possible to define a type where = is forbidden ?

possible to define a type where = is forbidden ?

[yoann padioleau]

I am making a program analysis tool  and in my AST I had previously tokens represented as strings,
but now I want also to associate the line numbers of those strings.

For instance I had

type expr = 
 Plus of expr * expr
 | ...
 | Constant of string
 | Int of int
and I want to pass to 

type expr = 
 Plus of expr * expr
 | ...
 | Constant of  string extended
 | Int of int extended

where type 'a extended = ('a * int)

The problem is that I have in many places some code such as   x = y   that works well
when the AST contain only the string of the token, but now that there is also the line,  the x = y  is not the good one.
I would like a  '= modulo I dont care about line'   such as for example  (Constant "toto",45)  =modulo_line= (Constant "toto", 61)  be true.

There is even some places where I use List.mem  that dont work too.

Is it possible to define a type  such as '=' is forbidden,  so that the compiler and the type system will
help me to spot all the occurences of = (and List.mem, ...)  that I have to change.

For the moment my partial solution  is to use functions to represented the extended type, 
because '=' fail at the execution on functionnal value.
I want to do: 

type line_number = int
type 'a extended = Abstract of 'a | Concrete of (unit -> ('a * line_number))

where Constant "toto"   becomes now  (Constant (Concrete (fun () -> "toto", 45))

The problem is that it dectects bad use of = only at run-time. I would prefer a solution that detects the "incrorrect" use
of = at compile time.



PS: if only caml had typeclasses :)

Re: [Caml-list] possible to define a type where = is forbidden ?

[Christian Lindig]

On Oct 10, 2005, at 2:32 PM, yoann padioleau wrote:

> I am making a program analysis tool  and in my AST I had previously  
> tokens represented as strings,
> but now I want also to associate the line numbers of those strings.

I like to suggest a different technique that does not require to change  
the representation of all branches in an AST. You can find it in an  
earlier posting here:

http://caml.inria.fr/pub/ml-archives/caml-list/2003/09/ 
f81c8063ed4878e06f1ddd8010256050.en.html

The posting is part of a thread that also discusses some disadvantages.

-- Christian

Re: [Caml-list] possible to define a type where = is forbidden ?

[yoann padioleau]

> On Oct 10, 2005, at 2:32 PM, yoann padioleau wrote:
> 
> > I am making a program analysis tool  and in my AST I had previously  
> > tokens represented as strings,
> > but now I want also to associate the line numbers of those strings.
> 
> I like to suggest a different technique that does not require to change  
> the representation of all branches in an AST. You can find it in an  
> earlier posting here:
> 
> http://caml.inria.fr/pub/ml-archives/caml-list/2003/09/ 
> f81c8063ed4878e06f1ddd8010256050.en.html

Interesting technique :)  but I think that it would not solve my problem. 

To be more precise my program analysis tool  try to compute "diff" between differents AST.
So we love to use pattern-patching and to (ab)use the polymorphic  =  to compare stuff.

I don't see how your ExprAt technique will make this easier (not to mention that it also makes 
the pattern matching ugly because I would have to wrap every  expression  with his ExprAt  because I want
that the resultting  Ast_diff  contain also line number.

Re: [Caml-list] possible to define a type where = is forbidden ?

[Thomas Fischbacher]

On Mon, 10 Oct 2005, yoann padioleau wrote:

> > http://caml.inria.fr/pub/ml-archives/caml-list/2003/09/ 
> > f81c8063ed4878e06f1ddd8010256050.en.html
> 
> Interesting technique :)  but I think that it would not solve my problem. 
> 
> To be more precise my program analysis tool  try to compute "diff" between differents AST.
> So we love to use pattern-patching and to (ab)use the polymorphic  =  to compare stuff.
> 
> I don't see how your ExprAt technique will make this easier (not to mention that it also makes 
> the pattern matching ugly because I would have to wrap every  expression  with his ExprAt  because I want
> that the resultting  Ast_diff  contain also line number.

When you want to associate extra data to stuff that should retain nice 
comparison properties, another technique which might be useful or not is 
to use a weak pointer hash, mapping subtrees to positions.

I would not go so far as to say that this is the favored approach one 
should use, but sometimes this idea may be useful. At least, it's nice to 
have that trick available.

-- 
regards,               tf@cip.physik.uni-muenchen.de              (o_
 Thomas Fischbacher -  http://www.cip.physik.uni-muenchen.de/~tf  //\
(lambda (n) ((lambda (p q r) (p p q r)) (lambda (g x y)           V_/_
(if (= x 0) y (g g (- x 1) (* x y)))) n 1))                  (Debian GNU)