ac> BS> for(typename std::set::iterator it = cont.begin(); 
ac> BS>     it != cont.end();
ac> BS>     ++it) {
ac> BS>     // do something with *it
ac> BS> }
ac>
ac>How does "typename" work here?  I'm not familiar with recent C++
ac>developments...

My error.. There should not be any typename there.. Typename should only
be used in relation to template arguments. typename C::iterator for a 
template argument C works. 

ac> BS> We can simply write:
ac>
ac> BS> for(auto it=cont.begin(); it != cont.end(); ++it) {
ac>
ac> BS> }
ac>
ac>Hmm.  I don't really know how that works, but I think I prefer this:

The compiler will deduce the type of 'it' from the return type of 
cont.begin() .. It's that simple.. :)

ac>    typedef std::set intset_t;
ac>    intset_t cont;
ac>
ac>    cont.insert(5);
ac>    cont.insert(7);
ac>
ac>    intset_t::iterator it = cont.begin();
ac>
ac>    while (it != cont.end())
ac>    {
ac>        // do something with *it
ac>        ++it;
ac>    }

Yes that's the generic way of doing it.. However we should use std::for_each
if we should be 100% generic. But then we are missing the lambda functions
where we can specify functions in one line.. :)

ac>Having said that, my personal preference is a bit academic - as these
ac>days I spend most of my programming time writing code in Python :-)

Heh, the STL/C++ is my academic preference.. I'm working at the cphstl.dk..

ac>  cont = set()
ac>  cont.add(5)
ac>  cont.add(7)
ac>
ac>  for it in cont:
ac>    print it
ac>

Yep that's nice.. You would probably use "dict" instead? 

I wonder if "dict" is also implemented as a balanced binary search tree
in python. In the STL you get O(lg n) guarantee for almost all your
operations on set/map. It's really a nice thing with the complexity
guarantees.

Bo

--- DayDream/Linux 2.15a