Re: [Boost-users] [variant] apply_visitor/static access to which()

Subject: Re: [Boost-users] [variant] apply_visitor/static access to which()
From: Stefan Strasser (strasser_at_[hidden])
Date: 2009-11-18 15:12:09

I spent quite some time figuring this out and 2 minutes after I sent this 
e-mail I discovered the obvious solution...

variant::types exposes the types of the variant as a MPL vector, which can be 
used to statically extract the vector index of the type visited.

sorry for the pointless mail.

Am Wednesday 18 November 2009 21:00:55 schrieben Sie:
> Hi,
>
> I have a problem regarding the static visitor of boost::variant.
>
> there is a variant and a mpl vector:
>
> variant<A1,A2,A3> var;
> typedef mpl::vector<B1,B2,B3> B_types;
>
> the variant and the vector have the same number of types, but not the same
> types.
>
> I need to "visit" the element of the variant, and access the B_types
> element at the same position as the element of the variant:
>
> struct visitor : static_visitor<void>{
>   template<class A>
>   void operator()(A a) const{
>      typedef typename mpl::at_c<B_types,var.which()>::type B; //error
>   }
> };
>
> so if var.which() == 1, I need a typedef of type B2. obviously the code
> above doesn't work because which() is a runtime function. but the value of
> which() is known statically at this point (that's the point of visitation).
>
> is there a way to access it, with boost::variant visitors?
>
> I like neither of the solutions I can come up with:
> 1. write my own visitation pattern that not only instantiates the visitor
> with the variant type but also with the vector type.
>
> 2.
> convert the variant to
>
> template<class A,class B>
> struct A_holder{
>   typedef B B_type;
>   A value;
> };
> variant<A_holder<A1,B1>,A_holder<A2,B2>,A_holder<A3,B3> > var;
>
> 3. do a seperate "visition" of the mpl vector inside the variant visitor:
> struct visitor : static_visitor<void>{
>   template<class A>
>   void operator()(A a) const{
>      switch(var.which()){
>        case...
>        case...
>      }
>   }
> };
>
> is there a better way to do this?