1 /*- 2 * Copyright (c) 1991 The Regents of the University of California. 3 * 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 char copyright[] = 13 "@(#) Copyright (c) 1991 The Regents of the University of California.\n\ 14 All rights reserved.\n"; 15 #endif /* not lint */ 16 17 #ifndef lint 18 static char sccsid[] = "@(#)look.c 5.2 (Berkeley) 06/24/92"; 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 void usage __P((void)); 66 67 main(argc, argv) 68 int argc; 69 char *argv[]; 70 { 71 struct stat sb; 72 int ch, fd; 73 char *back, *file, *front, *string; 74 75 file = _PATH_WORDS; 76 while ((ch = getopt(argc, argv, "df")) != EOF) 77 switch(ch) { 78 case 'd': 79 dflag = 1; 80 break; 81 case 'f': 82 fflag = 1; 83 break; 84 case '?': 85 default: 86 usage(); 87 } 88 argc -= optind; 89 argv += optind; 90 91 switch (argc) { 92 case 2: /* Don't set -df for user. */ 93 string = *argv++; 94 file = *argv; 95 break; 96 case 1: /* But set -df by default. */ 97 dflag = fflag = 1; 98 string = *argv; 99 break; 100 default: 101 usage(); 102 } 103 104 if ((fd = open(file, O_RDONLY, 0)) < 0 || fstat(fd, &sb)) 105 err("%s: %s", file, strerror(errno)); 106 if (sb.st_size > SIZE_T_MAX) 107 err("%s: %s", file, strerror(EFBIG)); 108 if ((front = mmap(NULL, 109 (size_t)sb.st_size, PROT_READ, 0, fd, (off_t)0)) == NULL) 110 err("%s: %s", file, strerror(errno)); 111 back = front + sb.st_size; 112 exit(look(string, front, back)); 113 } 114 115 look(string, front, back) 116 char *string, *front, *back; 117 { 118 register int ch; 119 register char *readp, *writep; 120 121 /* Reformat string string to avoid doing it multiple times later. */ 122 for (readp = writep = string; ch = *readp++;) { 123 if (fflag) 124 ch = FOLD(ch); 125 if (dflag) 126 ch = DICT(ch); 127 if (ch != NO_COMPARE) 128 *(writep++) = ch; 129 } 130 *writep = '\0'; 131 132 front = binary_search(string, front, back); 133 front = linear_search(string, front, back); 134 135 if (front) 136 print_from(string, front, back); 137 return (front ? 0 : 1); 138 } 139 140 141 /* 142 * Binary search for "string" in memory between "front" and "back". 143 * 144 * This routine is expected to return a pointer to the start of a line at 145 * *or before* the first word matching "string". Relaxing the constraint 146 * this way simplifies the algorithm. 147 * 148 * Invariants: 149 * front points to the beginning of a line at or before the first 150 * matching string. 151 * 152 * back points to the beginning of a line at or after the first 153 * matching line. 154 * 155 * Base of the Invariants. 156 * front = NULL; 157 * back = EOF; 158 * 159 * Advancing the Invariants: 160 * 161 * p = first newline after halfway point from front to back. 162 * 163 * If the string at "p" is not greater than the string to match, 164 * p is the new front. Otherwise it is the new back. 165 * 166 * Termination: 167 * 168 * The definition of the routine allows it return at any point, 169 * since front is always at or before the line to print. 170 * 171 * In fact, it returns when the chosen "p" equals "back". This 172 * implies that there exists a string is least half as long as 173 * (back - front), which in turn implies that a linear search will 174 * be no more expensive than the cost of simply printing a string or two. 175 * 176 * Trying to continue with binary search at this point would be 177 * more trouble than it's worth. 178 */ 179 #define SKIP_PAST_NEWLINE(p, back) \ 180 while (p < back && *p++ != '\n'); 181 182 char * 183 binary_search(string, front, back) 184 register char *string, *front, *back; 185 { 186 register char *p; 187 188 p = front + (back - front) / 2; 189 SKIP_PAST_NEWLINE(p, back); 190 191 while (p != back) { 192 if (compare(string, p, back) == GREATER) 193 front = p; 194 else 195 back = p; 196 p = front + (back - front) / 2; 197 SKIP_PAST_NEWLINE(p, back); 198 } 199 return (front); 200 } 201 202 /* 203 * Find the first line that starts with string, linearly searching from front 204 * to back. 205 * 206 * Return NULL for no such line. 207 * 208 * This routine assumes: 209 * 210 * o front points at the first character in a line. 211 * o front is before or at the first line to be printed. 212 */ 213 char * 214 linear_search(string, front, back) 215 char *string, *front, *back; 216 { 217 while (front < back) { 218 switch (compare(string, front, back)) { 219 case EQUAL: /* Found it. */ 220 return (front); 221 break; 222 case LESS: /* No such string. */ 223 return (NULL); 224 break; 225 case GREATER: /* Keep going. */ 226 break; 227 } 228 SKIP_PAST_NEWLINE(front, back); 229 } 230 return (NULL); 231 } 232 233 /* 234 * Print as many lines as match string, starting at front. 235 */ 236 void 237 print_from(string, front, back) 238 register char *string, *front, *back; 239 { 240 for (; front < back && compare(string, front, back) == EQUAL; ++front) { 241 for (; front < back && *front != '\n'; ++front) 242 if (putchar(*front) == EOF) 243 err("stdout: %s", strerror(errno)); 244 if (putchar('\n') == EOF) 245 err("stdout: %s", strerror(errno)); 246 } 247 } 248 249 /* 250 * Return LESS, GREATER, or EQUAL depending on how the string1 compares with 251 * string2 (s1 ??? s2). 252 * 253 * o Matches up to len(s1) are EQUAL. 254 * o Matches up to len(s2) are GREATER. 255 * 256 * Compare understands about the -f and -d flags, and treats comparisons 257 * appropriately. 258 * 259 * The string "s1" is null terminated. The string s2 is '\n' terminated (or 260 * "back" terminated). 261 */ 262 int 263 compare(s1, s2, back) 264 register char *s1, *s2, *back; 265 { 266 register int ch; 267 268 for (; *s1 && s2 < back && *s2 != '\n'; ++s1, ++s2) { 269 ch = *s2; 270 if (fflag) 271 ch = FOLD(ch); 272 if (dflag) 273 ch = DICT(ch); 274 275 if (ch == NO_COMPARE) { 276 ++s2; /* Ignore character in comparison. */ 277 continue; 278 } 279 if (*s1 != ch) 280 return (*s1 < ch ? LESS : GREATER); 281 } 282 return (*s1 ? GREATER : EQUAL); 283 } 284 285 static void 286 usage() 287 { 288 (void)fprintf(stderr, "usage: look [-df] string [file]\n"); 289 exit(2); 290 } 291 292 #if __STDC__ 293 #include <stdarg.h> 294 #else 295 #include <varargs.h> 296 #endif 297 298 void 299 #if __STDC__ 300 err(const char *fmt, ...) 301 #else 302 err(fmt, va_alist) 303 char *fmt; 304 va_dcl 305 #endif 306 { 307 va_list ap; 308 #if __STDC__ 309 va_start(ap, fmt); 310 #else 311 va_start(ap); 312 #endif 313 (void)fprintf(stderr, "look: "); 314 (void)vfprintf(stderr, fmt, ap); 315 va_end(ap); 316 (void)fprintf(stderr, "\n"); 317 exit(2); 318 /* NOTREACHED */ 319 } 320