Is there a reverse fn() for strstr

Question

All,

I am trying to find a similar function to 'strstr' that searches a substring starting from the end towards the beginning of the string.

Thanks

Answer 1

I don't believe there is in the c string lib, but it would be trivial to write your own, On one condition, you know the length of the string or it is properly terminated.

Answer 2

There isn't one in the standard C library. You may be able to find one on the web, or you may have to write your own.

Answer 3

I don't know of one. One of the nice things about C is that if you write your own function, it's just as fast and efficient as the library ones. (This is totally not the case in many other languages.)

You could reverse the string and the substring, and then search.

Finally, the other thing people often do when the string library isn't good enough is to move to regular expressions.

Ok, I wrote both reverse() and rstrstr(), which might work if we are lucky. Get rid of __restrict for C++. You also might want to make the parameters const, but then you will need to cast the return value. To answer your comment question, you can get the index from the address of the substring by just substracting the original string pointer from it. OK:

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

char *reverse(const char * __restrict const s)
{
  if (s == NULL)
    return NULL;
  size_t i, len = strlen(s);
  char *r = malloc(len + 1);

  for(i = 0; i < len; ++i)
    r[i] = s[len - i - 1];
  r[len] = 0;
  return r;
}

char *rstrstr(char *__restrict s1, char *__restrict s2)
{
  size_t  s1len = strlen(s1);
  size_t  s2len = strlen(s2);
  char *s;

  if (s2len > s1len)
    return NULL;
  for (s = s1 + s1len - s2len; s >= s1; --s)
    if (strncmp(s, s2, s2len) == 0)
      return s;
  return NULL;
}

Answer 4

If you can use C++, you can search strings like this:

std::string::iterator found=std::search(haystack.rbegin(), haystack.rend(), needle.rbegin(), needle.rend()).base();
// => yields haystack.begin() if not found, otherwise, an iterator past-the end of the occurence of needle

Answer 5

Long story short:

Nope - there is no function in the C-library that does what you need..

But as others have pointed out: It's not rocket-science to write such a function...

Answer 6

No. This is one of the places that the C++ std::string class has an obvious advantage -- along with std::string::find(), there's also std::string::rfind().

Answer 7

Here is one. Testing it is an exercise I'll leave to you :)

Answer 8

One possible, if not entirely elegant, implementation might look like:

#include "string.h"

const char* rstrstr(const char* haystack, const char* needle)
{
  int needle_length = strlen(needle);
  const char* haystack_end = haystack + strlen(haystack) - needle_length;
  const char* p;
  size_t i;

  for(p = haystack_end; p >= haystack; --p)
  {
    for(i = 0; i < needle_length; ++i) {
      if(p[i] != needle[i])
        goto next;
    }
    return p;

    next:;
  }
  return 0;
}

Answer 9

Is there a C Library function to find the index to the last occurrence of a substring within a string?

Edit: As @hhafez notes in a comment below, the first solution I posted for this was inefficient and incorrect (because I advanced the pointer by target_length which worked fine in my silly test). You can find that version in the edit history.

Here is an implementation that starts at the end and works back:

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

const char *
findlast(const char *source, const char *target) {
    const char *current;
    const char *found = NULL;

    size_t target_length = strlen(target);
    current = source + strlen(source) - target_length;

    while ( current >= source ) {
        if ( (found = strstr(current, target)) ) {
            break;
        }
        current -= 1;
    }

    return found;
}

int main(int argc, char *argv[]) {
    if ( argc != 3 ) {
        fputs("invoke with source and search strings as arguments", stderr);
        return EXIT_FAILURE;
    }

    const char *found = findlast(argv[1], argv[2]);

    if ( found ) {
        printf("Last occurence of '%s' in '%s' is at offset %d\n",
                argv[2], argv[1], found - argv[1]
                );
    }
    return 0;
}

Output:

C:\Temp> st "this is a test string that tests this" test
Last occurence of 'test' in 'this is a test string that tests this' is 
at offset 27

Answer 10

Thanks for your answers! There is one more way which came from the MSDN forum. http://social.msdn.microsoft.com/Forums/en-US/vclanguage/thread/ed0f6ef9-8911-4879-accb-b3c778a09d94

Answer 11

I think you can still do it using library funtions.

1.Use strrev funtion to reverse the string.

2.Use strstr funtion to do whatever you want to do.

3.You can find start index (from reverse ) of the search string by subtracting start index of the search string from the length of original string.

Answer 12

The standard C library does not have a "reverse strstr" function, so you have to find or write your own.

I came up with a couple of solutions of my own, and added some testing and benchmarking code together with the other functions in this thread. For those curious, running on my laptop (Ubuntu karmic, amd64 architecture) the output looks like this:

$ gcc -O2 --std=c99 strrstr.c && ./a.out
#1 0.123 us last_strstr
#2 0.440 us theo
#3 0.460 us cordelia
#4 1.690 us digitalross
#5 7.700 us backwards_memcmp
#6 8.600 us sinan

Your results may be different and, depending on your compiler and library, the ordering of the results may also be different.

To get the offset (index) of the match from the beginning of the string, use pointer arithmetic:

char *match = last_strstr(haystack, needle);
ptrdiff_t index;
if (match != NULL)
    index = match - haystack;
else
    index = -1;

And now, the larch (note that this is purely in C, I do not know C++ well enough to give an answer for it):

/*
 * In response to
 * http://stackoverflow.com/questions/1634359/is-there-a-reverse-fn-for-strstr
 *
 * Basically, strstr but return last occurence, not first.
 *
 * This file contains several implementations and a harness to test and
 * benchmark them.
 *
 * Some of the implementations of the actual function are copied from
 * elsewhere; they are commented with the location. The rest of the coe
 * was written by Lars Wirzenius ([email protected]) and is hereby released into
 * the public domain. No warranty. If it turns out to be broken, you get
 * to keep the pieces.
 */


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


/* By liw. */
static char *last_strstr(const char *haystack, const char *needle)
{
    if (*needle == '\0')
        return (char *) haystack;

    char *result = NULL;
    for (;;) {
        char *p = strstr(haystack, needle);
        if (p == NULL)
            break;
        result = p;
        haystack = p + 1;
    }

    return result;
}


/* By liw. */
static char *backwards_memcmp(const char *haystack, const char *needle)
{
    size_t haylen = strlen(haystack);

    if (*needle == '\0')
        return (char *) haystack;

    size_t needlelen = strlen(needle);
    if (needlelen > haylen)
        return NULL;

    const char *p = haystack + haylen - needlelen;
    for (;;) {
        if (memcmp(p, needle, needlelen) == 0)
            return (char *) p;
        if (p == haystack)
            return NULL;
        --p;
    }
}


/* From http://stuff.mit.edu/afs/sipb/user/cordelia/Diplomacy/mapit/strrstr.c
 */
static char *cordelia(const char *s1, const char *s2)
{
 const char *sc1, *sc2, *psc1, *ps1;

 if (*s2 == '\0')
  return((char *)s1);

 ps1 = s1 + strlen(s1);

 while(ps1 != s1) {
  --ps1;
  for (psc1 = ps1, sc2 = s2; ; )
   if (*(psc1++) != *(sc2++))
    break;
   else if (*sc2 == '\0')
    return ((char *)ps1);
 }
 return ((char *)NULL);
}


/* From http://stackoverflow.com/questions/1634359/
   is-there-a-reverse-fn-for-strstr/1634398#1634398 (DigitalRoss). */
static char *reverse(const char *s)
{
  if (s == NULL)
    return NULL;
  size_t i, len = strlen(s);
  char *r = malloc(len + 1);

  for(i = 0; i < len; ++i)
    r[i] = s[len - i - 1];
  r[len] = 0;
  return r;
}
char *digitalross(const char *s1, const char *s2)
{
  size_t  s1len = strlen(s1);
  size_t  s2len = strlen(s2);
  const char *s;

  if (s2len == 0)
    return (char *) s1;
  if (s2len > s1len)
    return NULL;
  for (s = s1 + s1len - s2len; s >= s1; --s)
    if (strncmp(s, s2, s2len) == 0)
      return (char *) s;
  return NULL;
}


/* From http://stackoverflow.com/questions/1634359/
  is-there-a-reverse-fn-for-strstr/1634487#1634487 (Sinan Ünür). */

char *sinan(const char *source, const char *target)
{
    const char *current;
    const char *found = NULL;

    if (*target == '\0')
        return (char *) source;

    size_t target_length = strlen(target);
    current = source + strlen(source) - target_length;

    while ( current >= source ) {
        if ( (found = strstr(current, target)) ) {
            break;
        }
        current -= 1;
    }

    return (char *) found;
}


/* From http://stackoverflow.com/questions/1634359/
  is-there-a-reverse-fn-for-strstr/1634441#1634441 (Theo Spears). */
char *theo(const char* haystack, const char* needle)
{
  int needle_length = strlen(needle);
  const char* haystack_end = haystack + strlen(haystack) - needle_length;
  const char* p;
  size_t i;

  if (*needle == '\0')
    return (char *) haystack;
  for(p = haystack_end; p >= haystack; --p)
  {
    for(i = 0; i < needle_length; ++i) {
      if(p[i] != needle[i])
        goto next;
    }
    return (char *) p;

    next:;
  }
  return 0;
}


/*
 * The rest of this code is a test and timing harness for the various
 * implementations above.
 */


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


/* Check that the given function works. */
static bool works(const char *name, char *(*func)(const char *, const char *))
{
    struct {
        const char *haystack;
        const char *needle;
        int offset;
    } tests[] = {
        { "", "", 0 },
        { "", "x", -1 },
        { "x", "", 0 },
        { "x", "x", 0 },
        { "xy", "x", 0 },
        { "xy", "y", 1 },
        { "xyx", "x", 2 },
        { "xyx", "y", 1 },
        { "xyx", "z", -1 },
        { "xyx", "", 0 },
    };
    const int num_tests = sizeof(tests) / sizeof(tests[0]);
    bool ok = true;

    for (int i = 0; i < num_tests; ++i) {
        int offset;
        char *p = func(tests[i].haystack, tests[i].needle);
        if (p == NULL)
            offset = -1;
        else
            offset = p - tests[i].haystack;
        if (offset != tests[i].offset) {
            fprintf(stderr, "FAIL %s, test %d: returned %d, haystack = '%s', "
                            "needle = '%s', correct return %d\n",
                            name, i, offset, tests[i].haystack, tests[i].needle,
                            tests[i].offset);
            ok = false;
        }
    }
    return ok;
}


/* Dummy function for calibrating the measurement loop. */
static char *dummy(const char *haystack, const char *needle)
{
    return NULL;
}


/* Measure how long it will take to call the given function with the
   given arguments the given number of times. Return clock ticks. */
static clock_t repeat(char *(*func)(const char *, const char *),
                       const char *haystack, const char *needle,
                       long num_times)
{
    clock_t start, end;

    start = clock();
    for (long i = 0; i < num_times; ++i) {
        func(haystack, needle);
    }
    end = clock();
    return end - start;
}


static clock_t min(clock_t a, clock_t b)
{
    if (a < b)
        return a;
    else
        return b;
}


/* Measure the time to execute one call of a function, and return the
   number of CPU clock ticks (see clock(3)). */
static double timeit(char *(*func)(const char *, const char *))
{
    /* The arguments for the functions to be measured. We deliberately
       choose a case where the haystack is large and the needle is in
       the middle, rather than at either end. Obviously, any test data
       will favor some implementations over others. This is the weakest
       part of the benchmark. */

    const char haystack[] = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "b"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
                            "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa";
    const char needle[] = "b";

    /* First we find out how many repeats we need to do to get a sufficiently
       long measurement time. These functions are so fast that measuring
       only a small number of repeats will give wrong results. However,
       we don't want to do a ridiculously long measurement, either, so 
       start with one repeat and multiply it by 10 until the total time is
       about 0.2 seconds. 

       Finally, we measure the dummy function the same number of times
       to get rid of the call overhead.

       */

    clock_t mintime = 0.2 * CLOCKS_PER_SEC;
    clock_t clocks;
    long repeats = 1;
    for (;;) {
        clocks = repeat(func, haystack, needle, repeats);
        if (clocks >= mintime)
            break;
        repeats *= 10;
    }

    clocks = min(clocks, repeat(func, haystack, needle, repeats));
    clocks = min(clocks, repeat(func, haystack, needle, repeats));

    clock_t dummy_clocks;

    dummy_clocks = repeat(dummy, haystack, needle, repeats);
    dummy_clocks = min(dummy_clocks, repeat(dummy, haystack, needle, repeats));
    dummy_clocks = min(dummy_clocks, repeat(dummy, haystack, needle, repeats));

    return (double) (clocks - dummy_clocks) / repeats / CLOCKS_PER_SEC;
}


/* Array of all functions. */
struct func {
    const char *name;
    char *(*func)(const char *, const char *);
    double secs;
} funcs[] = {
#define X(func) { #func, func, 0 }
    X(last_strstr),
    X(backwards_memcmp),
    X(cordelia),
    X(digitalross),
    X(sinan),
    X(theo),
#undef X
};
const int num_funcs = sizeof(funcs) / sizeof(funcs[0]);


/* Comparison function for qsort, comparing timings. */
int funcmp(const void *a, const void *b)
{
    const struct func *aa = a;
    const struct func *bb = b;

    if (aa->secs < bb->secs)
        return -1;
    else if (aa->secs > bb->secs)
        return 1;
    else
        return 0;
}


int main(void)
{

    bool ok = true;
    for (int i = 0; i < num_funcs; ++i) {
        if (!works(funcs[i].name, funcs[i].func)) {
            fprintf(stderr, "%s does not work\n", funcs[i].name);            
            ok = false;
        }
    }
    if (!ok)
        return EXIT_FAILURE;

    for (int i = 0; i < num_funcs; ++i)
        funcs[i].secs = timeit(funcs[i].func);
    qsort(funcs, num_funcs, sizeof(funcs[0]), funcmp);
    for (int i = 0; i < num_funcs; ++i)
        printf("#%d %.3f us %s\n", i+1, funcs[i].secs * 1e6, funcs[i].name);

    return 0;
}