Re: [Boost-users] [spirit] working with end_p

From: Manfred Doudar (manfred.doudar_at_[hidden])
Date: 2006-07-04 21:27:30

Hi Neil,

Neil Hunt wrote:
> Hello all,
> 
> Given a simple spirit grammar (simplified from my real grammer and 
> untested) such as:
> 
> file = *line;
> line = sensible_line >> ending >> *comment;
> sensible_line = confix_p( '\"', *c_escape_ch_p, '\"' );
> comment = !( ch_p(';') >> ( anychar_p - eol_p ) ) >> ending;
> ending = *blank_p >> ( eol_p | end_p );
> 
> 
> Note: any given line can be legitimately terminated by the end of the 
> file or end of line
> 
> And there can be arbitrary numbers of comment lines between sensible lines.
> 
> However since 'comment' is terminated with 'ending', the parser ends up 
> in an infinite loop matching the kleene star on end_p because it returns 
> zero length (doesn't advance the scanner). At least I think this is the 
> problem.
> 
> By using:
> ending = *blank_p >> eol_p ;
> 
> the parser runs to completion, but any given line must be terminated by 
> an end of line sequence.
> 
> Any suggestions on how to fix this problem?

Not looking to closely at the above code, but you should be specify your 
comment parser to be your skip parser:

line = sensible_line >> ending;

and your parse line would look something like:

parse(file_iterator_begin, file_iterator_end, grammar_class_instance, 
comment);

Note that 'comment' parser is your _skip_ parser here - so the effect 
being is that all comment types are ignored on parsing.

If I have been to succinct, email back and I'll flesh it out a bit more 
for you.

Cheers,
-- 
Manfred Doudar
05.07.2006