在Unix C ++中的文件夹根目录中列出文件和目录

Question

I've been given an assignment which requires us to do something with all files in a given directory. For each directory encountered, we're supposed to fork off another instance of our program. My initial approach uses opendir() and readdir() , but I've found that readdir() is enumerating all entries in the directory (not just the top level items), so some entries would be processed multiple times (once by the parent process, and once for each child process until the file is in a "root"). For example, given the following file system tree:

.
├── 1.txt
├── a
│   ├── 2.txt
│   ├── 3.txt
│   └── c
│       └── 4.txt
└── b
    └── 5.txt

If I invoke my program on . , it would process 1.txt , and create two child processes, one each for a and b and then wait for those processes to finish.

The first child process works on 2.txt and 3.txt , and creates another child process for c .

The third child process works on 5.txt

Put Simply : I have no clue how to read only one level of a directory

I could continue using my initial approach, but I feel like it would really inefficient to just ignore everything that's not in the immediate folder I'm currently inspecting.

---

EDIT 1 : Example code:

int process(std::string binpath, std::string src)
{
        DIR* root = nullptr;

        root = opendir(source.c_str());

        if(root == nullptr)
        {
            return -1;
        }

        struct dirent* details;
        struct stat file;
        std::string path;

        std::queue<int> childPids;

        bool error = false;

        while((details = readdir(root)) != nullptr)
        {
            path = source + details->d_name;
            lstat(path.c_str(), &file);

            if(IsDirectory(file.st_mode))
            {
                if(strcmp(details->d_name, ".") == 0 || strcmp(details->d_name, "..") == 0)
                {
                    continue;
                }

                // Fork and spawn new process and remember child pid
                auto pid = fork();
                if(pid == 0)
                {
                    char* args[] = {
                            (char*) "-q",
                            (char*)"-s", const_cast<char*>(path.c_str()),
                            NULL
                    };
                    char* env[] = {NULL};

                    execve(binpath.c_str(), args, env);
                }
                else
                {
                    childPids.push(pid);
                }
            }
            else
            {
                if(!DoWorkOnFile(path)) error = true;
            }
        }

        //Wait all child pids

        while(!childPids.empty())
        {
            auto pid = childPids.front();

            int status;
            if(waitpid(pid, &status, 0) == pid)
            {
                if(status != 0) error = true;
            }
            else
            {
                error = true;
            }

            childPids.pop();
        }

        closedir(root);

        return error ? -1 : 0;
}

Answer 1

So let's assume you use dirent and have a code like from following tutorial: http://www.dreamincode.net/forums/topic/59943-accessing-directories-in-cc-part-i/

You have to distinguish between file and directory which is offered by the dirent structure through d_type.

Take a look here: http://www.gnu.org/software/libc/manual/html_node/Directory-Entries.html

That way you can make him list only the files of one directory.