sscanf string format specifiers not works for '\t'

Question

#include <math.h>
#include <stdio.h>
#include <stdlib.h>
char *tokenstring = "first,25.5,second,15";
int result, i;
double fp;
char o[10], f[10], s[10], t[10];
void main()
{
   result = sscanf(tokenstring, "%[^','],%[^','],%[^','],%s", o, s, t, f);
   printf("%s\n %s\n %s\n %s\n", o, s, t, f);
   fp = atof(s);
   i  = atoi(f);
   printf("%s\n %lf\n %s\n %d\n", o, fp, t, i);
}

Codes above not works for '\t', why? It works for this I am using vc6.0

Not works

char *tokenstring = "first\t25.5\tsecond\t15";



   result = sscanf(tokenstring, "%[^'\t'],%[^'\t'],%[^'\t'],%s", o, s, t, f);

To begin with, don't enclose individual characters in quotes. The format for whitespace would be something like `"%[^ \r\n\t]"` — Some programmer dude, Apr 18 '14 at 16:29
`"%[^'\t'],%[^'\t']%s,%[^'\t'],%s"` ---> `"%[^\t] %[^\t] %[^\t] %s"` — BLUEPIXY, Apr 18 '14 at 16:32
`result = sscanf(tokenstring, "%[^\t]%[^\t]%[^\t]%s", o, s, t, f);` not work either. — Nick Dong, Apr 18 '14 at 16:56
I am using vc6.0. I have edit my question. There is only one `%s`. — Nick Dong, Apr 18 '14 at 16:59
You can post the document link on your answer. Me or other people can read the doc for more info when We meet similiar problem. @BLUEPIXY — Nick Dong, Apr 18 '14 at 17:06
It's just a simple misunderstanding. So I think that it is not even to answer. — BLUEPIXY, Apr 18 '14 at 17:08
I did not find any doc which to say that it need a space. @BLUEPIXY — Nick Dong, Apr 18 '14 at 17:09
Note: A format of `" %s"` and `"%s"` scan the _same_ way. `" "` directs to scan and not save any white-space. `"%s"` will scan and not save any white-space before it begins to scan and save non-white-space. This is not so with `" %[...]"` and `"%[...]"`. `"%[...]"` on its own, does not scan and not save leading white-space. — chux - Reinstate Monica, Apr 18 '14 at 19:05
docs for gnuc [Formatted-Output](http://www.gnu.org/software/libc/manual/html_node/Formatted-Input) | [Formatted-Output](http://www.gnu.org/software/libc/manual/html_node/Formatted-Output) — Nick Dong, Apr 19 '14 at 04:08

Chris Dodd · Accepted Answer · 2014-04-18T17:48:36.523

Look at what your format is matching:

"%[^'\t'],%[^'\t']
 ^     ^ ^
 \     | \- match a literal comma
  \    |
   \---+- match a sequence not containing tab or ' (single quote), up to the next
          tab or single quite.

So the first %[..] matches everything up to and not including the first tab in the input, and then it tries to match a comma, which doesn't match the tab, and so fails.

The easiest fix is to replace the commas in the string with spaces, which will skip whitespace (which include tabs). Using tabs will do the same thing, but will confuse people into thinking you're trying to match a tab rather than skip whitespace:

sscanf(tokenstring, "%[^\t] %[^\t] %[^\t]%s", o, s, t, f);

Note that you also probably don't want to treat ' characters specially in the matches, unless you want them to fail.

Now if you want to use just tabs for your separators (not just any whitespace), you need to use tab patterns:

sscanf(tokenstring, "%[^\t]%*1[\t\]%[^\t]%*1[\t]%[^\t]%s", o, s, t, f);

The pattern %*1[\t] will match exactly a single tab in the input and not store it anywhere.

This leads to another problem you may have noticed with your first (comma based) scanf -- a pattern like %[^,] or %[^\t] will not match an empty string -- if the next character on the input is a , (or \t in the second case), scanf will simply return without matching anything (or any of the following patterns), rather than storing an empty string.

In addition, if any of your strings are too long for the arrays, you'll overflow and crash (or worse). So whenever you use a scanf %s or %[ pattern into a buffer, you should ALWAYS specify the buffer size:

sscanf(tokenstring, "%9[^,],%9[^,],%9[^,],%9s", o, s, t, f);

Now instead of crashing or corrupting things, when the input is too long, the sscanf call will match only the first 9 characters of a field and return with the rest of the field yet to be read.

`sscanf(tokenstring, "%[^\t]%*1[\t]%[^\t]%*1[\t]%[^\t]%s", o, s, t, f);` — Nick Dong, Apr 19 '14 at 07:44

score 2 · Answer 2 · answered Apr 18 '14 at 17:34

When you use a comma to separate the fields, you'll have to add a , to the format string so as to skip it. Similarly for the \t.

#include <math.h>
#include <stdio.h>
#include <stdlib.h>

void test1()
{
   char *tokenstring = "first,25.5,second,15";
   int result, i;
   double fp;
   char o[10], f[10], s[10], t[10];

   // -----------------------------------------------------------
   // make sure you add a , between the string format specifiers
   // -----------------------------------------------------------
   result = sscanf(tokenstring, "%[^,],%[^,],%[^,],%s", o, s, t, f);
   printf("%s\n %s\n %s\n %s\n", o, s, t, f);
   fp = atof(s);
   i  = atoi(f);
   printf("%s\n %lf\n %s\n %d\n", o, fp, t, i);
}

void test2()
{
   char *tokenstring = "first\t25.5\tsecond\t15";
   int result, i;
   double fp;
   char o[10], f[10], s[10], t[10];

   // -----------------------------------------------------------
   // make sure you add a \t between the string format specifiers
   // -----------------------------------------------------------
   result = sscanf(tokenstring, "%[^\t]\t%[^\t]\t%[^\t]\t%s", o, s, t, f);
   printf("%s\n %s\n %s\n %s\n", o, s, t, f);
   fp = atof(s);
   i  = atoi(f);
   printf("%s\n %lf\n %s\n %d\n", o, fp, t, i);
}

void main()
{
   test1();
   test2();
}

sscanf string format specifiers not works for '\t'

2 Answers2