Re: gnubol: Re: Magic tokens

[RKRayhawk@aol.com]

Let me quote some material that went back and forth enough I don't know who 
said what, but it is important. Then if you will consider my comment below.


In a message dated 12/4/99 9:41:38 PM EST, mck@tivoli.mv.com writes:

<< >>>> "Mike" == Michael McKernan <mck@tivoli.mv.com>
 >>>>> wrote the following on Sat, 04 Dec 1999 21:23:50 -0500
 
   Mike> Tim, I don't know if this translates to bison very well, but
   Mike> I am thinking along these lines.
 
 add_imperative
    : add_common { size_err end_add | end_add }
    ;
 add_stmt
    : add_common { size_err } { end_add }
    ;
 add_common
    : ADD 
      ( CORR identifier TO identifier { ROUNDED }
      |  ( identifier | NUMERIC_LITERAL )+
         ( giving_phrase    
         ! TO ( identifier | NUMERIC_LITERAL ) giving_phrase
         | TO ( identifier { ROUNDED } )+
         )
      )
    ;
 
   Mike> A greedy parser will grab the size error for the appropriate
   Mike> verb.  (You'll remember that {} means optional.)
 
 Well, now that I look at it, it may be a reduce/reduce conflict since
 both rules are capable of deriving the same string.  Well, FWIW.
 
 
  >>


At any given time we need competition between

A)
  simple unterminated arithmetic statement
  simple explicitly terminated arithmetic statement
  sticky arithmetic with a conditional unterminated (A.3)
  arithmetic with a conditional terminated (thus no longer sticky) (A.4)
B)
  sticky arithmetic with both conditional branches yet unterminated
  arithmetic with both conditional branches terminated (thus nolonger sticky) 
(B.2)


The type B statements have blocks hanging on the first conditional that by 
definiton ended 
with the commencement of an arithmetic, that is how the second branch of the 
conditional shifted on and made them category B.  If the block that hangs on 
the first conditional branch does not have arithmetics, it can not be sticky 
and can not lead into the alternate clause. This can happen if the block has 
arithmetics that are closed up with explicit scope terminators or has 
arithmetics that are reduced by a subsequent verb (simple arithmetic 
imperatives).

Thus the inner blocks of type B above are distinct in kind, they must end 
with the leading porting of an arithmetic statement by definition, or they 
can not relate to a continuation hunted for with a type B rule.

Consequently type A and type B rules only competing in the sense that the 
competition is against the outer portion of type B rules. The inner rule of 
type B rules never reduces if the block does not end with an arithmetic that 
could be a candidate for the shift on of the alternate conditional. That 
difference actually helps eliminate some conflict.

The inner rules never look for a type A.3 for example, yet must allow closure 
and reduction of type A.4 anywhere in the inner block accept the end of the 
inner block. And harsh as it seems the inner block must also allow for 
reduction of type B.2 anywhere in the inner block accept the end of the inner 
block.

All of this applies to the other conditionals as well. 

I think life is complicated by permitting or planning for transposition of 
the alternate conditionals. Ditto the other conditionals.

The inner blocks must be specified distinctly for each alternating 
conditional verb.

Bob Rayhawk


--
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.