converting string to floating point numbers

[monkeyking]

I'm quite puzzled by this small program, can someone elaborate why the following program gives different results



#include <iostream>
#include <sstream>

///convert string to double
double to_double(std::string a){///@param a is the string to convert.
double str;
std::stringstream ss;
ss<<a;
if(!(ss>>str)){
printf("Error converting :\"%s\" will exit\n",a.c_str());
exit(0);
}
ss>>str;
return str;///@return The double value.
}

///Convert string to float.
float to_float(std::string a){///@param a is the string to convert.
float str;
std::stringstream ss;
ss<<a;
if(!(ss>>str)){
printf("Error converting :\"%s\" will exit\n",a.c_str());
exit(0);
}
ss>>str;
return str;///@return The float value.
}


int main(){
std::string var = "147541611";
printf("%s <-> %f\n",var.c_str(),to_double(var));
printf("%s <-> %f\n",var.c_str(),to_float(var));
return 0;
}


output


./a.out
147541611 <-> 147541611.000000
147541611 <-> 147541616.000000

The to_float function doesn't work but a 9 digit number should still be within the limit of a float?

thanks in advance

[jpkotta]

http://en.wikipedia.org/wiki/Single_precision

According to that, floats don't have enough precision for 9 decimal digits. Most of the time, you should use doubles. On some architectures, doubles are actually faster, because single precision calculations are done as double precision and then rounded.

[dwhitney67]

@ OP -

The problems you are encountering may stem from the generic use of the std::stringstream, versus using a specialized version, such as std::istringstream.

Try instead:


double to_double(std::string& str)
{
std::istringstream iss(str);
double dbl;
iss >> dbl;
return dbl;
}


------------------------------------------------------

EDIT: Scratch what I stated above... the problem appears to be with the precision difference between a double and a float.

------------------------------------------------------

Alternatively, you can use the C library and a more C++-ish way of doing things:


#include <string>
#include <iostream>
#include <iomanip>
#include <cstdlib>

class Converter
{
public:
Converter(const std::string& str)
: m_str(str)
{
}

operator double() const
{
return strtod(m_str.c_str(), 0);
}

operator float() const
{
return strtof(m_str.c_str(), 0);
}

private:
std::string m_str;
};

int main()
{
Converter conv("147541611");
double dbl = conv;
float flt = conv;

std::cout << "double value = " << dbl << std::endl;
std::cout << "float value = " << flt << std::endl;
std::cout << "double value = " << std::fixed << dbl << std::endl; // with these two statements, the
std::cout << "float value = " << std::fixed << flt << std::endl; // problem reappears
}


In lieu of strtod() and strtof(), feel free to substitute with the istringstream.