1/*	$NetBSD: tsort.c,v 1.26 2021/02/06 16:02:16 dbj 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. Neither the name of the University nor the names of its contributors
19 *    may be used to endorse or promote products derived from this software
20 *    without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 */
34
35#if HAVE_NBTOOL_CONFIG_H
36#include "nbtool_config.h"
37#endif
38
39#include <sys/cdefs.h>
40#if !defined(lint)
41__COPYRIGHT("@(#) Copyright (c) 1989, 1993, 1994\
42 The Regents of the University of California.  All rights reserved.");
43#if 0
44static char sccsid[] = "@(#)tsort.c	8.3 (Berkeley) 5/4/95";
45#endif
46__RCSID("$NetBSD: tsort.c,v 1.26 2021/02/06 16:02:16 dbj Exp $");
47#endif /* not lint */
48
49#include <sys/types.h>
50#include <ctype.h>
51#include <db.h>
52#include <err.h>
53#include <errno.h>
54#include <fcntl.h>
55#include <stdio.h>
56#include <stdlib.h>
57#include <string.h>
58#include <unistd.h>
59#if !HAVE_NBTOOL_CONFIG_H
60#include <util.h>
61#endif
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 compatible 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 than 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
85typedef struct node_str NODE;
86
87struct 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	size_t n_narcs;			/* number of arcs in n_arcs[] */
92	size_t n_arcsize;		/* size of n_arcs[] array */
93	size_t n_refcnt;		/* # of arcs pointing to this node */
94	int n_flags;			/* NF_* */
95	char n_name[1];			/* name of this node */
96};
97
98typedef struct _buf {
99	char *b_buf;
100	size_t b_bsize;
101} BUF;
102
103static DB *db;
104static NODE *graph, **cycle_buf, **longest_cycle;
105static int debug, longest, quiet, reverse;
106
107static void	 add_arc(const char *, const char *);
108static void	 clear_cycle(void);
109static size_t	 find_cycle(NODE *, NODE *, size_t, size_t);
110static NODE	*get_node(const char *);
111static void	 remove_node(NODE *);
112static void	 tsort(void);
113__dead static void	 usage(void);
114
115int
116main(int argc, char *argv[])
117{
118	BUF *b;
119	int c, n, ch;
120	FILE *fp;
121	size_t bsize, nused;
122	BUF bufs[2];
123
124	setprogname(argv[0]);
125
126	fp = NULL;
127	while ((ch = getopt(argc, argv, "dlqr")) != -1)
128		switch (ch) {
129		case 'd':
130			debug = 1;
131			break;
132		case 'l':
133			longest = 1;
134			break;
135		case 'q':
136			quiet = 1;
137			break;
138		case 'r':
139			reverse = 1;
140			break;
141		case '?':
142		default:
143			usage();
144		}
145	argc -= optind;
146	argv += optind;
147
148	switch (argc) {
149	case 0:
150		fp = stdin;
151		break;
152	case 1:
153		if ((fp = fopen(*argv, "r")) == NULL)
154			err(EXIT_FAILURE, "%s", *argv);
155		break;
156	default:
157		usage();
158	}
159
160	for (b = bufs, n = 2; n-- > 0; b++)
161		b->b_buf = erealloc(NULL, b->b_bsize = 1024);
162
163	/* parse input and build the graph */
164	for (n = 0, c = getc(fp);;) {
165		while (c != EOF && isspace(c))
166			c = getc(fp);
167		if (c == EOF)
168			break;
169
170		nused = 0;
171		b = &bufs[n];
172		bsize = b->b_bsize;
173		do {
174			b->b_buf[nused++] = (char)c;
175			if (nused == bsize)
176				b->b_buf = erealloc(b->b_buf, bsize *= 2);
177			c = getc(fp);
178		} while (c != EOF && !isspace(c));
179
180		b->b_buf[nused] = '\0';
181		b->b_bsize = bsize;
182		if (n) {
183			if (reverse)
184				add_arc(bufs[1].b_buf, bufs[0].b_buf);
185			else
186				add_arc(bufs[0].b_buf, bufs[1].b_buf);
187		}
188		n = !n;
189	}
190	(void)fclose(fp);
191	if (n)
192		errx(EXIT_FAILURE, "odd data count");
193
194	/* do the sort */
195	tsort();
196	return EXIT_SUCCESS;
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 */
203static void
204add_arc(const char *s1, const char *s2)
205{
206	NODE *n1;
207	NODE *n2;
208	size_t bsize, i;
209
210	n1 = get_node(s1);
211
212	if (strcmp(s1, s2) == 0)
213		return;
214
215	n2 = get_node(s2);
216
217	/*
218	 * Check if this arc is already here.
219	 */
220	for (i = 0; i < n1->n_narcs; i++)
221		if (n1->n_arcs[i] == n2)
222			return;
223	/*
224	 * Add it.
225	 */
226	if (n1->n_narcs == n1->n_arcsize) {
227		if (!n1->n_arcsize)
228			n1->n_arcsize = 10;
229		bsize = n1->n_arcsize * sizeof(*n1->n_arcs) * 2;
230		n1->n_arcs = erealloc(n1->n_arcs, bsize);
231		n1->n_arcsize = bsize / sizeof(*n1->n_arcs);
232	}
233	n1->n_arcs[n1->n_narcs++] = n2;
234	++n2->n_refcnt;
235}
236
237/* Find a node in the graph (insert if not found) and return a pointer to it. */
238static NODE *
239get_node(const char *name)
240{
241	DBT data, key;
242	NODE *n;
243
244	if (db == NULL &&
245	    (db = dbopen(NULL, O_RDWR, 0, DB_HASH, NULL)) == NULL)
246		err(EXIT_FAILURE, "db: %s", name);
247
248	key.data = __UNCONST(name);
249	key.size = strlen(name) + 1;
250
251	switch ((*db->get)(db, &key, &data, 0)) {
252	case 0:
253		(void)memmove(&n, data.data, sizeof(n));
254		return n;
255	case 1:
256		break;
257	default:
258	case -1:
259		err(EXIT_FAILURE, "db: %s", name);
260	}
261
262	n = emalloc(sizeof(NODE) + key.size);
263
264	n->n_narcs = 0;
265	n->n_arcsize = 0;
266	n->n_arcs = NULL;
267	n->n_refcnt = 0;
268	n->n_flags = 0;
269	(void)memmove(n->n_name, name, key.size);
270
271	/* Add to linked list. */
272	if ((n->n_next = graph) != NULL)
273		graph->n_prevp = &n->n_next;
274	n->n_prevp = &graph;
275	graph = n;
276
277	/* Add to hash table. */
278	data.data = &n;
279	data.size = sizeof(n);
280	if ((*db->put)(db, &key, &data, 0))
281		err(EXIT_FAILURE, "db: %s", name);
282	return n;
283}
284
285
286/*
287 * Clear the NODEST flag from all nodes.
288 */
289static void
290clear_cycle(void)
291{
292	NODE *n;
293
294	for (n = graph; n != NULL; n = n->n_next)
295		n->n_flags &= ~NF_NODEST;
296}
297
298/* do topological sort on graph */
299static void
300tsort(void)
301{
302	NODE *n, *next;
303	size_t cnt, i;
304
305	while (graph != NULL) {
306		/*
307		 * Keep getting rid of simple cases until there are none left,
308		 * if there are any nodes still in the graph, then there is
309		 * a cycle in it.
310		 */
311		do {
312			for (cnt = 0, n = graph; n != NULL; n = next) {
313				next = n->n_next;
314				if (n->n_refcnt == 0) {
315					remove_node(n);
316					++cnt;
317				}
318			}
319		} while (graph != NULL && cnt);
320
321		if (graph == NULL)
322			break;
323
324		if (!cycle_buf) {
325			/*
326			 * Allocate space for two cycle logs - one to be used
327			 * as scratch space, the other to save the longest
328			 * cycle.
329			 */
330			for (cnt = 0, n = graph; n != NULL; n = n->n_next)
331				++cnt;
332			cycle_buf = ecalloc(cnt, sizeof(*cycle_buf));
333			longest_cycle = ecalloc(cnt, sizeof(*longest_cycle));
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)) != 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) */
360static void
361remove_node(NODE *n)
362{
363	NODE **np;
364	size_t 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. */
377static size_t
378find_cycle(NODE *from, NODE *to, size_t longest_len, size_t depth)
379{
380	NODE **np;
381	size_t 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(longest_cycle, cycle_buf,
397				    longest_len * sizeof(NODE *));
398			}
399		} else {
400			if ((*np)->n_flags & (NF_MARK|NF_ACYCLIC|NF_NODEST))
401				continue;
402			len = find_cycle(*np, to, longest_len, depth + 1);
403
404			if (debug)
405				(void)printf("%*s %s->%s %zu\n", (int)depth, "",
406				    from->n_name, to->n_name, len);
407
408			if (len == 0)
409				(*np)->n_flags |= NF_NODEST;
410
411			if (len > longest_len)
412				longest_len = len;
413
414			if (len > 0 && !longest)
415				break;
416		}
417	}
418	from->n_flags &= ~NF_MARK;
419	return longest_len;
420}
421
422static void
423usage(void)
424{
425	(void)fprintf(stderr, "Usage: %s [-dlqr] [file]\n", getprogname());
426	exit(EXIT_FAILURE);
427}
428