Iterators as associative containers' key

[Ney André de Mello Zunino]

Hello.

A non-modifying algorithm I implemented uses two associative containers
from STL: set and map. The elements on those containers are supposed to
refer to actual elements which lie on another, bigger container.

I had the definition of my auxiliary containers based on pointers, but
considered using iterators instead. I know iterators might become
invalid upon certain operations, but as I said, the algorithm is
non-modifying. This is a snippet of what I was trying:

set<int> s;
typedef set<int>::itera tor ITER;

s.insert(1);

set<ITER> t;
t.insert(s.begi n());

In spite of my attempts, I could not get the code to work when using
iterators as keys for the associative containers. This probably has
good, obvious reasons, which are not clear to me at the moment. Would
anybody be so kind to explain that which I am missing?

Thank you,

--
Ney André de Mello Zunino

[Victor Bazarov]

Re: Iterators as associative containers' key

"Ney André de Mello Zunino" <zunino@unu.edu > wrote...[color=blue]
> A non-modifying algorithm I implemented uses two associative containers
> from STL: set and map. The elements on those containers are supposed to
> refer to actual elements which lie on another, bigger container.
>
> I had the definition of my auxiliary containers based on pointers, but
> considered using iterators instead. I know iterators might become
> invalid upon certain operations, but as I said, the algorithm is
> non-modifying. This is a snippet of what I was trying:
>
> set<int> s;
> typedef set<int>::itera tor ITER;
>
> s.insert(1);
>
> set<ITER> t;
> t.insert(s.begi n());
>
> In spite of my attempts, I could not get the code to work when using
> iterators as keys for the associative containers. This probably has
> good, obvious reasons, which are not clear to me at the moment. Would
> anybody be so kind to explain that which I am missing?[/color]

Set iterators do not define operator <, which is required if no
other comparison functor is provided for the Key type. You could
try defining your own comparison for type 'ITER' and pass it to
the set<ITER, myComparisonTyp e>. The simplest I could imagine is

struct myComparisonTyp e {
bool operator()(cons t ITER& i1, const ITER& i2) const {
return std::distance(i 1, i2) < 0;
}
};

(or something like it). Try it, I used it and it seems to compile
but will it work I am not sure.

Victor

[Ney André de Mello Zunino]

Re: Iterators as associative containers' key

Victor Bazarov wrote:
[color=blue]
> struct myComparisonTyp e {
> bool operator()(cons t ITER& i1, const ITER& i2) const {
> return std::distance(i 1, i2) < 0;
> }
> };
>
> (or something like it). Try it, I used it and it seems to compile
> but will it work I am not sure.[/color]

You are right and your solution works great. Thank you very much for the
insight.

Regards,

--
Ney André de Mello Zunino