1 /* $OpenBSD: mansearch.c,v 1.50 2016/07/09 15:23:36 schwarze Exp $ */
3 * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2013-2017 Ingo Schwarze <schwarze@openbsd.org>
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.
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.
21 #include <sys/types.h>
40 #include "mandoc_aux.h"
41 #include "mandoc_ohash.h"
43 #include "mansearch.h"
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;
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"
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 *);
85 mansearch(const struct mansearch *search,
86 const struct manpaths *paths,
87 int argc, char *argv[],
88 struct manpage **res, size_t *sz)
93 struct dbm_page *page;
94 struct manpage *mpage;
96 size_t cur, i, maxres, outkey;
98 int argi, chdir_status, getcwd_status, im;
101 if ((e = exprcomp(search, argc, argv, &argi)) == NULL) {
111 if (search->outkey != NULL)
112 for (im = 0; im < KEY_MAX; im++)
113 if (0 == strcasecmp(search->outkey,
114 mansearch_keynames[im])) {
120 * Remember the original working directory, if possible.
121 * This will be needed if the second or a later directory
122 * is given as a relative path.
123 * Do not error out if the current directory is not
124 * searchable: Maybe it won't be needed after all.
127 if (getcwd(buf, PATH_MAX) == NULL) {
129 (void)strlcpy(buf, strerror(errno), sizeof(buf));
134 * Loop over the directories (containing databases) for us to
136 * Don't let missing/bad databases/directories phase us.
137 * In each, try to open the resident database and, if it opens,
138 * scan it for our match expression.
142 for (i = 0; i < paths->sz; i++) {
143 if (chdir_status && paths->paths[i][0] != '/') {
144 if ( ! getcwd_status) {
145 warnx("%s: getcwd: %s", paths->paths[i], buf);
147 } else if (chdir(buf) == -1) {
152 if (chdir(paths->paths[i]) == -1) {
153 warn("%s", paths->paths[i]);
158 if (dbm_open(MANDOC_DB) == -1) {
160 warn("%s/%s", paths->paths[i], MANDOC_DB);
164 if ((htab = manmerge(e, NULL)) == NULL) {
169 for (rp = ohash_first(htab, &slot); rp != NULL;
170 rp = ohash_next(htab, &slot)) {
171 page = dbm_page_get(rp->page);
173 if (lstmatch(search->sec, page->sect) == 0 ||
174 lstmatch(search->arch, page->arch) == 0)
181 if (cur + 1 > maxres) {
183 *res = mandoc_reallocarray(*res,
184 maxres, sizeof(**res));
187 mandoc_asprintf(&mpage->file, "%s/%s",
188 paths->paths[i], page->file + 1);
189 mpage->names = buildnames(page);
190 mpage->output = buildoutput(outkey, page);
192 mpage->bits = rp->bits;
193 mpage->sec = *page->sect - '0';
194 if (mpage->sec < 0 || mpage->sec > 9)
196 mpage->form = *page->file;
205 * In man(1) mode, prefer matches in earlier trees
206 * over matches in later trees.
209 if (cur && search->firstmatch)
213 qsort(*res, cur, sizeof(struct manpage), manpage_compare);
214 if (chdir_status && getcwd_status && chdir(buf) == -1)
218 return res != NULL || cur;
222 * Merge the results for the expression tree rooted at e
223 * into the the result list htab.
225 static struct ohash *
226 manmerge(struct expr *e, struct ohash *htab)
230 return manmerge_term(e, htab);
232 return manmerge_or(e->child, htab);
234 return manmerge_and(e->child, htab);
240 static struct ohash *
241 manmerge_term(struct expr *e, struct ohash *htab)
243 struct dbm_res res, *rp;
249 htab = mandoc_malloc(sizeof(*htab));
250 mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
253 for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
254 if ((e->bits & ib) == 0)
259 dbm_page_byarch(&e->match);
262 dbm_page_bysect(&e->match);
265 dbm_page_byname(&e->match);
268 dbm_page_bydesc(&e->match);
271 dbm_page_bymacro(im - 2, &e->match);
276 * When hashing for deduplication, use the unique
277 * page ID itself instead of a hash function;
278 * that is quite efficient.
282 res = dbm_page_next();
285 slot = ohash_lookup_memory(htab,
286 (char *)&res, sizeof(res.page), res.page);
287 if ((rp = ohash_find(htab, slot)) != NULL) {
288 rp->bits |= res.bits;
291 rp = mandoc_malloc(sizeof(*rp));
293 ohash_insert(htab, slot, rp);
299 static struct ohash *
300 manmerge_or(struct expr *e, struct ohash *htab)
303 htab = manmerge(e, htab);
309 static struct ohash *
310 manmerge_and(struct expr *e, struct ohash *htab)
312 struct ohash *hand, *h1, *h2;
314 unsigned int slot1, slot2;
316 /* Evaluate the first term of the AND clause. */
318 hand = manmerge(e, NULL);
320 while ((e = e->next) != NULL) {
322 /* Evaluate the next term and prepare for ANDing. */
324 h2 = manmerge(e, NULL);
325 if (ohash_entries(h2) < ohash_entries(hand)) {
330 hand = mandoc_malloc(sizeof(*hand));
331 mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
333 /* Keep all pages that are in both result sets. */
335 for (res = ohash_first(h1, &slot1); res != NULL;
336 res = ohash_next(h1, &slot1)) {
337 if (ohash_find(h2, ohash_lookup_memory(h2,
338 (char *)res, sizeof(res->page),
342 ohash_insert(hand, ohash_lookup_memory(hand,
343 (char *)res, sizeof(res->page),
347 /* Discard the merged results. */
349 for (res = ohash_first(h2, &slot2); res != NULL;
350 res = ohash_next(h2, &slot2))
358 /* Merge the result of the AND into htab. */
363 for (res = ohash_first(hand, &slot1); res != NULL;
364 res = ohash_next(hand, &slot1)) {
365 slot2 = ohash_lookup_memory(htab,
366 (char *)res, sizeof(res->page), res->page);
367 if (ohash_find(htab, slot2) == NULL)
368 ohash_insert(htab, slot2, res);
373 /* Discard the merged result. */
381 mansearch_free(struct manpage *res, size_t sz)
385 for (i = 0; i < sz; i++) {
394 manpage_compare(const void *vp1, const void *vp2)
396 const struct manpage *mp1, *mp2;
397 const char *cp1, *cp2;
403 if ((diff = mp2->bits - mp1->bits) ||
404 (diff = mp1->sec - mp2->sec))
407 /* Fall back to alphabetic ordering of names. */
408 sz1 = strcspn(mp1->names, "(");
409 sz2 = strcspn(mp2->names, "(");
412 if ((diff = strncasecmp(mp1->names, mp2->names, sz1)))
415 /* For identical names and sections, prefer arch-dependent. */
416 cp1 = strchr(mp1->names + sz1, '/');
417 cp2 = strchr(mp2->names + sz2, '/');
418 return cp1 != NULL && cp2 != NULL ? strcasecmp(cp1, cp2) :
419 cp1 != NULL ? -1 : cp2 != NULL ? 1 : 0;
423 buildnames(const struct dbm_page *page)
428 sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
429 (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
430 buf = mandoc_malloc(sz);
432 lstcat(buf, &i, page->name, ", ");
434 lstcat(buf, &i, page->sect, ", ");
435 if (page->arch != NULL) {
437 lstcat(buf, &i, page->arch, ", ");
446 * Count the buffer space needed to print the NUL-terminated
447 * list of NUL-terminated strings, when printing sep separator
448 * characters between strings.
451 lstlen(const char *cp, size_t sep)
455 for (sz = 0;; sz++) {
460 } else if (cp[0] < ' ')
468 * Print the NUL-terminated list of NUL-terminated strings
469 * into the buffer, seperating strings with sep.
472 lstcat(char *buf, size_t *i, const char *cp, const char *sep)
483 } else if (cp[0] >= ' ')
490 * Return 1 if the string *want occurs in any of the strings
491 * in the NUL-terminated string list *have, or 0 otherwise.
492 * If either argument is NULL or empty, assume no filtering
493 * is desired and return 1.
496 lstmatch(const char *want, const char *have)
498 if (want == NULL || have == NULL || *have == '\0')
500 while (*have != '\0') {
501 if (strcasestr(have, want) != NULL)
503 have = strchr(have, '\0') + 1;
509 * Build a list of values taken by the macro im in the manual page.
512 buildoutput(size_t im, struct dbm_page *page)
514 const char *oldoutput, *sep, *input;
515 char *output, *newoutput, *value;
520 return mandoc_strdup(page->desc);
538 sz = lstlen(input, 3) + 1;
539 output = mandoc_malloc(sz);
541 lstcat(output, &i, input, " # ");
548 dbm_macro_bypage(im - 2, page->addr);
549 while ((value = dbm_macro_next()) != NULL) {
550 if (output == NULL) {
557 mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
565 * Compile a set of string tokens into an expression.
566 * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
567 * "(", "foo=bar", etc.).
570 exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
572 struct expr *parent, *child;
573 int needterm, nested;
575 if ((nested = *argi) == argc)
578 parent = child = NULL;
579 while (*argi < argc) {
580 if (strcmp(")", argv[*argi]) == 0) {
582 warnx("missing term "
583 "before closing parenthesis");
587 warnx("ignoring unmatched right parenthesis");
591 if (strcmp("-o", argv[*argi]) == 0) {
594 warnx("ignoring -o after %s",
597 warnx("ignoring initial -o");
605 child = expr_and(search, argc, argv, argi);
608 if (parent == NULL) {
609 parent = mandoc_calloc(1, sizeof(*parent));
610 parent->type = EXPR_OR;
612 parent->child = child;
614 child->next = expr_and(search, argc, argv, argi);
617 if (needterm && *argi)
618 warnx("ignoring trailing %s", argv[*argi - 1]);
619 return parent == NULL ? child : parent;
623 expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
625 struct expr *parent, *child;
629 parent = child = NULL;
630 while (*argi < argc) {
631 if (strcmp(")", argv[*argi]) == 0) {
633 warnx("missing term "
634 "before closing parenthesis");
638 if (strcmp("-o", argv[*argi]) == 0)
640 if (strcmp("-a", argv[*argi]) == 0) {
643 warnx("ignoring -a after %s",
646 warnx("ignoring initial -a");
655 child = exprterm(search, argc, argv, argi);
661 if (parent == NULL) {
662 parent = mandoc_calloc(1, sizeof(*parent));
663 parent->type = EXPR_AND;
665 parent->child = child;
667 child->next = exprterm(search, argc, argv, argi);
668 if (child->next != NULL) {
673 if (needterm && *argi)
674 warnx("ignoring trailing %s", argv[*argi - 1]);
675 return parent == NULL ? child : parent;
679 exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
687 if (strcmp("(", argv[*argi]) == 0) {
689 e = exprcomp(search, argc, argv, argi);
691 assert(strcmp(")", argv[*argi]) == 0);
694 warnx("unclosed parenthesis");
698 if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
704 e = mandoc_calloc(1, sizeof(*e));
710 if (search->argmode == ARG_NAME) {
712 e->match.type = DBM_EXACT;
713 e->match.str = argv[(*argi)++];
718 * Separate macro keys from search string.
719 * If needed, request regular expression handling.
722 if (search->argmode == ARG_WORD) {
724 e->match.type = DBM_REGEX;
726 mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
728 mandoc_asprintf(&val, "\\<%s\\>", argv[*argi]);
730 mandoc_asprintf(&val,
731 "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", argv[*argi]);
734 } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
735 e->bits = TYPE_Nm | TYPE_Nd;
736 e->match.type = DBM_SUB;
737 e->match.str = argv[*argi];
739 if (val == argv[*argi])
740 e->bits = TYPE_Nm | TYPE_Nd;
742 e->match.type = DBM_SUB;
743 e->match.str = val + 1;
745 e->match.type = DBM_REGEX;
747 if (strstr(argv[*argi], "arch") != NULL)
751 /* Compile regular expressions. */
753 if (e->match.type == DBM_REGEX) {
754 e->match.re = mandoc_malloc(sizeof(*e->match.re));
755 irc = regcomp(e->match.re, val,
756 REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
758 regerror(irc, e->match.re, errbuf, sizeof(errbuf));
759 warnx("regcomp /%s/: %s", val, errbuf);
761 if (search->argmode == ARG_WORD)
777 * Parse out all possible fields.
778 * If the field doesn't resolve, bail.
781 while (NULL != (key = strsep(&argv[*argi], ","))) {
784 for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
785 if (0 == strcasecmp(key, mansearch_keynames[i])) {
791 if (strcasecmp(key, "any"))
792 warnx("treating unknown key "
793 "\"%s\" as \"any\"", key);
803 exprfree(struct expr *e)
807 if (e->child != NULL)