xref: /netbsd-src/usr.bin/grep/fastgrep.c (revision 213144e1de7024d4193d04aa51005ba3a5ad95e7)
1 /*	$OpenBSD: util.c,v 1.36 2007/10/02 17:59:18 otto Exp $	*/
2 /*	$FreeBSD: head/usr.bin/grep/fastgrep.c 211496 2010-08-19 09:28:59Z des $ */
3 
4 /*-
5  * Copyright (c) 1999 James Howard and Dag-Erling Coïdan Smørgrav
6  * Copyright (C) 2008 Gabor Kovesdan <gabor@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28  * SUCH DAMAGE.
29  */
30 
31 /*
32  * XXX: This file is a speed up for grep to cover the defects of the
33  * regex library.  These optimizations should practically be implemented
34  * there keeping this code clean.  This is a future TODO, but for the
35  * meantime, we need to use this workaround.
36  */
37 
38 #include <sys/cdefs.h>
39 __RCSID("$NetBSD: fastgrep.c,v 1.3 2011/02/17 22:03:25 joerg Exp $");
40 
41 #include <limits.h>
42 #include <stdbool.h>
43 #include <stdlib.h>
44 #include <string.h>
45 #include <wchar.h>
46 #include <wctype.h>
47 
48 #include "grep.h"
49 
50 static inline int	grep_cmp(const unsigned char *, const unsigned char *, size_t);
51 static inline void	grep_revstr(unsigned char *, int);
52 
53 void
54 fgrepcomp(fastgrep_t *fg, const char *pat)
55 {
56 	unsigned int i;
57 
58 	/* Initialize. */
59 	fg->len = strlen(pat);
60 	fg->bol = false;
61 	fg->eol = false;
62 	fg->reversed = false;
63 
64 	fg->pattern = (unsigned char *)grep_strdup(pat);
65 
66 	/* Preprocess pattern. */
67 	for (i = 0; i <= UCHAR_MAX; i++)
68 		fg->qsBc[i] = fg->len;
69 	for (i = 1; i < fg->len; i++)
70 		fg->qsBc[fg->pattern[i]] = fg->len - i;
71 }
72 
73 /*
74  * Returns: -1 on failure, 0 on success
75  */
76 int
77 fastcomp(fastgrep_t *fg, const char *pat)
78 {
79 	unsigned int i;
80 	int firstHalfDot = -1;
81 	int firstLastHalfDot = -1;
82 	int hasDot = 0;
83 	int lastHalfDot = 0;
84 	int shiftPatternLen;
85 
86 	/* Initialize. */
87 	fg->len = strlen(pat);
88 	fg->bol = false;
89 	fg->eol = false;
90 	fg->reversed = false;
91 
92 	/* Remove end-of-line character ('$'). */
93 	if (fg->len > 0 && pat[fg->len - 1] == '$') {
94 		fg->eol = true;
95 		fg->len--;
96 	}
97 
98 	/* Remove beginning-of-line character ('^'). */
99 	if (pat[0] == '^') {
100 		fg->bol = true;
101 		fg->len--;
102 		pat++;
103 	}
104 
105 	if (fg->len >= 14 &&
106 	    memcmp(pat, "[[:<:]]", 7) == 0 &&
107 	    memcmp(pat + fg->len - 7, "[[:>:]]", 7) == 0) {
108 		fg->len -= 14;
109 		pat += 7;
110 		/* Word boundary is handled separately in util.c */
111 		wflag = true;
112 	}
113 
114 	/*
115 	 * pat has been adjusted earlier to not include '^', '$' or
116 	 * the word match character classes at the beginning and ending
117 	 * of the string respectively.
118 	 */
119 	fg->pattern = grep_malloc(fg->len + 1);
120 	memcpy(fg->pattern, pat, fg->len);
121 	fg->pattern[fg->len] = '\0';
122 
123 	/* Look for ways to cheat...er...avoid the full regex engine. */
124 	for (i = 0; i < fg->len; i++) {
125 		/* Can still cheat? */
126 		if (fg->pattern[i] == '.') {
127 			hasDot = i;
128 			if (i < fg->len / 2) {
129 				if (firstHalfDot < 0)
130 					/* Closest dot to the beginning */
131 					firstHalfDot = i;
132 			} else {
133 				/* Closest dot to the end of the pattern. */
134 				lastHalfDot = i;
135 				if (firstLastHalfDot < 0)
136 					firstLastHalfDot = i;
137 			}
138 		} else {
139 			/* Free memory and let others know this is empty. */
140 			free(fg->pattern);
141 			fg->pattern = NULL;
142 			return (-1);
143 		}
144 	}
145 
146 	/*
147 	 * Determine if a reverse search would be faster based on the placement
148 	 * of the dots.
149 	 */
150 	if ((!(lflag || cflag)) && ((!(fg->bol || fg->eol)) &&
151 	    ((lastHalfDot) && ((firstHalfDot < 0) ||
152 	    ((fg->len - (lastHalfDot + 1)) < (size_t)firstHalfDot)))) &&
153 	    !oflag && !color) {
154 		fg->reversed = true;
155 		hasDot = fg->len - (firstHalfDot < 0 ?
156 		    firstLastHalfDot : firstHalfDot) - 1;
157 		grep_revstr(fg->pattern, fg->len);
158 	}
159 
160 	/*
161 	 * Normal Quick Search would require a shift based on the position the
162 	 * next character after the comparison is within the pattern.  With
163 	 * wildcards, the position of the last dot effects the maximum shift
164 	 * distance.
165 	 * The closer to the end the wild card is the slower the search.  A
166 	 * reverse version of this algorithm would be useful for wildcards near
167 	 * the end of the string.
168 	 *
169 	 * Examples:
170 	 * Pattern	Max shift
171 	 * -------	---------
172 	 * this		5
173 	 * .his		4
174 	 * t.is		3
175 	 * th.s		2
176 	 * thi.		1
177 	 */
178 
179 	/* Adjust the shift based on location of the last dot ('.'). */
180 	shiftPatternLen = fg->len - hasDot;
181 
182 	/* Preprocess pattern. */
183 	for (i = 0; i <= (signed)UCHAR_MAX; i++)
184 		fg->qsBc[i] = shiftPatternLen;
185 	for (i = hasDot + 1; i < fg->len; i++) {
186 		fg->qsBc[fg->pattern[i]] = fg->len - i;
187 	}
188 
189 	/*
190 	 * Put pattern back to normal after pre-processing to allow for easy
191 	 * comparisons later.
192 	 */
193 	if (fg->reversed)
194 		grep_revstr(fg->pattern, fg->len);
195 
196 	return (0);
197 }
198 
199 int
200 grep_search(fastgrep_t *fg, const unsigned char *data, size_t len, regmatch_t *pmatch)
201 {
202 	unsigned int j;
203 	int ret = REG_NOMATCH;
204 
205 	if (pmatch->rm_so == (ssize_t)len)
206 		return (ret);
207 
208 	if (fg->bol && pmatch->rm_so != 0) {
209 		pmatch->rm_so = len;
210 		pmatch->rm_eo = len;
211 		return (ret);
212 	}
213 
214 	/* No point in going farther if we do not have enough data. */
215 	if (len < fg->len)
216 		return (ret);
217 
218 	/* Only try once at the beginning or ending of the line. */
219 	if (fg->bol || fg->eol) {
220 		/* Simple text comparison. */
221 		/* Verify data is >= pattern length before searching on it. */
222 		if (len >= fg->len) {
223 			/* Determine where in data to start search at. */
224 			j = fg->eol ? len - fg->len : 0;
225 			if (!((fg->bol && fg->eol) && (len != fg->len)))
226 				if (grep_cmp(fg->pattern, data + j,
227 				    fg->len) == -1) {
228 					pmatch->rm_so = j;
229 					pmatch->rm_eo = j + fg->len;
230 						ret = 0;
231 				}
232 		}
233 	} else if (fg->reversed) {
234 		/* Quick Search algorithm. */
235 		j = len;
236 		do {
237 			if (grep_cmp(fg->pattern, data + j - fg->len,
238 				fg->len) == -1) {
239 				pmatch->rm_so = j - fg->len;
240 				pmatch->rm_eo = j;
241 				ret = 0;
242 				break;
243 			}
244 			/* Shift if within bounds, otherwise, we are done. */
245 			if (j == fg->len)
246 				break;
247 			j -= fg->qsBc[data[j - fg->len - 1]];
248 		} while (j >= fg->len);
249 	} else {
250 		/* Quick Search algorithm. */
251 		j = pmatch->rm_so;
252 		do {
253 			if (grep_cmp(fg->pattern, data + j, fg->len) == -1) {
254 				pmatch->rm_so = j;
255 				pmatch->rm_eo = j + fg->len;
256 				ret = 0;
257 				break;
258 			}
259 
260 			/* Shift if within bounds, otherwise, we are done. */
261 			if (j + fg->len == len)
262 				break;
263 			else
264 				j += fg->qsBc[data[j + fg->len]];
265 		} while (j <= (len - fg->len));
266 	}
267 
268 	return (ret);
269 }
270 
271 /*
272  * Returns:	i >= 0 on failure (position that it failed)
273  *		-1 on success
274  */
275 static inline int
276 grep_cmp(const unsigned char *pat, const unsigned char *data, size_t len)
277 {
278 	size_t size;
279 	wchar_t *wdata, *wpat;
280 	unsigned int i;
281 
282 	if (iflag) {
283 		if ((size = mbstowcs(NULL, (const char *)data, 0)) ==
284 		    ((size_t) - 1))
285 			return (-1);
286 
287 		wdata = grep_malloc(size * sizeof(wint_t));
288 
289 		if (mbstowcs(wdata, (const char *)data, size) ==
290 		    ((size_t) - 1))
291 			return (-1);
292 
293 		if ((size = mbstowcs(NULL, (const char *)pat, 0)) ==
294 		    ((size_t) - 1))
295 			return (-1);
296 
297 		wpat = grep_malloc(size * sizeof(wint_t));
298 
299 		if (mbstowcs(wpat, (const char *)pat, size) == ((size_t) - 1))
300 			return (-1);
301 		for (i = 0; i < len; i++) {
302 			if ((towlower(wpat[i]) == towlower(wdata[i])) ||
303 			    ((grepbehave != GREP_FIXED) && wpat[i] == L'.'))
304 				continue;
305 			free(wpat);
306 			free(wdata);
307 				return (i);
308 		}
309 	} else {
310 		for (i = 0; i < len; i++) {
311 			if ((pat[i] == data[i]) || ((grepbehave != GREP_FIXED) &&
312 			    pat[i] == '.'))
313 				continue;
314 			return (i);
315 		}
316 	}
317 	return (-1);
318 }
319 
320 static inline void
321 grep_revstr(unsigned char *str, int len)
322 {
323 	int i;
324 	char c;
325 
326 	for (i = 0; i < len / 2; i++) {
327 		c = str[i];
328 		str[i] = str[len - i - 1];
329 		str[len - i - 1] = c;
330 	}
331 }
332