Result is Number1 / Number2which should be preferred for portability.
If both arguments are integers, then the result is of type float by default (coinciding with ISO-Prolog). This can be changed by switching the global flag 'prefer_rationals' to 'on': the result is then of rational type, and therefore precise. In practice, a better way to enforce a rational result is by explicitly converting one or both arguments to a rational before dividing, e.g. Z is rational(X)/Y.
The following table details the behaviour on zero-division, depending on the argument types. The exact result depends on the result's type ability to represent extreme values.
-3 / 0 -1.0Inf (negative infinity) 0 / 0 arithmetic exception 3 / 0 1.0Inf (positive infinity) -3.0 / 0.0 -1.0Inf (negative infinity) -0.0 / 0.0 arithmetic exception 0.0 / 0.0 arithmetic exception 3.0 / 0.0 1.0Inf (positive infinity) -3.0 / -0.0 1.0Inf (positive infinity) -0.0 / -0.0 arithmetic exception 0.0 / -0.0 arithmetic exception 3.0 / -0.0 -1.0Inf (negative infinity) rational(-3) / rational(0) representation error rational( 0) / rational(0) arithmetic exception rational( 3) / rational(0) representation error breal(-3) / breal(0) -1.0Inf__-1.0Inf (negative infinity) breal( 0) / breal(0) -1.0Inf__1.0Inf (undefined) breal( 3) / breal(0) 1.0Inf__1.0Inf (positive infinity)Dividing infinity by infinity yields the same result as 0/0.
In coroutining mode, if Number1 or Number2 are uninstantiated, the call to //3 is delayed until these variables are instantiated.
Success: /(10, 2, 5.0). /(10, -2.0, -5.0). /(9, 12, 3_4). (with set_flag(prefer_rationals, on)) Fail: /(1, 2, 1.0). /(5, 2, r). /(6, 2, 3). Error: /(A, 2, 6.0). (Error 4). /(2, 0, Result). (Error 20). /(4 + 2, 2, 12). (Error 24).