NULL value indicates an error. pathname must be a name of an existing directory.

After opening a directory, use readdir() to traverse it:

struct dirent * readdir (DIR * pdir);

pdir is the result of a previous opendir() call. readdir() returns a pointer to a dirent structure whose member d_name contains the name of the current file (the rest of dirent’s members depend on the specific file system installed on your computer). Each successive call advances to the next file in the directory. There is one quirk here, though. readdir() returns NULL under two conditions: if an error occurs or once you have traversed all the files in the directory. To distinguish between these two cases, be sure to examine errno after every readdir() call. Remember that readdir() doesn’t change errno unless an error has occurred. Therefore, reset errno explicitly before calling this function.

Figure 1. Typical Output: This is an example of the typical output of the program.

After viewing the contents of directory you need to close it explicitly by calling closedir():

int closedir (DIR * pdir);

pdir is the result of a previous opendir() call. Here is a complete program that lists the contents of the current directory:

#include  #include  #include  #include  int main() { DIR *pdir; struct dirent *pent; pdir=opendir("."); //"." refers to the current dir if (!pdir){  printf ("opendir() failure; terminating");  exit(1); } errno=0;  while ((pent=readdir(pdir))){   printf("%s", pent->d_name); } if (errno){  printf ("readdir() failure; terminating");  exit(1); } closedir(pdir);}

Figure 1 shows a typical output from this program.

Creating, Deleting, and Changing a Directory
Some of the directory-related operations such as creating, deleting, and changing a directory are grouped in another quasi-standard library called . In POSIX systems, these functions are declared in .

To create a new directory, use mkdir():

int mkdir(const char * dirname, [mode_t perm]);

dirname is the name of the directory to be created. If it’s an existing file name or an invalid directory string, mkdir() fails. The perm argument is used in POSIX systems to specify the directory’s permissions. It doesn’t exist in the Windows version of this function.

To delete an existing directory, use rmdir():

int rmdir(const char * dirname);

The directory must be empty. Otherwise, rmdir() fails.

The getcwd() function retrieves the full path name (including the drive) of the current working directory:

char *getcwd(char *path, int num);

This function is tricky and requires special attention. The path argument points to an array of num characters. If the full pathname length (including the terminating ‘’) is longer than num bytes, an error occurs. Alternatively, path can be NULL. In this case, getcwd() uses malloc() to allocate a buffer large enough to store the current directory. The only snag is that you have to call free(path) afterwards.

To set a different working directory, use chdir():

int chdir(const char * dirname);

chdir() sets dirname as the current working directory of the process. If dirname isn’t a valid path name or if the process doesn’t have the right permissions, chdir() fails.

Paths will Cross
You’re probably wondering why I have adhered to pure C code in this solution. My aim was to show that even state-of-the-art ISO C++ doesn’t always offer viable alternatives to vintage C libraries. However, these C libraries can be used as the underlying machinery of a higher level object-oriented interface since they lend themselves easily to various design idioms that I’ve discussed before.

For example, instead of the cumbersome errno mechanism, you can design a hierarchy of exception classes to automate error handling. Similarly, the opendir() and closedir() pair fits neatly into the well-known Resource Acquisition Is Initialization idiom.

Finally, instead of bare DIR* and dirent * pointers, use iterators. This way, you can use ++ to advance to the next file instead of calling readdir() repeatedly. Notational convenience not withstanding, the use of iterators also enables you to exert the power of STL algorithms. Boost already offers a file-system library designed according to these principles. Hopefully, it will be incorporated into standard C++ in the future. Until then, the and libraries presented here will serve you well if you’re aiming at portable and platform-neutral code.