skaller writes:
 > On Thu, 2005-01-27 at 21:20, Jean-Christophe Filliatre wrote:
 > > 
 > > What's wrong with this way of handling nested comments with ocamllex:
 > > 
 > > ======================================================================
 > > 
 > > | "(*" { comment lexbuf; ... }
 > > ...
 > > 
 > > and comment = parse
 > > | "(*" { comment lexbuf; comment lexbuf }
 > > | "*)" { () }
 > > | _    { comment lexbuf }
 > > 
 > 
 > Well it doesn't handle (* or *) in strings ..

I was only focusing on nested comments, but handling strings is rather
trivial.  See the ocaml  lexer for  instance (where  you can  see that
introducing a  new lexing function "string" is  definitely better than
trying to write a regular expression for strings literals).

 > However, whilst this code (given a fix
 > to handle strings) would probably work,
 > it isn't a plain lexer, but a mix of lexing
 > and client code. 

Sure,  it is  clearly more  powerful than  an automaton  (since nested
comments are not regular), if this  is what you mean by "isn't a plain
lexer".

But that's precisely why ocamllex is so a powerful tool. You only need
to  know  that  ocamllex  is  building a  set  of  mutually  recusvive
functions with the lexbuf as argument  and then you are not limited in
what you can do in the actions. You can even pass additional arguments
to the lexing functions. 

I like to  think about ocamllex as a general-purpose  tool to insert a
bit  of  regular  expressions  in  ocaml programs  (lexers,  but  also
filters, file  formats readers,  line counters, code  indenters, etc.)
and not only  as a tool to write lexers. With  the header and trailer,
it  is even  easy  to build  a  whole ocaml  program  within a  single
ocamllex file.  For  instance, two programs of mine  I use intensively
are a program to count lines  of code and comment in my ocaml programs
(a 173 lines  long ocamllex file) and a preprocessor  for my web pages
(a 129 lines long ocamllex file).

-- 
Jean-Christophe Filliâtre (http://www.lri.fr/~filliatr)