xref: /freebsd/lib/libc/string/strstr.c (revision f0cfa1b168014f56c02b83e5f28412cc5f78d117)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 2005-2014 Rich Felker, et al.
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining
7  * a copy of this software and associated documentation files (the
8  * "Software"), to deal in the Software without restriction, including
9  * without limitation the rights to use, copy, modify, merge, publish,
10  * distribute, sublicense, and/or sell copies of the Software, and to
11  * permit persons to whom the Software is furnished to do so, subject to
12  * the following conditions:
13  *
14  * The above copyright notice and this permission notice shall be
15  * included in all copies or substantial portions of the Software.
16  *
17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
18  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
19  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
20  * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
21  * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
22  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
23  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
24  */
25 #include <sys/cdefs.h>
26 __FBSDID("$FreeBSD$");
27 
28 #include <string.h>
29 #include <stdint.h>
30 
31 static char *twobyte_strstr(const unsigned char *h, const unsigned char *n)
32 {
33 	uint16_t nw = n[0]<<8 | n[1], hw = h[0]<<8 | h[1];
34 	for (h++; *h && hw != nw; hw = hw<<8 | *++h);
35 	return *h ? (char *)h-1 : 0;
36 }
37 
38 static char *threebyte_strstr(const unsigned char *h, const unsigned char *n)
39 {
40 	uint32_t nw = n[0]<<24 | n[1]<<16 | n[2]<<8;
41 	uint32_t hw = h[0]<<24 | h[1]<<16 | h[2]<<8;
42 	for (h+=2; *h && hw != nw; hw = (hw|*++h)<<8);
43 	return *h ? (char *)h-2 : 0;
44 }
45 
46 static char *fourbyte_strstr(const unsigned char *h, const unsigned char *n)
47 {
48 	uint32_t nw = n[0]<<24 | n[1]<<16 | n[2]<<8 | n[3];
49 	uint32_t hw = h[0]<<24 | h[1]<<16 | h[2]<<8 | h[3];
50 	for (h+=3; *h && hw != nw; hw = hw<<8 | *++h);
51 	return *h ? (char *)h-3 : 0;
52 }
53 
54 #define MAX(a,b) ((a)>(b)?(a):(b))
55 #define MIN(a,b) ((a)<(b)?(a):(b))
56 
57 #define BITOP(a,b,op) \
58  ((a)[(size_t)(b)/(8*sizeof *(a))] op (size_t)1<<((size_t)(b)%(8*sizeof *(a))))
59 
60 /*
61  * Two Way string search algorithm, with a bad shift table applied to the last
62  * byte of the window. A bit array marks which entries in the shift table are
63  * initialized to avoid fully initializing a 1kb/2kb table.
64  *
65  * Reference: CROCHEMORE M., PERRIN D., 1991, Two-way string-matching,
66  * Journal of the ACM 38(3):651-675
67  */
68 static char *twoway_strstr(const unsigned char *h, const unsigned char *n)
69 {
70 	const unsigned char *z;
71 	size_t l, ip, jp, k, p, ms, p0, mem, mem0;
72 	size_t byteset[32 / sizeof(size_t)] = { 0 };
73 	size_t shift[256];
74 
75 	/* Computing length of needle and fill shift table */
76 	for (l=0; n[l] && h[l]; l++)
77 		BITOP(byteset, n[l], |=), shift[n[l]] = l+1;
78 	if (n[l]) return 0; /* hit the end of h */
79 
80 	/* Compute maximal suffix */
81 	ip = -1; jp = 0; k = p = 1;
82 	while (jp+k<l) {
83 		if (n[ip+k] == n[jp+k]) {
84 			if (k == p) {
85 				jp += p;
86 				k = 1;
87 			} else k++;
88 		} else if (n[ip+k] > n[jp+k]) {
89 			jp += k;
90 			k = 1;
91 			p = jp - ip;
92 		} else {
93 			ip = jp++;
94 			k = p = 1;
95 		}
96 	}
97 	ms = ip;
98 	p0 = p;
99 
100 	/* And with the opposite comparison */
101 	ip = -1; jp = 0; k = p = 1;
102 	while (jp+k<l) {
103 		if (n[ip+k] == n[jp+k]) {
104 			if (k == p) {
105 				jp += p;
106 				k = 1;
107 			} else k++;
108 		} else if (n[ip+k] < n[jp+k]) {
109 			jp += k;
110 			k = 1;
111 			p = jp - ip;
112 		} else {
113 			ip = jp++;
114 			k = p = 1;
115 		}
116 	}
117 	if (ip+1 > ms+1) ms = ip;
118 	else p = p0;
119 
120 	/* Periodic needle? */
121 	if (memcmp(n, n+p, ms+1)) {
122 		mem0 = 0;
123 		p = MAX(ms, l-ms-1) + 1;
124 	} else mem0 = l-p;
125 	mem = 0;
126 
127 	/* Initialize incremental end-of-haystack pointer */
128 	z = h;
129 
130 	/* Search loop */
131 	for (;;) {
132 		/* Update incremental end-of-haystack pointer */
133 		if (z-h < l) {
134 			/* Fast estimate for MIN(l,63) */
135 			size_t grow = l | 63;
136 			const unsigned char *z2 = memchr(z, 0, grow);
137 			if (z2) {
138 				z = z2;
139 				if (z-h < l) return 0;
140 			} else z += grow;
141 		}
142 
143 		/* Check last byte first; advance by shift on mismatch */
144 		if (BITOP(byteset, h[l-1], &)) {
145 			k = l-shift[h[l-1]];
146 			//printf("adv by %zu (on %c) at [%s] (%zu;l=%zu)\n", k, h[l-1], h, shift[h[l-1]], l);
147 			if (k) {
148 				if (mem0 && mem && k < p) k = l-p;
149 				h += k;
150 				mem = 0;
151 				continue;
152 			}
153 		} else {
154 			h += l;
155 			mem = 0;
156 			continue;
157 		}
158 
159 		/* Compare right half */
160 		for (k=MAX(ms+1,mem); n[k] && n[k] == h[k]; k++);
161 		if (n[k]) {
162 			h += k-ms;
163 			mem = 0;
164 			continue;
165 		}
166 		/* Compare left half */
167 		for (k=ms+1; k>mem && n[k-1] == h[k-1]; k--);
168 		if (k <= mem) return (char *)h;
169 		h += p;
170 		mem = mem0;
171 	}
172 }
173 
174 char *strstr(const char *h, const char *n)
175 {
176 	/* Return immediately on empty needle */
177 	if (!n[0]) return (char *)h;
178 
179 	/* Use faster algorithms for short needles */
180 	h = strchr(h, *n);
181 	if (!h || !n[1]) return (char *)h;
182 	if (!h[1]) return 0;
183 	if (!n[2]) return twobyte_strstr((void *)h, (void *)n);
184 	if (!h[2]) return 0;
185 	if (!n[3]) return threebyte_strstr((void *)h, (void *)n);
186 	if (!h[3]) return 0;
187 	if (!n[4]) return fourbyte_strstr((void *)h, (void *)n);
188 
189 	return twoway_strstr((void *)h, (void *)n);
190 }
191