1 /* $NetBSD: tsort.c,v 1.9 1994/12/07 01:06:25 jtc Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993, 1994 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Michael Rendell of Memorial University of Newfoundland. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 */ 38 39 #ifndef lint 40 static char copyright[] = 41 "@(#) Copyright (c) 1989, 1993, 1994\n\ 42 The Regents of the University of California. All rights reserved.\n"; 43 #endif /* not lint */ 44 45 #ifndef lint 46 #if 0 47 static char sccsid[] = "@(#)tsort.c 8.2 (Berkeley) 3/30/94"; 48 #endif 49 static char rcsid[] = "$NetBSD: tsort.c,v 1.9 1994/12/07 01:06:25 jtc Exp $"; 50 #endif /* not lint */ 51 52 #include <sys/types.h> 53 54 #include <ctype.h> 55 #include <db.h> 56 #include <err.h> 57 #include <errno.h> 58 #include <fcntl.h> 59 #include <stdio.h> 60 #include <stdlib.h> 61 #include <string.h> 62 63 /* 64 * Topological sort. Input is a list of pairs of strings separated by 65 * white space (spaces, tabs, and/or newlines); strings are written to 66 * standard output in sorted order, one per line. 67 * 68 * usage: 69 * tsort [-l] [inputfile] 70 * If no input file is specified, standard input is read. 71 * 72 * Should be compatable with AT&T tsort HOWEVER the output is not identical 73 * (i.e. for most graphs there is more than one sorted order, and this tsort 74 * usually generates a different one then the AT&T tsort). Also, cycle 75 * reporting seems to be more accurate in this version (the AT&T tsort 76 * sometimes says a node is in a cycle when it isn't). 77 * 78 * Michael Rendell, michael@stretch.cs.mun.ca - Feb 26, '90 79 */ 80 #define HASHSIZE 53 /* doesn't need to be big */ 81 #define NF_MARK 0x1 /* marker for cycle detection */ 82 #define NF_ACYCLIC 0x2 /* this node is cycle free */ 83 #define NF_NODEST 0x4 /* Unreachable */ 84 85 typedef struct node_str NODE; 86 87 struct node_str { 88 NODE **n_prevp; /* pointer to previous node's n_next */ 89 NODE *n_next; /* next node in graph */ 90 NODE **n_arcs; /* array of arcs to other nodes */ 91 int n_narcs; /* number of arcs in n_arcs[] */ 92 int n_arcsize; /* size of n_arcs[] array */ 93 int n_refcnt; /* # of arcs pointing to this node */ 94 int n_flags; /* NF_* */ 95 char n_name[1]; /* name of this node */ 96 }; 97 98 typedef struct _buf { 99 char *b_buf; 100 int b_bsize; 101 } BUF; 102 103 DB *db; 104 NODE *graph, **cycle_buf, **longest_cycle; 105 int debug, longest; 106 107 void add_arc __P((char *, char *)); 108 int find_cycle __P((NODE *, NODE *, int, int)); 109 NODE *get_node __P((char *)); 110 void *grow_buf __P((void *, int)); 111 void remove_node __P((NODE *)); 112 void tsort __P((void)); 113 void usage __P((void)); 114 115 int 116 main(argc, argv) 117 int argc; 118 char *argv[]; 119 { 120 register BUF *b; 121 register int c, n; 122 FILE *fp; 123 int bsize, ch, nused; 124 BUF bufs[2]; 125 126 while ((ch = getopt(argc, argv, "dl")) != EOF) 127 switch (ch) { 128 case 'd': 129 debug = 1; 130 break; 131 case 'l': 132 longest = 1; 133 break; 134 case '?': 135 default: 136 usage(); 137 } 138 argc -= optind; 139 argv += optind; 140 141 switch (argc) { 142 case 0: 143 fp = stdin; 144 break; 145 case 1: 146 if ((fp = fopen(*argv, "r")) == NULL) 147 err(1, "%s", *argv); 148 break; 149 default: 150 usage(); 151 } 152 153 for (b = bufs, n = 2; --n >= 0; b++) 154 b->b_buf = grow_buf(NULL, b->b_bsize = 1024); 155 156 /* parse input and build the graph */ 157 for (n = 0, c = getc(fp);;) { 158 while (c != EOF && isspace(c)) 159 c = getc(fp); 160 if (c == EOF) 161 break; 162 163 nused = 0; 164 b = &bufs[n]; 165 bsize = b->b_bsize; 166 do { 167 b->b_buf[nused++] = c; 168 if (nused == bsize) 169 b->b_buf = grow_buf(b->b_buf, bsize *= 2); 170 c = getc(fp); 171 } while (c != EOF && !isspace(c)); 172 173 b->b_buf[nused] = '\0'; 174 b->b_bsize = bsize; 175 if (n) 176 add_arc(bufs[0].b_buf, bufs[1].b_buf); 177 n = !n; 178 } 179 (void)fclose(fp); 180 if (n) 181 errx(1, "odd data count"); 182 183 /* do the sort */ 184 tsort(); 185 exit(0); 186 } 187 188 /* double the size of oldbuf and return a pointer to the new buffer. */ 189 void * 190 grow_buf(bp, size) 191 void *bp; 192 int size; 193 { 194 if ((bp = realloc(bp, (u_int)size)) == NULL) 195 err(1, NULL); 196 return (bp); 197 } 198 199 /* 200 * add an arc from node s1 to node s2 in the graph. If s1 or s2 are not in 201 * the graph, then add them. 202 */ 203 void 204 add_arc(s1, s2) 205 char *s1, *s2; 206 { 207 register NODE *n1; 208 NODE *n2; 209 int bsize, i; 210 211 n1 = get_node(s1); 212 213 if (!strcmp(s1, s2)) 214 return; 215 216 n2 = get_node(s2); 217 218 /* 219 * Check if this arc is already here. 220 */ 221 for (i = 0; i < n1->n_narcs; i++) 222 if (n1->n_arcs[i] == n2) 223 return; 224 /* 225 * Add it. 226 */ 227 if (n1->n_narcs == n1->n_arcsize) { 228 if (!n1->n_arcsize) 229 n1->n_arcsize = 10; 230 bsize = n1->n_arcsize * sizeof(*n1->n_arcs) * 2; 231 n1->n_arcs = grow_buf(n1->n_arcs, bsize); 232 n1->n_arcsize = bsize / sizeof(*n1->n_arcs); 233 } 234 n1->n_arcs[n1->n_narcs++] = n2; 235 ++n2->n_refcnt; 236 } 237 238 /* Find a node in the graph (insert if not found) and return a pointer to it. */ 239 NODE * 240 get_node(name) 241 char *name; 242 { 243 DBT data, key; 244 NODE *n; 245 246 if (db == NULL && 247 (db = dbopen(NULL, O_RDWR, 0, DB_HASH, NULL)) == NULL) 248 err(1, "db: %s", name); 249 250 key.data = name; 251 key.size = strlen(name) + 1; 252 253 switch ((*db->get)(db, &key, &data, 0)) { 254 case 0: 255 bcopy(data.data, &n, sizeof(n)); 256 return (n); 257 case 1: 258 break; 259 default: 260 case -1: 261 err(1, "db: %s", name); 262 } 263 264 if ((n = malloc(sizeof(NODE) + key.size)) == NULL) 265 err(1, NULL); 266 267 n->n_narcs = 0; 268 n->n_arcsize = 0; 269 n->n_arcs = NULL; 270 n->n_refcnt = 0; 271 n->n_flags = 0; 272 bcopy(name, n->n_name, key.size); 273 274 /* Add to linked list. */ 275 if ((n->n_next = graph) != NULL) 276 graph->n_prevp = &n->n_next; 277 n->n_prevp = &graph; 278 graph = n; 279 280 /* Add to hash table. */ 281 data.data = &n; 282 data.size = sizeof(n); 283 if ((*db->put)(db, &key, &data, 0)) 284 err(1, "db: %s", name); 285 return (n); 286 } 287 288 289 /* 290 * Clear the NODEST flag from all nodes. 291 */ 292 void 293 clear_cycle() 294 { 295 NODE *n; 296 297 for (n = graph; n != NULL; n = n->n_next) 298 n->n_flags &= ~NF_NODEST; 299 } 300 301 /* do topological sort on graph */ 302 void 303 tsort() 304 { 305 register NODE *n, *next; 306 register int cnt, i; 307 308 while (graph != NULL) { 309 /* 310 * Keep getting rid of simple cases until there are none left, 311 * if there are any nodes still in the graph, then there is 312 * a cycle in it. 313 */ 314 do { 315 for (cnt = 0, n = graph; n != NULL; n = next) { 316 next = n->n_next; 317 if (n->n_refcnt == 0) { 318 remove_node(n); 319 ++cnt; 320 } 321 } 322 } while (graph != NULL && cnt); 323 324 if (graph == NULL) 325 break; 326 327 if (!cycle_buf) { 328 /* 329 * Allocate space for two cycle logs - one to be used 330 * as scratch space, the other to save the longest 331 * cycle. 332 */ 333 for (cnt = 0, n = graph; n != NULL; n = n->n_next) 334 ++cnt; 335 cycle_buf = malloc((u_int)sizeof(NODE *) * cnt); 336 longest_cycle = malloc((u_int)sizeof(NODE *) * cnt); 337 if (cycle_buf == NULL || longest_cycle == NULL) 338 err(1, NULL); 339 } 340 for (n = graph; n != NULL; n = n->n_next) 341 if (!(n->n_flags & NF_ACYCLIC)) 342 if (cnt = find_cycle(n, n, 0, 0)) { 343 warnx("cycle in data"); 344 for (i = 0; i < cnt; i++) 345 warnx("%s", 346 longest_cycle[i]->n_name); 347 remove_node(n); 348 clear_cycle(); 349 break; 350 } else { 351 /* to avoid further checks */ 352 n->n_flags |= NF_ACYCLIC; 353 clear_cycle(); 354 } 355 356 if (n == NULL) 357 errx(1, "internal error -- could not find cycle"); 358 } 359 } 360 361 /* print node and remove from graph (does not actually free node) */ 362 void 363 remove_node(n) 364 register NODE *n; 365 { 366 register NODE **np; 367 register int i; 368 369 (void)printf("%s\n", n->n_name); 370 for (np = n->n_arcs, i = n->n_narcs; --i >= 0; np++) 371 --(*np)->n_refcnt; 372 n->n_narcs = 0; 373 *n->n_prevp = n->n_next; 374 if (n->n_next) 375 n->n_next->n_prevp = n->n_prevp; 376 } 377 378 379 /* look for the longest? cycle from node from to node to. */ 380 int 381 find_cycle(from, to, longest_len, depth) 382 NODE *from, *to; 383 int depth, longest_len; 384 { 385 register NODE **np; 386 register int i, len; 387 388 /* 389 * avoid infinite loops and ignore portions of the graph known 390 * to be acyclic 391 */ 392 if (from->n_flags & (NF_NODEST|NF_MARK|NF_ACYCLIC)) 393 return (0); 394 from->n_flags |= NF_MARK; 395 396 for (np = from->n_arcs, i = from->n_narcs; --i >= 0; np++) { 397 cycle_buf[depth] = *np; 398 if (*np == to) { 399 if (depth + 1 > longest_len) { 400 longest_len = depth + 1; 401 (void)memcpy((char *)longest_cycle, 402 (char *)cycle_buf, 403 longest_len * sizeof(NODE *)); 404 } 405 } else { 406 if ((*np)->n_flags & (NF_MARK|NF_ACYCLIC|NF_NODEST)) 407 continue; 408 len = find_cycle(*np, to, longest_len, depth + 1); 409 410 if (debug) 411 (void)printf("%*s %s->%s %d\n", depth, "", 412 from->n_name, to->n_name, len); 413 414 if (len == 0) 415 (*np)->n_flags |= NF_NODEST; 416 417 if (len > longest_len) 418 longest_len = len; 419 420 if (len > 0 && !longest) 421 break; 422 } 423 } 424 from->n_flags &= ~NF_MARK; 425 return (longest_len); 426 } 427 428 void 429 usage() 430 { 431 (void)fprintf(stderr, "usage: tsort [-l] [file]\n"); 432 exit(1); 433 } 434