1 /* $Id: mansearch.c,v 1.80 2018/12/13 11:55:46 schwarze Exp $ */
3 * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2013-2018 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>
39 #include "mandoc_aux.h"
40 #include "mandoc_ohash.h"
42 #include "mansearch.h"
47 struct dbm_match match; /* Match type and expression. */
48 uint64_t bits; /* Type mask. */
49 /* Used for OR and AND groups: */
50 struct expr *next; /* Next child in the parent group. */
51 struct expr *child; /* First child in this group. */
52 enum { EXPR_TERM, EXPR_OR, EXPR_AND } type;
55 const char *const mansearch_keynames[KEY_MAX] = {
56 "arch", "sec", "Xr", "Ar", "Fa", "Fl", "Dv", "Fn",
57 "Ic", "Pa", "Cm", "Li", "Em", "Cd", "Va", "Ft",
58 "Tn", "Er", "Ev", "Sy", "Sh", "In", "Ss", "Ox",
59 "An", "Mt", "St", "Bx", "At", "Nx", "Fx", "Lk",
60 "Ms", "Bsx", "Dx", "Rs", "Vt", "Lb", "Nm", "Nd"
64 static struct ohash *manmerge(struct expr *, struct ohash *);
65 static struct ohash *manmerge_term(struct expr *, struct ohash *);
66 static struct ohash *manmerge_or(struct expr *, struct ohash *);
67 static struct ohash *manmerge_and(struct expr *, struct ohash *);
68 static char *buildnames(const struct dbm_page *);
69 static char *buildoutput(size_t, struct dbm_page *);
70 static size_t lstlen(const char *, size_t);
71 static void lstcat(char *, size_t *, const char *, const char *);
72 static int lstmatch(const char *, const char *);
73 static struct expr *exprcomp(const struct mansearch *,
74 int, char *[], int *);
75 static struct expr *expr_and(const struct mansearch *,
76 int, char *[], int *);
77 static struct expr *exprterm(const struct mansearch *,
78 int, char *[], int *);
79 static void exprfree(struct expr *);
80 static int manpage_compare(const void *, const void *);
84 mansearch(const struct mansearch *search,
85 const struct manpaths *paths,
86 int argc, char *argv[],
87 struct manpage **res, size_t *sz)
92 struct dbm_page *page;
93 struct manpage *mpage;
95 size_t cur, i, maxres, outkey;
97 int argi, chdir_status, getcwd_status, im;
100 if ((e = exprcomp(search, argc, argv, &argi)) == NULL) {
110 if (search->outkey != NULL)
111 for (im = 0; im < KEY_MAX; im++)
112 if (0 == strcasecmp(search->outkey,
113 mansearch_keynames[im])) {
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.
126 if (getcwd(buf, PATH_MAX) == NULL) {
128 (void)strlcpy(buf, strerror(errno), sizeof(buf));
133 * Loop over the directories (containing databases) for us to
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.
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);
146 } else if (chdir(buf) == -1) {
151 if (chdir(paths->paths[i]) == -1) {
152 warn("%s", paths->paths[i]);
157 if (dbm_open(MANDOC_DB) == -1) {
159 warn("%s/%s", paths->paths[i], MANDOC_DB);
163 if ((htab = manmerge(e, NULL)) == NULL) {
168 for (rp = ohash_first(htab, &slot); rp != NULL;
169 rp = ohash_next(htab, &slot)) {
170 page = dbm_page_get(rp->page);
172 if (lstmatch(search->sec, page->sect) == 0 ||
173 lstmatch(search->arch, page->arch) == 0 ||
174 (search->argmode == ARG_NAME &&
175 rp->bits <= (int32_t)(NAME_SYN & NAME_MASK)))
182 if (cur + 1 > maxres) {
184 *res = mandoc_reallocarray(*res,
185 maxres, sizeof(**res));
188 mandoc_asprintf(&mpage->file, "%s/%s",
189 paths->paths[i], page->file + 1);
190 if (access(chdir_status ? page->file + 1 :
191 mpage->file, R_OK) == -1) {
192 warn("%s", mpage->file);
193 warnx("outdated mandoc.db contains "
194 "bogus %s entry, run makewhatis %s",
195 page->file + 1, paths->paths[i]);
200 mpage->names = buildnames(page);
201 mpage->output = buildoutput(outkey, page);
203 mpage->sec = *page->sect - '0';
204 if (mpage->sec < 0 || mpage->sec > 9)
206 mpage->form = *page->file;
215 * In man(1) mode, prefer matches in earlier trees
216 * over matches in later trees.
219 if (cur && search->firstmatch)
223 qsort(*res, cur, sizeof(struct manpage), manpage_compare);
224 if (chdir_status && getcwd_status && chdir(buf) == -1)
228 return res != NULL || cur;
232 * Merge the results for the expression tree rooted at e
233 * into the the result list htab.
235 static struct ohash *
236 manmerge(struct expr *e, struct ohash *htab)
240 return manmerge_term(e, htab);
242 return manmerge_or(e->child, htab);
244 return manmerge_and(e->child, htab);
250 static struct ohash *
251 manmerge_term(struct expr *e, struct ohash *htab)
253 struct dbm_res res, *rp;
259 htab = mandoc_malloc(sizeof(*htab));
260 mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
263 for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
264 if ((e->bits & ib) == 0)
269 dbm_page_byarch(&e->match);
272 dbm_page_bysect(&e->match);
275 dbm_page_byname(&e->match);
278 dbm_page_bydesc(&e->match);
281 dbm_page_bymacro(im - 2, &e->match);
286 * When hashing for deduplication, use the unique
287 * page ID itself instead of a hash function;
288 * that is quite efficient.
292 res = dbm_page_next();
295 slot = ohash_lookup_memory(htab,
296 (char *)&res, sizeof(res.page), res.page);
297 if ((rp = ohash_find(htab, slot)) != NULL)
299 rp = mandoc_malloc(sizeof(*rp));
301 ohash_insert(htab, slot, rp);
307 static struct ohash *
308 manmerge_or(struct expr *e, struct ohash *htab)
311 htab = manmerge(e, htab);
317 static struct ohash *
318 manmerge_and(struct expr *e, struct ohash *htab)
320 struct ohash *hand, *h1, *h2;
322 unsigned int slot1, slot2;
324 /* Evaluate the first term of the AND clause. */
326 hand = manmerge(e, NULL);
328 while ((e = e->next) != NULL) {
330 /* Evaluate the next term and prepare for ANDing. */
332 h2 = manmerge(e, NULL);
333 if (ohash_entries(h2) < ohash_entries(hand)) {
338 hand = mandoc_malloc(sizeof(*hand));
339 mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
341 /* Keep all pages that are in both result sets. */
343 for (res = ohash_first(h1, &slot1); res != NULL;
344 res = ohash_next(h1, &slot1)) {
345 if (ohash_find(h2, ohash_lookup_memory(h2,
346 (char *)res, sizeof(res->page),
350 ohash_insert(hand, ohash_lookup_memory(hand,
351 (char *)res, sizeof(res->page),
355 /* Discard the merged results. */
357 for (res = ohash_first(h2, &slot2); res != NULL;
358 res = ohash_next(h2, &slot2))
366 /* Merge the result of the AND into htab. */
371 for (res = ohash_first(hand, &slot1); res != NULL;
372 res = ohash_next(hand, &slot1)) {
373 slot2 = ohash_lookup_memory(htab,
374 (char *)res, sizeof(res->page), res->page);
375 if (ohash_find(htab, slot2) == NULL)
376 ohash_insert(htab, slot2, res);
381 /* Discard the merged result. */
389 mansearch_free(struct manpage *res, size_t sz)
393 for (i = 0; i < sz; i++) {
402 manpage_compare(const void *vp1, const void *vp2)
404 const struct manpage *mp1, *mp2;
405 const char *cp1, *cp2;
411 if ((diff = mp1->sec - mp2->sec))
414 /* Fall back to alphabetic ordering of names. */
415 sz1 = strcspn(mp1->names, "(");
416 sz2 = strcspn(mp2->names, "(");
419 if ((diff = strncasecmp(mp1->names, mp2->names, sz1)))
422 /* For identical names and sections, prefer arch-dependent. */
423 cp1 = strchr(mp1->names + sz1, '/');
424 cp2 = strchr(mp2->names + sz2, '/');
425 return cp1 != NULL && cp2 != NULL ? strcasecmp(cp1, cp2) :
426 cp1 != NULL ? -1 : cp2 != NULL ? 1 : 0;
430 buildnames(const struct dbm_page *page)
435 sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
436 (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
437 buf = mandoc_malloc(sz);
439 lstcat(buf, &i, page->name, ", ");
441 lstcat(buf, &i, page->sect, ", ");
442 if (page->arch != NULL) {
444 lstcat(buf, &i, page->arch, ", ");
453 * Count the buffer space needed to print the NUL-terminated
454 * list of NUL-terminated strings, when printing sep separator
455 * characters between strings.
458 lstlen(const char *cp, size_t sep)
462 for (sz = 0; *cp != '\0'; cp++) {
464 /* Skip names appearing only in the SYNOPSIS. */
465 if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
471 /* Skip name class markers. */
475 /* Print a separator before each but the first string. */
479 /* Copy one string. */
480 while (*cp != '\0') {
489 * Print the NUL-terminated list of NUL-terminated strings
490 * into the buffer, seperating strings with sep.
493 lstcat(char *buf, size_t *i, const char *cp, const char *sep)
498 for (i_start = *i; *cp != '\0'; cp++) {
500 /* Skip names appearing only in the SYNOPSIS. */
501 if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
507 /* Skip name class markers. */
511 /* Print a separator before each but the first string. */
518 /* Copy one string. */
526 * Return 1 if the string *want occurs in any of the strings
527 * in the NUL-terminated string list *have, or 0 otherwise.
528 * If either argument is NULL or empty, assume no filtering
529 * is desired and return 1.
532 lstmatch(const char *want, const char *have)
534 if (want == NULL || have == NULL || *have == '\0')
536 while (*have != '\0') {
537 if (strcasestr(have, want) != NULL)
539 have = strchr(have, '\0') + 1;
545 * Build a list of values taken by the macro im in the manual page.
548 buildoutput(size_t im, struct dbm_page *page)
550 const char *oldoutput, *sep, *input;
551 char *output, *newoutput, *value;
556 return mandoc_strdup(page->desc);
574 sz = lstlen(input, 3) + 1;
575 output = mandoc_malloc(sz);
577 lstcat(output, &i, input, " # ");
584 dbm_macro_bypage(im - 2, page->addr);
585 while ((value = dbm_macro_next()) != NULL) {
586 if (output == NULL) {
593 mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
601 * Compile a set of string tokens into an expression.
602 * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
603 * "(", "foo=bar", etc.).
606 exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
608 struct expr *parent, *child;
609 int needterm, nested;
611 if ((nested = *argi) == argc)
614 parent = child = NULL;
615 while (*argi < argc) {
616 if (strcmp(")", argv[*argi]) == 0) {
618 warnx("missing term "
619 "before closing parenthesis");
623 warnx("ignoring unmatched right parenthesis");
627 if (strcmp("-o", argv[*argi]) == 0) {
630 warnx("ignoring -o after %s",
633 warnx("ignoring initial -o");
641 child = expr_and(search, argc, argv, argi);
644 if (parent == NULL) {
645 parent = mandoc_calloc(1, sizeof(*parent));
646 parent->type = EXPR_OR;
648 parent->child = child;
650 child->next = expr_and(search, argc, argv, argi);
653 if (needterm && *argi)
654 warnx("ignoring trailing %s", argv[*argi - 1]);
655 return parent == NULL ? child : parent;
659 expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
661 struct expr *parent, *child;
665 parent = child = NULL;
666 while (*argi < argc) {
667 if (strcmp(")", argv[*argi]) == 0) {
669 warnx("missing term "
670 "before closing parenthesis");
674 if (strcmp("-o", argv[*argi]) == 0)
676 if (strcmp("-a", argv[*argi]) == 0) {
679 warnx("ignoring -a after %s",
682 warnx("ignoring initial -a");
691 child = exprterm(search, argc, argv, argi);
697 if (parent == NULL) {
698 parent = mandoc_calloc(1, sizeof(*parent));
699 parent->type = EXPR_AND;
701 parent->child = child;
703 child->next = exprterm(search, argc, argv, argi);
704 if (child->next != NULL) {
709 if (needterm && *argi)
710 warnx("ignoring trailing %s", argv[*argi - 1]);
711 return parent == NULL ? child : parent;
715 exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
723 if (strcmp("(", argv[*argi]) == 0) {
725 e = exprcomp(search, argc, argv, argi);
727 assert(strcmp(")", argv[*argi]) == 0);
730 warnx("unclosed parenthesis");
734 if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
740 e = mandoc_calloc(1, sizeof(*e));
746 if (search->argmode == ARG_NAME) {
748 e->match.type = DBM_EXACT;
749 e->match.str = argv[(*argi)++];
754 * Separate macro keys from search string.
755 * If needed, request regular expression handling.
758 if (search->argmode == ARG_WORD) {
760 e->match.type = DBM_REGEX;
762 mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
764 mandoc_asprintf(&val, "\\<%s\\>", argv[*argi]);
766 mandoc_asprintf(&val,
767 "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", argv[*argi]);
770 } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
771 e->bits = TYPE_Nm | TYPE_Nd;
772 e->match.type = DBM_REGEX;
776 if (val == argv[*argi])
777 e->bits = TYPE_Nm | TYPE_Nd;
779 e->match.type = DBM_SUB;
780 e->match.str = val + 1;
782 e->match.type = DBM_REGEX;
784 if (strstr(argv[*argi], "arch") != NULL)
788 /* Compile regular expressions. */
790 if (e->match.type == DBM_REGEX) {
791 e->match.re = mandoc_malloc(sizeof(*e->match.re));
792 irc = regcomp(e->match.re, val,
793 REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
795 regerror(irc, e->match.re, errbuf, sizeof(errbuf));
796 warnx("regcomp /%s/: %s", val, errbuf);
798 if (search->argmode == ARG_WORD)
814 * Parse out all possible fields.
815 * If the field doesn't resolve, bail.
818 while (NULL != (key = strsep(&argv[*argi], ","))) {
821 for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
822 if (0 == strcasecmp(key, mansearch_keynames[i])) {
828 if (strcasecmp(key, "any"))
829 warnx("treating unknown key "
830 "\"%s\" as \"any\"", key);
840 exprfree(struct expr *e)
844 if (e->child != NULL)