1

I'm trying to implement a rule in my boost spirit qi grammar that will behave like QUOTE in a Lisp-like language.

Something like: QUOTE(a b c)

The idea is that anything between QUOTE's opening and closing parenthesis will be captured into a string literal, rather than being parsed any further.

Since the '(' or ')' character may be present inside of a QUOTE, I can't use: '\"' >> *~char_('\"') >> '\"' and instead need to keep track of parenthesis to determine the end-of-quote.

For example: QUOTE(a ( b c ) d) Here, I would want to parse the contents "a ( b c ) d" as a string literal.

I can think of a few ways to do this, the general idea being to keep a local variable to increment/decrement the parenthesis depth, etc.

Due to the inefficiency of semantic actions, I was hoping someone might comment an efficient way to approach this problem.

Thanks!

pt3dNyc
  • 369
  • 1
  • 16
  • How are semantic actions _inefficient_? They may be cumbersome and slow at compile time. But they're also usually very efficient. (If not, it's often the grammar). However [see here](http://stackoverflow.com/a/8259585/85371) for some downsides of Semantic Actions – sehe Jul 11 '14 at 22:35

1 Answers1

3

There's qi::raw[] for this.

It exposes the source iterator range, so you can decide to copy or construct, e.g., a boost::string_ref from it. It doesn't get much more efficient than this.

For example:

qi::rule<It, std::string()> demo =
      qi::raw [
          qi::int_ >> '{' >> (qi::double_ % ',') >> '}'
      ];
sehe
  • 374,641
  • 47
  • 450
  • 633