Question

I’m writing a DSL and using a Boost Spirit lexer to tokenize my input. In my grammar, I want a rule similar to this (where tok is the lexer):

header_block =
    tok.name >> ':' >> tok.stringval > ';' >>
    tok.description >> ':' >> tok.stringval > ';'
  ;

Rather than specifying reserved words for the language (e.g. “name”, “description”) and deal with synchronizing these between the lexer and grammar, I want to just tokenize everything that matches [a-zA-Z_]\w* as a single token type (e.g. tok.symbol), and let the grammar sort it out. If I weren’t using a lexer, I might do something like this:

stringval = lexeme['"' >> *(char_ - '"') >> '"'];
header_block =
    lit("name") >> ':' >> stringval > ';' >>
    lit("description") >> ':' >> stringval > ';'
  ;

With a lexer in the mix, I can compile the following rule, but of course it matches more than I want — it doesn’t care about the particular symbol values “name” and “description”:

header_block =
    tok.symbol >> ':' >> tok.stringval > ';' >>
    tok.symbol >> ':' >> tok.stringval > ';'
  ;

What I’m looking for is something like this:

header_block =
    specific_symbol_matcher("name") >> ':' >> tok.stringval > ';' >>
    specific_symbol_matcher("description") >> ':' >> tok.stringval > ';'
  ;

Does Qi provide anything I can use instead of my specific_symbol_matcher hand-waving, there? I’d rather not write my own matcher if I can get close using stuff that’s provided. If I must write my own matcher, can anyone suggest how to do that?

Answer 1

If the token exposes a std::string, you should just be able to do:

 statement =
      ( tok.keyword [ qi::_pass = (_1 == "if")   ] >> if_stmt )
    | ( tok.keyword [ qi::_pass = (_1 == "while) ] >> while_stmt );

If I understood you right, this is, more or less, what you were asking.

While you are at it, do look at qi::symbol<> and an especially nifty application of that, known as the Nabialek Trick.

---

Bonus material

In case you’re just struggling to make an existing grammar work with a lexer, here’s what I just did with the calc_utree_ast.cpp example to make it work with a lexer.

It shows

• how you can directly consume the exposed attributes
• how you can still parse based on char-literals, as long as these char literals are registered as (anonymous) tokens
• how the (simple) expression gammar was minimally changed
• how the skipping behaviour was moved into the lexer

///////////////////////////////////////////////////////////////////////////////
//
//  Plain calculator example demonstrating the grammar. The parser is a
//  syntax checker only and does not do any semantic evaluation.
//
//  [ JDG May 10, 2002 ]        spirit1
//  [ JDG March 4, 2007 ]       spirit2
//  [ HK November 30, 2010 ]    spirit2/utree
//  [ SH July 17, 2012 ]        use a lexer
//
///////////////////////////////////////////////////////////////////////////////

#define BOOST_SPIRIT_DEBUG

#include <boost/config/warning_disable.hpp>
#include <boost/spirit/include/lex_lexertl.hpp>
#include <boost/spirit/include/support_utree.hpp>
#include <boost/spirit/include/qi.hpp>
#include <boost/spirit/include/phoenix_operator.hpp>
#include <boost/spirit/include/phoenix_function.hpp>

#include <iostream>
#include <string>

namespace lex    = boost::spirit::lex;
namespace qi     = boost::spirit::qi;
namespace spirit = boost::spirit;
namespace phx    = boost::phoenix;

// base iterator type
typedef std::string::const_iterator BaseIteratorT;

// token type
typedef lex::lexertl::token<BaseIteratorT, boost::mpl::vector<char, uint32_t> > TokenT;

// lexer type
typedef lex::lexertl::actor_lexer<TokenT> LexerT;

template <typename LexerT_>
struct Tokens: public lex::lexer<LexerT_> {
   Tokens() {
      // literals
      uint_ = "[0-9]+";
      space = " \t\r\n";

      // literal rules
      this->self += uint_;
      this->self += '+';
      this->self += '-';
      this->self += '*';
      this->self += '/';
      this->self += '(';
      this->self += ')';

      using lex::_pass;
      using lex::pass_flags;
      this->self += space [ _pass = pass_flags::pass_ignore ];
   }

   lex::token_def<uint32_t> uint_;
   lex::token_def<lex::omit> space;
};

namespace client
{
    namespace qi     = boost::spirit::qi;
    namespace ascii  = boost::spirit::ascii;
    namespace spirit = boost::spirit;

    struct expr
    {
        template <typename T1, typename T2 = void>
        struct result { typedef void type; };

        expr(char op) : op(op) {}

        void operator()(spirit::utree& expr, spirit::utree const& rhs) const
        {
            spirit::utree lhs;
            lhs.swap(expr);
            expr.push_back(spirit::utf8_symbol_range_type(&op, &op+1));
            expr.push_back(lhs);
            expr.push_back(rhs);
        }

        char const op;
    };
    boost::phoenix::function<expr> const plus   = expr('+');
    boost::phoenix::function<expr> const minus  = expr('-');
    boost::phoenix::function<expr> const times  = expr('*');
    boost::phoenix::function<expr> const divide = expr('/');

    struct negate_expr
    {
        template <typename T1, typename T2 = void>
        struct result { typedef void type; };

        void operator()(spirit::utree& expr, spirit::utree const& rhs) const
        {
            char const op = '-';
            expr.clear();
            expr.push_back(spirit::utf8_symbol_range_type(&op, &op+1));
            expr.push_back(rhs);
        }
    };
    boost::phoenix::function<negate_expr> neg;

    ///////////////////////////////////////////////////////////////////////////////
    //  Our calculator grammar
    ///////////////////////////////////////////////////////////////////////////////
    template <typename Iterator>
    struct calculator : qi::grammar<Iterator, spirit::utree()>
    {
        template <typename Tokens>
        calculator(Tokens const& toks) : calculator::base_type(expression)
        {
            using qi::_val;
            using qi::_1;

            expression =
                term                            [_val = _1]
                >> *(   ('+' >> term            [plus(_val, _1)])
                    |   ('-' >> term            [minus(_val, _1)])
                    )
                ;

            term =
                factor                          [_val = _1]
                >> *(   ('*' >> factor          [times(_val, _1)])
                    |   ('/' >> factor          [divide(_val, _1)])
                    )
                ;

            factor =
                    toks.uint_                  [_val = _1]
                |   '(' >> expression           [_val = _1] >> ')'
                |   ('-' >> factor              [neg(_val, _1)])
                |   ('+' >> factor              [_val = _1])
                ;

            BOOST_SPIRIT_DEBUG_NODE(expression);
            BOOST_SPIRIT_DEBUG_NODE(term);
            BOOST_SPIRIT_DEBUG_NODE(factor);
        }

        qi::rule<Iterator, spirit::utree()> expression, term, factor;
    };
}

///////////////////////////////////////////////////////////////////////////////
//  Main program
///////////////////////////////////////////////////////////////////////////////
int main()
{
    std::cout << "/////////////////////////////////////////////////////////\n\n";
    std::cout << "Expression parser...\n\n";
    std::cout << "/////////////////////////////////////////////////////////\n\n";
    std::cout << "Type an expression...or [q or Q] to quit\n\n";

    using boost::spirit::utree;
    typedef std::string::const_iterator iterator_type;
    typedef Tokens<LexerT>::iterator_type IteratorT;
    typedef client::calculator<IteratorT> calculator;

    Tokens<LexerT> l;
    calculator calc(l); // Our grammar

    std::string str;
    while (std::getline(std::cin, str))
    {
        if (str.empty() || str[0] == 'q' || str[0] == 'Q')
            break;

        std::string::const_iterator iter = str.begin();
        std::string::const_iterator end  = str.end();
        utree ut;
        bool r = lex::tokenize_and_parse(iter, end, l, calc, ut);

        if (r && iter == end)
        {
            std::cout << "-------------------------\n";
            std::cout << "Parsing succeeded: " << ut << "\n";
            std::cout << "-------------------------\n";
        }
        else
        {
            std::string rest(iter, end);
            std::cout << "-------------------------\n";
            std::cout << "Parsing failed\n";
            std::cout << "stopped at: \"" << rest << "\"\n";
            std::cout << "-------------------------\n";
        }
    }

    std::cout << "Bye... :-) \n\n";
    return 0;
}

For the input

8*12312*(4+5)

It prints (without debug info)

Parsing succeeded: ( * ( * 8 12312 ) ( + 4 5 ) )