]> git.cameronkatri.com Git - mandoc.git/blob - mansearch.c
quoted words are not parsed for defined keys
[mandoc.git] / mansearch.c
1 /* $OpenBSD: mansearch.c,v 1.50 2016/07/09 15:23:36 schwarze Exp $ */
2 /*
3 * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2013-2017 Ingo Schwarze <schwarze@openbsd.org>
5 *
6 * Permission to use, copy, modify, and distribute this software for any
7 * purpose with or without fee is hereby granted, provided that the above
8 * copyright notice and this permission notice appear in all copies.
9 *
10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
13 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17 */
18 #include "config.h"
19
20 #include <sys/mman.h>
21 #include <sys/types.h>
22
23 #include <assert.h>
24 #if HAVE_ERR
25 #include <err.h>
26 #endif
27 #include <errno.h>
28 #include <fcntl.h>
29 #include <glob.h>
30 #include <limits.h>
31 #include <regex.h>
32 #include <stdio.h>
33 #include <stdint.h>
34 #include <stddef.h>
35 #include <stdlib.h>
36 #include <string.h>
37 #include <unistd.h>
38
39 #include "mandoc.h"
40 #include "mandoc_aux.h"
41 #include "mandoc_ohash.h"
42 #include "manconf.h"
43 #include "mansearch.h"
44 #include "dbm.h"
45
46 struct expr {
47 /* Used for terms: */
48 struct dbm_match match; /* Match type and expression. */
49 uint64_t bits; /* Type mask. */
50 /* Used for OR and AND groups: */
51 struct expr *next; /* Next child in the parent group. */
52 struct expr *child; /* First child in this group. */
53 enum { EXPR_TERM, EXPR_OR, EXPR_AND } type;
54 };
55
56 const char *const mansearch_keynames[KEY_MAX] = {
57 "arch", "sec", "Xr", "Ar", "Fa", "Fl", "Dv", "Fn",
58 "Ic", "Pa", "Cm", "Li", "Em", "Cd", "Va", "Ft",
59 "Tn", "Er", "Ev", "Sy", "Sh", "In", "Ss", "Ox",
60 "An", "Mt", "St", "Bx", "At", "Nx", "Fx", "Lk",
61 "Ms", "Bsx", "Dx", "Rs", "Vt", "Lb", "Nm", "Nd"
62 };
63
64
65 static struct ohash *manmerge(struct expr *, struct ohash *);
66 static struct ohash *manmerge_term(struct expr *, struct ohash *);
67 static struct ohash *manmerge_or(struct expr *, struct ohash *);
68 static struct ohash *manmerge_and(struct expr *, struct ohash *);
69 static char *buildnames(const struct dbm_page *);
70 static char *buildoutput(size_t, struct dbm_page *);
71 static size_t lstlen(const char *, size_t);
72 static void lstcat(char *, size_t *, const char *, const char *);
73 static int lstmatch(const char *, const char *);
74 static struct expr *exprcomp(const struct mansearch *,
75 int, char *[], int *);
76 static struct expr *expr_and(const struct mansearch *,
77 int, char *[], int *);
78 static struct expr *exprterm(const struct mansearch *,
79 int, char *[], int *);
80 static void exprfree(struct expr *);
81 static int manpage_compare(const void *, const void *);
82
83
84 int
85 mansearch(const struct mansearch *search,
86 const struct manpaths *paths,
87 int argc, char *argv[],
88 struct manpage **res, size_t *sz)
89 {
90 char buf[PATH_MAX];
91 struct dbm_res *rp;
92 struct expr *e;
93 struct dbm_page *page;
94 struct manpage *mpage;
95 struct ohash *htab;
96 size_t cur, i, maxres, outkey;
97 unsigned int slot;
98 int argi, chdir_status, getcwd_status, im;
99
100 argi = 0;
101 if ((e = exprcomp(search, argc, argv, &argi)) == NULL) {
102 *sz = 0;
103 return 0;
104 }
105
106 cur = maxres = 0;
107 *res = NULL;
108
109 outkey = KEY_Nd;
110 if (search->outkey != NULL)
111 for (im = 0; im < KEY_MAX; im++)
112 if (0 == strcasecmp(search->outkey,
113 mansearch_keynames[im])) {
114 outkey = im;
115 break;
116 }
117
118 /*
119 * Remember the original working directory, if possible.
120 * This will be needed if the second or a later directory
121 * is given as a relative path.
122 * Do not error out if the current directory is not
123 * searchable: Maybe it won't be needed after all.
124 */
125
126 if (getcwd(buf, PATH_MAX) == NULL) {
127 getcwd_status = 0;
128 (void)strlcpy(buf, strerror(errno), sizeof(buf));
129 } else
130 getcwd_status = 1;
131
132 /*
133 * Loop over the directories (containing databases) for us to
134 * search.
135 * Don't let missing/bad databases/directories phase us.
136 * In each, try to open the resident database and, if it opens,
137 * scan it for our match expression.
138 */
139
140 chdir_status = 0;
141 for (i = 0; i < paths->sz; i++) {
142 if (chdir_status && paths->paths[i][0] != '/') {
143 if ( ! getcwd_status) {
144 warnx("%s: getcwd: %s", paths->paths[i], buf);
145 continue;
146 } else if (chdir(buf) == -1) {
147 warn("%s", buf);
148 continue;
149 }
150 }
151 if (chdir(paths->paths[i]) == -1) {
152 warn("%s", paths->paths[i]);
153 continue;
154 }
155 chdir_status = 1;
156
157 if (dbm_open(MANDOC_DB) == -1) {
158 if (errno != ENOENT)
159 warn("%s/%s", paths->paths[i], MANDOC_DB);
160 continue;
161 }
162
163 if ((htab = manmerge(e, NULL)) == NULL) {
164 dbm_close();
165 continue;
166 }
167
168 for (rp = ohash_first(htab, &slot); rp != NULL;
169 rp = ohash_next(htab, &slot)) {
170 page = dbm_page_get(rp->page);
171
172 if (lstmatch(search->sec, page->sect) == 0 ||
173 lstmatch(search->arch, page->arch) == 0)
174 continue;
175
176 if (cur + 1 > maxres) {
177 maxres += 1024;
178 *res = mandoc_reallocarray(*res,
179 maxres, sizeof(**res));
180 }
181 mpage = *res + cur;
182 mandoc_asprintf(&mpage->file, "%s/%s",
183 paths->paths[i], page->file + 1);
184 mpage->names = buildnames(page);
185 mpage->output = buildoutput(outkey, page);
186 mpage->ipath = i;
187 mpage->bits = rp->bits;
188 mpage->sec = *page->sect - '0';
189 if (mpage->sec < 0 || mpage->sec > 9)
190 mpage->sec = 10;
191 mpage->form = *page->file;
192 free(rp);
193 cur++;
194 }
195 ohash_delete(htab);
196 free(htab);
197 dbm_close();
198
199 /*
200 * In man(1) mode, prefer matches in earlier trees
201 * over matches in later trees.
202 */
203
204 if (cur && search->firstmatch)
205 break;
206 }
207 qsort(*res, cur, sizeof(struct manpage), manpage_compare);
208 if (chdir_status && getcwd_status && chdir(buf) == -1)
209 warn("%s", buf);
210 exprfree(e);
211 *sz = cur;
212 return 1;
213 }
214
215 /*
216 * Merge the results for the expression tree rooted at e
217 * into the the result list htab.
218 */
219 static struct ohash *
220 manmerge(struct expr *e, struct ohash *htab)
221 {
222 switch (e->type) {
223 case EXPR_TERM:
224 return manmerge_term(e, htab);
225 case EXPR_OR:
226 return manmerge_or(e->child, htab);
227 case EXPR_AND:
228 return manmerge_and(e->child, htab);
229 default:
230 abort();
231 }
232 }
233
234 static struct ohash *
235 manmerge_term(struct expr *e, struct ohash *htab)
236 {
237 struct dbm_res res, *rp;
238 uint64_t ib;
239 unsigned int slot;
240 int im;
241
242 if (htab == NULL) {
243 htab = mandoc_malloc(sizeof(*htab));
244 mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
245 }
246
247 for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
248 if ((e->bits & ib) == 0)
249 continue;
250
251 switch (ib) {
252 case TYPE_arch:
253 dbm_page_byarch(&e->match);
254 break;
255 case TYPE_sec:
256 dbm_page_bysect(&e->match);
257 break;
258 case TYPE_Nm:
259 dbm_page_byname(&e->match);
260 break;
261 case TYPE_Nd:
262 dbm_page_bydesc(&e->match);
263 break;
264 default:
265 dbm_page_bymacro(im - 2, &e->match);
266 break;
267 }
268
269 /*
270 * When hashing for deduplication, use the unique
271 * page ID itself instead of a hash function;
272 * that is quite efficient.
273 */
274
275 for (;;) {
276 res = dbm_page_next();
277 if (res.page == -1)
278 break;
279 slot = ohash_lookup_memory(htab,
280 (char *)&res, sizeof(res.page), res.page);
281 if ((rp = ohash_find(htab, slot)) != NULL) {
282 rp->bits |= res.bits;
283 continue;
284 }
285 rp = mandoc_malloc(sizeof(*rp));
286 *rp = res;
287 ohash_insert(htab, slot, rp);
288 }
289 }
290 return htab;
291 }
292
293 static struct ohash *
294 manmerge_or(struct expr *e, struct ohash *htab)
295 {
296 while (e != NULL) {
297 htab = manmerge(e, htab);
298 e = e->next;
299 }
300 return htab;
301 }
302
303 static struct ohash *
304 manmerge_and(struct expr *e, struct ohash *htab)
305 {
306 struct ohash *hand, *h1, *h2;
307 struct dbm_res *res;
308 unsigned int slot1, slot2;
309
310 /* Evaluate the first term of the AND clause. */
311
312 hand = manmerge(e, NULL);
313
314 while ((e = e->next) != NULL) {
315
316 /* Evaluate the next term and prepare for ANDing. */
317
318 h2 = manmerge(e, NULL);
319 if (ohash_entries(h2) < ohash_entries(hand)) {
320 h1 = h2;
321 h2 = hand;
322 } else
323 h1 = hand;
324 hand = mandoc_malloc(sizeof(*hand));
325 mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
326
327 /* Keep all pages that are in both result sets. */
328
329 for (res = ohash_first(h1, &slot1); res != NULL;
330 res = ohash_next(h1, &slot1)) {
331 if (ohash_find(h2, ohash_lookup_memory(h2,
332 (char *)res, sizeof(res->page),
333 res->page)) == NULL)
334 free(res);
335 else
336 ohash_insert(hand, ohash_lookup_memory(hand,
337 (char *)res, sizeof(res->page),
338 res->page), res);
339 }
340
341 /* Discard the merged results. */
342
343 for (res = ohash_first(h2, &slot2); res != NULL;
344 res = ohash_next(h2, &slot2))
345 free(res);
346 ohash_delete(h2);
347 free(h2);
348 ohash_delete(h1);
349 free(h1);
350 }
351
352 /* Merge the result of the AND into htab. */
353
354 if (htab == NULL)
355 return hand;
356
357 for (res = ohash_first(hand, &slot1); res != NULL;
358 res = ohash_next(hand, &slot1)) {
359 slot2 = ohash_lookup_memory(htab,
360 (char *)res, sizeof(res->page), res->page);
361 if (ohash_find(htab, slot2) == NULL)
362 ohash_insert(htab, slot2, res);
363 else
364 free(res);
365 }
366
367 /* Discard the merged result. */
368
369 ohash_delete(hand);
370 free(hand);
371 return htab;
372 }
373
374 void
375 mansearch_free(struct manpage *res, size_t sz)
376 {
377 size_t i;
378
379 for (i = 0; i < sz; i++) {
380 free(res[i].file);
381 free(res[i].names);
382 free(res[i].output);
383 }
384 free(res);
385 }
386
387 static int
388 manpage_compare(const void *vp1, const void *vp2)
389 {
390 const struct manpage *mp1, *mp2;
391 int diff;
392
393 mp1 = vp1;
394 mp2 = vp2;
395 return (diff = mp2->bits - mp1->bits) ? diff :
396 (diff = mp1->sec - mp2->sec) ? diff :
397 strcasecmp(mp1->names, mp2->names);
398 }
399
400 static char *
401 buildnames(const struct dbm_page *page)
402 {
403 char *buf;
404 size_t i, sz;
405
406 sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
407 (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
408 buf = mandoc_malloc(sz);
409 i = 0;
410 lstcat(buf, &i, page->name, ", ");
411 buf[i++] = '(';
412 lstcat(buf, &i, page->sect, ", ");
413 if (page->arch != NULL) {
414 buf[i++] = '/';
415 lstcat(buf, &i, page->arch, ", ");
416 }
417 buf[i++] = ')';
418 buf[i++] = '\0';
419 assert(i == sz);
420 return buf;
421 }
422
423 /*
424 * Count the buffer space needed to print the NUL-terminated
425 * list of NUL-terminated strings, when printing sep separator
426 * characters between strings.
427 */
428 static size_t
429 lstlen(const char *cp, size_t sep)
430 {
431 size_t sz;
432
433 for (sz = 0;; sz++) {
434 if (cp[0] == '\0') {
435 if (cp[1] == '\0')
436 break;
437 sz += sep - 1;
438 } else if (cp[0] < ' ')
439 sz--;
440 cp++;
441 }
442 return sz;
443 }
444
445 /*
446 * Print the NUL-terminated list of NUL-terminated strings
447 * into the buffer, seperating strings with sep.
448 */
449 static void
450 lstcat(char *buf, size_t *i, const char *cp, const char *sep)
451 {
452 const char *s;
453
454 for (;;) {
455 if (cp[0] == '\0') {
456 if (cp[1] == '\0')
457 break;
458 s = sep;
459 while (*s != '\0')
460 buf[(*i)++] = *s++;
461 } else if (cp[0] >= ' ')
462 buf[(*i)++] = cp[0];
463 cp++;
464 }
465 }
466
467 /*
468 * Return 1 if the string *want occurs in any of the strings
469 * in the NUL-terminated string list *have, or 0 otherwise.
470 * If either argument is NULL or empty, assume no filtering
471 * is desired and return 1.
472 */
473 static int
474 lstmatch(const char *want, const char *have)
475 {
476 if (want == NULL || have == NULL || *have == '\0')
477 return 1;
478 while (*have != '\0') {
479 if (strcasestr(have, want) != NULL)
480 return 1;
481 have = strchr(have, '\0') + 1;
482 }
483 return 0;
484 }
485
486 /*
487 * Build a list of values taken by the macro im in the manual page.
488 */
489 static char *
490 buildoutput(size_t im, struct dbm_page *page)
491 {
492 const char *oldoutput, *sep, *input;
493 char *output, *newoutput, *value;
494 size_t sz, i;
495
496 switch (im) {
497 case KEY_Nd:
498 return mandoc_strdup(page->desc);
499 case KEY_Nm:
500 input = page->name;
501 break;
502 case KEY_sec:
503 input = page->sect;
504 break;
505 case KEY_arch:
506 input = page->arch;
507 if (input == NULL)
508 input = "all\0";
509 break;
510 default:
511 input = NULL;
512 break;
513 }
514
515 if (input != NULL) {
516 sz = lstlen(input, 3) + 1;
517 output = mandoc_malloc(sz);
518 i = 0;
519 lstcat(output, &i, input, " # ");
520 output[i++] = '\0';
521 assert(i == sz);
522 return output;
523 }
524
525 output = NULL;
526 dbm_macro_bypage(im - 2, page->addr);
527 while ((value = dbm_macro_next()) != NULL) {
528 if (output == NULL) {
529 oldoutput = "";
530 sep = "";
531 } else {
532 oldoutput = output;
533 sep = " # ";
534 }
535 mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
536 free(output);
537 output = newoutput;
538 }
539 return output;
540 }
541
542 /*
543 * Compile a set of string tokens into an expression.
544 * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
545 * "(", "foo=bar", etc.).
546 */
547 static struct expr *
548 exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
549 {
550 struct expr *parent, *child;
551 int needterm, nested;
552
553 if ((nested = *argi) == argc)
554 return NULL;
555 needterm = 1;
556 parent = child = NULL;
557 while (*argi < argc) {
558 if (strcmp(")", argv[*argi]) == 0) {
559 if (needterm)
560 warnx("missing term "
561 "before closing parenthesis");
562 needterm = 0;
563 if (nested)
564 break;
565 warnx("ignoring unmatched right parenthesis");
566 ++*argi;
567 continue;
568 }
569 if (strcmp("-o", argv[*argi]) == 0) {
570 if (needterm) {
571 if (*argi > 0)
572 warnx("ignoring -o after %s",
573 argv[*argi - 1]);
574 else
575 warnx("ignoring initial -o");
576 }
577 needterm = 1;
578 ++*argi;
579 continue;
580 }
581 needterm = 0;
582 if (child == NULL) {
583 child = expr_and(search, argc, argv, argi);
584 continue;
585 }
586 if (parent == NULL) {
587 parent = mandoc_calloc(1, sizeof(*parent));
588 parent->type = EXPR_OR;
589 parent->next = NULL;
590 parent->child = child;
591 }
592 child->next = expr_and(search, argc, argv, argi);
593 child = child->next;
594 }
595 if (needterm && *argi)
596 warnx("ignoring trailing %s", argv[*argi - 1]);
597 return parent == NULL ? child : parent;
598 }
599
600 static struct expr *
601 expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
602 {
603 struct expr *parent, *child;
604 int needterm;
605
606 needterm = 1;
607 parent = child = NULL;
608 while (*argi < argc) {
609 if (strcmp(")", argv[*argi]) == 0) {
610 if (needterm)
611 warnx("missing term "
612 "before closing parenthesis");
613 needterm = 0;
614 break;
615 }
616 if (strcmp("-o", argv[*argi]) == 0)
617 break;
618 if (strcmp("-a", argv[*argi]) == 0) {
619 if (needterm) {
620 if (*argi > 0)
621 warnx("ignoring -a after %s",
622 argv[*argi - 1]);
623 else
624 warnx("ignoring initial -a");
625 }
626 needterm = 1;
627 ++*argi;
628 continue;
629 }
630 if (needterm == 0)
631 break;
632 if (child == NULL) {
633 child = exprterm(search, argc, argv, argi);
634 if (child != NULL)
635 needterm = 0;
636 continue;
637 }
638 needterm = 0;
639 if (parent == NULL) {
640 parent = mandoc_calloc(1, sizeof(*parent));
641 parent->type = EXPR_AND;
642 parent->next = NULL;
643 parent->child = child;
644 }
645 child->next = exprterm(search, argc, argv, argi);
646 if (child->next != NULL) {
647 child = child->next;
648 needterm = 0;
649 }
650 }
651 if (needterm && *argi)
652 warnx("ignoring trailing %s", argv[*argi - 1]);
653 return parent == NULL ? child : parent;
654 }
655
656 static struct expr *
657 exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
658 {
659 char errbuf[BUFSIZ];
660 struct expr *e;
661 char *key, *val;
662 uint64_t iterbit;
663 int cs, i, irc;
664
665 if (strcmp("(", argv[*argi]) == 0) {
666 ++*argi;
667 e = exprcomp(search, argc, argv, argi);
668 if (*argi < argc) {
669 assert(strcmp(")", argv[*argi]) == 0);
670 ++*argi;
671 } else
672 warnx("unclosed parenthesis");
673 return e;
674 }
675
676 if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
677 cs = 0;
678 ++*argi;
679 } else
680 cs = 1;
681
682 e = mandoc_calloc(1, sizeof(*e));
683 e->type = EXPR_TERM;
684 e->bits = 0;
685 e->next = NULL;
686 e->child = NULL;
687
688 if (search->argmode == ARG_NAME) {
689 e->bits = TYPE_Nm;
690 e->match.type = DBM_EXACT;
691 e->match.str = argv[(*argi)++];
692 return e;
693 }
694
695 /*
696 * Separate macro keys from search string.
697 * If needed, request regular expression handling.
698 */
699
700 if (search->argmode == ARG_WORD) {
701 e->bits = TYPE_Nm;
702 e->match.type = DBM_REGEX;
703 #if HAVE_REWB_BSD
704 mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
705 #elif HAVE_REWB_SYSV
706 mandoc_asprintf(&val, "\\<%s\\>", argv[*argi]);
707 #else
708 mandoc_asprintf(&val,
709 "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", argv[*argi]);
710 #endif
711 cs = 0;
712 } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
713 e->bits = TYPE_Nm | TYPE_Nd;
714 e->match.type = DBM_SUB;
715 e->match.str = argv[*argi];
716 } else {
717 if (val == argv[*argi])
718 e->bits = TYPE_Nm | TYPE_Nd;
719 if (*val == '=') {
720 e->match.type = DBM_SUB;
721 e->match.str = val + 1;
722 } else
723 e->match.type = DBM_REGEX;
724 *val++ = '\0';
725 if (strstr(argv[*argi], "arch") != NULL)
726 cs = 0;
727 }
728
729 /* Compile regular expressions. */
730
731 if (e->match.type == DBM_REGEX) {
732 e->match.re = mandoc_malloc(sizeof(*e->match.re));
733 irc = regcomp(e->match.re, val,
734 REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
735 if (irc) {
736 regerror(irc, e->match.re, errbuf, sizeof(errbuf));
737 warnx("regcomp /%s/: %s", val, errbuf);
738 }
739 if (search->argmode == ARG_WORD)
740 free(val);
741 if (irc) {
742 free(e->match.re);
743 free(e);
744 ++*argi;
745 return NULL;
746 }
747 }
748
749 if (e->bits) {
750 ++*argi;
751 return e;
752 }
753
754 /*
755 * Parse out all possible fields.
756 * If the field doesn't resolve, bail.
757 */
758
759 while (NULL != (key = strsep(&argv[*argi], ","))) {
760 if ('\0' == *key)
761 continue;
762 for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
763 if (0 == strcasecmp(key, mansearch_keynames[i])) {
764 e->bits |= iterbit;
765 break;
766 }
767 }
768 if (i == KEY_MAX) {
769 if (strcasecmp(key, "any"))
770 warnx("treating unknown key "
771 "\"%s\" as \"any\"", key);
772 e->bits |= ~0ULL;
773 }
774 }
775
776 ++*argi;
777 return e;
778 }
779
780 static void
781 exprfree(struct expr *e)
782 {
783 if (e->next != NULL)
784 exprfree(e->next);
785 if (e->child != NULL)
786 exprfree(e->child);
787 free(e);
788 }