boost-spirit-qi

I'm working on a compiler and I would like to improve its performances. I found that about 50% of the time is spent parsing the source files. As the source file are quite small and I do quite a lot of transformations after that, it seems to me that it is perfectible. My parser is a Boost Spirit parser with a lexer (with lexer::pos_iterator) and I have a middle sized grammar. I'm parsing the source into an AST. My problem is that I have no idea what takes the most time during the parsing: copies of AST nodes, lexer, parser rules or memory. I don't think that it is I/O problem since I'm working

I want to parse a float, but not allow NaN values, so I generate a policy which inherits from the default policy and create a real_parser with it: // using boost::spirit::qi::{real_parser,real_policies, // phrase_parse,double_,char_}; template <typename T> struct no_nan_policy : real_policies<T> { template <typename I, typename A> static bool parse_nan(I&, I const&, A&) { return false; } }; real_parser<double, no_nan_policy<double> > no_nan; // then I can use no_nan to parse, as in the following grammar bool ok = phrase_parse(first, last, no_nan[ref(valA) = _1] >> char_('@') >> double_[ref(b)

I'm looking for a way to parse a string as an int or a double, the parser should try both alternatives and choose the one matching the longest portion of the input stream. There is a deprecated directive (longest_d) that does exactly what I'm looking for: number = longest_d[ integer | real ]; ...since it's deprecated, there are any other alternatives? In case it's necessary to implement a semantic action to achieve the desired behavior, does anyone have a suggestion? Firstly, do switch to Spirit V2 - which has superseded classical spirit for years now. Second, you need to make sure an int gets

I'm trying to parse a sequence of the syntax: < direction > < type > < name >. For example: in float foo where the direction can be either in , out , or in_out . I've succeeded in parsing correct text by using a qi::symbols class to convert the direction keywords to an enum. However, the problem shows when I don't have correct text. Take the example: int foo The symbol table parser will except the 'in' part of the 'int' type and so the results will be: direction: in type: t name: foo And the error is not detected. What's the best way to be able to parse the in, out and in_out reserved words