27239687
From: dave@boostpro.com
on Wed Mar 14 2012, Gene Bushuyev wrote:
> On Mar 2, 2:01 pm, Dave Abrahams wrote:
>> on Fri Mar 02 2012, Gene Bushuyev wrote:
>> > I was looking at the C++11 standard docs and didn't see iterators for
>> > rvalue containers. For example, std::vector has only these begin()
>> > member functions:
>>
>
>> > iterator begin();
>> > const_iterator begin() const;
>>
>> > Where is "iterator begin() &&;" ?
>>
>> > Say, I want to construct my container by moving from a vector,
>>
>> > MyContainer::MyContainer(vector&& v)
>> > {
>> > reserve(v.size());
>> > uninitialized_copy(move(v).begin(), move(v).end(), data);
>> > size = v.size();
>> > }
>>
>> > But that would copy instead of moving, because containers have no
>> > rvalue iterators. Am I missing something or the standard missed them?
>>
>> I don't think you really want implicit move iterators for container
>> rvalues. Move iterators only work for single-read traversals. Replace
>> with some algorithm that makes multiple passes over the input data and
>> you could have a disaster on your hands.
>>
>
> Yes, I would like move iterators for rvalue containers,
Trust me, you really don't want that :-)
> because the perfect forwarding would work.
The problem is that too many other things would "work" (until runtime)
as well.
> And if algorithms makes multiple passes it shouldn't take container by
> r-value reference, or at least know when to move.
The standard library algorithms don't operate on containers; they
operate on iterators. You'd have to, /at least/, make sure the "rvalue
iterators" were classified no more refined-ly than InputIterator to
avoid them being used in multiple-read situations. And even that isn't
right; you can read an input iterator multiple times. We really don't
have a proper concept for read-once, random-access iterators like
move_iterator, and until we do, you're better off being really explicit
about it when you want one. move_iterator is best thought of as a
low-level convenience adaptor for implementing various moving algorithms
in terms of their non-moving cousins.
--
Dave Abrahams
BoostPro Computing
http://www.boostpro.com
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
