Re: [boost] [proto][doc.tree_transformations]"Given expression of form" meaning?

From: Eric Niebler (eric_at_[hidden])
Date: 2007-05-30 11:26:29

Larry Evans wrote:
> The doc file:
>  
> libs/xpressive/proto/doc/html/boost_proto/user_s_guide/tree_transformations.html
> 
> contains:
> 
>   Second, unary_expr<  _, CalculatorGrammar >, has a default transform
>   associated with it. It is a pass-through  transform. Given an
>   expression of the form expr< T, arg1< X > >, the transform will invoke
>   the CalculatorGrammar  transform (which we haven't completely defined
>   yet -- patience) on X resulting in Y, and then reassemble the
>   expression as expr< T, arg1< Y > >.
> 
> yet unary_expr<_, CalculatorGrammar >, defined in traits.hpp, is:
> 
>          template<typename Tag, typename T>
>          struct unary_expr
>          : has_pass_through_transform<unary_expr<Tag, T> >
>          {...};
> 
> so it isn't of the form, expr<T,arg1<X> > for any T or X.  OTOH,
> the nested unary_expr typedef:
> 
>    typedef expr<Tag, args1<T> > type;
> 
> is of this form; so, I guess what's meant by "Given an expression of the 
> form.." is "Given an expression of type, E, where E::type is of the 
> form..."?

What this is trying to say is that unary_expr<  _, CalculatorGrammar > 
IS-A transform, in addition to being a pattern. unary_expr<> also is a 
meta-function for constructing expr<> types. The three roles are related 
as follows:

1) unary_expr<T, X>::type is a typedef for expr<T, args1<X> >.
2) unary_expr<T', X'> is a pattern that matches expr<T, args1<X> >
    IFF T' is T or proto::_, and X' is a pattern that matches X.
3) unary_expr<T', X'>::apply<expr<T, args1<X> >, S, V>::type applies
    unary_expr<>'s pass-through transform to expr<T, args1<X> > with
    state S and visitor V. The result is:
      expr<T, args1< X'::apply<X, S, V>::type> >

By "given an expression of the form ...," I mean "when and expression of 
form ... is passed to the transform's apply<> member template."

Sorry for the confusion. I'll make this clearer.

-- 
Eric Niebler
Boost Consulting
www.boost-consulting.com