Re: Bottom-up versus Top-down

I'd think with the advent of predicated LL(k) (used by PCCTS/ANTLR)
the "power of LALR" advantage disappears. In many areas, LL(k) has a
distinct advantage, especially when it comes to error repair and
recovery due to its "knowing where I've been" context.


I've yet to see a _real_ programming problem that could be solved with
LALR but could not be solved with LL(k).


To go a step further, there are even some languages that are more
difficult to implement in LALR (C++...)




The trick is to cast the mindset from the "LALR" way of writing a
grammar to the "LL(k)" way of writing a grammar. Yes, you have to
deal with left recursion and left-factoring, but it's not _that_ big
of a deal (see http://www.scruz.net/~thetick/lalrtoll.html)


The bigger problem seems to be along the lines of "BetaMax vs VHS" --
VHS had more software; there are more readily-available LALR
grammars... You usually get stuck converting LALR grammars to LL(k)


(Don't tell Frank DeRemer & Tom Pennello I just compared LALR to
VHS... I think I'm still friends with them ;)


-- Scott


=============================================
Scott Stanchfield - http://www.scruz.net/~thetick
MageLang Institute - http://www.magelang.com










--

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