Set precision of std::to_string when converting floating point values [duplicate]

Question

In C++11, std::to_string defaults to 6 decimal places when given an input value of type float or double. What is the recommended, or most elegant, method for changing this precision?

Malaya

Answer 1

There is no way to change the precision via to_string() but the setprecision IO manipulator could be used instead:

#include <sstream>

template <typename T>
std::string to_string_with_precision(const T a_value, const int n = 6)
{
    std::ostringstream out;
    out.precision(n);
    out << std::fixed << a_value;
    return std::move(out).str();
}

Consubstantial

Nice, but it'd be great to be able to do this without creating a temp string. :/ Especially in a very, very tight inner loop. – Among

isn't the inner string implicitly moved by being a returned rval? – Execratory

I get "3e+01" for the 33.33~ case, even only using n = 1. – Bagwig

To ensure the number of digits, also std::fixed must be passed. See sample in cplusplus.com/reference/iomanip/setprecision – Bradbury

Using string at all in a tight inner loop is a bad idea. Memory allocations are super slow. You are trying to use 1 memory allocation instead of having 2 memory allocations by avoiding a temporary. You try to avoid any memory allocations if you want performance by using sprintf and a stack allocated buffer. :-) – Thoer

You don't need iomanip for this: just use out.precision(n); out << a_value;. Using iomanip isn't even less typing. – Kerby

If you automatically want to choose the best available precision you can use const unsigned int digits = std::numeric_limits<T>::digits10; instead of the integer n – Mindexpanding

When I tried float x = 107.2; cout<<to_string(x) gives 107.199997. Is this normal? – Leventhal

Could please someone confirm this: " isn't the inner string implicitly moved by being a returned rval? – Jules G.M. Aug 18, 2014 at 22:40 "? It's an interesting statement and I think it's correct? Could please someone confirm? – Carolyncarolyne