[antlr-interest] Lemon parser. Any comments?

[Ruslan Zasukhin]

On 12/19/05 5:41 PM, "Paul Johnson" <gt54-antlr at cyconix.com> wrote:

Hi Paul,

>> * no $n values, instead easy named parameters
> 
> I've never really seen this as a problem; it seems natural enough in
> normal use.
> 
>> * no global variables, so can be used in threads
> 
> '%pure-parser' will give you a re-entrant parser; all the globals are
> turned into locals. I haven't tried this.

Do you mean option of YACC ?
 
>> * can run several parsers in the same time.
> 
> I already do this, and I know of others doing the same. You can do this
> even without the re-entrant option by changing the prefix on the global
> names (-p option on the command line??)

Again, I assume you talk about YACC, not ANTLR, right ?
 
>> * faster of YACC (claimed)
> 
> No idea, I'm afraid. However, parsing is a small fraction of my total
> compilation time.

You are lucky 

I have spend today again the whole day doing different benches and
profiling.  

In loop with INSERT of million records, we get ANTLR pure time 470 seconds,
Our db engine do all rest work on low level in 58 second.

SqlLite with that Lemon parser do all in 132 seconds.

So we can allow parser to eat maximum (132 - 58) = 70 seconds.
But not 470 seconds.
 
>> * also they have auto-destructors for terminal symbols.
> 
> Not sure that I understand this. However, I've had no memory leak problems.
> 
> Note also that Bison has had multi-token lookahead for several years now
> (it uses GLR).

Aha. Ok. Thank you.

-- 
Best regards,

Ruslan Zasukhin
VP Engineering and New Technology
Paradigma Software, Inc

Valentina - Joining Worlds of Information
http://www.paradigmasoft.com

[I feel the need: the need for speed]