On Tue, Jul 27, 2004 at 11:07:50PM +0100, Jon Harrop wrote:
> 
> > the unasked question here is "Is this the right way to go?" I mean, is
> > Camlp4 the tool I should be using?  I think it is, because the top down,
> > recursive descent parsing is what Spirit does, but maybe I've just confused
> > things.
> 
> I am no expert on parser technology but I've written a few interpreters and 
> compilers in OCaml now and it is an excellent language for this kind of 
> stuff.

that's what I've read, but I've yet to be able to do work myself.

>Coming from a C++ background I can also say that C++ would be an awful 
> language for this and, as for template metaprogramming, well... :-)

Spirit is great.  The template metaprograming, from a user-developers
standpoint isn't so bad.  But it is one of the reasons I'd like to move 
all of my developement to Ocaml.  

> I've only used yacc, so only LALR(1) grammars. I'm not sure if this is 
> formally correct, but in a lot of cases you can get extended functionality by 
> parsing what you can and then manipulating the resulting data structures. For 
> example, in my latest compiler I need a parser which can understand not only 
> "a<b" but also "a<b<c" and so on. You can't use the same approach as for 
> addition ("a+b+c" = "(a+b)+c") so I parse into an intermediate representation 
> "Inequality[a, Less, b, Less, c]" which I postprocess into the equivalent of 
> "a<b<c".
> 
> You may find that, with similar tricks, you can parse what you need using 
> yacc.

this may be exactly what I need to do.  You don't have to do this in Spirit.
Not that that changes anything, I'm just noting it.

> Do you have a specification for the grammar? Can you give examples of input 
> and the corresponding data structure you would like?

Kinda.  I don't have an EBNF for it, but I do have input and desired output
for one of the sets (this is the simplest, but it's indicitive of
what is ussually there).

Input:
a23fassssadfj4532|10,2;13,3;20;20|10|20|30|B

Desired output (in Oaml'ish):

type origination = B | S | T | A;;
type pattern = {score:int;weight:int};;

type item = {id:string;patterns:pattern
list;max:int;min:int;time:int;origin:origination }

So, a23fassssadfj4532|10,2;13,3;20;20|10|20|30|B

would be 

id:a23fassssadfj4532
patterns:[{10,2};{13,3};{20,20}]
max:10
min:20
time:30
origin:B

It has also been noted that I could just use a combination of splits to
accomplish this, but I've found that approch to be fragile and slow.  

I've also got some ocamlyacc/ocamllex code that works (it parses a subset of
the desired input), but doesn't do what I want yet (it's mostly just
cut/paste from the manual).  I've that at http://www.bad-poetry.net/ocaml/
if anyone is interested.

Thanks for the help.

-jbs

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners