A tiny class to compare floating point variables the right way with desired precision.
template<typename T>class FloatingPoint { public: FloatingPoint() {Precision(sizeof(T));} int Compare(const T x, const T y) {return (((fabs(x-y)<min))?0:((x-y>=min))?1:-1);} void Precision(int digits) {min=pow(10,-fabs(digits));} protected: double min; };
Let's say you have two doubles with a value of 1.15.
When you compare them, you expect them to be equal.
But doubles in computer behave the strange way.
1.15 may be stored as 1.14999999999.
So comparing will fail your expectations.
The proper way to compare floating point variables, is by checking their subtraction with some minimum value.
FlatingPoint must be initialized with double or float datatype.
Comparer method returns 0 if x = y; 1 if x > y; and -1 if x < y.
Precision can be changed.
Deafult precision is 8 digits after fp for doubles, and 4 for floats.
Example of FloatingPoint class usage.
void OnStart() { FloatingPoint<double>fp; double a=1.15,b=1.14999999999; for(int i=15; i>5; i--) { fp.Precision(i); int compare=fp.Compare(a,b); string res=(compare==0)?"=":(compare==1)?">":"<"; PrintFormat("%g %s %g at %d digit precision",a,res,b,i); } }
Output:
/** you do not expect this: 1.15 > 1.15 at 15 digit precision 1.15 > 1.15 at 14 digit precision 1.15 > 1.15 at 13 digit precision 1.15 > 1.15 at 12 digit precision 1.15 > 1.15 at 11 digit precision you expect this: 1.15 = 1.15 at 10 digit precision 1.15 = 1.15 at 9 digit precision 1.15 = 1.15 at 8 digit precision 1.15 = 1.15 at 7 digit precision 1.15 = 1.15 at 6 digit precision /**/ 
