[lug] OT: C++ question, const_iterator design

D. Stimits stimits at attbi.com
Sun Jun 1 10:17:36 MDT 2003


Tkil wrote:

> >>>>>"DS" == D Stimits  writes:
>
>
> DS> When I do this it works:
>
> DS>    iterator it_non_const = foo.begin();
>
> DS> When I do this, it fails:
>
> DS>    const_iterator it_const = foo.begin();
>
> How about:
>
>    const_iterator it_const = const_cast(foo).begin();
>
> This is based on the idea that, unless foo is declared to be const,
> then foo.begin is resolved to a call to the plain begin() (not
> "begin() const").


Nope, this is a failure if I must do it that way. The object foo should 
return a plain iterator if lhs is iterator, and should return 
const_iterator if lhs is const_iterator. I want it to work the way it 
should, exactly as STL does. To do it by casting would make it violate 
STL container principles. To do it by inheritance is a work-around. I 
want to know the root cause, and not just "make it work".

>
> If that doesn't work, a copy of the actual diagnostic message would be
> helpful.

I'm not looking for something that just works, I'm looking for the way 
it should be done to do it exactly as a perfect STL container does it. 
The message is a bit more complicated because foo is a template, but for 
example foo<string> would say there is no match for
  foo<string>::const_iterator & = foo<string>::iterator

Then it names candidates, and tells me I *can* use
  foo<string>::const_iterator operator= (foo<string>::const_iterator)

It sees the operator= overload that accepts const_iterator, but it is 
calling "iterator begin()" instead of "const_iterator begin() const". 
This is wrong, and on the STL it works right, so I know it must be 
something I did wrong instead of a g++ bug.

D. Stimits, stimits AT attbi DOT com

>
> t.
>
>
> _______________________________________________
> Web Page:  http://lug.boulder.co.us
> Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug
> Join us on IRC: lug.boulder.co.us port=6667 channel=#colug
>





More information about the LUG mailing list