1 /* c-strstr.c -- substring search in C locale
2 Copyright (C) 2005-2007 Free Software Foundation, Inc.
3 Written by Bruno Haible <bruno@clisp.org>, 2005, 2007.
5 This program is free software: you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 the Free Software Foundation; either version 3 of the License, or
8 (at your option) any later version.
10 This program is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 GNU General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program. If not, see <http://www.gnu.org/licenses/>. */
29 /* Knuth-Morris-Pratt algorithm.
30 See http://en.wikipedia.org/wiki/Knuth-Morris-Pratt_algorithm
31 Return a boolean indicating success. */
33 knuth_morris_pratt (const char *haystack, const char *needle,
36 size_t m = strlen (needle);
38 /* Allocate the table. */
39 size_t *table = (size_t *) nmalloca (m, sizeof (size_t));
44 0 < table[i] <= i is defined such that
45 forall 0 < x < table[i]: needle[x..i-1] != needle[0..i-1-x],
46 and table[i] is as large as possible with this property.
50 needle[table[i]..i-1] = needle[0..i-1-table[i]].
52 rhaystack[0..i-1] == needle[0..i-1]
53 and exists h, i <= h < m: rhaystack[h] != needle[h]
55 forall 0 <= x < table[i]: rhaystack[x..x+m-1] != needle[0..m-1].
56 table[0] remains uninitialized. */
60 /* i = 1: Nothing to verify for x = 0. */
64 for (i = 2; i < m; i++)
66 /* Here: j = i-1 - table[i-1].
67 The inequality needle[x..i-1] != needle[0..i-1-x] is known to hold
68 for x < table[i-1], by induction.
69 Furthermore, if j>0: needle[i-1-j..i-2] = needle[0..j-1]. */
70 unsigned char b = (unsigned char) needle[i - 1];
74 /* Invariants: The inequality needle[x..i-1] != needle[0..i-1-x]
75 is known to hold for x < i-1-j.
76 Furthermore, if j>0: needle[i-1-j..i-2] = needle[0..j-1]. */
77 if (b == (unsigned char) needle[j])
79 /* Set table[i] := i-1-j. */
83 /* The inequality needle[x..i-1] != needle[0..i-1-x] also holds
84 for x = i-1-j, because
85 needle[i-1] != needle[j] = needle[i-1-x]. */
88 /* The inequality holds for all possible x. */
92 /* The inequality needle[x..i-1] != needle[0..i-1-x] also holds
93 for i-1-j < x < i-1-j+table[j], because for these x:
95 = needle[x-(i-1-j)..j-1]
96 != needle[0..j-1-(x-(i-1-j))] (by definition of table[j])
98 hence needle[x..i-1] != needle[0..i-1-x].
100 needle[i-1-j+table[j]..i-2]
101 = needle[table[j]..j-1]
102 = needle[0..j-1-table[j]] (by definition of table[j]). */
105 /* Here: j = i - table[i]. */
109 /* Search, using the table to accelerate the processing. */
112 const char *rhaystack;
113 const char *phaystack;
117 rhaystack = haystack;
118 phaystack = haystack;
119 /* Invariant: phaystack = rhaystack + j. */
120 while (*phaystack != '\0')
121 if ((unsigned char) needle[j] == (unsigned char) *phaystack)
127 /* The entire needle has been found. */
128 *resultp = rhaystack;
134 /* Found a match of needle[0..j-1], mismatch at needle[j]. */
135 rhaystack += table[j];
140 /* Found a mismatch at needle[0] already. */
150 /* Find the first occurrence of NEEDLE in HAYSTACK. */
152 c_strstr (const char *haystack, const char *needle)
154 /* Be careful not to look at the entire extent of haystack or needle
155 until needed. This is useful because of these two cases:
156 - haystack may be very long, and a match of needle found early,
157 - needle may be very long, and not even a short initial segment of
158 needle may be found in haystack. */
161 /* Minimizing the worst-case complexity:
162 Let n = strlen(haystack), m = strlen(needle).
163 The naïve algorithm is O(n*m) worst-case.
164 The Knuth-Morris-Pratt algorithm is O(n) worst-case but it needs a
166 To achieve linear complexity and yet amortize the cost of the memory
167 allocation, we activate the Knuth-Morris-Pratt algorithm only once
168 the naïve algorithm has already run for some time; more precisely,
170 - the outer loop count is >= 10,
171 - the average number of comparisons per outer loop is >= 5,
172 - the total number of comparisons is >= m.
173 But we try it only once. If the memory allocation attempt failed,
174 we don't retry it. */
176 size_t outer_loop_count = 0;
177 size_t comparison_count = 0;
178 size_t last_ccount = 0; /* last comparison count */
179 const char *needle_last_ccount = needle; /* = needle + last_ccount */
181 /* Speed up the following searches of needle by caching its first
183 unsigned char b = (unsigned char) *needle;
188 if (*haystack == '\0')
192 /* See whether it's advisable to use an asymptotically faster
195 && outer_loop_count >= 10
196 && comparison_count >= 5 * outer_loop_count)
198 /* See if needle + comparison_count now reaches the end of
200 if (needle_last_ccount != NULL)
202 needle_last_ccount +=
203 strnlen (needle_last_ccount, comparison_count - last_ccount);
204 if (*needle_last_ccount == '\0')
205 needle_last_ccount = NULL;
206 last_ccount = comparison_count;
208 if (needle_last_ccount == NULL)
210 /* Try the Knuth-Morris-Pratt algorithm. */
213 knuth_morris_pratt (haystack, needle - 1, &result);
215 return (char *) result;
222 if ((unsigned char) *haystack == b)
223 /* The first character matches. */
225 const char *rhaystack = haystack + 1;
226 const char *rneedle = needle;
228 for (;; rhaystack++, rneedle++)
230 if (*rneedle == '\0')
232 return (char *) haystack;
233 if (*rhaystack == '\0')
237 if ((unsigned char) *rhaystack != (unsigned char) *rneedle)
238 /* Nothing in this round. */
245 return (char *) haystack;