Re: [RFA] Algorithms vs operator* and operator==

[chris jefferson <caj at cs dot york dot ac dot uk>]

Paolo Carlini wrote:

> Matt Austern wrote:
>
> > We're reinventing what Howard Hinnant calls "move semantics".
> > If we're going down this route, let's build on his work.
>
>
> I can only agree, in principle: some time ago I pointed Chris to some 
> background literature about "move semantics" ;)
>
> However (comments very appreciated!) I gather that the clean and neat 
> and general way to "move semantics" involves compiler support and 
> Chris wants something rather simple, for the short/medium term. I'm 
> under the impression that the improvement could be rather impressive 
> in some cases, without too much added complexity.
Also, in the case of algorithms, I'm not convinced that move symantices 
would help that much.

Under my understanding, move symantics can be split into two broad sections:
1) Some things involving temporaries that I'll admit I don't really 
understand
2) The case where we have two variables x and y and we want to do "x=y" 
without caring what happens to y, but require that y ends up in a 
consistent state.

Looking at the classes I have access to, the most efficent way of 
implementing (2) looks like it is usually either just doing x=y, or 
doing swap(x,y). therefore move symantics are overkill for this when you 
simply want to either do swap(x,y) or x=y.

Firther, if we know that we are performing swap(x,y) to "move y to x", 
then we can often further optimise algorithms by never creating 
temporary variables (as we don't have to bother storing x's value, it's 
now in y).

Therefore, I argue that just specialising for using either swap or 
assignment is a) simpler and b) at least as, and sometimes more efficent 
than a general move operator.

However, I'm happy to be proved wrong by someone who knows more than me :)

Chris