←back to thread

Don’t blindly prefer emplace_back to push_back

(quuxplusone.github.io)
178 points todsacerdoti | 3 comments | 04 Mar 21 08:02 UTC | HN request time: 0.001s | source
Show context
MauranKilom ◴[04 Mar 21 12:01 UTC] No.26341332[source]▶
Random emplace_back hate: I really dislike how you can't use it for objects that you'd normally use aggregate initialization for.

See https://godbolt.org/z/6qW7q3

    struct SimpleData
    {
        int a;
        double b;
    };

    SimpleData someData = {1, 2.0}; // ok

    std::vector<SimpleData> data;

    // Why can I not do this?
    data.emplace_back(1, 2.0);
replies(3): >>26341692 #>>26344690 #>>26347384 #
1. Koshkin ◴[04 Mar 21 12:51 UTC] No.26341692[source]▶
To be fair, push_back won’t help you here, either.
replies(1): >>26341778 #
2. MauranKilom ◴[04 Mar 21 13:01 UTC] No.26341778[source]▶
It does, actually. In fact, this is the most concise one (see godbolt link):

    data.push_back({1, 2.0});
That does not work with emplace_back (not that it would be useful) because the template arguments can't be deduced from the initializer list (whereas for push_back the type is known so you can initialize it with the initializer list).
replies(1): >>26343145 #
3. Koshkin ◴[04 Mar 21 14:53 UTC] No.26343145[source]▶
You are correct, the problem is exclusively with emplace_back; incidentally, there is an "official" discussion of this at https://cplusplus.github.io/LWG/issue2089.