[eclipse-clp-users] IC limitations

[Chris M. <Chr...@se...>]

[using ECLiPSe 7.0]
I have a query about IC floats limitations.

I understand some of the limitations of interval arithmetic and that not all decimals can be represented accurately by floats.
But practically, I need to know what is the largest float range that can be used without generating aberrant results.

?- lib(ic).
Yes (0.08s cpu)
?- R $:: -5 .. 5, R $= 1.6.
R = 1.6__1.6000000000000005                           % That's ok
Yes (0.00s cpu)

?- R $:: -3.4028235 * 10 ^ 38 .. 3.4028235 * 10 ^ 38, R $= 1.6.    % i.e. single precision 32 bits C float range
R = 0.0__3.7778931862957162e+22                       % aberrant
Yes (0.00s CPU)

?- R $:: -3.4028235 * 10 ^ 3 .. 3.4028235 * 10 ^ 3, R $= 1.6.
R = 1.5999999999999091__1.6000000000003638            % I can live with that
Yes (0.00s cpu)

?- R $= 1.6.
R = 1.6__1.6                                          % but that would be nicer all the time
Yes (0.00s cpu)

?- R $:: -2 ^ 31 - 1 .. 2 ^ 31 - 1, R $= 1.6.                     % using the C long range: -2147483647 .. 2147483647
R = 1.5999999046325684__1.6000003814697266            % that's ok
Yes (0.00s cpu)

?- R $:: -3.4028235 * 10 ^ 20 .. 3.4028235 * 10 ^ 20, R $= 1.6.
R = 0.0__65536.0                                      % starting to be silly...
Yes (0.00s cpu)

It's the unpredictability that I have difficulties with.

What's the best way to deal with this? Not imposing bounds? Using a safe range, but which one?

Many thanks for your efforts

Regards,
Chris
--
Dr Chris Meudec<http://www.echancrure.eu/> (he/him) Maitrise, PhD, MA (T&L)
Lecturer in Software Development
E chr...@se...<mailto:chr...@se...> |  setu.ie<https://setu.ie/>
[turn on images]<https://setu.ie/>

Kilkenny Road Campus, Kilkenny Road, Carlow, R93 V960, Ireland
Campas Bhóthar Chill Chainnigh, Bóthar Chill Chainnigh, Ceatharlach, R93 V960, Éire