How do I parse out the fields in a comma separated string using sscanf while supporting empty fields?

Question

I have a comma separated string which might contain empty fields. For example:

1,2,,4

Using a basic

sscanf(string,"%[^,],%[^,],%[^,],%[^,],%[^,]", &val1, &val2, &val3, &val4);

I get all the values prior to the empty field, and unexpected results from the empty field onwards.

When I remove the expression for the empty field from the sscanf(),

sscanf(string,"%[^,],%[^,],,%[^,],%[^,]", &val1, &val2, &val3, &val4);

everything works out fine.

Since I don't know when I'm going to get an empty field, is there a way to rewrite the expression to handle empty fields elegantly?

score 12 · Accepted Answer · edited May 23 '17 at 12:10

12

If you use strtok with the comma as your separator character you'll get a list of strings one or more of which will be null/zero length.

Have a look at my answer here for more information.

edited May 23 '17 at 12:10

Community

1
1

answered Oct 02 '09 at 10:24

ChrisF

134,786
31
255
325

Yes, that is the route I'll go if I can't find a way to adapt the sscanf. I'm still hoping I can just adapt my sscanf, but as far as I can tell [^,] doesn't match empty sequences. – Belrog Oct 02 '09 at 10:28
@Belrog, it seems that `sscanf`'s internal counter get stuck if the string contains `,,`. So better use `strtok`. – Nick Dandoulakis Oct 02 '09 at 10:53
I was really hoping to not have to do that. I have a long list of values of different types (chars/ints/floats). That would mean a field counter and a large switch statement to assign it correctly. *sigh* – Belrog Oct 02 '09 at 11:19
1

strtok is now deprecated, use strsep instead. http://www.manpagez.com/man/3/strsep/ – manash Mar 09 '15 at 09:10

score 5 · Answer 2 · answered Oct 02 '09 at 12:43

5

man sscanf:

[ Matches a nonempty sequence of characters from the specified set of accepted characters;

(emphasis added).

answered Oct 02 '09 at 12:43

Sinan Ünür

116,958
15
196
339

2

Quite true. Is there no flag to change this behaviour? – Craig McQueen Jul 20 '12 at 05:32

score 1 · Answer 3 · answered Sep 23 '16 at 14:30

There are some problems with strtok() listed here: http://benpfaff.org/writings/clc/strtok.html

Hence, it is better to avoid strtok.

Now, consider a string containing a empty field as follows:

char myCSVString[101] = "-1.4,2.6,,-0.24,1.26"; // specify input here

You can use simple function to be able convert String in CSV format to read them to a float Array:

int strCSV2Float(float *strFloatArray , char *myCSVStringing);

Please find the Usage below:

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



int strCSV2Float(float *strFloatArray , char *myCSVStringing);

  void main()
 {

    char myCSVString[101] = "-1.4,2.6,,-0.24,1.26"; // specify input here
    float floatArr[10]; // specify size here 
    int totalValues = 0;

    printf("myCSVString == %s \n",&myCSVString[0]);

    totalValues = strCSV2Float(&floatArr[0] , &myCSVString[0]); // call the function here 

    int floatValueCount = 0;

    for (floatValueCount = 0 ; floatValueCount < totalValues ; floatValueCount++)
    {

      printf("floatArr[%d] = %f\n",floatValueCount , floatArr[floatValueCount]);

    }

 }




int strCSV2Float(float *strFloatArray , char *myCSVStringing)
{

int strLen = 0;
int commaCount =0; // count the number of commas
int commaCountOld =0; // count the number of commas
int wordEndChar = 0;
int wordStartChar = -1;
int wordLength =0;


   for(strLen=0; myCSVStringing[strLen] != '\0'; strLen++) // first get the string length
   {

       if ( (myCSVStringing[strLen] == ',')  || ( myCSVStringing[strLen+1] == '\0' ))
        {
           commaCount++;
           wordEndChar = strLen;
        }
       if ( (commaCount - commaCountOld) > 0 )
        {
          int aIter =0;
          wordLength = (wordEndChar - wordStartChar);
          char word[55] = "";
          for (aIter = 0;  aIter < wordLength; aIter++)
          {
            word[aIter] = myCSVStringing[strLen-wordLength+aIter+1];
          }

          if (word[aIter-1] == ',') 
           word[aIter-1] = '\0';

          //  printf("\n");
          word[wordLength] = '\0';
          strFloatArray[commaCount-1] = atof(&word[0]);

          wordLength = 0;
          wordStartChar = wordEndChar;
          commaCountOld = commaCount;

        }  
  }

  return commaCount;

}

Output is as follows :

myCSVString == -1.4,2.6,,-0.24,1.26 
floatArr[0] = -1.400000
floatArr[1] = 2.600000
floatArr[2] = 0.000000
floatArr[3] = -0.240000
floatArr[4] = 1.260000

score 0 · Answer 4 · edited Oct 30 '12 at 08:53

0

Put a '*' after the '%' to skip reading. In addition it is possible to read only 3 characters noting '%3s' for example.

edited Oct 30 '12 at 08:53

Andrew Barber

39,603
20
94
123

answered Oct 30 '12 at 05:41

hanoo

4,175
2
26
19

Your article isn't about the topic being asked about here. I have removed the link to prevent your answer from being flagged as Spam. – Andrew Barber Oct 30 '12 at 08:53

score 0 · Answer 5 · answered Oct 02 '09 at 13:26

This looks like you are currently dealing with CSV values. If you need to extend it to handle quoted strings (so that fields can contain commas, for example), you will find that the scanf-family can't handle all the complexities of the format. Thus, you will need to use code specifically designed to handle (your variant of) CSV-format.

You will find a discussion of a set CSV library implementations in 'The Practice of Programming' - in C and C++. No doubt there are many others available.

sambowry · Answer 6 · 2009-10-02T14:28:39.697

Here is my version to scan comma separated int values. The code detect empty and non-integer fields.

#include <stdio.h> 
#include <string.h> 

int main(){
  char str[] = " 1 , 2 x, , 4 ";
  printf("str: '%s'\n", str );

  for( char *s2 = str; s2; ){
    while( *s2 == ' ' || *s2 == '\t' ) s2++;
    char *s1 = strsep( &s2, "," );
    if( !*s1 ){
      printf("val: (empty)\n" );
    }
    else{
      int val;
      char ch;
      int ret = sscanf( s1, " %i %c", &val, &ch );
      if( ret != 1 ){
        printf("val: (syntax error)\n" );
      }
      else{
        printf("val: %i\n", val );
      }
    }
  }

  return 0;
}

Result:

str: ' 1 , 2 x, , 4 '
val: 1
val: (syntax error)
val: (empty)
val: 4

score 0 · Answer 7 · answered Jan 29 '14 at 12:11

I arrived here looking for answers to the same question. I didn't want to leave behind the scanf funcion either. In the end, I build a zsscanf myself, where I parsed the format, sscanf'ed every data one by one and checked the return of sscanf to see if I got an empty read in any. This was somewhat my particular case: I wanted only some of the fields, some of which could be empty, and could not assume the separator.

#include <stdarg.h>
#include <stdio.h>

int zsscanf(char *data, char *format, ...)
{
    va_list argp;
    va_start(argp, format);
    int fptr = 0, sptr = 0, iptr = 0, isptr = 0, ok, saved = 0;
    char def[32];
    while (1)
    {
        if (format[fptr] != '%')
        {
            ok = sscanf(&format[fptr], "%28[^%]%n", def, &iptr);
            if (!ok) break;
            fptr += iptr;
            def[iptr] = '%';
            def[iptr+1] = 'n';
            def[iptr+2] = 0;
            ok = sscanf(&data[sptr], def, &isptr);
            if (!ok) break;
            sptr += isptr;
        }
        else
            if (format[fptr+1] == '%')
            {
                if (data[sptr] == '%')
                {
                    fptr += 2;
                    sptr += 1;
                }
                else
                {
                    ok = -1;
                    break;
                }
            }
            else
            {
                void *savehere = NULL;
                ok = sscanf(&format[fptr], "%%%28[^%]%n", &def[1], &iptr);
                if (!ok) break;
                fptr += iptr;
                def[0] = '%';
                def[iptr] = '%';
                def[iptr+1] = 'n';
                def[iptr+2] = 0;
                isptr = 0;
                if (def[1] != '*')
                {
                    savehere = va_arg(argp, void*);
                    ok = sscanf(&data[sptr], def, savehere, &isptr);
                    if (ok == 0 && isptr == 0)
                    {
                        // Let's assume only char types. Won't hurt in other cases.
                        ((char*)savehere)[0] = 0;
                        ok = 1;
                    }
                    if (ok > 0)
                    {
                        saved++;
                    }
                }
                else
                {
                    ok = sscanf(&data[sptr], def, &isptr) == 0;
                }
                if (ok < 0) break;
                sptr += isptr;
            }
    }
    va_end(argp);
    return saved == 0 ? ok : saved;
}

int main()
{
    char *format = "%15[^\t;,]%*1[\t;,]" // NameId
                   "%*[^\t;,]%*1[\t;,]" // Name
                   "%*[^\t;,]%*1[\t;,]" // Abbreviation
                   "%*[^\t;,]%*1[\t;,]" // Description
                   "%31[^\t;,]"; // Electrical Line
    char nameId[16];
    char elect[32];
    char *line1 = "TVC-CCTV-0002\tTVC-CCTV-0002\tTVC-CCTV-0002\tCCTV DOMO CAMERA 21-32-29\tELECTRICAL_TopoLine_823\tfoo\tbar";
    char *line2 = "TVC-CCTV-0000;;;;;foo;bar;";

    int ok = zsscanf(line1, format, nameId, elect);
    printf ("%d: |%s|%s|\n", ok, nameId, elect);
    ok = zsscanf(line2, format, nameId, elect);
    printf ("%d: |%s|%s|\n", ok, nameId, elect);
    return 0;
}

Output:

    2: |TVC-CCTV-0002|ELECTRICAL_TopoLine_823|
    2: |TVC-CCTV-0000||

Be warned, it's not fully tested and has severe limitations (the most obvious ones: accepts only %...s, %...c, %...[...] and requires separators as %...[...]; otherwise I'd really hade to care about the format string, this way I only care about %).

xerostomus · Answer 8 · 2014-04-29T21:02:10.593

I had to modify this code a bit to work properly:

//rm token_pure;gcc -Wall -O3 -o token_pure token_pure.c; ./token_pure 
#include <stdio.h>
#include <string.h>

int main ()
{
    char str[] = " 1 , 2 x, , 4 ";
    char *s1;
    char *s2;
    s2=(void*)&str; //this is here to avoid warning of assignment from incompatible pointer type 
        do {
            while( *s2 == ' ' || *s2 == '\t' )  s2++;
            s1 = strsep( &s2, "," );
            if( !*s1 ){
                printf("val: (empty)\n" );
            }
            else{
                int val;
                char ch;
                int ret = sscanf( s1, " %i %c", &val, &ch );
                if( ret != 1 ){
                    printf("val: (syntax error)\n" );
                }
                else{
                    printf("val: %i\n", val );
                }
            }
        } while (s2!=0 );
        return 0;
    }

and the output:

val: 1
val: (syntax error)
val: (empty)
val: 4

xerostomus · Answer 9 · 2014-04-29T21:24:04.390

I made a modification for tab delimited TSV files, hopefully it may help:

//rm token_tab;gcc -Wall -O3 -o token_tab token_tab.c; ./token_tab 
#include <stdio.h>
#include <string.h>

int main ()
{
//  char str[] = " 1     2 x         text   4 ";
    char str[] = " 1\t 2 x\t\t text\t4 ";
    char *s1;
    char *s2;
    s2=(void*)&str; //this is here to avoid warning of assignment from incompatible pointer type 
        do {
            while( *s2 == ' ')  s2++;
            s1 = strsep( &s2, "\t" );
            if( !*s1 ){
                printf("val: (empty)\n" );
            }
            else{
                int val;
                char ch;
                int ret = sscanf( s1, " %i %c", &val, &ch );
                if( ret != 1 ){
                    printf("val: (syntax error or string)=%s\n", s1 );
                }
                else{
                    printf("val: %i\n", val );
                }
            }
        } while (s2!=0 );
        return 0;
    }

And the ouput:

val: 1
val: (syntax error or string)=2 x
val: (empty)
val: (syntax error or string)=text
val: 4

pmg · Answer 10 · 2009-10-02T13:53:45.527

scanf() returns the number of items assigned. Maybe you can use that info ...

char *data = "1, 2,,, 5, 6";
int a[6];
int assigned = sscanf(data, "%d,%d,%d,%d,%d,%d", a, a+1, a+2, a+3, a+4, a+5);
if (assigned < 6) {
    char fmt[18];
    switch (assigned) {
        default: assert(0 && "this did not happen"); break;
        case 0: fmt = ",%d,%d,%d,%d,%d"; break;
        case 1: fmt = "%d,,%d,%d,%d,%d"; break;
        case 2: fmt = "%d,%d,,%d,%d,%d"; break;
        case 3: fmt = "%d,%d,%d,,%d,%d"; break;
        case 4: fmt = "%d,%d,%d,%d,,%d"; break;
        case 5: fmt = "%d,%d,%d,%d,%d,"; break;
    }
    sscanf(data, fmt, a+(assigned<=0), a+1+(assigned<=1), a+2+(assigned<=2),
                      a+3+(assigned<=3), a+4+(assigned<=4));
}

Ugh! And that's only for 1 missing value
As has been pointed out by other answers, you're much better off parsing the string in the 'usual' way: fgets() and strtok().

score -2 · Answer 11 · answered Oct 11 '20 at 19:54

If the inputstring has no embedded blanks, and no embedded spaces, pseudo code: scan the string replacing comma with ' '; follow up with sscanf with format paramter &s

... char inputstring[]="this,is,a,parameter,string"; after converting , to ' '

char inputstring[]="this is a parameter string";

int i=sscanf(inputstring,"&s&s&s&s&s",ptr1,ptr2,ptr3,ptr4,ptr5);

where i=4 and ptr1 has "this" and ptr5 has "strong".

How do I parse out the fields in a comma separated string using sscanf while supporting empty fields?

11 Answers11

Linked

Related