Re: [Boost-users] [Xpressive] Nested patterns and sub-matches' semantic values

Subject: Re: [Boost-users] [Xpressive] Nested patterns and sub-matches' semantic values
From: Simon (turner25_at_[hidden])
Date: 2013-05-30 16:01:39

Yes, I've been reading on Spirit for a little while now.
I think I'll be reading a while more before I can do what I want though...

Let me know if you have any ideas using Xpressive.

Thanks,
  Simon

On Thu, May 30, 2013 at 3:47 PM, Eric Niebler <eniebler_at_[hidden]> wrote:

> On 13-05-30 08:35 AM, Simon wrote:
> > Hi,
> >   For one part of a large project, I need to parse math and turn the
> > "mathematical objects" into a C++ abstraction of them.  Is it possible
> > that a nested pattern hold semantic values within itself and can be
> > queried for that value somehow?  As if a match were a function of its
> > sub-patterns' matches, if that makes sense...
> >
> > Code might speak better than my english:
> >
> >   class math_number {...};
> >   class math_addition {...};
> >   class math_arglist {...};
> >   class math_funcdef {...};
> >
> >   sregex integer = (s1= +_d)
> >      [ return math_number( as<int>(s1) ) ];
> >
> >   sregex addition = (s1= integer) >> '+' >> (s2= integer)
> >      [ return math_addition( as<math_number>(s1), as<math_number>(s2) )
> ];
> >
> >   sregex arglist = (s1= identifier) >> *( ',' >> (s2= arglist))
> >     [ std::vector<math_id> ids( as<math_arglist>(s2).begin(),
> > as<math_arglist>(s2).end() );
> >        ids.push_back(as<math_id>(s1));
> >        return ids; ]
> >
> >   sregex funcdef = ((s1= identifier) >> '(' >> (s2= arglist) >> ")=" >>
> > (s3= addition))
> >      [ return math_funcdef( as<math_id>(s1), as<math_arglist>(s2),
> > as<math_addition>(s3) ) ];
> >
> >   I just want to avoid the funcdef above to be responsible for the
> > construction of the various math_* objects the funcdef will hold (some
> > of which will be complex to construct).  And if I really must store the
> > objects in an external stack, how should the parent-objects find their
> > child-objects?
>
> It really sounds to me that this problem would be better addressed by
> Boost.Spirit and its attributed parsers.
>
> --
> Eric Niebler
> Boost.org
> _______________________________________________
> Boost-users mailing list
> Boost-users_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/boost-users
>