Simple expression with boost::spirit

寵の児 提交于 2020-02-25 05:13:46

问题


I need to parse simple_expression ::= limit int_number (days | hours | minutes). I wrote code for grammar

struct Parser: grammar<std::string::const_iterator, boost::spirit::ascii::space_type>
{
public:
  Parser(ConditionTree& a_lTree):
    Parser::base_type(limit_expression),
    m_lTree(a_lTree)
  {
    using boost::spirit::qi::uint_;

    using boost::spirit::qi::_1;
    using boost::spirit::qi::_2;

    limit_expression = limit_days_operator | limit_hours_operator | limit_minutes_operator ;

    limit_days_operator = ( string("limit") > uint_ > string("days") )[ phoenix::bind( &ConditionTree::AddDaysLimitOperator, m_lTree, _2) ]  ;
    limit_hours_operator = ( string("limit") > uint_ > string("hours") )[ phoenix::bind( &ConditionTree::AddHoursLimitOperator, m_lTree, _2) ]  ;
    limit_minutes_operator = ( string("limit") > uint_ > string("minutes") )[ phoenix::bind( &ConditionTree::AddMinutesLimitOperator, m_lTree, _2) ] ;

    BOOST_SPIRIT_DEBUG_NODE(limit_expression);

    BOOST_SPIRIT_DEBUG_NODE(limit_days_operator);
    BOOST_SPIRIT_DEBUG_NODE(limit_hours_operator);
    BOOST_SPIRIT_DEBUG_NODE(limit_minutes_operator);
  }

  rule<std::string::const_iterator, boost::spirit::ascii::space_type> limit_expression;
  rule<std::string::const_iterator, boost::spirit::ascii::space_type> limit_days_operator;
  rule<std::string::const_iterator, boost::spirit::ascii::space_type> limit_hours_operator;
  rule<std::string::const_iterator, boost::spirit::ascii::space_type> limit_minutes_operator;

  ConditionTree& m_lTree;
}


void main()
{

  ConditionTree oTree;

  Parser parser(oTree);

  std::string strTest("limit5minutes");

  std::string::const_iterator it_begin(strTest.begin());
  std::string::const_iterator it_end(strTest.end());

  bool result = phrase_parse(it_begin, it_end, parser, space);
}

But it can't compile with following 2 errors: /usr/include/boost/spirit/home/support/argument.hpp:103: ошибка: no matching function for call to 'assertion_failed(mpl_::failed************ (boost::spirit::result_of::get_arg<boost::fusion::vector1<unsigned int&>, 1>::index_is_out_of_bounds::************)())'

/usr/include/boost/spirit/home/phoenix/bind/detail/member_function_ptr.hpp:103: ошибка: invalid initialization of reference of type 'const unsigned int&' from expression of type 'mpl_::void_'

on line limit_days_operator = ( string("limit") > uint_ > string("days") )[ phoenix::bind( &ConditionTree::AddDaysLimitOperator, m_lTree, _2) ] ;

I tryed to move semantic action to uint_ :

limit_days_operator = string("limit") > uint_ [ phoenix::bind( &ConditionTree::AddDaysLimitOperator, m_lTree, _1) ] > string("days")  ;
limit_hours_operator = string("limit") > uint_ [ phoenix::bind( &ConditionTree::AddHoursLimitOperator, m_lTree, _1) ] > string("hours")  ;

Then the parser correctly read limit5days, but not correctly limit5minutes, because, as I see, it can't differ limit5days from limit5hours.


回答1:


There's a lot going on here. However, by the time I was done fixing up little things trying to read it and make it complete (SSCCE), it compiled too: Live On Coliru

I briefly illuminate some individual points here:

  • the bind expressions in the semantic action copy m_lTree by value, this way you can't mutate the member. phx::ref was required here
  • the expectation points made it impossible to parse anything other than "days" limits successfully
  • use lit unless you actually want to consume the value of the string as an attribute. Better yet, just write "limit" >> uint_ >> "days" because overload overloading does the rest

As you've noticed there are number of anti-patterns in this grammar, making it complex:

  • Boost Spirit: "Semantic actions are evil"?
  • the reference-type member m_lTree leads to Action At A Distance and tight coupling.

However:

That all said, I'd simplify things by avoiding semantic actions, and separating parsing and evaluation. Simply parsing into a Limit struct:

qi::rule<Iterator, ConditionTree::Limit(), Skipper> limit_expression;

limit_expression = "limit" >> qi::uint_ >> unit_;

Handles the units without separate expression branches:

unit_.add("days",    ConditionTree::Limit::days)
         ("hours",   ConditionTree::Limit::hours)
         ("minutes", ConditionTree::Limit::minutes);

Because, now, the grammar just parses a Limit object you can do the evaluation atomically after the parse:

ConditionTree::Limit limit;
if (phrase_parse(iter, end, parser, ascii::space, limit))
{
    AddLimit(oTree, limit);
}

Separating parsing from evaluation enables you to add things like

  • repeated evaluation of the same expression (without repeated parsing)
  • simplifying the expression tree before evaluation
  • validating before evaluation
  • debugging
  • slightly more interestingly, allows you to write ApplyLimit in a functional style, not mutating an object:

    ConditionTree ApplyLimit(ConditionTree const& ct, Limit limit) { 
        return ct + limit; // do something here
    }
    
  • but most importantly: it hugely simplifies the grammar and saves you hours of your life that were better spent otherwise

See this Live On Coliru, outputting:

void AddLimit(ConditionTree&, ConditionTree::Limit): 5 minutes

Listing

#include <boost/fusion/adapted/struct.hpp>
#include <boost/spirit/include/qi.hpp>

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

struct ConditionTree {
    struct Limit {
        unsigned value;
        enum unit_t { days, hours, minutes } unit;
    };

    friend void AddLimit(ConditionTree& ct, Limit limit) {
        std::cout << "AddLimit: " << limit.value;
        switch (limit.unit) {
            case Limit::days:    std::cout << " days\n"; break;
            case Limit::hours:   std::cout << " hours\n"; break;
            case Limit::minutes: std::cout << " minutes\n"; break;
        }
    }
};

BOOST_FUSION_ADAPT_STRUCT(ConditionTree::Limit, (unsigned,value)(ConditionTree::Limit::unit_t,unit))

template <typename Iterator = std::string::const_iterator, typename Skipper = ascii::space_type>
struct Parser: qi::grammar<Iterator, ConditionTree::Limit(), ascii::space_type>
{
    public:
        Parser() : Parser::base_type(limit_expression)
    {
        unit_.add("days", ConditionTree::Limit::days)
                 ("hours", ConditionTree::Limit::hours)
                 ("minutes", ConditionTree::Limit::minutes);

        limit_expression = "limit" >> qi::uint_ >> unit_;
    }

    qi::symbols<char, ConditionTree::Limit::unit_t> unit_;
    qi::rule<Iterator, ConditionTree::Limit(), Skipper> limit_expression;
};

int main()
{
    ConditionTree oTree;

    Parser<> parser;

    std::string strTest("limit5minutes");

    std::string::const_iterator iter(strTest.begin()), end(strTest.end());

    ConditionTree::Limit limit;
    if (phrase_parse(iter, end, parser, ascii::space, limit))
    {
        AddLimit(oTree, limit);
    }
}


来源:https://stackoverflow.com/questions/24933455/simple-expression-with-boostspirit

易学教程内所有资源均来自网络或用户发布的内容,如有违反法律规定的内容欢迎反馈
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!