Re: [Boost-users] Understanding MPL and using '::type'

From: Edward Diener (eldiener_at_[hidden])
Date: 2008-05-06 23:30:42

Ovanes Markarian wrote:
> Edward,
> 
> Just wanted to add my 2 cent. That might help. You need ::type when you 
> do type calculations.
> For example:
> 
> Access the type stored in a type sequence -> ::type

OK

> transform type sequence to another type sequence -> ::type

OK

> So meta-algorithms operate on types and result in new types -> ::type

There is nothing in the book or documentation reference about 
"meta-algorithms", whatever they are.

> 
> You don't need ::type when you pass parameters or transport the 
> contained type(s), e.g.

What do you mean by the above ?

> Placeholder transports type parameter to a metafunction -> no ::type

OK.

> vector of type passed to a metafunction also transports type(s) -> no ::type

OK

> 
> Hope that helps.

It helps a little but it does not break the two general cases in my mind 
into all the specific instances of using mpl constructs in the reference.

I am gathering that the correct specific way to determine whether or not 
to append "::type" is to look at the mpl construct and if it is a 
metafunction, which has its own nested type, always append "::type". 
More general rules regarding categories of decision about appending 
"::type", as you have suggested above, seem very vague to me except in 
specific cases which Steve and you have mentioned.

Eddie

> 
> Ovanes
> 
> P.S. sorry that I did not remove unrelated posting lines.
> 
> On Tue, May 6, 2008 at 9:44 PM, Steven Watanabe <watanabesj_at_[hidden] 
> <mailto:watanabesj_at_[hidden]>> wrote:
> 
>     AMDG
> 
>     Edward Diener wrote:
>      > I understand the explanation about how placeholders work. The
>     rule you
>      > are specifying appears to be that one never adds the '::type' to
>     the end
>      > of a placeholder expression when it is being passed to an mpl
>     construct.
>      > Is that correct ?
>      >
> 
>     You should never add ::type to a placeholder expression at all.
>     Otherwise, the metafunction will probably get very confused.
> 
>      > I wish I knew what the is an mpl function as opposed to what is
>     not an
>      > mpl function. Perhaps the mpl algorithms are the mpl functions and
>      > everything which is not an algorithm is not an mpl function.
>      >
> 
>     Inserters, Sequences, and the types listed under "Data Types" are not
>     metafunctions.
>     Everything listed in a section with Metafunctions of Algorithms in its
>     name are metafunctions.
> 
>      >>  If your not
>      >> sure, the
>      >> only thing to do is look it up.
>      >>
>      >
>      > Look up what ? Where in the reference documentation does it tell me
>      > whether I need to add '::type' to the end of an mpl construct ?
>      >
> 
>     Here are two examples from the reference:
> 
>     template< typename Sequence , typename State , typename ForwardOp >
>     struct accumulate <./accumulate.html> {
>        typedef /unspecified/ type;
>     };
> 
>     template< typename Seq >
>     struct back_inserter <./back-inserter.html> {
>        // /unspecified/
>        // /.../
>     };
> 
>     In Christ,
>     Steven Watanabe
> 
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Boost-users mailing list
> Boost-users_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/boost-users