On 9/28/05, David Abrahams wrote:

:) Yes, that's expected. When dereferenced, indirect_iterator gives you a reference onto an existing value.

...

Why isn't this working?

It wasn't designed to do that.

But it is possible to do input with an indirect_iterator no? I mean, there's the example from the documentation (somewhat summarized): char* pointers_to_chars[N]; char* pointers_to_mutable_chars[N]; boost::indirect_iterator mutable_indirect_first(pointers_to_mutable_chars), mutable_indirect_last(pointers_to_mutable_chars + N); boost::indirect_iterator const_indirect_first(pointers_to_chars), const_indirect_last(pointers_to_chars + N); std::transform(const_indirect_first, const_indirect_last, mutable_indirect_first, std::bind1st(std::plus<char>(), 1)); There you do use mutable_indirect_first to copy data into an array of pointers. Ok, my approach using the helper function was most likely too naive, though I'm not entirely sure why that one doesn't work. I tried doing the same with a vector of smartpointers, but I can't get that to work. I don't really get how to do the "char * const *" equivalent, substituting "const iterator" doesn't compile. Making it non-const compiles, but crashes. This is the code: typedef boost::shared_ptr< double > d_ptr; typedef std::vector< d_ptr >::iterator const d_it; std::vector< d_ptr > v( 10 ); std::vector< d_ptr > w( 10 ); boost::indirect_iterator< d_it > mutable_indirect_first( w.begin() ), mutable_indirect_last( w.end() ); boost::indirect_iterator< d_it, double const> const_indirect_first( v.begin() ), const_indirect_last( v.end() ); std::copy( c_indirect_first, c_indirect_last, mut_indirect_first ); And this is the error message: /usr/include/boost/shared_ptr.hpp:247: typename boost::detail::shared_ptr_traits<T>::reference boost::shared_ptr<T>::operator*() const [with T = double]: Assertion `px != 0' failed. I must say it's all a bit beyond me right now. Could you explain why things go wrong here?

Also, some way to get it to work as a back_insert_iterator would be

...

interesting, does this exist (aside from writing your own)?

Neither exists, but for either one, writing your own is possible using iterator_facade. The tutorial is at

http://www.boost.org/libs/iterator/doc/iterator_facade.html#tutorial-example

Ok, I'll look into it, thanks. -- Alex Borghgraef

On 9/29/05, David Abrahams wrote:

Try compiling and running the example, then try compiling and running your "somewhat summarized" version. Notice anything? When your version crashes, it will reveal to you the important things you left out: our example initializes all those pointers to refer to valid locations first.

Aha, indeed. And when I initialize the second 'empty' vector with zeros, it works indeed. Now I get it, thanks. P.S. I ordered your book and it arrived in the mail yesterday, so maybe I won't be posing so many stupid questions here any more in the near future :-) -- Alex Borghgraef