On Jun 30, 5:31 pm, Kai-Uwe Bux <jkherci...@gmx.net> wrote:
> abir wrote:
> > On Jun 30, 4:52 pm, Kai-Uwe Bux <jkherci...@gmx.net> wrote:
> >> abir wrote:
> >> > I am matching a template, and specializing based of a template, rather
> >> > than a single class.
>
> >> > The codes are like,
> >> > template<template<typename T,typename Alloc = std::allocator<T>
> >> >>class C>
> >> > class pix{
> >> > };
>
> >> > template<>
> >> > class pix<vector>{
> >> > };
> >> > The matching works perfectly.
> >> > Now how will i get the template parameters in the specialized class?
> >> > I want to do things like,
> >> > template<>
> >> > class pix<std::vector>{
> >> > private:
> >> > std::vector<T>* v_;
> >> > };
>
> >> Why a pointer to a vector. That looks fishy.
>
> >> > From where will I get T & Alloc fro which it is a match?
>
> >> You won't get them from anywhere. You have to supply them.
>
> >> BTW: this is not related to the class being specialized. If you consider
> >> the unspecialized pix, you will find that you cannot get at T and Alloc
> >> either because they are just placeholders for something you still need to
> >> specify.
>
> >> Best
>
> >> Kai-Uwe Bux
>
> > It can be reference to a vector instead of pointer. I want to have an
> > external iterator kind of thing, which i name pix, which is portable,
> > i.e can be stored as long as the element at the location exists,
> > irrespective of whether container resizes or not. (Of course they can
> > be made immune to pop_front,push_front, insert, erase etc also)
> > The pointer to a vector is not at all fishy in this case, almost all
> > (not all) iterators has a pointer to the container.
> > I want something like,
> > typedef vector<int> V;
> > V v; v.push_back() a few elements ...
> > pix<V> p(v);
> > ++p; *p; get element
> > v.push_back() a few more.
> > *p to get element.
>
> > so for a container C , i wanted pix as,
> > template<typename C>
> > class pix{
> > C& c_;
> > C::size_type index_;
> > };
> > when C is a vector, but another implementation for another C. Only
> > problem is C itself is a template.
>
> I see. In this case, I think, you want a pointer and not a reference. That
> way, you won't run into trouble implementing an assignment operator for
> pix.
>
Yes, that is the only reason I had to use pointer instead of
reference.
> > Why the template template parameters are "just placeholders" when the
> > instance is a full class?
>
> This is too vague. The instance of what is a class?
>
> Anyway, when you have a template template parameter, what you pass as a
> parameter is a template. In you case, you pass std::vector. Note that you
> do not pass an instantiation of that template, i.e., you do not pass
> something like std::vector<int>. You just pass std::vector. The pix class
> would be able to use the parameter C to do stuff like C<int>.
>
> > Any other way to do it?
>
> You don't want to use template template parameters. Use an ordinary typename
> parameter:
>
> template < typename Sequence >
> class pix;
>
> and then specialize for vectors:
>
> template < typename T, typename A >
> class pix< std::vector< T, A > > {
> ...
> };
>
> for deques:
>
> template < typename T, typename A >
> class pix< std::deque< T, A > > {
> ...
> };
>
> and for lists:
>
> template < typename T, typename A >
> class pix< std::list< T, A > > {
> ...
> };
>
> (not run by a compiler)
>
Got it. It runs perfectly & works according to my wishes!
Thanks a lot.
BTW, is there any already existing code which can refer to a slice of
sequence which doesn't get invalidated due to memory capacity related
issues? (I know blitz array boost numerics etc have slicing , but the
containers are non re sizable there ). I am very much in need for
that.
> Best
>
> Kai-Uwe Bux
abir
|
Similar Threads
