Version française
Home     About     Download     Resources     Contact us    
Browse thread
[camlp4 extension] pa_refutable : request for comments
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: -- (:)
From: Nicolas Pouillard <nicolas.pouillard@g...>
Subject: Re: [Caml-list] [camlp4 extension] pa_refutable : request for comments
Excerpts from bluestorm.dylc's message of Tue Dec 25 18:36:27 +0100 2007:
> Hi,
Hi,

> I just created my first "serious" camlp4 (3.10) extension. I'm looking
> for comments, and have some questions too.
> 
> The extension (to the classical syntax) enables an explicit use of
> non-irrefutable pattern matching in "let" declarations :
> let refutable hd::tl = ...
> let refutable [a; b; c] = List.map foo ['a'; 'b'; 'c']
> let rec refutable func (Some thing) = ...

Looks useful indeed.

> The code is available here : http://bluestorm.info/tmp/pa_refutable.ml.html
> (side question : are 80 lines of code short enough to be included in
> my message ? what is the mailing-list recommended behaviour here ?)

I think that's short enough, but colored html page is perhaps more readable.

> The actual error-reporting is rather naive : the extension generate a
> string containing the location of the refutable pattern in the
> original source, wich is raised at runtime using "failwith".. Is there
> a more elegant way to do that ?

Seems sufficient.

> There are 6 repetitive lines of code in my refutable_mono function :
>     let binds_patt binds _loc =
>       let patt_of_id (id, _loc) = <:patt< $lid:id$ >> in
>       Ast.PaTup _loc (Ast.paCom_of_list (List.map patt_of_id binds)) in
>     let binds_expr binds _loc =
>       let expr_of_id (id, _loc) = <:expr< $lid:id$ >> in
>       Ast.ExTup _loc (Ast.exCom_of_list (List.map expr_of_id binds)) in
> Is there an better way to do this ?

In  fact  I  think this part is wrong. What's happen if you get no variable or
just   one   in  the  pattern?  You  will  try  to  build  a  nil-tuple  or  a
singleton-tuple and that's forbidden.

Here is the untested code for patterns:

let binds_patt binds _loc =
  let patt_of_id (id, _loc) = <:patt< $lid:id$ >> in
  match binds with
  | [] -> <:patt< () >>
  | [c] -> patt_of_id c
  | c :: cs -> <:patt< ( $patt_of_id c$, $list:List.map patt_of_id cs$ ) >>
in

The  cool  thing is that by making explicit the tuple (by having more than one
element)  one avoid resorting to concrete constructors and can use the $list:$
special antiquotation that will insert the paCom_of_list call.

> Lastly, i had to duplicate work in the grammar-modification part,
> because i didn't manage to use the "opt_rec" rule. I'm not aware of
> the subtleties of camlp4 parsing; is there another way ?

I don't think there is any subtlety here, just replace:

| "let"; "rec"; "refutable"; bi = binding ->
    <:str_item< value rec $refutable bi$ >>
| "let"; "refutable"; bi = binding ->
    <:str_item< value $refutable bi$ >>

by something like:

| "let"; r = opt_rec; "refutable"; bi = binding ->
    <:str_item< value $rec:r$ $refutable bi$ >>

Best regards,
-- 
Nicolas Pouillard aka Ertai