Re: auto and const_iterators

[John W. Eaton]

On 05/02/2018 06:22 PM, Rik wrote:
> On 05/02/2018 03:10 PM, John W. Eaton wrote:
> > On 05/02/2018 06:06 PM, Rik wrote:
> > > On 05/02/2018 02:52 PM, John W. Eaton wrote:
> > > > Either solution is OK with me, but I see more lines that remove
> > > > const_iterator than iterator.  Are the only ones we have to change the
> > > > ones where the object itself is not const but we were accessing elements
> > > > with a const_iterator?
> > >
> > > That's right.  It's only when we want to confirm to the compiler that the
> > > access should be read-only which is only necessary for a mutable object.
> > > If the entire object is const already then all accesses are necessarily
> > > read-only.
> >
> > OK, I can fix this, but is there an easier way to find them other than
> > manually looking at each of the instances where a const_iterator
> > declaration was replaced by auto?
>
> Unfortunately, I don't think so :(

I pushed the following changeset:

  http://hg.savannah.gnu.org/hgweb/octave/rev/416856765a55

If the object was already const or could be made const (directly, not 
with a second const reference), I used auto.  Otherwise, I made a note 
that auto could be used in combination with cbegin, cend, crbegin, or 
crend in the future.

Then I checked and found that cbegin, cend, crbegin, and crend methods 
were already present in C++11 standard containers.  If I understand 
correctly, C++14 adds a template that can be used to create these from 
objects that don't provide them.  So I also pushed this changeset:

  http://hg.savannah.gnu.org/hgweb/octave/rev/4d7790d9793f

jwe