C program doesn't recognize '\n' in file exported from MATLAB

Question

I have a matrix G in MATLAB that I have printed into a text file using:

file = fopen('G.dat','w');
fprintf(file, [repmat('%f\t', 1, size(G, 2)) '\n'], G'); 
fclose(file);

The dimension of this matrix is 100 x 500. If I count rows and columns using awk, for instance, using

cat G.dat | awk '{print NF}END{print NR}'

I see that the dimensions correspond to the original one.

Now, I want to read this file, G.dat, from a C program that counts the columns of the first row just to understand the columns' dimension as in:

    while (!feof(file) && (fscanf(file, "%lf%c", &k, &c) == 2) ) {
        Ng++; 
        if (c == '\n') 
            break;
}

Unfortunately it gives me Ng = 50000 and it doesn't recognize any of the '\n'. Instead, if I create the text file just by copying and pasting the data, it works. Can you explain me why? Thanks!

Use a binary editor/viewer to view what is different between the files. — mah, Jul 06 '15 at 14:57
Shouldn't the second if be inside the first one?. Also, [don't use `feof`](http://stackoverflow.com/questions/5431941/why-is-while-feof-file-always-wrong) — Eugene Sh., Jul 06 '15 at 14:58
Compose some small test matrix and step through in debugger. — Eugene Sh., Jul 06 '15 at 14:59
Yes, ok. I have changed the loop but it doesn't change the result.. I will try with small matrices.. — Anna, Jul 06 '15 at 15:02
I get a warning: "comparison is always false due to limited range of data type" and it doesn't affect the output. — Anna, Jul 06 '15 at 15:07
You can't compare a character to `\r\n` as it is two characters... — Eugene Sh., Jul 06 '15 at 15:09
@Anna Post your solution as an answer and accept it instead of editing the question. — chux - Reinstate Monica, Jul 06 '15 at 16:27

score 1 · Answer 1 · answered Jul 06 '15 at 15:10

1

Are you working in Windows? Try opening your output file in text mode:

file = fopen('G.dat','wt');

This will automatically insert a carriage return before each newline when writing to the file.

answered Jul 06 '15 at 15:10

Max

2,121
3
16
20

score 0 · Answer 2 · 2015-07-06T21:19:10.467

0

if I understand the matlab syntax correctly, this expands to a format string like %f\t%f\t%f\t\n\%f\t%f\t%f\t\n for a 3x2 matrix. Note the extra \t at the end of each line. If this assumption is correct, the last fscanf() call in a line will assign the last \t to &c. The next fscanf() call just skips the \n because it doesn't match your format.

I'd propose you use fgets() instead for reading each line and then loop over the fields using strtok(), reading the values with atof() e.g.

char buf[8192];
if (fgets(buf, 8192, file))
{
    if (strtok(buf, '\t'))
    {
        ++Ng;
        while (strtok(0, '\t')) ++Ng;
    }
}
else
{
    /* error reading ... */
}

edited Jul 06 '15 at 21:19

answered Jul 06 '15 at 15:13

It is ok but vector "buf" needs a dimension and I often have to read matrix with columns of order from 1e2 to 1e6.. – Anna Jul 06 '15 at 16:14
Well, in this case, read the file one character at a time, counting the `\t` characters until you hit a `\n` and use `malloc()` for the line buffer with a size `(Ng * (##maxlen(float)## +1) +2)`, the maxlen being the maximum length of a float in text representation... – Jul 06 '15 at 17:00

chux - Reinstate Monica · Accepted Answer · 2015-07-06T16:59:33.843

Code's approach is too fragile to "counts the columns of the first row just to understand the columns' dimension". fscanf(file, "%lf%c"... is too susceptible to variant white-space delimiters and EOL to detect the '\n'.

Recommend explicitly examine white-space to determine width:

// return 0 on success, 1 on error
int GetWidth(FILE *file, size_t *width) {
  *width = 0;
  for (;;) {
    int ch;
    while (isspace(ch = fgetc(file))) {
      if (ch == '\n') return 0;
    }
    if (ch == EOF) return 0;
    ungetc(ch, file);
    double d;
    if (fscanf(file, "%lf", &d) != 1)) {
      return 1;  // unexpected non convertible text
    }
    (*width)++;
  }
}

//Sample, usage
size_t width;
if (GetWidth(file, &width)) return 1;

// read entire file
rewind(file);
for (size_t line = 0; foo(); line++)
  for (size_t column = 0; column<width; column++) {
    double d;
    if (fscanf(file, "%lf", &d) != 1)) {
      break;  // EOF, unexpected non convertible text or input error
    }
  }
  ...
}

score 0 · Answer 4 · answered Jul 06 '15 at 17:35

0

Matlab writes rows as

 %f\t%\f .. %f\t\n

which is a problem. I have used

dlmwrite('G.dat', G, '\t');

and it is fine!

answered Jul 06 '15 at 17:35

Anna

67
1
1
5

C program doesn't recognize '\n' in file exported from MATLAB

4 Answers4