Re: LL(1) vs LALR(1) parsers

[This is a corrected version of 95-11-122 -John]


Saileshwar Krishnamurthy (krish@cs.purdue.edu) wrote:
: They can be parsed perhaps, but the final LL(1) grammar that
: corresponds to the language will probably not be so intuitively
: obvious.


Just thought I'd throw in my 2 cents on LL(1). If I'm building
a recognizer only, LALR(1) clearly wins. The decision is not
so clear when you talk about the debugging, adding actions, etc...
LL(k>1) is actually a nice solution [LALR(k>1) would be nice too].
The syntactic predicates of PCCTS allow arbitrary lookahead
through selective backtracking. Kinda nice. The famous C++
expressions vs declaration ambiguity can be solved via:


stat : (decl)? // if it looks like a declaration, it is
          | expr // else it's an expression.
          ;


Note that we are smart about this: "a=3;" immediately jumps to
the expr alternative as it's obviously not a declaration.


Please see my upcoming paper with Russell Quong in SIGPLAN Notices
(Jan or Feb 96) called: ``LL and LR Translators Need k>1 Lookahead''.
PS can be had of it now at


ftp://ftp.parr-research.com/pub/pccts/papers/needlook.ps


Best regards,
Terence Parr
http://www.parr-research.com/~parrt/
--

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