[dmd-beta] rvalue references

[Andrei Alexandrescu]

On 4/13/12 2:36 AM, Dmitry Olshansky wrote:
> To clarify my stance on this somewhat - I believe it just shows 2 things:
> a) Sealed conatiners with rvalue returns are rapidly becoming extinct
> beasts with this new ref-binds-rvalues rule. It arguably might be viewed
> as an antipattern like declaring C++ operator[] returning non-ref where
> you obviously didn't meant it.
> b) Apparently the whole trend of "I don't care if this an rvalue or
> lvalue" comes to a stop as it turns out that you do care in a lot of cases.

Agreed to both. I think the no-escape rule for ref is great, but we need 
to work some kinks out off it. The rvalue->lvalue needs a fair amount of 
work.

Andrei