Knuth-Morris-Pratt implementation in pure C

Question

I have the next KMP-implementation:

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

int kmp(char substr[], char str[])
{
   int i, j, N, M;

   N = strlen(str);
   M = strlen(substr);

   int *d = (int*)malloc(M * sizeof(int));
   d[0] = 0;

   for(i = 0, j = 0; i < M; i++)
   {
      while(j > 0 && substr[j] != substr[i])
      {
         j = d[j - 1];
      }

      if(substr[j] == substr[i])
      {
         j++;
         d[i] = j;
      }
   }

   for(i = 0, j = 0; i < N; i++)
   {
      while(j > 0 && substr[j] != str[i])
      {
         j = d[j - 1];
      }

      if(substr[j] == str[i])
      {
         j++;
      }

      if(j == M)
      {
         free(d);
         return i - j + 1;
      }
   }

   free(d);

   return -1;
}

int main(void)
{
   char substr[] = "World",
      str[] = "Hello World!";

   int pos = kmp(substr, str);

   printf("position starts at: %i\r\n", pos);

   return 0;
}

You can test it here: http://liveworkspace.org/code/d2e7b3be72083c72ed768720f4716f80

It works well on small strings, and I have tested it with a large loop, on this way all is fine.

But if I change the substring I'm searching for and the complete string to these:

char substr[] = "%end%",
str[] = "<h1>The result is: <%lua% oleg = { x = 0xa }
         table.insert(oleg, y) oleg.y = 5 print(oleg.y) %end%></h1>";

Only after first try, this implementation fails...

Please, could you help me with repairing implementation of KMP to make the algorithm work with such data in strings...

Answer 1

In one place you deviate from your source, the source has

while(j>0 && p[j]!=p[i]) j = d[j-1];
    if(p[j]==p[i])
        j++;
        d[i]=j;

while you have

while(j > 0 && substr[j] != substr[i])
{
    j = d[j - 1];
}
if(substr[j] == substr[i])
{
    j++;
    d[i] = j;
}

being deceived by the source's indentation. In the source, there are no braces around the if() branch, so only the increment j++; is controlled by the if; d[i] = j; is unconditional.

Then, the source has an error, probably due to the unusual use of indices. The correct way to set up the array is

int *d = (int*)malloc(M * sizeof(int));
d[0] = 0;

for(i = 1, j = 0; i < M; i++)
{
    while(j > 0 && substr[j-1] != substr[i-1])
    {
        j = d[j - 1];
    }

    if(substr[j] == substr[i])
        j++;
    d[i] = j;
}

But it's confusing, since the setup here uses the indices i-1 and j-1 as well as i and j to determine d[i]. The usual way to implement it is different; the way it is implemented in C#. Since that's the form you find in most sources, it's far easier to convince yourself of the correctness of that.