Size of produced parser: menhir vs ocamlyacc

Size of produced parser: menhir vs ocamlyacc

[Joel Reymont]

_build/easy_parser.cmo -- ocamlyacc: 98,339,  menhir: 1,542,108
_build/easy_parser.ml  -- ocamlyacc: 323,729, menhir: 3,550,332

Compiling a parser with produced by ocamlyacc also takes a fraction  
of a second vs several seconds with menhir.

I hope parsers produced by menhir is just as fast as those produce by  
ocamlyacc during execution!


--
http://wagerlabs.com/

Re: [Caml-list] Size of produced parser: menhir vs ocamlyacc

[Francois Pottier]

Hello,

On Fri, Mar 30, 2007 at 02:14:53PM +0100, Joel Reymont wrote:
> _build/easy_parser.cmo -- ocamlyacc: 98,339,  menhir: 1,542,108
> _build/easy_parser.ml  -- ocamlyacc: 323,729, menhir: 3,550,332
> 
> Compiling a parser with produced by ocamlyacc also takes a fraction  
> of a second vs several seconds with menhir.

This is because we compile automata down to code, whereas ocamlyacc compiles
them down to tables. Ideally, we should provide an option to compile down to
tables instead of code, but we never found time to do so.

> I hope parsers produced by menhir is just as fast as those produce by  
> ocamlyacc during execution!

We did not perform any serious measurements, but we believe that we are
competitive. If you run a comparison, we would be interested in knowing
about the results.

-- 
François Pottier
Francois.Pottier@inria.fr
http://cristal.inria.fr/~fpottier/

Re: [Caml-list] Size of produced parser: menhir vs ocamlyacc

[Markus Mottl]

On 3/30/07, Francois Pottier <Francois.Pottier@inria.fr> wrote:
> > I hope parsers produced by menhir is just as fast as those produce by
> > ocamlyacc during execution!
>
> We did not perform any serious measurements, but we believe that we are
> competitive. If you run a comparison, we would be interested in knowing
> about the results.

I also haven't done any serious benchmarks, but I remember seeing
menhir beat ocamlyacc on some of my parsers.

Regards,
Markus

-- 
Markus Mottl        http://www.ocaml.info        markus.mottl@gmail.com

Re: [Caml-list] Size of produced parser: menhir vs ocamlyacc

[Joel Reymont]

On Mar 30, 2007, at 2:45 PM, Markus Mottl wrote:

> I also haven't done any serious benchmarks, but I remember seeing
> menhir beat ocamlyacc on some of my parsers.

How complex were they?

--
http://wagerlabs.com/

Re: [Caml-list] Size of produced parser: menhir vs ocamlyacc

[Markus Mottl]

On 3/30/07, Joel Reymont <joelr1@gmail.com> wrote:
> How complex were they?

Not very.  S-expressions, etc.  But I wouldn't be surprised if Menhir
performs better on more complex parsers, too.  Generating parser code
rather than tables may boost performance.

Regards,
Markus

-- 
Markus Mottl        http://www.ocaml.info        markus.mottl@gmail.com

Re: [Caml-list] Size of produced parser: menhir vs ocamlyacc

[Joel Reymont]

On Mar 30, 2007, at 3:26 PM, Markus Mottl wrote:

> Not very.  S-expressions, etc.  But I wouldn't be surprised if Menhir
> performs better on more complex parsers, too.  Generating parser code
> rather than tables may boost performance.

Why did you pick menhir over ocamlyacc? I assume it was to use it in  
server code but thought I would ask anyway.

	Thanks, Joel

--
http://wagerlabs.com/

Re: [Caml-list] Size of produced parser: menhir vs ocamlyacc

[skaller]

On Fri, 2007-03-30 at 15:28 +0100, Joel Reymont wrote:
> On Mar 30, 2007, at 3:26 PM, Markus Mottl wrote:
> 
> > Not very.  S-expressions, etc.  But I wouldn't be surprised if Menhir
> > performs better on more complex parsers, too.  Generating parser code
> > rather than tables may boost performance.
> 
> Why did you pick menhir over ocamlyacc? I assume it was to use it in  
> server code but thought I would ask anyway.

Note Ocamlyacc is LALR(1) whilst Menhir is LR(1).

-- 
John Skaller <skaller at users dot sf dot net>
Felix, successor to C++: http://felix.sf.net

Re: [Caml-list] Size of produced parser: menhir vs ocamlyacc

[Markus Mottl]

On 3/30/07, Joel Reymont <joelr1@gmail.com> wrote:
> Why did you pick menhir over ocamlyacc? I assume it was to use it in
> server code but thought I would ask anyway.

I just wanted to try it out.  I still use ocamlyacc by default, and
have actually implemented a hand-written parser for optimum
performance and flexibility.

Regards,
Markus

-- 
Markus Mottl        http://www.ocaml.info        markus.mottl@gmail.com