1/*
2 * Copyright (c) 1989, 1993, 1994
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Michael Rendell of Memorial University of Newfoundland.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33#ifndef lint
34static const char copyright[] =
35"@(#) Copyright (c) 1989, 1993, 1994\n\
36	The Regents of the University of California.  All rights reserved.\n";
37#endif /* not lint */
38
39#ifndef lint
40static const char sccsid[] = "@(#)tsort.c	8.3 (Berkeley) 5/4/95";
41#endif /* not lint */
42
43#include <sys/cdefs.h>
44__FBSDID("$FreeBSD$");
45
46#include <sys/types.h>
47
48#include <ctype.h>
49#include <db.h>
50#include <err.h>
51#include <errno.h>
52#include <fcntl.h>
53#include <stdio.h>
54#include <stdlib.h>
55#include <string.h>
56#include <unistd.h>
57
58/*
59 *  Topological sort.  Input is a list of pairs of strings separated by
60 *  white space (spaces, tabs, and/or newlines); strings are written to
61 *  standard output in sorted order, one per line.
62 *
63 *  usage:
64 *     tsort [-dlq] [inputfile]
65 *  If no input file is specified, standard input is read.
66 *
67 *  Should be compatible with AT&T tsort HOWEVER the output is not identical
68 *  (i.e. for most graphs there is more than one sorted order, and this tsort
69 *  usually generates a different one then the AT&T tsort).  Also, cycle
70 *  reporting seems to be more accurate in this version (the AT&T tsort
71 *  sometimes says a node is in a cycle when it isn't).
72 *
73 *  Michael Rendell, michael@stretch.cs.mun.ca - Feb 26, '90
74 */
75
76#define	NF_MARK		0x1		/* marker for cycle detection */
77#define	NF_ACYCLIC	0x2		/* this node is cycle free */
78#define	NF_NODEST	0x4		/* Unreachable */
79
80
81typedef struct node_str NODE;
82
83struct node_str {
84	NODE **n_prevp;			/* pointer to previous node's n_next */
85	NODE *n_next;			/* next node in graph */
86	NODE **n_arcs;			/* array of arcs to other nodes */
87	int n_narcs;			/* number of arcs in n_arcs[] */
88	int n_arcsize;			/* size of n_arcs[] array */
89	int n_refcnt;			/* # of arcs pointing to this node */
90	int n_flags;			/* NF_* */
91	char n_name[1];			/* name of this node */
92};
93
94typedef struct _buf {
95	char *b_buf;
96	int b_bsize;
97} BUF;
98
99DB *db;
100NODE *graph, **cycle_buf, **longest_cycle;
101int debug, longest, quiet;
102
103void	 add_arc(char *, char *);
104int	 find_cycle(NODE *, NODE *, int, int);
105NODE	*get_node(char *);
106void	*grow_buf(void *, size_t);
107void	 remove_node(NODE *);
108void	 clear_cycle(void);
109void	 tsort(void);
110void	 usage(void);
111
112int
113main(int argc, char *argv[])
114{
115	BUF *b;
116	int c, n;
117	FILE *fp;
118	int bsize, ch, nused;
119	BUF bufs[2];
120
121	fp = NULL;
122	while ((ch = getopt(argc, argv, "dlq")) != -1)
123		switch (ch) {
124		case 'd':
125			debug = 1;
126			break;
127		case 'l':
128			longest = 1;
129			break;
130		case 'q':
131			quiet = 1;
132			break;
133		case '?':
134		default:
135			usage();
136		}
137	argc -= optind;
138	argv += optind;
139
140	switch (argc) {
141	case 0:
142		fp = stdin;
143		break;
144	case 1:
145		if ((fp = fopen(*argv, "r")) == NULL)
146			err(1, "%s", *argv);
147		break;
148	default:
149		usage();
150	}
151
152	for (b = bufs, n = 2; --n >= 0; b++)
153		b->b_buf = grow_buf(NULL, b->b_bsize = 1024);
154
155	/* parse input and build the graph */
156	for (n = 0, c = getc(fp);;) {
157		while (c != EOF && isspace(c))
158			c = getc(fp);
159		if (c == EOF)
160			break;
161
162		nused = 0;
163		b = &bufs[n];
164		bsize = b->b_bsize;
165		do {
166			b->b_buf[nused++] = c;
167			if (nused == bsize)
168				b->b_buf = grow_buf(b->b_buf, bsize *= 2);
169			c = getc(fp);
170		} while (c != EOF && !isspace(c));
171
172		b->b_buf[nused] = '\0';
173		b->b_bsize = bsize;
174		if (n)
175			add_arc(bufs[0].b_buf, bufs[1].b_buf);
176		n = !n;
177	}
178	(void)fclose(fp);
179	if (n)
180		errx(1, "odd data count");
181
182	/* do the sort */
183	tsort();
184	exit(0);
185}
186
187/* double the size of oldbuf and return a pointer to the new buffer. */
188void *
189grow_buf(void *bp, size_t size)
190{
191	if ((bp = realloc(bp, size)) == NULL)
192		err(1, NULL);
193	return (bp);
194}
195
196/*
197 * add an arc from node s1 to node s2 in the graph.  If s1 or s2 are not in
198 * the graph, then add them.
199 */
200void
201add_arc(char *s1, char *s2)
202{
203	NODE *n1;
204	NODE *n2;
205	int bsize, i;
206
207	n1 = get_node(s1);
208
209	if (!strcmp(s1, s2))
210		return;
211
212	n2 = get_node(s2);
213
214	/*
215	 * Check if this arc is already here.
216	 */
217	for (i = 0; i < n1->n_narcs; i++)
218		if (n1->n_arcs[i] == n2)
219			return;
220	/*
221	 * Add it.
222	 */
223	if (n1->n_narcs == n1->n_arcsize) {
224		if (!n1->n_arcsize)
225			n1->n_arcsize = 10;
226		bsize = n1->n_arcsize * sizeof(*n1->n_arcs) * 2;
227		n1->n_arcs = grow_buf(n1->n_arcs, bsize);
228		n1->n_arcsize = bsize / sizeof(*n1->n_arcs);
229	}
230	n1->n_arcs[n1->n_narcs++] = n2;
231	++n2->n_refcnt;
232}
233
234/* Find a node in the graph (insert if not found) and return a pointer to it. */
235NODE *
236get_node(char *name)
237{
238	DBT data, key;
239	NODE *n;
240
241	if (db == NULL &&
242	    (db = dbopen(NULL, O_RDWR, 0, DB_HASH, NULL)) == NULL)
243		err(1, "db: %s", name);
244
245	key.data = name;
246	key.size = strlen(name) + 1;
247
248	switch ((*db->get)(db, &key, &data, 0)) {
249	case 0:
250		bcopy(data.data, &n, sizeof(n));
251		return (n);
252	case 1:
253		break;
254	default:
255	case -1:
256		err(1, "db: %s", name);
257	}
258
259	if ((n = malloc(sizeof(NODE) + key.size)) == NULL)
260		err(1, NULL);
261
262	n->n_narcs = 0;
263	n->n_arcsize = 0;
264	n->n_arcs = NULL;
265	n->n_refcnt = 0;
266	n->n_flags = 0;
267	bcopy(name, n->n_name, key.size);
268
269	/* Add to linked list. */
270	if ((n->n_next = graph) != NULL)
271		graph->n_prevp = &n->n_next;
272	n->n_prevp = &graph;
273	graph = n;
274
275	/* Add to hash table. */
276	data.data = &n;
277	data.size = sizeof(n);
278	if ((*db->put)(db, &key, &data, 0))
279		err(1, "db: %s", name);
280	return (n);
281}
282
283
284/*
285 * Clear the NODEST flag from all nodes.
286 */
287void
288clear_cycle(void)
289{
290	NODE *n;
291
292	for (n = graph; n != NULL; n = n->n_next)
293		n->n_flags &= ~NF_NODEST;
294}
295
296/* do topological sort on graph */
297void
298tsort(void)
299{
300	NODE *n, *next;
301	int cnt, i;
302
303	while (graph != NULL) {
304		/*
305		 * Keep getting rid of simple cases until there are none left,
306		 * if there are any nodes still in the graph, then there is
307		 * a cycle in it.
308		 */
309		do {
310			for (cnt = 0, n = graph; n != NULL; n = next) {
311				next = n->n_next;
312				if (n->n_refcnt == 0) {
313					remove_node(n);
314					++cnt;
315				}
316			}
317		} while (graph != NULL && cnt);
318
319		if (graph == NULL)
320			break;
321
322		if (!cycle_buf) {
323			/*
324			 * Allocate space for two cycle logs - one to be used
325			 * as scratch space, the other to save the longest
326			 * cycle.
327			 */
328			for (cnt = 0, n = graph; n != NULL; n = n->n_next)
329				++cnt;
330			cycle_buf = malloc(sizeof(NODE *) * cnt);
331			longest_cycle = malloc(sizeof(NODE *) * cnt);
332			if (cycle_buf == NULL || longest_cycle == NULL)
333				err(1, NULL);
334		}
335		for (n = graph; n != NULL; n = n->n_next)
336			if (!(n->n_flags & NF_ACYCLIC)) {
337				if ((cnt = find_cycle(n, n, 0, 0))) {
338					if (!quiet) {
339						warnx("cycle in data");
340						for (i = 0; i < cnt; i++)
341							warnx("%s",
342							    longest_cycle[i]->n_name);
343					}
344					remove_node(n);
345					clear_cycle();
346					break;
347				} else {
348					/* to avoid further checks */
349					n->n_flags  |= NF_ACYCLIC;
350					clear_cycle();
351				}
352			}
353
354		if (n == NULL)
355			errx(1, "internal error -- could not find cycle");
356	}
357}
358
359/* print node and remove from graph (does not actually free node) */
360void
361remove_node(NODE *n)
362{
363	NODE **np;
364	int i;
365
366	(void)printf("%s\n", n->n_name);
367	for (np = n->n_arcs, i = n->n_narcs; --i >= 0; np++)
368		--(*np)->n_refcnt;
369	n->n_narcs = 0;
370	*n->n_prevp = n->n_next;
371	if (n->n_next)
372		n->n_next->n_prevp = n->n_prevp;
373}
374
375
376/* look for the longest? cycle from node from to node to. */
377int
378find_cycle(NODE *from, NODE *to, int longest_len, int depth)
379{
380	NODE **np;
381	int i, len;
382
383	/*
384	 * avoid infinite loops and ignore portions of the graph known
385	 * to be acyclic
386	 */
387	if (from->n_flags & (NF_NODEST|NF_MARK|NF_ACYCLIC))
388		return (0);
389	from->n_flags |= NF_MARK;
390
391	for (np = from->n_arcs, i = from->n_narcs; --i >= 0; np++) {
392		cycle_buf[depth] = *np;
393		if (*np == to) {
394			if (depth + 1 > longest_len) {
395				longest_len = depth + 1;
396				(void)memcpy((char *)longest_cycle,
397				    (char *)cycle_buf,
398				    longest_len * sizeof(NODE *));
399			}
400		} else {
401			if ((*np)->n_flags & (NF_MARK|NF_ACYCLIC|NF_NODEST))
402				continue;
403			len = find_cycle(*np, to, longest_len, depth + 1);
404
405			if (debug)
406				(void)printf("%*s %s->%s %d\n", depth, "",
407				    from->n_name, to->n_name, len);
408
409			if (len == 0)
410				(*np)->n_flags |= NF_NODEST;
411
412			if (len > longest_len)
413				longest_len = len;
414
415			if (len > 0 && !longest)
416				break;
417		}
418	}
419	from->n_flags &= ~NF_MARK;
420	return (longest_len);
421}
422
423void
424usage(void)
425{
426	(void)fprintf(stderr, "usage: tsort [-dlq] [file]\n");
427	exit(1);
428}
429