Thread: Rvalues: They help, how??

I was reading up on the new C++11 standard, and noticed that it's been consistently a big deal that it now supports "rvalue references". But how does this help, exactly?
The example I've seen so far (in multiple places), is the "better way" of swapping objects. But in all those instances, simply swapping them via pointers would have just as efficient.
Can anyone think of a _real_ example? Maybe one not so lame? As it stands, it seems like a pretty useless feature to me. But hey, 'splain to me otherwise

Originally Posted by Yarin

The example I've seen so far (in multiple places), is the "better way" of swapping objects. But in all those instances, simply swapping them via pointers would have just as efficient.
Can anyone think of a _real_ example? Maybe one not so lame? As it stands, it seems like a pretty useless feature to me. But hey, 'splain to me otherwise

There are lots of real examples in this long article about rvalue references.

Originally Posted by whiteflags

There are lots of real examples in this long article about rvalue references.

So far, this one is better than what I've found. Thanks.

Originally Posted by laserlight

If you want to swap via pointers, you need to work with pointers even when you don't otherwise need to, so your counterexample is lame.

I see your point. Though... what you "need" to do is subjective. You don't "need" to do a lot of different things when programming, but you do it anyway, for what ever reason. Usually, it's cleaner, easier, etc., which is the same with rvalue references. I'm just not catching where that would actually be the case, and it would be much more useful than using a pointer. i.e., why add this new feature, when an existing one already does it just as well?

Well, I don't know if you've read this far or not, but it implies that a non-copyable object can be used in the STL, because of rvalue references and std::move. For me that is enormous, because if there has ever been an idea that I am more skeptical of, it's smart pointers, and I hated how you had to manage a non-copyable object in a smart pointer just to get what you want.

And there is the whole section on perfect forwarding, which is important if you believe in design patterns, and in particular, the factory pattern. Rvalue references work out to a lot less code for that.

There's probably an argument somewhere to be made about how many times you needlessly make temporary objects in C++ and rvalue references mitigate that.

Originally Posted by whiteflags

Well, I don't know if you've read this far or not, but it implies that a non-copyable object can be used in the STL, because of rvalue references and std::move. For me that is enormous, because if there has ever been an idea that I am more skeptical of, it's smart pointers, and I hated how you had to manage a non-copyable object in a smart pointer just to get what you want.

Dunno, I've never really felt a problem here. And smart pointers seem, to me, contrary to the C++ way of doing things. That's what other languages (who do it better) are for :P

Originally Posted by whiteflags

And there is the whole section on perfect forwarding, which is important if you believe in design patterns, and in particular, the factory pattern. Rvalue references work out to a lot less code for that.

Yep, just got there. I have to admit, I do say some possible advantages for that.

Originally Posted by whiteflags

There's probably an argument somewhere to be made about how many times you needlessly make temporary objects in C++ and rvalue references mitigate that.

Yeah I read about that in other articles. Though, I don't find that argument very convincing, as a good optimizer should be able to take care of that for you in most cases.


I just found this article, specifically, see the section on the "Bulb Paradox". He makes some interesting points. I suppose one could consider these new features in a similar manner.