Hello,

On 01-07-2009, Andreas Rossberg <rossberg@mpi-sws.org> wrote:
> Mike Lin wrote:
>> OK, now I'm curious :) how does your lexer match balanced parentheses,
>> or in this case comments?
>>   
>
> Easily, with a bit of side effects (I think that's roughly how all ML 
> compilers do it):
>
> ------------------------------------------------
> let error l s = (* ... *)
> let commentDepth = ref 0
> let start = ref 0
> let loc length = let pos = !start in (pos, pos+length)
>
> rule lex =
>     parse eof            { EOF }
>     (* | ... *)
>     | "{-"            { start := pos lexbuf;
>                   lexNestComment lexbuf }
>
> and lexNestComment =
>     parse eof            { error (loc 2) "unterminated comment" }
>     | "(*"            { incr commentDepth;
>                   lexNestComment lexbuf }
>     | "*)"            { decr commentDepth;
>                   if !commentDepth > 0
>                   then lexNestComment lexbuf
>                   else lex lexbuf }
>     | _            { lexNestComment lexbuf }
> ------------------------------------------------
>
> If you also want to treat strings in comments specially (like OCaml), 
> then you need to do a bit more work, but it's basically the same idea.
>

May I recommend you to write this in a more simple way:

-------------------------------------------------------------------------
rule lex =
  parse eof    { () }
  | "(*"       { start := pos lexbuf; lexNestComment lexbuf; lex lexbuf }

and lexNestComment =
  parse eof    { error (loc 2) "unterminated comment" }
| "(*"         { lexNestComment lexbuf }
| "*)"         { () }
| _            { lexNestComment lexbuf }
-------------------------------------------------------------------------

I think it works the same way, except that it uses less global
variables.

Regards,
Sylvain Le Gall