Re: regular LALR parsers

simonh@swidev.demon.co.uk writes:
>How does an LALR parser with regular expressions (eg. *, +, []) handle the
>attribute stack? Obviously, the attributes on the stack depend on the number
>of iterations of constructs such as *, +.


I take closure to mean a list of attributes and optional to mean an
attribute or null. It's pretty easy in my application since I'm
generating Icon code, which has built-in list and null types.


For the implementation, I avoid all your questions about shift
conflicts by rewriting all the extended BNF to ordinary
BNF. For example, I rewrite closure `(a b c {semantics})*' to


    foo: { $$ = [] } # empty list
          | foo foo_1 { $$ = put($1, $2) } # append element to list
          ;
    foo_1 : a b c {semantics};


When I get around to generating C code, I'll probably define a
simple list type using cons cells.
--

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