Subject: Re: [boost] [Range] Non-mutating algorithm on non const qualified range
From: Steven Watanabe (watanabesj_at_[hidden])
Date: 2014-03-24 12:00:12
AMDG
On 03/24/2014 12:06 AM, Albert Yiamakis wrote:
> In the users mailing list I have remarked that this piece of code:
>
> https://gist.github.com/Kojirion/9714131
>
> will fail to compile, without providing 'mutable' iterators.
> I have since found out it will also compile by simply const qualifying
> the dummy object. (ie const Dummy dummy;)
>
> The reason is that without it this:
>
> is_const<Dummy> gets evaluated to false, resulting in
>
> boost::mpl::eval_if_c<false, boost::range_const_iterator<Dummy>,
> boost::range_mutable_iterator<Dummy> >
>
> Surely this is not intended behavior and it should be possible to use a
> range that has only const iterators in a non-mutating algorithm without
> const qualifying it?
>
This is the expected behavior. A range must provide
both const and non-const iterators. For an immutable
range, they can be the same, but you still need to
provide both. (Note that you don't need to provide
overloads for range_begin/range_end as the implicit
conversion will work there.)
http://www.boost.org/libs/range/doc/html/range/concepts/single_pass_range.html
http://www.boost.org/libs/range/doc/html/range/reference/extending/method_2.html
In Christ,
Steven Watanabe