1 /*- 2 * Copyright (c) 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * David Hitz of Auspex Systems, Inc. 7 * 8 * %sccs.include.redist.c% 9 */ 10 11 #ifndef lint 12 static char copyright[] = 13 "@(#) Copyright (c) 1991, 1993\n\ 14 The Regents of the University of California. All rights reserved.\n"; 15 #endif /* not lint */ 16 17 #ifndef lint 18 static char sccsid[] = "@(#)look.c 8.1 (Berkeley) 06/14/93"; 19 #endif /* not lint */ 20 21 /* 22 * look -- find lines in a sorted list. 23 * 24 * The man page said that TABs and SPACEs participate in -d comparisons. 25 * In fact, they were ignored. This implements historic practice, not 26 * the manual page. 27 */ 28 29 #include <sys/types.h> 30 #include <sys/mman.h> 31 #include <sys/stat.h> 32 33 #include <limits.h> 34 #include <errno.h> 35 #include <fcntl.h> 36 #include <stdio.h> 37 #include <stdlib.h> 38 #include <string.h> 39 #include <ctype.h> 40 #include "pathnames.h" 41 42 /* 43 * FOLD and DICT convert characters to a normal form for comparison, 44 * according to the user specified flags. 45 * 46 * DICT expects integers because it uses a non-character value to 47 * indicate a character which should not participate in comparisons. 48 */ 49 #define EQUAL 0 50 #define GREATER 1 51 #define LESS (-1) 52 #define NO_COMPARE (-2) 53 54 #define FOLD(c) (isascii(c) && isupper(c) ? tolower(c) : (c)) 55 #define DICT(c) (isascii(c) && isalnum(c) ? (c) : NO_COMPARE) 56 57 int dflag, fflag; 58 59 char *binary_search __P((char *, char *, char *)); 60 int compare __P((char *, char *, char *)); 61 void err __P((const char *fmt, ...)); 62 char *linear_search __P((char *, char *, char *)); 63 int look __P((char *, char *, char *)); 64 void print_from __P((char *, char *, char *)); 65 66 static void usage __P((void)); 67 68 main(argc, argv) 69 int argc; 70 char *argv[]; 71 { 72 struct stat sb; 73 int ch, fd, termchar; 74 char *back, *file, *front, *string, *p; 75 76 file = _PATH_WORDS; 77 termchar = '\0'; 78 while ((ch = getopt(argc, argv, "dft:")) != EOF) 79 switch(ch) { 80 case 'd': 81 dflag = 1; 82 break; 83 case 'f': 84 fflag = 1; 85 break; 86 case 't': 87 termchar = *optarg; 88 break; 89 case '?': 90 default: 91 usage(); 92 } 93 argc -= optind; 94 argv += optind; 95 96 switch (argc) { 97 case 2: /* Don't set -df for user. */ 98 string = *argv++; 99 file = *argv; 100 break; 101 case 1: /* But set -df by default. */ 102 dflag = fflag = 1; 103 string = *argv; 104 break; 105 default: 106 usage(); 107 } 108 109 if (termchar != '\0' && (p = strchr(string, termchar)) != NULL) 110 *++p = '\0'; 111 112 if ((fd = open(file, O_RDONLY, 0)) < 0 || fstat(fd, &sb)) 113 err("%s: %s", file, strerror(errno)); 114 if (sb.st_size > SIZE_T_MAX) 115 err("%s: %s", file, strerror(EFBIG)); 116 if ((front = mmap(NULL, 117 (size_t)sb.st_size, PROT_READ, 0, fd, (off_t)0)) == NULL) 118 err("%s: %s", file, strerror(errno)); 119 back = front + sb.st_size; 120 exit(look(string, front, back)); 121 } 122 123 look(string, front, back) 124 char *string, *front, *back; 125 { 126 register int ch; 127 register char *readp, *writep; 128 129 /* Reformat string string to avoid doing it multiple times later. */ 130 for (readp = writep = string; ch = *readp++;) { 131 if (fflag) 132 ch = FOLD(ch); 133 if (dflag) 134 ch = DICT(ch); 135 if (ch != NO_COMPARE) 136 *(writep++) = ch; 137 } 138 *writep = '\0'; 139 140 front = binary_search(string, front, back); 141 front = linear_search(string, front, back); 142 143 if (front) 144 print_from(string, front, back); 145 return (front ? 0 : 1); 146 } 147 148 149 /* 150 * Binary search for "string" in memory between "front" and "back". 151 * 152 * This routine is expected to return a pointer to the start of a line at 153 * *or before* the first word matching "string". Relaxing the constraint 154 * this way simplifies the algorithm. 155 * 156 * Invariants: 157 * front points to the beginning of a line at or before the first 158 * matching string. 159 * 160 * back points to the beginning of a line at or after the first 161 * matching line. 162 * 163 * Base of the Invariants. 164 * front = NULL; 165 * back = EOF; 166 * 167 * Advancing the Invariants: 168 * 169 * p = first newline after halfway point from front to back. 170 * 171 * If the string at "p" is not greater than the string to match, 172 * p is the new front. Otherwise it is the new back. 173 * 174 * Termination: 175 * 176 * The definition of the routine allows it return at any point, 177 * since front is always at or before the line to print. 178 * 179 * In fact, it returns when the chosen "p" equals "back". This 180 * implies that there exists a string is least half as long as 181 * (back - front), which in turn implies that a linear search will 182 * be no more expensive than the cost of simply printing a string or two. 183 * 184 * Trying to continue with binary search at this point would be 185 * more trouble than it's worth. 186 */ 187 #define SKIP_PAST_NEWLINE(p, back) \ 188 while (p < back && *p++ != '\n'); 189 190 char * 191 binary_search(string, front, back) 192 register char *string, *front, *back; 193 { 194 register char *p; 195 196 p = front + (back - front) / 2; 197 SKIP_PAST_NEWLINE(p, back); 198 199 /* 200 * If the file changes underneath us, make sure we don't 201 * infinitely loop. 202 */ 203 while (p < back && back > front) { 204 if (compare(string, p, back) == GREATER) 205 front = p; 206 else 207 back = p; 208 p = front + (back - front) / 2; 209 SKIP_PAST_NEWLINE(p, back); 210 } 211 return (front); 212 } 213 214 /* 215 * Find the first line that starts with string, linearly searching from front 216 * to back. 217 * 218 * Return NULL for no such line. 219 * 220 * This routine assumes: 221 * 222 * o front points at the first character in a line. 223 * o front is before or at the first line to be printed. 224 */ 225 char * 226 linear_search(string, front, back) 227 char *string, *front, *back; 228 { 229 while (front < back) { 230 switch (compare(string, front, back)) { 231 case EQUAL: /* Found it. */ 232 return (front); 233 break; 234 case LESS: /* No such string. */ 235 return (NULL); 236 break; 237 case GREATER: /* Keep going. */ 238 break; 239 } 240 SKIP_PAST_NEWLINE(front, back); 241 } 242 return (NULL); 243 } 244 245 /* 246 * Print as many lines as match string, starting at front. 247 */ 248 void 249 print_from(string, front, back) 250 register char *string, *front, *back; 251 { 252 for (; front < back && compare(string, front, back) == EQUAL; ++front) { 253 for (; front < back && *front != '\n'; ++front) 254 if (putchar(*front) == EOF) 255 err("stdout: %s", strerror(errno)); 256 if (putchar('\n') == EOF) 257 err("stdout: %s", strerror(errno)); 258 } 259 } 260 261 /* 262 * Return LESS, GREATER, or EQUAL depending on how the string1 compares with 263 * string2 (s1 ??? s2). 264 * 265 * o Matches up to len(s1) are EQUAL. 266 * o Matches up to len(s2) are GREATER. 267 * 268 * Compare understands about the -f and -d flags, and treats comparisons 269 * appropriately. 270 * 271 * The string "s1" is null terminated. The string s2 is '\n' terminated (or 272 * "back" terminated). 273 */ 274 int 275 compare(s1, s2, back) 276 register char *s1, *s2, *back; 277 { 278 register int ch; 279 280 for (; *s1 && s2 < back && *s2 != '\n'; ++s1, ++s2) { 281 ch = *s2; 282 if (fflag) 283 ch = FOLD(ch); 284 if (dflag) 285 ch = DICT(ch); 286 287 if (ch == NO_COMPARE) { 288 ++s2; /* Ignore character in comparison. */ 289 continue; 290 } 291 if (*s1 != ch) 292 return (*s1 < ch ? LESS : GREATER); 293 } 294 return (*s1 ? GREATER : EQUAL); 295 } 296 297 static void 298 usage() 299 { 300 (void)fprintf(stderr, "usage: look [-df] [-t char] string [file]\n"); 301 exit(2); 302 } 303 304 #if __STDC__ 305 #include <stdarg.h> 306 #else 307 #include <varargs.h> 308 #endif 309 310 void 311 #if __STDC__ 312 err(const char *fmt, ...) 313 #else 314 err(fmt, va_alist) 315 char *fmt; 316 va_dcl 317 #endif 318 { 319 va_list ap; 320 #if __STDC__ 321 va_start(ap, fmt); 322 #else 323 va_start(ap); 324 #endif 325 (void)fprintf(stderr, "look: "); 326 (void)vfprintf(stderr, fmt, ap); 327 va_end(ap); 328 (void)fprintf(stderr, "\n"); 329 exit(2); 330 /* NOTREACHED */ 331 } 332