Some seeming context sensitivity, and shift/reduce errors

[jakykong]

Greetings!

I am fairly new at using Bison and Flex. I've googled this a bit and read
most of the Bison manual. Here's my problem:

test.l returns these tokens:
WORD - series of alphanumeric characters
WHITE - series of [:blank:] characters
NL - newline character
EF - end-of-file character
'*' - an asterisk, for bolds and lists

I am trying to make a simple markup language, that will take a text file and
change it into an HTML file (eventually, thiss will be part of a Wiki i'm
writing, but for now, it's just outputting to stdout).

test.y contains:
///////////////////////////////////////////////
%{
#include "./scan.yy.h"
%}

  /* because defining YYSTYPE to char * gave me compiler errors; didn't want
to fix that right now */
%union {
   char *str
}

 /* %token <str> X for the tokens that flex returns; skipped for brevity */
%type <str> PAGE LINE SENTENCE BOLD
%%
 /* note that I've done some memory allocation and concatenation that I've
shorthanded in actions here;
  * the problem isn't memory allocation; i've not gotten that far yet. */
PAGE: LINE {$$=$1;}
         | PAGE LINE {$$=$1+$2;} /* as I said, I shorthanded concatenation
and memory management */
         | PAGE EF {$$=$1+$2; YYACCEPT; }
         ;

LINE: SENTENCE NL {$$=$1;}
       ;

SENTENCE: WORD {$$=$1;}
                  | SENTENCE WORD {$$=$1+$2;}
                  | SENTENCE SPACE {$$=$1+$2;}
                  | BOLD {$$=$1;}
                  | SENTENCE BOLD {$$=$1+$2;}
                  ;

BOLD: '*' SENTENCE '*' {$$=""+$2+"";}
         ;

///////////////////////////////////////

Now, when I try to run this through bison, it complains because there is an
ambiguity:

*WORD WHITE WORD * * WORD WHITE WORD* WHITE * WHITE

right smack in the middle, there are 2 *'s, the second of which may mean an
embedded BOLD, or it may mean the end of one BOLD and the start of another.
(a shift/reduce conflict) My intention is to make this be 2 bolds, one after
the other, interpreted in that way. Also, the final * (surrounded entirely
by whitespaces) should simply be output. But that is contextual, as far as I
can tell, and I can't figure out how to represent that in YACC.

I'm stuck trying to get the 2 '*'s to be attached to the stuff in between
them, and interpreted as though nesting were not allowed. I'm also stuck
trying to get white-surrounded '*'s to be interpreted as simple asterisks.

Could anyone help?
-- 
View this message in context: 
http://www.nabble.com/Some-seeming-context-sensitivity%2C-and-shift-reduce-errors-tf4958973.html#a14202401
Sent from the Gnu - Bison - Help mailing list archive at Nabble.com.