recursive folder scanning in c++

Question

I want to scan a directory tree and list all files and folders inside each directory. I created a program that downloads images from a webcamera and saves them locally. This program creates a filetree based on the time the picture is downloaded. I now want to scan these folders and upload the images to a webserver but I´m not sure how I can scan the directories to find the images. If anyone could post some sample code it would be very helpful.

edit: I´m running this on an embedded linux system and don´t want to use boost

Solution 1

See man ftw for a simple "file tree walk". I also used fnmatch in this example.

#include <ftw.h>
#include <fnmatch.h>

static const char *filters[] = {
    "*.jpg", "*.jpeg", "*.gif", "*.png"
};

static int callback(const char *fpath, const struct stat *sb, int typeflag) {
    /* if it's a file */
    if (typeflag == FTW_F) {
        int i;
        /* for each filter, */
        for (i = 0; i < sizeof(filters) / sizeof(filters[0]); i++) {
            /* if the filename matches the filter, */
            if (fnmatch(filters[i], fpath, FNM_CASEFOLD) == 0) {
                /* do something */
                printf("found image: %s\n", fpath);
                break;
            }
        }
    }

    /* tell ftw to continue */
    return 0;
}

int main() {
    ftw(".", callback, 16);
}

(Not even compile-tested, but you get the idea.)

This is much simpler than dealing with DIRENTs and recursive traversal yourself.

---

For greater control over traversal, there's also fts. In this example, dot-files (files and directories with names starting with ".") are skipped, unless explicitly passed to the program as a starting point.

#include <fts.h>
#include <string.h>

int main(int argc, char **argv) {
    char *dot[] = {".", 0};
    char **paths = argc > 1 ? argv + 1 : dot;

    FTS *tree = fts_open(paths, FTS_NOCHDIR, 0);
    if (!tree) {
        perror("fts_open");
        return 1;
    }

    FTSENT *node;
    while ((node = fts_read(tree))) {
        if (node->fts_level > 0 && node->fts_name[0] == '.')
            fts_set(tree, node, FTS_SKIP);
        else if (node->fts_info & FTS_F) {
            printf("got file named %s at depth %d, "
                "accessible via %s from the current directory "
                "or via %s from the original starting directory\n",
                node->fts_name, node->fts_level,
                node->fts_accpath, node->fts_path);
            /* if fts_open is not given FTS_NOCHDIR,
             * fts may change the program's current working directory */
        }
    }
    if (errno) {
        perror("fts_read");
        return 1;
    }

    if (fts_close(tree)) {
        perror("fts_close");
        return 1;
    }

    return 0;
}

Again, it's neither compile-tested nor run-tested, but I thought I'd mention it.

Solution 2

Boost.Filesystem allows you to do that. Check out the docs!

EDIT:
If you are using Linux and you don't want to use Boost, you will have to use the Linux native C functions. This page shows many examples on how to do just that.

Solution 3

You can also use glob/globfree.

Comments

[ephemient]

Easier than reading directories and performing matches yourself, but still not recursive. Granted, using GLOB_ONLYDIR means that you could avoid dealing with DIRENTs altogether, but it's still not as handy as ftw, and purely name-based traversals are racy.

[Hortitude]

The FTS sample worked great. The only changes I had to make were "ftsread" --> "fts_read" and I had to cast the result of fts_read to (FTSENT*). I would have thought finding code like this would have been easier on the web, but this was definitely the cleanest example I found. Thanks!

[darron]

Values for fts_info are not single bits... options FTS_D through FTS_W are defined as sequential values 1 through 14. It is probably confusing to some to have the code as "& FTS_F" without a comment explaining what's up. It's POSSIBLE people intended to grab FTS_F, FTS_INIT, FTS_NS, FTS_NSOK, FTS_SL, FTS_SLNONE, and FTS_W here... but FTS_NS or FTS_NSOK would be a bit odd. Also, I see code on the web doing "& FTS_D", which is almost certainly not what they intended (gets FTS_ERR, FTS_DEFAULT).