Re: [Boost-users] Boost.Ranges 2.0 algorithms accepting a const reference

Subject: Re: [Boost-users] Boost.Ranges 2.0 algorithms accepting a const reference
From: Stefano (pelux_at_[hidden])
Date: 2010-07-10 13:05:29

Mathias Gaunard <mathias.gaunard <at> ens-lyon.org> writes:

> 
> Le 08/07/2010 22:30, Stefano Peluchetti a écrit :
> > Hi,
> >
> > the Boost.Ranges (v 1.43) documentation suggests that for each
> > algorithm there is a version accepting a reference to a range and a
> > version accepting a const reference to a range (which in my case would
> > be useful for the "non-const reference to temporary object problem"
> > when dealing for example with a
> > boost::iterator_range<std::vector<double>::iterator>  range).
> >
> > However the code (take for instance the reverse algorithm) does not
> > seems to include the const reference version.
> >
> > Am I missing something, is the documentation incorrect, or it has been
> > decided against the const reference version?
> 
> reverse modifies its input, therefore it only takes a non-const range.
> 

Ok, but for instance find does not necessarly modify the input, and still there
is only the version accepting the non-const reference.
 
This is in contrast with the "Introduction" section of the documentation which
stresss that both versions of find have to be implemented due to the forwarding
problem.

Moreover, iterator_range beheaves like stl iterators (and const_iterators,
depending on the template argument used) and as such a const
iterator_range<std::vector<double>::iterator> will behave like a mutable range
(and the range_iterator class knows this).

Please notice this is in contrast with the sub_range class (something I
personally find a bit confusing :P).