TITLE: converting doubles to strings with precision

(Newsgroups: comp.lang.c++.moderated, 24 Sept 96)

KANZE: kanze@gabi-soft.fr (J. Kanze)

CADman wrote:

> The answer comes out fine but I would like it to
> round off to 2 decimal points.

> The program takes a string then converts the string to a double
> does the calculation then converts it back to a string.
> 
> The answer comes out say 386.1567752 and I would like the
> answer to read 386.16

To which Mark McIntyre <Mark@garthorn.demon.co.uk> answered:

> char str[10;
> double z = 386.1567752;
> sprintf(str, "%.2f", z);
> 
> str now contains "386.16"

and "Per 'Pedro' Olsson" <pedro@algonet.se> answered:

> double answer;
> char output[40];
> 
> answer = 386.1567752;
> sprintf( output, "%.2Lf", answer ); /* output = "386.16" */

and Igor Siemienowicz <igor@adacel.com.au> answered:

> Use the 'sprintf' function to print to the string like this:
> 
> char pszAnswer[20];
> sprintf(pszTemp, "%5.2f", nAnswer);
> 
> Where 'nAnswer' is the result of your calculation.

I'm curious about the number of wrong answers here.  Obviously, the
value given by the original poster is an example; if the real value were
a constant, he could just use a string literal and be done with it.
Given a variable, all of the above programs fail radically ; to see why,
suppose that the variable happened to contain 1.23e300, instead of the
example results.  There is *NO* way to write a correct program using a
"%f" format to sprintf without having restricted the range of the
variable first.

The correct answer is, of course:

	ostrstream          s ;
	s.setf( ios::fixed , ios::floatfield ) ;
    s.precision( 2 ) ;
    s << value << ends ;
    //  Use s.str(), perhaps by making a copy of it.
    s.rdbuf()->freeze( 0 ) ;