sequence parser, revised syntax

sequence parser, revised syntax

[dsb]

I have a noobie question about parsers.

The output of
  Grammar.Entry.print Pcaml.expr;;

contains several things that appear to be parsers
(this example in the revised syntax):

  | "let"; "module"; UIDENT; module_binding; "in"; SELF
  | "let"; OPT "rec"; LIST1 let_binding SEP "and"; "in"; SELF
  | "fun"; "["; LIST0 match_case SEP "|"; "]"
  | "fun"; ipatt; fun_def
  | "while"; SELF; "do"; "{"; sequence; "}"
  | "while"; SELF; "do"; LIST0 [ expr; ";" ]; warning_sequence; "done"
  | "object"; OPT class_self_patt; class_structure; "end" ]

but don't seem to be documented.  My current interest happens to be
sequence:

  open Pcaml;
  EXTEND
    expr: LEVEL "top" [
  [ "when"; test = expr; "do"; "{"; seq = sequence; "}"
    ->
      <:expr< if $test$ then ($list:seq$) else () >>
  ]];
  END;

but it doesn't seem to work:

  $ make stdlibr.cmo
  ocamlc -pp "camlp4r pa_extend.cmo q_MLast.cmo" \
          -I +camlp4 -c stdlibr.ml
  File "stdlibr.ml", line 20, characters 40-48:
  Unbound value Pcaml.sequence
  make: *** [stdlibr.cmo] Error 2

How are these parsers supposed to be used?

Thanks.
--Dan Bensen
www.prairienet.org/~dsb/

Re: [Caml-list] sequence parser, revised syntax

[Daniel de Rauglaudre]

Hi,

On Mon, Jul 09, 2007 at 09:22:32AM -0500, dsb@prairienet.org wrote:

>   $ make stdlibr.cmo
>   ocamlc -pp "camlp4r pa_extend.cmo q_MLast.cmo" \
>           -I +camlp4 -c stdlibr.ml
>   File "stdlibr.ml", line 20, characters 40-48:
>   Unbound value Pcaml.sequence
>   make: *** [stdlibr.cmo] Error 2

Some of the grammar sub-entries used in the files defining the syntax
(e.g. "sequence") are indeed not accessible in the Pcaml module.

The problem is that there are many of them, many small entries useful
only in specifics parts of the grammar.

These small entries may change, appear, disappear, from a version of
camlp4 to another when new things are added in the language syntax, or
to fix possible bugs. It is often difficult to "standardize" them.

Moreover, some of these entries may be used in revised syntax and not
in normal syntax and vice-versa, or have a different type and semantics.

The solution is to rewrite, yourself, the missing entry in your sources.
You can see how "sequence" is implemented in the revised syntax and copy
it (it is short).

Perhaps the function Grammar.Entry.print should show differently the
"accessible" entries and the "non accessible" ones. I am going to see
if it is possible to implement that.

-- 
Daniel de Rauglaudre
http://pauillac.inria.fr/~ddr/

Re: [Caml-list] sequence parser, revised syntax

[Nicolas Pouillard]

On 7/10/07, Daniel de Rauglaudre <daniel.de_rauglaudre@inria.fr> wrote:
> Hi,

Hi,

>
> On Mon, Jul 09, 2007 at 09:22:32AM -0500, dsb@prairienet.org wrote:
>
> >   $ make stdlibr.cmo
> >   ocamlc -pp "camlp4r pa_extend.cmo q_MLast.cmo" \
> >           -I +camlp4 -c stdlibr.ml
> >   File "stdlibr.ml", line 20, characters 40-48:
> >   Unbound value Pcaml.sequence
> >   make: *** [stdlibr.cmo] Error 2

FYI in ocaml 3.10...

> Some of the grammar sub-entries used in the files defining the syntax
> (e.g. "sequence") are indeed not accessible in the Pcaml module.

In 3.10 sequence is exported (so extensible).

> The problem is that there are many of them, many small entries useful
> only in specifics parts of the grammar.

In 3.10 most of them are exported.

> These small entries may change, appear, disappear, from a version of
> camlp4 to another when new things are added in the language syntax, or
> to fix possible bugs. It is often difficult to "standardize" them.

Even if it's difficult we tend to reach a stability point and then
prefer expose these things.

> Moreover, some of these entries may be used in revised syntax and not
> in normal syntax and vice-versa, or have a different type and semantics.

In 3.10 both syntaxes share a common grammar (the revised one),
therefore there are closer in types and semantics.

Cheers,

-- 
Nicolas Pouillard