[Caml-list] Alpha release of DeCaP (combinator library + extensible OCaml parser)

[Caml-list] Alpha release of DeCaP (combinator library + extensible OCaml parser)

[Christophe Raffalli]

[-- Attachment #1: Type: text/plain, Size: 3161 bytes --]

Dear Caml-list,

We are proud to announce the release of a new parser combinator library called
DeCaP. It has been used to implement a new extensible parser for OCaml called
pa_ocaml, together with a syntax extension for writing parsers using a format
similar to Backus-Naur Form (BNF). Through this syntax extension, parsers are
compiled down to calls of DeCaP combinators. In this way, we manage to obtain
a user-friendly syntax for writing parsers, while preserving the advantages of
functional combinators. To give you an idea, we give bellow the example of a
small calculator.

## calc_prio.ml ##############################################################
open Decap

type calc_prio = Sum | Prod | Atom
let expr, set_expr = grammar_family "expr" 

let float_num =
  let float_re = "[0-9]+\\([.][0-9]+\\)?\\([eE][-+]?[0-9]+\\)?" in
  parser
  | f:RE(float_re) -> float_of_string f

let prod_sym =
  parser
  | CHR('*') -> ( *. )
  | CHR('/') -> ( /. )

let sum_sym =
  parser
  | CHR('+') -> ( +. )
  | CHR('-') -> ( -. )

let _ = set_expr (fun prio ->
  parser
  | f:float_num                    when prio = Atom -> f
  | CHR('(') e:(expr Sum) CHR(')') when prio = Atom -> e
  | CHR('-') e:(expr Atom)         when prio = Atom -> -. e
  | CHR('+') e:(expr Atom)         when prio = Atom -> e
  | e:(expr Atom) l:{fn:prod_sym e':(expr Atom)}*
                                   when prio = Prod ->
      List.fold_left (fun acc (fn, e') -> fn acc e') e l
  | e:(expr Prod) l:{fn:sum_sym  e':(expr Prod)}*
                                   when prio = Sum  ->
      List.fold_left (fun acc (fn, e') -> fn acc e') e l)

(* The main loop *)
let _ =
  let blank = blank_regexp (Str.regexp "[ \t]*") in
  try while true do
    Printf.printf ">> %!";
    let l = input_line stdin in
    let r = handle_exception (parse_string (expr Sum) blank) l in
    Printf.printf "%f\n%!" r
  done with End_of_file -> ()
##############################################################################

The tools to write syntax extensions are similar to those of Camlp4 (quotation
and anti-quotation) but are not complete yet (quotation patterns are missing).
The corresponding part of the documentation is also unfinished.

On the side of performances, pa_ocaml is on average five times slower than the
original OCaml parser (implemented using Ocamlyacc) and two times faster than
Camlp4. The web-page of the project can be found here:

http://lama.univ-savoie.fr/decap/

As DeCaP and pa_ocaml are still under development, we would greatly
appreciate feedback. We are particularly interested in bug in the
OCaml parser. At the moment, the grammar is supposed to be complete
from OCaml 3.12.1 to 4.01.0 while some of the extension of 4.02.0 are
missing. Our target is to produce identical parse trees compare to
those of the original ocamlyacc parser. On our test files, this is
achieved up to positions and we are working on positions currently.

Bug report on mantis: http://lama.univ-savoie.fr/mantis


Rodolphe Lepigre & Christophe Raffalli
LAMA, UMR 5127 CNRS - Université Savoie Mont Blanc

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]