C中strtok和strsep有什么区别

Question

Could someone explain me what differences there are between strtok() and strsep() ? What are the advantages and disadvantages of them? And why would I pick one over the other one.

Answer 1

One major difference between strtok() and strsep() is that strtok() is standardized (by the C standard, and hence also by POSIX) but strsep() is not standardized (by C or POSIX; it is available in the GNU C Library, and originated on BSD). Thus, portable code is more likely to use strtok() than strsep() .

Another difference is that calls to the strsep() function on different strings can be interleaved, whereas you cannot do that with strtok() (though you can with strtok_r() ). So, using strsep() in a library doesn't break other code accidentally, whereas using strtok() in a library function must be documented because other code using strtok() at the same time cannot call the library function.

The manual page for strsep() at kernel.org says:

The strsep() function was introduced as a replacement for strtok(3), since the latter cannot handle empty fields.

Thus, the other major difference is the one highlighted by George Gaál in his answer; strtok() permits multiple delimiters between a single token, whereas strsep() expects a single delimiter between tokens, and interprets adjacent delimiters as an empty token.

Both strsep() and strtok() modify their input strings and neither lets you identify which delimiter character marked the end of the token (because both write a NUL '\\0' over the separator after the end of the token).

When to use them?

• You would use strsep() when you want empty tokens rather than allowing multiple delimiters between tokens, and when you don't mind about portability.
• You would use strtok_r() when you want to allow multiple delimiters between tokens and you don't want empty tokens (and POSIX is sufficiently portable for you).
• You would only use strtok() when someone threatens your life if you don't do so. And you'd only use it for long enough to get you out of the life-threatening situation; you would then abandon all use of it once more. It is poisonous; do not use it. It would be better to write your own strtok_r() or strsep() than to use strtok() .

Why is strtok() poisonous?

The strtok() function is poisonous if used in a library function. If your library function uses strtok() , it must be documented clearly.

That's because:

1. If any calling function is using strtok() and calls your function that also uses strtok() , you break the calling function.
2. If your function calls any function that calls strtok() , that will break your function's use of strtok() .
3. If your program is multithreaded, at most one thread can be using strtok() at any given time — across a sequence of strtok() calls.

The root of this problem is the saved state between calls that allows strtok() to continue where it left off. There is no sensible way to fix the problem other than "do not use strtok() ".

• You can use strsep() if it is available.
• You can use POSIX's strtok_r() if it is available.
• You can use Microsoft's strtok_s() if it is available.
• Nominally, you could use the ISO/IEC 9899:2011 Annex K.3.7.3.1 function strtok_s() , but its interface is different from both strtok_r() and Microsoft's strtok_s() .

BSD strsep() :

char *strsep(char **stringp, const char *delim);

POSIX strtok_r() :

char *strtok_r(char *restrict s, const char *restrict sep, char **restrict state);

Microsoft strtok_s() :

char *strtok_s(char *strToken, const char *strDelimit, char **context);

Annex K strtok_s() :

char *strtok_s(char * restrict s1, rsize_t * restrict s1max,
               const char * restrict s2, char ** restrict ptr);

Note that this has 4 arguments, not 3 as in the other two variants on strtok() .

Answer 2

From The GNU C Library manual - Finding Tokens in a String :

One difference between strsep and strtok_r is that if the input string contains more than one character from delimiter in a row strsep returns an empty string for each pair of characters from delimiter. This means that a program normally should test for strsep returning an empty string before processing it.

Answer 3

First difference in strtok() and strsep() is the way they handle contiguous delimiter characters in the input string.

Contiguous delimiter characters handling by strtok() :

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

int main(void) {
    const char* teststr = "aaa-bbb --ccc-ddd"; //Contiguous delimiters between bbb and ccc sub-string
    const char* delims = " -";  // delimiters - space and hyphen character
    char* token;
    char* ptr = strdup(teststr);

    if (ptr == NULL) {
        fprintf(stderr, "strdup failed");
        exit(EXIT_FAILURE);
    }

    printf ("Original String: %s\n", ptr);

    token = strtok (ptr, delims);
    while (token != NULL) {
        printf("%s\n", token);
        token = strtok (NULL, delims);
    }

    printf ("Original String: %s\n", ptr);
    free (ptr);
    return 0;
}

Output:

# ./example1_strtok
Original String: aaa-bbb --ccc-ddd
aaa
bbb
ccc
ddd
Original String: aaa

In the output, you can see the token "bbb" and "ccc" one after another. strtok() does not indicate the occurrence of contiguous delimiter characters . Also, the strtok() modify the input string .

Contiguous delimiter characters handling by strsep() :

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

int main(void) {
    const char* teststr = "aaa-bbb --ccc-ddd"; //Contiguous delimiters between bbb and ccc sub-string
    const char* delims = " -";  // delimiters - space and hyphen character
    char* token;
    char* ptr1;
    char* ptr = strdup(teststr);

    if (ptr == NULL) {
        fprintf(stderr, "strdup failed");
        exit(EXIT_FAILURE);
    }

    ptr1 = ptr;

    printf ("Original String: %s\n", ptr);
    while ((token = strsep(&ptr1, delims)) != NULL) {
        if (*token == '\0') {
            token = "<empty>";
        }
        printf("%s\n", token);
    }

    if (ptr1 == NULL) // This is just to show that the strsep() modifies the pointer passed to it
        printf ("ptr1 is NULL\n");
    printf ("Original String: %s\n", ptr);
    free (ptr);
    return 0;
}

Output:

# ./example1_strsep
Original String: aaa-bbb --ccc-ddd
aaa
bbb
<empty>             <==============
<empty>             <==============
ccc
ddd
ptr1 is NULL
Original String: aaa

In the output, you can see the two empty string (indicated through <empty> ) between bbb and ccc . Those two empty strings are for "--" between "bbb" and "ccc" . When strsep() found a delimiter character ' ' after "bbb" , it replaced delimiter character with '\\0' character and returned "bbb" . After this, strsep() found another delimiter character '-' . Then it replaced delimiter character with '\\0' character and returned the empty string. Same is for the next delimiter character.

Contiguous delimiter characters are indicated when strsep() returns a pointer to a null character (that is, a character with the value '\\0' ).

The strsep() modify the input string as well as the pointer whose address passed as first argument to strsep() .

Second difference is, strtok() relies on a static variable to keep track of the current parse location within a string. This implementation requires to completely parse one string before beginning a second string . But this is not the case with strsep() .

Calling strtok() when another strtok() is not finished:

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

void another_function_callng_strtok(void)
{
    char str[] ="ttt -vvvv";
    char* delims = " -";
    char* token;

    printf ("Original String: %s\n", str);
    token = strtok (str, delims);
    while (token != NULL) {
        printf ("%s\n", token);
        token = strtok (NULL, delims);
    }
    printf ("another_function_callng_strtok: I am done.\n");
}

void function_callng_strtok ()
{
    char str[] ="aaa --bbb-ccc";
    char* delims = " -";
    char* token;

    printf ("Original String: %s\n", str);
    token = strtok (str, delims);
    while (token != NULL)
    {
        printf ("%s\n",token);
        another_function_callng_strtok();
        token = strtok (NULL, delims);
    }
}

int main(void) {
    function_callng_strtok();
    return 0;
}

Output:

# ./example2_strtok
Original String: aaa --bbb-ccc
aaa
Original String: ttt -vvvv
ttt
vvvv
another_function_callng_strtok: I am done.

The function function_callng_strtok() only print token "aaa" and does not print the rest of the tokens of input string because it calls another_function_callng_strtok() which in turn call strtok() and it set the static pointer of strtok() to NULL when it finishes with extracting all the tokens. The control comes back to function_callng_strtok() while loop, strtok() returns NULL due to the static pointer pointing to NULL and which make the loop condition false and loop exits.

Calling strsep() when another strsep() is not finished:

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

void another_function_callng_strsep(void)
{
    char str[] ="ttt -vvvv";
    const char* delims = " -";
    char* token;
    char* ptr = str;

    printf ("Original String: %s\n", str);
    while ((token = strsep(&ptr, delims)) != NULL) {
        if (*token == '\0') {
            token = "<empty>";
        }
        printf("%s\n", token);
    }
    printf ("another_function_callng_strsep: I am done.\n");
}

void function_callng_strsep ()
{
    char str[] ="aaa --bbb-ccc";
    const char* delims = " -";
    char* token;
    char* ptr = str;

    printf ("Original String: %s\n", str);
    while ((token = strsep(&ptr, delims)) != NULL) {
        if (*token == '\0') {
            token = "<empty>";
        }
        printf("%s\n", token);
        another_function_callng_strsep();
    }
}

int main(void) {
    function_callng_strsep();
    return 0;
}

Output:

# ./example2_strsep
Original String: aaa --bbb-ccc
aaa
Original String: ttt -vvvv
ttt
<empty>
vvvv
another_function_callng_strsep: I am done.
<empty>
Original String: ttt -vvvv
ttt
<empty>
vvvv
another_function_callng_strsep: I am done.
<empty>
Original String: ttt -vvvv
ttt
<empty>
vvvv
another_function_callng_strsep: I am done.
bbb
Original String: ttt -vvvv
ttt
<empty>
vvvv
another_function_callng_strsep: I am done.
ccc
Original String: ttt -vvvv
ttt
<empty>
vvvv
another_function_callng_strsep: I am done.

Here you can see, calling strsep() before completely parse one string doesn't makes any difference.

So, the disadvantage of strtok() and strsep() is that both modify the input string but strsep() has couple of advantages over strtok() as illustrated above.

From strsep :

The strsep() function is intended as a replacement for the strtok() function. While the strtok() function should be preferred for portability reasons (it conforms to ISO/IEC 9899:1990 (``ISO C90'')) it is unable to handle empty fields, ie, detect fields delimited by two adjacent delimiter characters, or to be used for more than a single string at a time. The strsep() function first appeared in 4.4BSD.

---

For reference:

• strtok() and strtok_r()
• strsep()