Chris said:
Agreed, as long as ISO C don't specify threads and the current (lack of)
signal definition exist, the "_r" functions is better left to POSIX.
Well, even without threads, strtok() remains an "accident waiting
to happen":
static const char ws[] = " \t\n"; /* or even " \t\b\f\v\r\n" */
char *tok;
/* do some work on the whitespace-separated keywords */
for (tok = strtok(str, ws); tok != NULL; tok = strtok(NULL, ws))
operate_on(tok);
If operate_on() is a function, and if operate_on() itself calls
strtok() with a non-NULL first parameter, the code breaks. The
POSIX strtok_r() fixes this, although that is the only silly part
of strtok() that it fixes. (The BSD strsep() fixes two "bad
behaviors", at the cost of making the caller merge adjacent separators
if that is desired.
I really don't know why your strsep() was left out of C99, perhaps
strtok() was regarded as a minor problem, as experienced C programmers
are aware of strtok() and avoid using it. At least I haven't used it for
10+ years, after some surprising parsing results.
The Linux man page is clear:
Name
strtok, strtok_r - extract tokens from strings
[...]
Bugs
Avoid using these functions. If you do use them, note that:
These functions modify their first argument.
These functions cannot be used on constant strings.
The identity of the delimiting character is lost.
The strtok() function uses a static buffer while parsing, so it's not
thread safe. Use strtok_r() if this matters to you.
So I think adding strsep() would be better than adding strtok_r().
Now, strtok_s() has arrived too.
After reading this:
#define __STDC_WANT_LIB_EXT1__ 1
#include <string.h>
static char str1[] = "?a???b,,,#c";
static char str2[] = "\t \t";
char *t, *ptr1, *ptr2;
rsize_t max1 = sizeof(str1);
rsize_t max2 = sizeof(str2);
t = strtok_s(str1, &max1, "?", &ptr1); // t points to the token "a"
t = strtok_s(NULL, &max1, ",", &ptr1); // t points to the token "??b"
t = strtok_s(str2, &max2, " \t", &ptr2); // t is a null pointer
t = strtok_s(NULL, &max1, "#,", &ptr1); // t points to the token "c"
t = strtok_s(NULL, &max1, "?", &ptr1); // t is a null pointer
I think this interface fails on simplicity, and yes, I agree that
strsep() is the better alternative, but Linux man page say:
Name
strsep - extract token from string
[...]
Bugs
This function suffers from the same problems as strtok(). In particular,
it modifies the original string. Avoid it.
I wouldn't put it as strong as "Avoid it", as a simple duplicate string
beforehand do.
