Parrot lacks a boolean type

[lucian]

And instead, boolean contexts fall back to expecting ints. This is not semantically correct, and leads to issues, especially when trying to expose a HLL's object system to the rest of parrot.

In my specific case, I implemented get_bool on Python objects, but Rosella's is_true and winxed's if coerce to int. Rosella and Winxed are doing the least wrong thing possible with that they've got.

Implementing get_integer to return 0 for false, 1 for true and the value for Python ints is simply incorrect. At least in python 0 is falsy, which makes it possible to have this workaround, in other languages it would be required to simply not implement get_bool or get_integer, to not break things.

The workaround in Winxed is the ternary operator (bla ? 1 : 0), which ends up calling bla.get_bool.

[pmichaud]

On Tue, Aug 16, 2011 at 12:36:55PM -0000, Parrot wrote:
> #2178: Parrot lacks a boolean type
>  And instead, boolean contexts fall back to expecting ints. This is not
>  semantically correct, and leads to issues, especially when trying to
>  expose a HLL's object system to the rest of parrot.

I don't understand this ticket.  Parrot *has* a boolean type, both in 
the "Boolean" data type, and in the "get_bool" vtable function (which 
is separate from "get_integer").  I'm not aware of Parrot requiring
"boolean contexts fall back to expecting ints", nor does it do so in
many cases.

It's true that the get_bool vtable expects to return an integer
indicating true/false, but it's still a very separate vtable from
get_integer.  Also, most (if not all) of the boolean opcodes use get_bool
and not get_integer:

    if $P0, label
    unless $P0, label
    and $P0, $P1, $P2
    or $P0, $P1, $P2
    not $P0, $P1
    istrue $I0, $P0
    isfalse $I0, $P0

That said, it wouldn't bug me if we had a get_bool_pmc vtable to
go along with get_bool.

>  In my specific case, I implemented get_bool on Python objects, but
>  Rosella's is_true and winxed's if coerce to int. 

That sounds like a HLL/library issue to me, not a Parrot one.

>  Rosella and Winxed are
>  doing the least wrong thing possible with that they've got.

Note that none of NQP, Rakudo, nor PCT perform casting to ints in
order to provide boolean context.

Pm

[NotFound]

One problem in the parrot side is the lack of a way to call vtable set_bool from pir.

On the winxed side, I think the problem lies in the && and operators, not in if or other conditionals.