Re: gnubol: New bison Grammar available (long)

[RKRayhawk@aol.com]

In a message dated 12/26/99 5:11:41 PM EST, tej@melbpc.org.au writes:

<< 
 > The following is not intended to harp on a point, or to make anything out 
of
 > it more than the technical interaction already posted about the 
shift/reduce
 > conflict on the ELSE clauses.
 
 Bob can you suggest some test cases and I will run them as a
 priority.
  >>

actually just two deep will do it.

IF this
  THEN
    IF that
      THEN do some things
       ELSE do some different things
    ELSE do some very different things
END-IF.

about the only thing special is to avoid explicit termination of the inner 
IF, thus presenting your parser with an opportunity to yield innappropriately 
to the second ELSE.
 
Your %prec PREC_LOWER_THAN_END_IF, as initially published, is also lower that 
ELSE!
Your extended statement IF-THEN-ELSE will not reduce when confronted with 
ELSE (because it thinks it is hoping for an END-IF), since it will not 
reduce, the parser _will_ shift! There is no such rule as an 
IF-THEN-ELSE-ELSE, so the code example will produce a parse error.

by coding %prec PREC_LOWER_THAN_END_IF lower than ELSE, and tacking that onto 
the  end of the extended IF-THEN-ELSE, you commanded the parser to make a 
mistake.

It will take a more elaborate interaction to reach this but, surely you have 
the intellectual skills necessary to see that the alternating error 
conditionals of the various verbs that sport that have the same issue.  You 
will dive head long into sequences like
   ARITH SIZE_ERROR ... NOT_SIZE_ERROR ... NOT_SIZE_ERROR 
shift parse errors because the %prec at the end of the fully elaborate dual 
phrase form will 
not be high enough to kick off the third phrase. It will be named right, 
something impressive, like

 %prec THIS_HAS_JUST_GOTTA_SHIFT_END_ARITH_BUT_REDUCE_ON_CONDITIONAL

but it will not, if it is defined low.  In fact it will shift every thing. In 
fact it should shift all of the following garbage (technical term):

   ARITH SIZE_ERROR ... NOT_SIZE_ERROR ... TO
   ARITH SIZE_ERROR ... NOT_SIZE_ERROR ... BY
   ARITH SIZE_ERROR ... NOT_SIZE_ERROR ... GIVING
   ARITH SIZE_ERROR ... NOT_SIZE_ERROR ... PICTURE
   ARITH SIZE_ERROR ... NOT_SIZE_ERROR ... SELECT

The algorithm will shift these, and then commence error recovery. That seems 
almost okay.

but when an enclosing statement is

   ARITH SIZE_ERROR ... 
you do not want the interior statement to keep shifting
   ARITH SIZE_ERROR ... NOT_SIZE_ERROR ... NOT_SIZE_ERROR 

Instead when the enclosing statement is 

   ARITH SIZE_ERROR ... 
you want the interior statement to reduce
   ARITH SIZE_ERROR ... NOT_SIZE_ERROR ... 
that reduction will be  amongst statements of a stmt_block_recurse, as
we pop out of there, the lookahead now finds a better bind point, the 
enclosing staement ...
   NOT_SIZE_ERROR 

In transposition, you can do all that the other way around
   ARITH NOT_SIZE_ERROR ... 
      ARITH SIZE_ERROR ... NOT_SIZE_ERROR ... 
   SIZE_ERROR 

The technical issue is the same as on the far right edge of the fully 
populated IF-THEN-ELSE, out there you do not want another shift, you want a 
reduction when you see an ELSE (although you do want a shift when you see 
END-IF).

Piece of cake.


Best Wishes,
Bob Rayhawk
RKRayhawk@aol.com

--
This message was sent through the gnu-cobol mailing list.  To remove yourself
from this mailing list, send a message to majordomo@lusars.net with the
words "unsubscribe gnu-cobol" in the message body.  For more information on
the GNU COBOL project, send mail to gnu-cobol-owner@lusars.net.