Re: LALR(1)- but not LR(1)-conflict

holzmuel@kafka.informatik.uni-stuttgart.de (Bernd Holzmueller) writes:


>Does anybody know of a concrete example of an LALR(1)-conflict in an existing
>(or hypothetical but semantically meaningful) programming language grammar
>which is *exactly* LALR(1), i.e. the conflict is solved by moving to LR(1)?


Here's a real example I came across. It's from the "Arden Syntax",
the formal specification (ASTM E1460) of which I am currently editing:


expr_function ::=
expr_factor
| of_func_op opt_OF expr_function
| from_of_func_op index_modifier opt_OF expr_function
| from_of_func_op index_modifier expr_factor FROM expr_function
| from_func_op index_modifier expr_factor FROM expr_function


opt_OF ::= OF | /* empty */


YACC (and BISON) didn't like this; I had to replace the 3rd rule by 2
lines (one with OF and one without). Note that the 2nd rule still
has opt_OF:


expr_function ::=
expr_factor
| of_func_op opt_OF expr_function
| from_of_func_op index_modifier expr_function
| from_of_func_op index_modifier OF expr_function
| from_of_func_op index_modifier expr_factor FROM expr_function
| from_func_op index_modifier expr_factor FROM expr_function


--
Peter Ludemann ludemann@netcom.com
--

Return to the comp.compilers page.
Search the comp.compilers archives again.