Re: [Boost-users] [iterator] [iterator_facade] Why is advance not defined for bidirectional iterators?

Subject: Re: [Boost-users] [iterator] [iterator_facade] Why is advance not defined for bidirectional iterators?
From: Ovanes Markarian (om_boost_at_[hidden])
Date: 2011-08-22 04:51:11

Hi!
Is someone responsible for iterator_facade? Would be great if I'd be able to
clarify this with one of the authors...

Many thanks,
Ovanes

On Wed, Aug 17, 2011 at 2:58 PM, Ovanes Markarian <om_boost_at_[hidden]>wrote:

> Hello *!
>
> I defined a custom bidi-iterator and assumed it was going to support
> std::advance(iterator, Difference n), where n can be negative for random
> access and bidirectional iterators.
>
> The definition of my iterator's facade base looks like:
> struct my_iter
>   : boost::iterator_facade
>       < my_iter
>       , const byte
>       , boost::bidirectional_traversal_tag
>       , const byte
>       >
> {
>
> // some impl stuff follows...
> };
>
>
> My tests show that calling:
> advance(instance_of_my_iter, N) works fine (increment implementation is
> called)
>
> advance(instance_of_my_iter, -N) does not work, because increment (instead
> of decrement) is called as well
>
> While reading through the docs for iterator facade at
> http://www.boost.org/doc/libs/1_46_0/libs/iterator/doc/iterator_facade.html I
> found the section iterator_facade Core Operations.
>
> For the advance implementation is stated there that it is only gets called
> for Random Access Iterators. C++ Standard clearly defines that N-Parameter
> to std::advance can be negative to random access and bidi iterators.
> Implementing void advance member in my_iter still results in the call to
> increment member. Can someone give me any suggestions or is it a bug in the
> iterator lib?
>
> My compiler is gcc 4.3 under Debian Linux with Kernal 2.6.26. and Boost
> 1.42.
>
>
> With Kind Regards,
> Ovanes
>
>