[antlr-interest] Re: full LL(K) vs linear approximation?
lgcraymer
lgc at mail1.jpl.nasa.gov
Tue Mar 18 10:36:44 PST 2003
"Distraction" may be a better description than "laziness", as I'm
sure Monty will agree: ANTLR 2 was implemented hurriedly, and Ter was
immediately thereafter very actively involved in getting first
MageLang and then jGuru going. There are a few PCCTS features--like
full LL(k) and semantic predicate hoisting--that didn't make it into
ANTLR 2. Ter has "promised" to make sure that they make it into ANTLR
3.
--Loring
--- In antlr-interest at yahoogroups.com, mzukowski at y... wrote:
> It also takes longer to do the analysis for LL(K) than for LALL(K).
The
> main reason it isn't done is because of laziness; LALL(K) works most
of the
> time. The ideal is to detect those situations where LALL(K) fails
and then
> do full LL(K) analysis. This has some pretty big repercussions on
the
> current implementation of the LALL(K) analysis algorithm, which
needs to be
> reworked anyhow to accomodate hoisting of semantic predicates.
>
> Speed is also a major issue in the generated parser. The hybrid
approach
> will be the best win. Generating LL(K) decision code is
straightforward.
>
> Monty
>
> -----Original Message-----
> From: Albert Huh [mailto:albert.huh at e...]
> Sent: Monday, March 17, 2003 11:03 AM
> To: antlr-interest at yahoogroups.com
> Subject: [antlr-interest] full LL(K) vs linear approximation?
>
>
> does anybody know exactly why full LL(k) isn't done in antlr?
>
> is it due to speed with larger k-values? (each decision will take
k^2
> comparisions as opposed to just k comparisions?)
>
> is it harder to generate full LL(k) analysis code?
>
>
>
> Your use of Yahoo! Groups is subject to
http://docs.yahoo.com/info/terms/
Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
More information about the antlr-interest
mailing list