[RFC] CEIL, FLOOR, ROUND returning REAL

[Pietro Monteiro]

On Mon, May 18, 2026, at 1:06 AM, Pietro Monteiro wrote:
> On Mon, May 18, 2026, at 12:39 AM, chris hermansen wrote:
>> Piero and everyone,
>>
>> On Sun, May 17, 2026, 21:27 Pietro Monteiro <pietro@sociotechnical.xyz> wrote:
>>> While working on implementing CEIL I realized that CEIL, FLOOR, ROUND
>>> return an integer of same SIZETY as the real operand. I fortunately
>>> don't much involving floating point math, but I think that returning a
>>> real of same SIZETY as the operand might be usefull. What do y'all
>>> think?
>>
>> I think the length of the result should be the same as that of the operand.
>>
>> But I'm not sure I'm following you completely; ENTIER returns an INT, 
>> LONG INT OR LONG LONG INT, not reals. But your subject line suggests 
>> you plan to return real results from CEIL, FLOOR, ROUND?
>
> I was trying to propose adding CEIL, FLOOR/ENTIER and ROUND that would return
> a real of the same length of the operand in addition to the ones that
> return an integer.
>
>> I guess an error is thrown if the operand is outside the range 
>> representable by INT, LONG INT, LONG LONG INT respectively?
>
> I'm not sure. I'll try to test it soonish.

$ cat <<EOF > foo.a68
begin int foo  = ENTIER 3.40282346639e+38;
puts ("foo: " + REPR foo + "'n")
end
EOF
$ ga68 foo.a68 -o foo
$ ./foo
foo.a68:7: runtime error: -2147483648 is not a valid character point
Aborted (core dumped)

The error is about the "REPR foo" not the fact that the the operand is outside the range of INT.

pietro