warning converting to int from double

[simons]

Hi,

I am trying to write a Fraction class and getting the following warning when compiling my code :

Fraction.cpp: In constructor 'Fraction::Fraction(double)':
Fraction.cpp:8: warning :converting to 'int' from 'double'

My Fraction.cpp class looks like :

#include "Fraction.h"

Fraction::Fraction(int n, int d):num(n),den(d)
{
   cout << This is double param constructor <<endl;
}

Fraction::Fraction(double d):num(d),den(0)
{
   cout <<This is single param constructor <<endl;
}

My Header Fraction.h is like :

class Fraction 
{
  int num;
  unsigned int den;

  public:

  Fraction(int = 1,int =1);
  Fraction(double);
};

And the application main.cpp is as follows:

int main()
{
  Fraction f1(3,9);
  Fraction f2(8);


  return 0;
}

How can I get rid of the warning ? Please help.

Thanks.
Simons

[Peter87]

Are you sure you want to have a Fraction constructor taking a double as argument?

[vlad from moscow]

The range of values that the type double can hold is greater than the range of values for the type int. So when a double value is converted to int you can get an incorrect result.

[simons]

Thanks , it helped

[JLBorges]

Something like this (approximate conversion):

#include <cmath>

// to a precision of three digits after the decimal
Fraction::Fraction( double d ) : den(1000)  // set denominator to 1000
{
    // set numerator to 1000 * d converted to an int
    num = std::round( dbl * den ) ; // C++11
    num = int( dbl * den ) ; // C++98
}

[andywestken]

How can I get rid of the warning

It's easy enough to shut the warning up using a cast or a compiler directive. But that would just hide the problem.

What do you expect the double constructor to do?

Fraction f2(0.25); // a fraction with num = 1, den = 4 ??

and

Fraction f2(3.141592654); // ??

How do you real with the fact that not all floating point numbers are rational?

Andy

[JLBorges]

> What do you expect the double constructor to do?
> Fraction f2(0.25); // a fraction with num = 1, den = 4 ??

Sets num = 250 and den = 1000 which is equivalent.


> the fact that not all floating point numbers are rational?

All floating point numbers (except infinity and NaN) are rational numbers.
(Adding a check for NaN or infinity would be a good idea.)

[simons]

Well, that Fraction (double); in the header file is just a constructor for explicit conversion, double to fraction conversion.

How can I write the body of it in the implementation ?

[Peter87]

Are you aware that floating point types like float and double are just approximations? Not all values can be stored exactly. One such value is 0.1.

std::cout << std::setprecision(20) << 0.1 << std::endl; // prints 0.10000000000000000555 

If you do not round or handle it in a somehow special way and just treat 0.1 exactly the way it is you will not get the nice fraction 1/10 but something with much larger values.