1 /* Copyright (C) 1991,92,93,94,96,97,98,2000,2004,2007 Free Software Foundation, Inc.
2 This file is part of the GNU C Library.
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 2, or (at your option)
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License along
15 with this program; if not, write to the Free Software Foundation,
16 Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */
29 # define __builtin_expect(expr, val) (expr)
32 /* Knuth-Morris-Pratt algorithm.
33 See http://en.wikipedia.org/wiki/Knuth-Morris-Pratt_algorithm
34 Return a boolean indicating success. */
37 knuth_morris_pratt (const char *haystack, const char *last_haystack,
38 const char *needle, size_t m,
41 /* Allocate the table. */
42 size_t *table = (size_t *) nmalloca (m, sizeof (size_t));
47 0 < table[i] <= i is defined such that
48 forall 0 < x < table[i]: needle[x..i-1] != needle[0..i-1-x],
49 and table[i] is as large as possible with this property.
53 needle[table[i]..i-1] = needle[0..i-1-table[i]].
55 rhaystack[0..i-1] == needle[0..i-1]
56 and exists h, i <= h < m: rhaystack[h] != needle[h]
58 forall 0 <= x < table[i]: rhaystack[x..x+m-1] != needle[0..m-1].
59 table[0] remains uninitialized. */
63 /* i = 1: Nothing to verify for x = 0. */
67 for (i = 2; i < m; i++)
69 /* Here: j = i-1 - table[i-1].
70 The inequality needle[x..i-1] != needle[0..i-1-x] is known to hold
71 for x < table[i-1], by induction.
72 Furthermore, if j>0: needle[i-1-j..i-2] = needle[0..j-1]. */
73 unsigned char b = (unsigned char) needle[i - 1];
77 /* Invariants: The inequality needle[x..i-1] != needle[0..i-1-x]
78 is known to hold for x < i-1-j.
79 Furthermore, if j>0: needle[i-1-j..i-2] = needle[0..j-1]. */
80 if (b == (unsigned char) needle[j])
82 /* Set table[i] := i-1-j. */
86 /* The inequality needle[x..i-1] != needle[0..i-1-x] also holds
87 for x = i-1-j, because
88 needle[i-1] != needle[j] = needle[i-1-x]. */
91 /* The inequality holds for all possible x. */
95 /* The inequality needle[x..i-1] != needle[0..i-1-x] also holds
96 for i-1-j < x < i-1-j+table[j], because for these x:
98 = needle[x-(i-1-j)..j-1]
99 != needle[0..j-1-(x-(i-1-j))] (by definition of table[j])
101 hence needle[x..i-1] != needle[0..i-1-x].
103 needle[i-1-j+table[j]..i-2]
104 = needle[table[j]..j-1]
105 = needle[0..j-1-table[j]] (by definition of table[j]). */
108 /* Here: j = i - table[i]. */
112 /* Search, using the table to accelerate the processing. */
115 const char *rhaystack;
116 const char *phaystack;
120 rhaystack = haystack;
121 phaystack = haystack;
122 /* Invariant: phaystack = rhaystack + j. */
123 while (phaystack != last_haystack)
124 if ((unsigned char) needle[j] == (unsigned char) *phaystack)
130 /* The entire needle has been found. */
131 *resultp = rhaystack;
137 /* Found a match of needle[0..j-1], mismatch at needle[j]. */
138 rhaystack += table[j];
143 /* Found a mismatch at needle[0] already. */
153 /* Return the first occurrence of NEEDLE in HAYSTACK. Return HAYSTACK
154 if NEEDLE_LEN is 0, otherwise NULL if NEEDLE is not found in
157 memmem (const void *haystack_start, size_t haystack_len,
158 const void *needle_start, size_t needle_len)
160 /* Operating with void * is awkward. */
161 const char *haystack = (const char *) haystack_start;
162 const char *needle = (const char *) needle_start;
163 const char *last_haystack = haystack + haystack_len;
164 const char *last_needle = needle + needle_len;
167 /* The first occurrence of the empty string is deemed to occur at
168 the beginning of the string. */
169 return (void *) haystack;
171 /* Sanity check, otherwise the loop might search through the whole
173 if (__builtin_expect (haystack_len < needle_len, 0))
176 /* Use optimizations in memchr when possible. */
177 if (__builtin_expect (needle_len == 1, 0))
178 return memchr (haystack, (unsigned char) *needle, haystack_len);
180 /* Minimizing the worst-case complexity:
181 Let n = haystack_len, m = needle_len.
182 The naïve algorithm is O(n*m) worst-case.
183 The Knuth-Morris-Pratt algorithm is O(n) worst-case but it needs a
185 To achieve linear complexity and yet amortize the cost of the
186 memory allocation, we activate the Knuth-Morris-Pratt algorithm
187 only once the naïve algorithm has already run for some time; more
189 - the outer loop count is >= 10,
190 - the average number of comparisons per outer loop is >= 5,
191 - the total number of comparisons is >= m.
192 But we try it only once. If the memory allocation attempt failed,
193 we don't retry it. */
196 size_t outer_loop_count = 0;
197 size_t comparison_count = 0;
199 /* Speed up the following searches of needle by caching its first
205 if (haystack == last_haystack)
209 /* See whether it's advisable to use an asymptotically faster
212 && outer_loop_count >= 10
213 && comparison_count >= 5 * outer_loop_count)
215 /* See if needle + comparison_count now reaches the end of
217 if (comparison_count >= needle_len)
219 /* Try the Knuth-Morris-Pratt algorithm. */
221 if (knuth_morris_pratt (haystack, last_haystack,
222 needle - 1, needle_len, &result))
223 return (void *) result;
231 /* The first byte matches. */
233 const char *rhaystack = haystack + 1;
234 const char *rneedle = needle;
236 for (;; rhaystack++, rneedle++)
238 if (rneedle == last_needle)
240 return (void *) haystack;
241 if (rhaystack == last_haystack)
245 if (*rhaystack != *rneedle)
246 /* Nothing in this round. */