src/string/memmem.c

   1 #define _GNU_SOURCE
   2 #include <string.h>
   3 #include <stdint.h>
   4
   5 static char *twobyte_memmem(const unsigned char *h, size_t k, const unsigned char *n)
   6 {
   7         uint16_t nw = n[0]<<8 | n[1], hw = h[0]<<8 | h[1];
   8         for (h+=2, k-=2; k; k--, hw = hw<<8 | *h++)
   9                 if (hw == nw) return (char *)h-2;
  10         return hw == nw ? (char *)h-2 : 0;
  11 }
  12
  13 static char *threebyte_memmem(const unsigned char *h, size_t k, const unsigned char *n)
  14 {
  15         uint32_t nw = (uint32_t)n[0]<<24 | n[1]<<16 | n[2]<<8;
  16         uint32_t hw = (uint32_t)h[0]<<24 | h[1]<<16 | h[2]<<8;
  17         for (h+=3, k-=3; k; k--, hw = (hw|*h++)<<8)
  18                 if (hw == nw) return (char *)h-3;
  19         return hw == nw ? (char *)h-3 : 0;
  20 }
  21
  22 static char *fourbyte_memmem(const unsigned char *h, size_t k, const unsigned char *n)
  23 {
  24         uint32_t nw = (uint32_t)n[0]<<24 | n[1]<<16 | n[2]<<8 | n[3];
  25         uint32_t hw = (uint32_t)h[0]<<24 | h[1]<<16 | h[2]<<8 | h[3];
  26         for (h+=4, k-=4; k; k--, hw = hw<<8 | *h++)
  27                 if (hw == nw) return (char *)h-4;
  28         return hw == nw ? (char *)h-4 : 0;
  29 }
  30
  31 #define MAX(a,b) ((a)>(b)?(a):(b))
  32 #define MIN(a,b) ((a)<(b)?(a):(b))
  33
  34 #define BITOP(a,b,op) \
  35  ((a)[(size_t)(b)/(8*sizeof *(a))] op (size_t)1<<((size_t)(b)%(8*sizeof *(a))))
  36
  37 static char *twoway_memmem(const unsigned char *h, const unsigned char *z, const unsigned char *n, size_t l)
  38 {
  39         size_t i, ip, jp, k, p, ms, p0, mem, mem0;
  40         size_t byteset[32 / sizeof(size_t)] = { 0 };
  41         size_t shift[256];
  42
  43         /* Computing length of needle and fill shift table */
  44         for (i=0; i<l; i++)
  45                 BITOP(byteset, n[i], |=), shift[n[i]] = i+1;
  46
  47         /* Compute maximal suffix */
  48         ip = -1; jp = 0; k = p = 1;
  49         while (jp+k<l) {
  50                 if (n[ip+k] == n[jp+k]) {
  51                         if (k == p) {
  52                                 jp += p;
  53                                 k = 1;
  54                         } else k++;
  55                 } else if (n[ip+k] > n[jp+k]) {
  56                         jp += k;
  57                         k = 1;
  58                         p = jp - ip;
  59                 } else {
  60                         ip = jp++;
  61                         k = p = 1;
  62                 }
  63         }
  64         ms = ip;
  65         p0 = p;
  66
  67         /* And with the opposite comparison */
  68         ip = -1; jp = 0; k = p = 1;
  69         while (jp+k<l) {
  70                 if (n[ip+k] == n[jp+k]) {
  71                         if (k == p) {
  72                                 jp += p;
  73                                 k = 1;
  74                         } else k++;
  75                 } else if (n[ip+k] < n[jp+k]) {
  76                         jp += k;
  77                         k = 1;
  78                         p = jp - ip;
  79                 } else {
  80                         ip = jp++;
  81                         k = p = 1;
  82                 }
  83         }
  84         if (ip+1 > ms+1) ms = ip;
  85         else p = p0;
  86
  87         /* Periodic needle? */
  88         if (memcmp(n, n+p, ms+1)) {
  89                 mem0 = 0;
  90                 p = MAX(ms, l-ms-1) + 1;
  91         } else mem0 = l-p;
  92         mem = 0;
  93
  94         /* Search loop */
  95         for (;;) {
  96                 /* If remainder of haystack is shorter than needle, done */
  97                 if (z-h < l) return 0;
  98
  99                 /* Check last byte first; advance by shift on mismatch */
 100                 if (BITOP(byteset, h[l-1], &)) {
 101                         k = l-shift[h[l-1]];
 102                         if (k) {
 103                                 if (k < mem) k = mem;
 104                                 h += k;
 105                                 mem = 0;
 106                                 continue;
 107                         }
 108                 } else {
 109                         h += l;
 110                         mem = 0;
 111                         continue;
 112                 }
 113
 114                 /* Compare right half */
 115                 for (k=MAX(ms+1,mem); k<l && n[k] == h[k]; k++);
 116                 if (k < l) {
 117                         h += k-ms;
 118                         mem = 0;
 119                         continue;
 120                 }
 121                 /* Compare left half */
 122                 for (k=ms+1; k>mem && n[k-1] == h[k-1]; k--);
 123                 if (k <= mem) return (char *)h;
 124                 h += p;
 125                 mem = mem0;
 126         }
 127 }
 128
 129 void *memmem(const void *h0, size_t k, const void *n0, size_t l)
 130 {
 131         const unsigned char *h = h0, *n = n0;
 132
 133         /* Return immediately on empty needle */
 134         if (!l) return (void *)h;
 135
 136         /* Return immediately when needle is longer than haystack */
 137         if (k<l) return 0;
 138
 139         /* Use faster algorithms for short needles */
 140         h = memchr(h0, *n, k);
 141         if (!h || l==1) return (void *)h;
 142         k -= h - (const unsigned char *)h0;
 143         if (k<l) return 0;
 144         if (l==2) return twobyte_memmem(h, k, n);
 145         if (l==3) return threebyte_memmem(h, k, n);
 146         if (l==4) return fourbyte_memmem(h, k, n);
 147
 148         return twoway_memmem(h, h+k, n, l);
 149 }