这是在linux编程中同时处理文件的有效方法吗？

Question

I want to process different files simultaneously. My program will get the first argument which is a file which contains names( absolute paths ) of different files on my pc and the second argument the name of the program to run(say anotherProgram for now). So one such file could be something like this

/usr/home/username/Desktop/folder/file.txt
/usr/home/username/Desktop/file2.txt
/usr/home/username/Desktop/folder/directory/anotherfile.txt
/usr/home/username/Desktop/folder/file3.txt
/usr/home/username/Documents/folder/file5.txt
/usr/home/username/Desktop/file10.txt
/usr/home/username/Desktop/folder/file9.txt

Now my C program will read each line and then open that particular file using low-level open system call. Now I will fork of a process and in the child I will perform the task( anotherProgram in our example) on that particular file descriptor whereas the parent goes back and opens the second file(on second line) and forks again and so on and so forth. (see question 2)

This anotherProgram will be exec'd with the file descriptor as the parameter. This way in my exec'd child, I can still use the same file descriptor to refer to the same file even after forking right? and hence process concurrently? This anotherProgram could be something that just reads from a file descriptor etc.

So now that I have said all this, I have a few questions.

1. Is this an efficient way to process files concurrently? I don't want threads, but I want to use fork someway to do this.

2. Also since I run fork in a loop in parent, every new child inherits previous file descriptors, like for example, the first child would inherit '3' and the second child would inherit '3' and '4' and the third child would inherit '3', '4' and '5' file descriptors. So is there a way to avoid this so that child 'i' inherits only 'ith' file descriptor?

TLDR; I have a list of files and a program which I want to apply to all of them concurrently using fork, dup, exec or piping, without any threads. How do I do this?

Answer 1

I have a list of files and a program which I want to apply to all of them concurrently using fork, dup, exec or piping, without any threads. How do I do this?

Your outlined design will work, but can be improved if the parent doesn't need to do anything else with a file descriptor after passing it to the child. In this case, the parent can close the file, so that the next opened file can use the same descriptor and there's no danger to use up the descriptors. Moreover, a well-known descriptor can be used, so that there's no need to pass it to the child task. E. g.:

#include <stdio.h>
#include <string.h>
#include <limits.h>

main(int argc, char *argv[])
{
    if (argc != 3)
        return printf("usage: %s 'namesFile' 'anotherProgram'\n", *argv), 1;
    FILE *names = fopen(argv[1], "r");
    if (!names) return perror(argv[1]), 1;
    char path[PATH_MAX];
    while (fgets(path, sizeof path, names))
    {
        char *np = strchr(path, '\n');
        if (np) *np = '\0';
        FILE *input = freopen(path, "r", stdin);    // reuse well-known 'stdin'
        if (!input) { perror(path); continue; }
        if (fork()) return execvp(argv[2], argv+2), perror(argv[2]), 1;
    }
}