problem of reading a double from .mat file in C

Hi All,

I have a .mat file named "test.mat" which contains an array D=[1,0.2]. I wrote a C program to read the second element (0.2) into a double variable x, and then to store 1000*x into another int variable n. My code follows:

      double x;
      int n;

      pmat = matOpen("test.mat", "r");
      pa = matGetVariable(pmat, "D");

      x = *(mxGetPr(pa) +1);
      cout << "x = " << x << endl;
      n = (int)(1000*x);
      cout << "n = " <<n << endl;
      mxDestroyArray(pa);
      return 0;

Surprsingly, the output is:

x = 0.2

n = 199

Why n is 199 instead of 200? What may be the cause?

Many Thanks!

Answer 1

[James Tursa]

n = (int)(1000*x) is truncating the expression instead of rounding it, and your 0.2 is probably slightly less than 0.2 (since 0.2 can't be represented exactly in floating point). Try using num2strexact on x and on 1000*x to see what their exact decimal representation is. You can find num2strexact here:

http://www.mathworks.com/matlabcentral/fileexchange/22239-num2strexact-exact-version-of-num2str

[Dillon Geo]

Thanks to James Tursa and Chirag Gupta.

I am also suspecting it is related to the way a floating number is stored. I used round() to eliminate this problem, and it seems working well.

I just did another experiment. I examined the meomry content of a new double variable X=1000*x (x still get its value from .mat file), and the memory content of another double variable Y which is assigned a value of 200 in C. Their memory contents ARE different, HOWEVER, they both print out as 200, either via cout or printf("%f"). Really hard to catch such a problem!

Answer 2

[Chirag Gupta]

It might be something to do with precision and and how the floating number was stored. On my MATLAB (R2011a) on mac (64 bit), I get: x =0.2000 n =200

My code is as follows:

#include <stdio.h>
#include "mat.h"

int main() 
{
  MATFile *pmat;
  double x;
  mxArray *pa;
  int n;
  pmat = matOpen("test.mat", "r");
  pa = matGetVariable(pmat, "D");
  x = *(mxGetPr(pa) +1);
  printf("\nx=%lf",x);
  n = (int)(1000*x);
  printf("\nn = %d\n",n);
  mxDestroyArray(pa);
  return 0; 
}