3

I have a string containg spaces and tags like:

<note label="description">sp|P02671|FIBA_HUMAN Fibrinogen alpha chain OS=Homo sapiens GN=FGA PE=1 SV=2</note>

I want to capture only the part after the description tag and before 'OS=' and was wondering if it makes sense to do with sscanf with a custom notation (see my current work in progress at bottom of page) or if it's better to use a second strstr to parse for OS.

Thanks in advance

-- Additional info --

if ( (p_str = (char*) strstr(buffer,"\"description\">")) ) {
  sscanf(p_str+14,"%[^OS]",(file+teller)->description);
}

PS: %[^OS] breaks the string at the first occurance of a 'O', i would love to know how to enter a set of characters (if that's possible).

Bas Jansen
  • 3,273
  • 5
  • 30
  • 66

1 Answers1

1

I would rather use strstr twice, as follow:

#include <assert.h>
#include <stdlib.h>
#include <string.h>

char *f (const char *s) {
#define START_SYM   "\"description\""
#define END_SYM     "OS="

    char *dst   = NULL;
    char *start = strstr(s, START_SYM);
    char *end   = strstr(s, END_SYM);

    if (start != NULL && end != NULL) {
        ptrdiff_t diff = end - start;

        assert(diff > 0);
        dst = malloc(diff + 1);

        if (dst != NULL) {
            memcpy(dst, start + sizeof START_SYM, diff);
            dst[diff] = '\0';
        }
    }

    return dst;
}
md5
  • 23,373
  • 3
  • 44
  • 93
  • What are your thoughts about the performance impact of performing a second strstr and the memcpy in comparison to using the single strstr with a sscanf (if it's even possible)? – Bas Jansen Feb 25 '13 at 17:23
  • According to the length of your string, it should be negligeable. `sscanf` is a formatted function, whereas `memcpy` is a builtin on many compilers, perhaps optimized with vector operations which are very fast on modern processors. Use a profiler if you really care about such details, although I'm quite sure there are bigger bottlenecks in your application. ;-) – md5 Feb 25 '13 at 17:34