Can anyone explain why the flag value is returning False?
double a = 1.0;
double b = 0.0;
double c = a / b;
boolean flag = Double.isNaN(c);
System.out.println(flag); // False?
System.out.println(c); // Infinity
Asked
Active
Viewed 228 times
1
Grs007
- 129
- 1
- 1
- 7
-
I've reopened this because the previously linked duplicate is answering a different question and doesn't cover this case. – ajb Jul 05 '17 at 05:36
-
1see http://javaconceptoftheday.com/isnan-and-isinfinite-methods-java/ – Scary Wombat Jul 05 '17 at 05:40
-
2Because infinity isn't a `NaN`? – user207421 Jul 05 '17 at 05:41
-
see also: https://stackoverflow.com/questions/1007586/what-do-these-three-special-floating-point-values-mean-positive-infinity-negat – Jul 05 '17 at 05:44
2 Answers
2
This is due to the definition of floating-point representation in the IEEE 754 standard. The standard has representations for both "infinity" and "NaN". "Infinity" is for operations that either produce an infinite result (such as 1/0 or tan(pi/2) [*]) or produce a result whose absolute value is larger than the largest possible number the format can represent. More precisely, in math, there really isn't such a thing as an infinite result; rather, it's defined in terms of limits. Thus, 1/0 doesn't exist, but the limit of 1/x as x approaches 0 is infinite.
NaN is returned for "indeterminate forms". In math, these are cases such as 0/0 when the limit can't be determined just by looking at the numerator and denominator. (If you have two functions f(x) and g(x) where the values are both 0 at some point f(a)=g(a)=0, then you can't determine the limit of f(x)/g(x) as x approaches a, without extra work such as L'Hopital's rule.) NaN is also returned for things like taking the square root of a negative number.
In Java, isNan returns true only for actual NaN's, not for infinities. Even though infinity really is "not a number", it doesn't meet the definition of a NaN according to the IEEE standard.
See here for a definition of which operations generate NaN.
[*] Note that "pi" can't be represented exactly in a floating-point number, so this really isn't an operation that could produce an "infinite result" on a computer.
ajb
- 31,309
- 3
- 58
- 84
1
That's what the IEEE-754 standard says. When you divide a non-zero floating point number by zero, it does not return a number, but infinity.
Take a look at this: http://grouper.ieee.org/groups/754/faq.html#exceptions
Tavo
- 3,087
- 5
- 29
- 44
-
"When you divide a floating point number by zero" should be "When you divide a non-zero floating point number by zero". 0/0 will return a NaN. – ajb Jul 05 '17 at 05:49
-