1 /* Case-insensitive searching in a string.
2 Copyright (C) 2005-2007 Free Software Foundation, Inc.
3 Written by Bruno Haible <bruno@clisp.org>, 2005.
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 2, or (at your option)
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, write to the Free Software Foundation,
17 Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */
26 #include <stddef.h> /* for NULL, in case a nonstandard string.h lacks it */
30 #define TOLOWER(Ch) (isupper (Ch) ? tolower (Ch) : (Ch))
32 /* Knuth-Morris-Pratt algorithm.
33 See http://en.wikipedia.org/wiki/Knuth-Morris-Pratt_algorithm
34 Return a boolean indicating success. */
36 knuth_morris_pratt (const char *haystack, const char *needle,
39 size_t m = strlen (needle);
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 = TOLOWER ((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 == TOLOWER ((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 != '\0')
124 if (TOLOWER ((unsigned char) needle[j])
125 == TOLOWER ((unsigned char) *phaystack))
131 /* The entire needle has been found. */
132 *resultp = rhaystack;
138 /* Found a match of needle[0..j-1], mismatch at needle[j]. */
139 rhaystack += table[j];
144 /* Found a mismatch at needle[0] already. */
154 /* Find the first occurrence of NEEDLE in HAYSTACK, using case-insensitive
156 Note: This function may, in multibyte locales, return success even if
157 strlen (haystack) < strlen (needle) ! */
159 strcasestr (const char *haystack, const char *needle)
163 /* Minimizing the worst-case complexity:
164 Let n = strlen(haystack), m = strlen(needle).
165 The naïve algorithm is O(n*m) worst-case.
166 The Knuth-Morris-Pratt algorithm is O(n) worst-case but it needs a
168 To achieve linear complexity and yet amortize the cost of the memory
169 allocation, we activate the Knuth-Morris-Pratt algorithm only once
170 the naïve algorithm has already run for some time; more precisely,
172 - the outer loop count is >= 10,
173 - the average number of comparisons per outer loop is >= 5,
174 - the total number of comparisons is >= m.
175 But we try it only once. If the memory allocation attempt failed,
176 we don't retry it. */
178 size_t outer_loop_count = 0;
179 size_t comparison_count = 0;
180 size_t last_ccount = 0; /* last comparison count */
181 const char *needle_last_ccount = needle; /* = needle + last_ccount */
183 /* Speed up the following searches of needle by caching its first
185 unsigned char b = TOLOWER ((unsigned char) *needle);
190 if (*haystack == '\0')
194 /* See whether it's advisable to use an asymptotically faster
197 && outer_loop_count >= 10
198 && comparison_count >= 5 * outer_loop_count)
200 /* See if needle + comparison_count now reaches the end of
202 if (needle_last_ccount != NULL)
204 needle_last_ccount +=
205 strnlen (needle_last_ccount, comparison_count - last_ccount);
206 if (*needle_last_ccount == '\0')
207 needle_last_ccount = NULL;
208 last_ccount = comparison_count;
210 if (needle_last_ccount == NULL)
212 /* Try the Knuth-Morris-Pratt algorithm. */
215 knuth_morris_pratt (haystack, needle - 1, &result);
217 return (char *) result;
224 if (TOLOWER ((unsigned char) *haystack) == b)
225 /* The first character matches. */
227 const char *rhaystack = haystack + 1;
228 const char *rneedle = needle;
230 for (;; rhaystack++, rneedle++)
232 if (*rneedle == '\0')
234 return (char *) haystack;
235 if (*rhaystack == '\0')
239 if (TOLOWER ((unsigned char) *rhaystack)
240 != TOLOWER ((unsigned char) *rneedle))
241 /* Nothing in this round. */
248 return (char *) haystack;