Is it safe to use -DBL_MAX in code?

Question

I want extreme -ve value .

#include <iostream> 
using namespace std; 
#include <math.h>
#include <limits.h>
#include <values.h>

#define THRESHOLD 2*DBL_MIN

#define FEQ(x, y) (fabs((x) - (y)) < THRESHOLD)

int main( )
{ 
double  a = -DBL_MAX; // I want here minimum value of double
if(FEQ(a,-DBL_MAX))
cout<<"Equal " <<endl;
else
cout<<"NOt equal"<<endl;
return 0;



}

So , is it safe to use -DBL_MAX in code? If anybody know better approach please share here.

Answer 1

DBL_MAX is part of the standard library. As such it's not particularly unsafe.

However, comparing non-integral floating point values for equality is inherently unsafe, in the sense of unreliable.

That's because floating point arithmetic may be performed with higher precision than your chosen type, with results converted back and forth, and not always as you'd think.

---

in passing, defining a macro like:

#define FEQ(x, y) (fabs((x) - (y)) < THRESHOLD)

… is unsafe: you can easily become victim of undesired text substitution and besides it's an eyesore.

Instead use an inline function, even if inline does not guarantee optimization:

inline bool feq( double const x, double const y ) { return fabs( x - y ) < THRESHOLD; }

and then the same goes for the constant THRESHOLD, it should simply be a constant in C++:

double const threshold = 2*DBL_MIN;