Don’t blindly prefer emplace_back to push_back

Don’t blindly prefer emplace_back to push_back*

*when using it incorrectly. The premise of of emplace_back is that you use it for calling the constructor of the object in place. Obviously it won't help you if you call a copy constructor instead. I find this article a bit pointless. Clang's suggestion was spot-on and emplace_back would have (potentially) helped if the suggestion was actually followed correctly.

The article goes even further than that: When all else is equal, prefer push_back to emplace_back. It still shows the example of using emplace_back with a pre-existing object, which as you say, isn't the point of emplace_back. It also states that it's more work for the compiler, stating that the template resolution seems more complex, and suggesting it's likely to bloat compile times. That's a more persuasive point (hard numbers are provided at the end of the article), I wonder if more work has been done on this.

The author seems to be an advanced C++ programmer, but doesn't seem to properly acknowledge that, as you say, The premise of of emplace_back is that you use it for calling the constructor of the object in place. The only instance of proper use of emplace_back is in

    widgets.emplace_back(foo, bar, baz);

I was surprised to see no follow-up for this example:

    auto w = Widget(1,2,3);
    widgets.emplace_back(w);  // Fixed? Nope!

No mention of the obvious fix:

    widgets.emplace_back(1,2,3);

I don't mean to 'pile on', but, a pet peeve of mine: emplace_back is not magic C++11 pixie dust. This isn't saying anything. You could say garbage collection isn't magic, or error-correcting codes aren't magic, or sound type systems aren't magic. To say that something isn't magic is an empty statement, especially when it's something like emplace_back which introduces an important new ability.

> It also states that it's more work for the compiler, stating that the template resolution seems more complex, and suggesting it's likely to bloat compile times. That's a more persuasive point (hard numbers are provided at the end of the article)

I take the author's point about the cost of compiling a template, but the benchmark is odd: a thousand functions with a single call to emplace_back(). C++ programmers are (in)famously willing to trade compile-time performance for run-time performance. Complement the compile benchmark with a realistic program benchmark. In a typical program, I would expect emplace_back() to be called many more times than it's compiled.