Re: [Boost-users] [Spirit] question about alternative list expressions

Subject: Re: [Boost-users] [Spirit] question about alternative list expressions
From: philip tucker (phhht_at_[hidden])
Date: 2010-05-14 22:43:22

On 13 May 2010, at 4:39 PM, Hartmut Kaiser wrote:

>> /*
>> Hello,
>>
>> This code (second copy attached) compiles to recognize dotlists,  
>> ie "a",
>> "a.b", etc.
>> I want to report on incomplete list errors of the following kind
>>
>>      a.
>>      a.#
>>
>> The parser now detects such errors by failing to parse them  
>> completely.
>> Instead I want to expect an id after a '.'.
>>
>> In dotlist_grammar() I wanted to replace this expression
>>
>>      start %= (id % '.') ;
>>
>> with this one
>>
>>      start %= (id >> (*('.' > id))) ;  // XXX
>>
>
> '>>' has difference operator precedence than '>', which means that  
> Spirit is
> not able to flatten the whole expression leading to the attribute type
> mismatch you're seeing.
>
>
>> but it won't compile. Neither will this
>>
>>      start %= (id >> (*('.' >> id))) ; // XXX
>>
>
> Attribute-wise this will work. The problem is that the function  
> signature
> you're attaching doesn't match the attribute type exposed by the  
> expression
> it is attached to. It now exposes a fusion::vector<std::string,
> std::vector<std::string> >.
>
>
>> What's wrong?  Is there a difference between the synthesized  
>> attributes of
>> the two alternative list expressions?  Or is it a different bug?
>>
>> P.S.  Both expressions seem to work for int_ instead of id.
>>
>
> HTH
> Regards Hartmut
>
> ---------------
> Meet me at BoostCon
> www.boostcon.com
>
>
/* This code (compilable copy attached) compiles to recognize  
dotlists, ie
    (string % '.'), with the added expectation that an id will follow  
a '.'.

    The solution is unsatisfactory because it requires a structure to
    contain a dotlist, and because it uses explicit semantic actions.

    Do you have any suggestions about how to obviate the semantic  
actions?

    The following pseudo-code illustrates something I don't know how  
to do,
    although I am pretty sure the solution is obvious to Spirit/ 
Fusion users.
    If that's true, I think it will solve the problem of the extraneous
    structure.

    This pseudo-grammar exposes attribute vector<string> instead of some
    structure.  What I want is to push a recognized string onto that  
vector,
    rather than onto a part of a structure.  That is what I mean  
below by

                 "_val.push_back(_1)"

    If you can show me the expression which does what I need, I thank  
you.

    template <typename Iterator>
    struct slotchain_grammar : qi::grammar<Iterator, vector<string> 
(), space>
    {
      slotchain_grammar() : slotchain_grammar::base_type(start)
      {
        using namespace qi::labels;

        using phoenix::at_c;
        using phoenix::push_back;

        start
        =  id         [ "_val.push_back(_1)" ]
        >> *('.' > id [ "_val.push_back(_1)" ])
        ;
      }

      id_grammar<Iterator> id;
      qi::rule<Iterator, vector<string>(), space> start;
    };

*/