camlp4 3.10 and ints

camlp4 3.10 and ints

[Richard Jones]

I'm getting some odd errors with camlp4 (3.10) and integers.

Firstly, INT returns a string, rather than an int, and similarly
$int:...$ expects a string instead of an int.  I can live with this,
but it seems a bit strange, and unless I'm mistaken is different from
what camlp4 <= 3.09 did.

Secondly, I get the error 'Failure: "Integer literal exceeds the range
of representable integers of type int"' if I try to pass a string
containing any int32 or int64 literal, eg. $int:"0x3ffL"$, even ones
which obviously do not exceed the range of representable integers.

To demonstrate this second one, save the following to files:

-------------------------------------------------- pa_test.ml ---
open Camlp4.PreCast
open Syntax

let expand_test _loc () =
  <:str_item< value test = $int:"0x3ffL"$ >>

EXTEND Gram
  GLOBAL: str_item;

  str_item: LEVEL "top" [
  [ "TEST" -> expand_test _loc () ]
  ];
END;;
--------------------------------------------------

-------------------------------------------------- test.ml ---
TEST
--------------------------------------------------

And then compile with:

  $ ocamlc -pp "camlp4o pa_extend.cmo q_MLast.cmo" -I +camlp4 -c pa_test.ml 
  $ ocamlc -pp "camlp4o ./pa_test.cmo" test.ml
  File "test.ml", line 1, characters 0-4:
  Failure: "Integer literal exceeds the range of representable integers of type int"
  Preprocessor error

Are these bugs or am I missing something?

Rich.

-- 
Richard Jones
Red Hat

Re: [Caml-list] camlp4 3.10 and ints

[Jeremy Yallop]

Richard Jones wrote:
> $int:...$ expects a string instead of an int.  I can live with this,
> but it seems a bit strange, and unless I'm mistaken is different from
> what camlp4 <= 3.09 did.

The old camlp4 did this as well.  The new one has an extra feature, 
though: there's an antiquotation

    $`int:...$

which expects an int.

> Secondly, I get the error 'Failure: "Integer literal exceeds the range
> of representable integers of type int"' if I try to pass a string
> containing any int32 or int64 literal, eg. $int:"0x3ffL"$, even ones
> which obviously do not exceed the range of representable integers.

For int32 and int64 there are antiquotations $int32:...$, $`int3:...$, etc.

Jeremy.

Re: [Caml-list] camlp4 3.10 and ints

[Nicolas Pouillard]

On 6/13/07, Richard Jones <rich@annexia.org> wrote:
> I'm getting some odd errors with camlp4 (3.10) and integers.

Thanks to Jeremy Yallop for responding to this. I then just add a few bits...

>
> Firstly, INT returns a string, rather than an int, and similarly
> $int:...$ expects a string instead of an int.  I can live with this,
> but it seems a bit strange, and unless I'm mistaken is different from
> what camlp4 <= 3.09 did.

Not it's not different and the reason is simple. Camlp4 wants to keep
the representation of your integer and turning it to int will
normalize it ("42", "4_2", "0000042", "0b0101010"...).

With 3.10 you also have active antiquotations $`int:...$ $`int32:...$ for that.

Conversely on the parser side you can also get integers instead of
string by using real patterns in your production rule.

old_prod:
[[ s = INT -> (* s is a string *) ... ]];

new_prod:
[[ `INT(i, s) -> (* i is an integer and s its representation *) ... ]];

-- 
Nicolas Pouillard