Re: Fusion-ifying proto parse trees

Eric Niebler wrote:
> Joel de Guzman wrote:
> > Eric Niebler wrote:
> > > Joel de Guzman wrote:
> > > > The issue of flattening is based on extensive deliberations on
> > > > and off list that will take lots of effort for me to explain in
> > > > one post. The discussions go back years.
> > >
> > > It's crucial that I understand the context. Can you point me to the 
> > > discussions?
> >
> > Oh jeez, that will require a bit of searching. I intend to do that
> > anyway when I rewrite the design strategies pages, but I'll have to
> > defer that work for later.
> >
> > Anyway, one definite rationale is ease of transformation and
> > attribute processing. Given a>>b>>c>>d, the attribute is
> > tuple<RA,RB,RC,RD> where RA...RD are the attributes of a...d
> > that you can obtain through traits::attribute_of<P>::type.
> > It's definitely more convenient to process seq(a,b,c,d) than,
> > say, seq(seq(seq(a,b),c),d) and all its permutations (e.g.
> > seq(seq(a,b),seq(c,d)). It is a must for semantic handling too.
> > Imagine semantic actions receiving different permutations of
> > tuple A..D in various grouped forms.
>
>
> But you'll never have to worry about all the permutations, because you 
> will have a view that makes it look flat.
>
> tuple<RA,RB,RC,RD> attrs =
>   generate(transform_view((a>>b)>>(c>>d), attribute_of());
>
> The grouping won't matter -- you'll get the same tuple each time. 
> Likewise with semantic actions -- you'll never see any permutations.
>
> But you already know this. I'm still very confused -- what is the problem?

Oh man, we're running around in circles.

1) I mentioned the need for flat sequences on associative operators
2) I realized that proto-ET can be a view or generate a view that
   gives this flat impression. You wrote some code.
3) We realized that there's a performance penalty compile-time/runtime
   however you implement it (subject to testing of course)
4) I noted that this elaborate dance wouldn't be needed
   where if proto had generated the flat sequences where appropriate
   in the first place.
5) Now you go back to 1 and ask me why I need flat sequences.

Regards,
--
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net



-------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642