Correct Way to set maximum double precision for ostringstream

Question

By referring to

http://stackoverflow.com/questions/554063/how-do-i-print-a-double-value-with-full-precision-using-cout

and

http://stackoverflow.com/questions/747470/what-is-the-meaning-of-numeric-limitsdoubledigits10

I am slightly confused.

Shall I used :

ss.precision(std::numeric_limits<double>::digits10);

or

ss.precision(std::numeric_limits<double>::digits10 + 2);

to get maximum double precision?

Answer 1

it depend on number of "decimal component" you have in data.

std::numeric_limits::digits10 will give the maximum number of "floating component"

ios_base::precision specifies the number of digits (decimal + float components) to display.

if your decimal component is always less than 100 (-99 to 99), code below always give maximum precision.

ss.precision(std::numeric_limits<double>::digits10 + 2);

Answer 2

According to http://stackoverflow.com/questions/747470/what-is-the-meaning-of-numeric-limitsdoubledigits10, specifically the highest voted answer, std::numericlimits<double>::digits10 is:

the maximum number of digits you can represent without loss of precision.

That means if you set ss.precision to that, these floating point digits will be accurate; if you set it to that +2 you will still display that+2 digits but the last two aren't guaranteed to be accurately calculated.

For arbitray precision, if that's what you need, take a look at mpir, gmp, mpfr