Round in C with N digits after decimal point

Question

Possible Duplicate:
Rounding Number to 2 Decimal Places in C

I have not found a function with a signature double round(double d, int digits) like here in c. When i try to build i get a error:

error: too many arguments to function 'round'

How can I round in C with N digits after the decimal point?

@melpomene, thanks i have found http://stackoverflow.com/questions/994764/rounding-doubles-5-sprintf but i need to get return value not only print it — testCoder, Dec 22 '12 at 14:08
Decimal point is relevant for display. One could round to the nearest negative power of ten - but - why would you want such a thing? — Pavel Radzivilovsky, Dec 22 '12 at 14:11
**Voting to re-open**. The cited "duplicate" is a special case, requiring rounding to only 2 decimal places. This question asks the more general case of rounding to an arbitrary number of decimal places. — andand, Dec 23 '12 at 02:35

andand · Accepted Answer · 2012-12-22T14:53:24.643

4

Using recursion (which is going to be slow for some values of digits)

#include <math.h>
double my_round(double x, unsigned int digits) {
  if (digits > 0) {
    return my_round(x*10.0, digits-1)/10.0;
  }
  else {
    return round(x);
  }
}

A method likely to be somewhat faster, but which relies on a single call to the slow pow function:

#include <math.h>

double my_round(double x, unsigned int digits) {
    double fac = pow(10, digits);
    return round(x*fac)/fac;
}

An even faster method is to precompute a lookup table with the likely powers and use that instead of pow.

#include <math.h>

double fac[];  // population of this is left as an exercise for the reader

double my_round(double x, unsigned int digits) {
    return round(x*fac[digits])/fac[digits];
}

edited Dec 22 '12 at 14:53

answered Dec 22 '12 at 14:15

andand

17,134
11
53
79

2

I'm guessing that this doesn't actually compile as C, but would as C++. – Mats Petersson Dec 22 '12 at 14:16
3

using recursion where simple arithmetics would suffice? – Pavel Radzivilovsky Dec 22 '12 at 14:16
@PavelRadzivilovsky: which is why in my original response I also included the more efficient method. – andand Dec 22 '12 at 14:22
I would remove the recursion. It is like going to a chemist thru the moon.. – Pavel Radzivilovsky Dec 22 '12 at 14:24
I'm very surprised this even compiles in C, much less "nicely work[s]". [To be specific, I'm surprised the compiler didn't object to `round` having two different call signatures, and thus conflicting types.] – DSM Dec 22 '12 at 14:25
Yes definitely works, i have tested first and second methods. – testCoder Dec 22 '12 at 14:30
@DSM: fixed the naming reference... it's been a while since I've done straight C programming. – andand Dec 22 '12 at 14:34
In the recursive version `my_round(1,1)` gives result `10`. – Henry Dec 22 '12 at 14:38
@Henry: Forgot to divide by 10... fixed. – andand Dec 22 '12 at 14:53

score 1 · Answer 2 · answered Dec 22 '12 at 14:15

Whilst "answerd" gives a decent answer, here's one that works for arbitrarily large numbers:

double round1(double num, int N) {
      ASSERT(N > 0);
      double p10 = pow(10,N);
      return round(num* p10) / p10;
}

Of course, as stated, floating point numbers don't have a set number of decimal digits, and this is NOT guaranteed to PRINT as 3.70000 if you call printf("%8.5f", round1(3.7519, 1)); for example.

score 0 · Answer 3 · answered Dec 22 '12 at 14:10

0

here's a (very) simple function,

double round1(double num, int N) {
      int temp=(int) num*pow(10,N); 
      double roundedN= temp/pow(10,N);
      return roundedN;
}

answered Dec 22 '12 at 14:10

Muhammad

89
1
5

This does not round, it truncates. Moreover `pow` is an expensive operation, store the result instead of invoking it twice. – Henry Dec 22 '12 at 14:12
May be i'm doing something wroing but i test it repeatedly, it not work. – testCoder Dec 22 '12 at 14:21
@Henry, `pow` for integer power could be implemented efficiently (with O(log n) complexity). So in this case it's cheap. – Artem Sobolev Dec 22 '12 at 14:23
@Barmaley.exe, you are right, but thats still more than O(1) for an assignment. – Henry Dec 22 '12 at 14:29

score 0 · Answer 4 · answered Dec 22 '12 at 14:14

In C standard, such function does not exist. Anyway, you can write your own.

#include <math.h>

/* Round `n` with `c` digits after decimal point. */

double nround (double n, unsigned int c)
{
    double marge = pow (10, c);
    double up    = n * marge;
    double ret   = round (up) / marge;

    return ret;
}

See also the comments above about floating points "decimal point".

Round in C with N digits after decimal point

4 Answers4