Glib::Regex returns junk, but equivalent C functions work fine

Question

I'm trying to use the Glib::Regex, but it keeps returning junk.

Here is a simplified version of the code:

void testGetPos(std::string fileName){
    auto regEx = Glib::Regex::create(
        "^sprite_[0-9]+__x(-?[0-9]+)_y(-?[0-9]+)\\.tif$",
        Glib::REGEX_CASELESS
    )

    Glib::MatchInfo match;

    if(!regEx->match(fileName, match)){
        continue;
    }

    if(!match.matches()){
        continue;
    }

    auto posX = match.fetch(1);
    auto posY = match.fetch(2);

    // ... Use posX and posY
}

int main(){
    testGetPos("sprite_000__x-28_y-32.tif");
}

After this has run, posX and posY are filled with junk. However, using C functions on the wrapped objects:

void testGetPos(std::string fileName){
    auto regEx = Glib::Regex::create(
        "^sprite_[0-9]+__x(-?[0-9]+)_y(-?[0-9]+)\\.tif$",
        Glib::REGEX_CASELESS
    )

    GMatchInfo *match = nullptr;
    if(!g_regex_match(regEx->gobj(), fileName.c_str(), (GRegexMatchFlags)0, &match)){
        if(match != nullptr){
            g_match_info_free(match);
        }
        return;
    }

    auto posX = g_match_info_fetch(match, 1);
    auto posY = g_match_info_fetch(match, 2);

    // ... Use posX and posY

    g_free(posX);
    g_free(posY);
    g_match_info_free(match);
}

int main(){
    testGetPos("sprite_000__x-28_y-32.tif");
}

Works fine. Am I doing something wrong or is this broken.

score 1 · Accepted Answer · answered Jul 09 '21 at 10:40

So, after writing this question, I tried one more thing and that fixed it. Thought I'd better document it here in case someone else hits the same issue.

I changed the equivalent of:

void testGetPos(std::string fileName){

to something like this:

void testGetPos(std::string _fileName){
    Glib::ustring fileName = Glib::filename_to_utf8(_fileName);

TL;DR: Was passing an implicitly created temporary to regEx->match and Glib::MatchInfo needs to access the Glib::ustring reference.

Turns out that the problem is this: regEx->match(fileName, match) takes const Glib::ustring & as its first parameter, but I was passing it const std::string & which was being implicitly converted. In most cases, this would be fine, however, the GMatchInfo object under the hood of Glib::MatchInfo does not copy the string that is passed to the match function and it needs that data to be available until the object is freed. When I call regEx->match with a std::string parameter, a temporary Glib::ustring object is created while regEx->match is executing and is destroyed after it finishes. This means that the data that Glib::MatchInfo is accessing is now invalid and therefore junk is returned. By using Glib::filename_to_utf8, I create a variable with a lifetime that exceeds the use of the Glib::MatchInfo object that is using it, and use and appropriate conversion function.

Hopefully, this helps anyone else that encounters this.

Glib::Regex returns junk, but equivalent C functions work fine

1 Answers1