Annotation of src/usr.bin/mandoc/mansearch.c, Revision 1.66
1.66 ! schwarze 1: /* $OpenBSD: mansearch.c,v 1.65 2019/07/01 22:43:03 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.61 schwarze 4: * Copyright (c) 2013-2018 Ingo Schwarze <schwarze@openbsd.org>
1.1 schwarze 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: *
1.43 schwarze 10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
1.1 schwarze 11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
1.43 schwarze 12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
1.1 schwarze 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: */
1.33 schwarze 18:
1.19 schwarze 19: #include <sys/mman.h>
1.33 schwarze 20: #include <sys/types.h>
21:
1.1 schwarze 22: #include <assert.h>
1.46 schwarze 23: #include <err.h>
1.39 schwarze 24: #include <errno.h>
1.1 schwarze 25: #include <fcntl.h>
1.41 schwarze 26: #include <glob.h>
1.1 schwarze 27: #include <limits.h>
28: #include <regex.h>
29: #include <stdio.h>
30: #include <stdint.h>
31: #include <stddef.h>
32: #include <stdlib.h>
33: #include <string.h>
34: #include <unistd.h>
35:
1.14 schwarze 36: #include "mandoc_aux.h"
1.47 schwarze 37: #include "mandoc_ohash.h"
1.43 schwarze 38: #include "manconf.h"
1.1 schwarze 39: #include "mansearch.h"
1.51 schwarze 40: #include "dbm.h"
1.1 schwarze 41:
42: struct expr {
1.51 schwarze 43: /* Used for terms: */
44: struct dbm_match match; /* Match type and expression. */
45: uint64_t bits; /* Type mask. */
46: /* Used for OR and AND groups: */
47: struct expr *next; /* Next child in the parent group. */
48: struct expr *child; /* First child in this group. */
49: enum { EXPR_TERM, EXPR_OR, EXPR_AND } type;
1.1 schwarze 50: };
51:
1.51 schwarze 52: const char *const mansearch_keynames[KEY_MAX] = {
53: "arch", "sec", "Xr", "Ar", "Fa", "Fl", "Dv", "Fn",
54: "Ic", "Pa", "Cm", "Li", "Em", "Cd", "Va", "Ft",
55: "Tn", "Er", "Ev", "Sy", "Sh", "In", "Ss", "Ox",
56: "An", "Mt", "St", "Bx", "At", "Nx", "Fx", "Lk",
57: "Ms", "Bsx", "Dx", "Rs", "Vt", "Lb", "Nm", "Nd"
1.1 schwarze 58: };
59:
1.51 schwarze 60:
61: static struct ohash *manmerge(struct expr *, struct ohash *);
62: static struct ohash *manmerge_term(struct expr *, struct ohash *);
63: static struct ohash *manmerge_or(struct expr *, struct ohash *);
64: static struct ohash *manmerge_and(struct expr *, struct ohash *);
65: static char *buildnames(const struct dbm_page *);
1.53 schwarze 66: static char *buildoutput(size_t, struct dbm_page *);
67: static size_t lstlen(const char *, size_t);
68: static void lstcat(char *, size_t *, const char *, const char *);
1.51 schwarze 69: static int lstmatch(const char *, const char *);
1.24 schwarze 70: static struct expr *exprcomp(const struct mansearch *,
1.51 schwarze 71: int, char *[], int *);
72: static struct expr *expr_and(const struct mansearch *,
73: int, char *[], int *);
74: static struct expr *exprterm(const struct mansearch *,
75: int, char *[], int *);
1.1 schwarze 76: static void exprfree(struct expr *);
1.29 schwarze 77: static int manpage_compare(const void *, const void *);
1.19 schwarze 78:
1.24 schwarze 79:
1.19 schwarze 80: int
1.1 schwarze 81: mansearch(const struct mansearch *search,
1.3 schwarze 82: const struct manpaths *paths,
83: int argc, char *argv[],
1.1 schwarze 84: struct manpage **res, size_t *sz)
85: {
86: char buf[PATH_MAX];
1.51 schwarze 87: struct dbm_res *rp;
88: struct expr *e;
89: struct dbm_page *page;
1.1 schwarze 90: struct manpage *mpage;
1.51 schwarze 91: struct ohash *htab;
92: size_t cur, i, maxres, outkey;
93: unsigned int slot;
94: int argi, chdir_status, getcwd_status, im;
1.44 schwarze 95:
1.51 schwarze 96: argi = 0;
97: if ((e = exprcomp(search, argc, argv, &argi)) == NULL) {
1.44 schwarze 98: *sz = 0;
1.45 schwarze 99: return 0;
1.44 schwarze 100: }
1.1 schwarze 101:
1.44 schwarze 102: cur = maxres = 0;
1.57 schwarze 103: if (res != NULL)
104: *res = NULL;
1.1 schwarze 105:
1.51 schwarze 106: outkey = KEY_Nd;
107: if (search->outkey != NULL)
108: for (im = 0; im < KEY_MAX; im++)
1.33 schwarze 109: if (0 == strcasecmp(search->outkey,
1.51 schwarze 110: mansearch_keynames[im])) {
111: outkey = im;
1.3 schwarze 112: break;
113: }
114:
1.1 schwarze 115: /*
1.44 schwarze 116: * Remember the original working directory, if possible.
117: * This will be needed if the second or a later directory
118: * is given as a relative path.
119: * Do not error out if the current directory is not
120: * searchable: Maybe it won't be needed after all.
1.1 schwarze 121: */
122:
1.44 schwarze 123: if (getcwd(buf, PATH_MAX) == NULL) {
124: getcwd_status = 0;
125: (void)strlcpy(buf, strerror(errno), sizeof(buf));
126: } else
127: getcwd_status = 1;
1.1 schwarze 128:
129: /*
130: * Loop over the directories (containing databases) for us to
131: * search.
132: * Don't let missing/bad databases/directories phase us.
133: * In each, try to open the resident database and, if it opens,
134: * scan it for our match expression.
135: */
136:
1.44 schwarze 137: chdir_status = 0;
1.1 schwarze 138: for (i = 0; i < paths->sz; i++) {
1.44 schwarze 139: if (chdir_status && paths->paths[i][0] != '/') {
140: if ( ! getcwd_status) {
1.46 schwarze 141: warnx("%s: getcwd: %s", paths->paths[i], buf);
1.44 schwarze 142: continue;
143: } else if (chdir(buf) == -1) {
1.49 schwarze 144: warn("%s", buf);
1.44 schwarze 145: continue;
146: }
147: }
148: if (chdir(paths->paths[i]) == -1) {
1.49 schwarze 149: warn("%s", paths->paths[i]);
1.1 schwarze 150: continue;
1.24 schwarze 151: }
1.44 schwarze 152: chdir_status = 1;
1.1 schwarze 153:
1.51 schwarze 154: if (dbm_open(MANDOC_DB) == -1) {
1.56 schwarze 155: if (errno != ENOENT)
156: warn("%s/%s", paths->paths[i], MANDOC_DB);
1.1 schwarze 157: continue;
158: }
159:
1.51 schwarze 160: if ((htab = manmerge(e, NULL)) == NULL) {
161: dbm_close();
162: continue;
1.1 schwarze 163: }
164:
1.51 schwarze 165: for (rp = ohash_first(htab, &slot); rp != NULL;
166: rp = ohash_next(htab, &slot)) {
167: page = dbm_page_get(rp->page);
1.1 schwarze 168:
1.51 schwarze 169: if (lstmatch(search->sec, page->sect) == 0 ||
1.59 schwarze 170: lstmatch(search->arch, page->arch) == 0 ||
171: (search->argmode == ARG_NAME &&
172: rp->bits <= (int32_t)(NAME_SYN & NAME_MASK)))
1.1 schwarze 173: continue;
174:
1.57 schwarze 175: if (res == NULL) {
176: cur = 1;
177: break;
178: }
1.1 schwarze 179: if (cur + 1 > maxres) {
180: maxres += 1024;
1.26 schwarze 181: *res = mandoc_reallocarray(*res,
1.51 schwarze 182: maxres, sizeof(**res));
1.1 schwarze 183: }
184: mpage = *res + cur;
1.51 schwarze 185: mandoc_asprintf(&mpage->file, "%s/%s",
186: paths->paths[i], page->file + 1);
1.60 schwarze 187: if (access(chdir_status ? page->file + 1 :
188: mpage->file, R_OK) == -1) {
189: warn("%s", mpage->file);
190: warnx("outdated mandoc.db contains "
1.65 schwarze 191: "bogus %s entry, run makewhatis %s",
1.60 schwarze 192: page->file + 1, paths->paths[i]);
193: free(mpage->file);
194: free(rp);
195: continue;
196: }
1.51 schwarze 197: mpage->names = buildnames(page);
1.53 schwarze 198: mpage->output = buildoutput(outkey, page);
1.64 schwarze 199: mpage->bits = search->firstmatch ? rp->bits : 0;
1.34 schwarze 200: mpage->ipath = i;
1.51 schwarze 201: mpage->sec = *page->sect - '0';
202: if (mpage->sec < 0 || mpage->sec > 9)
203: mpage->sec = 10;
204: mpage->form = *page->file;
205: free(rp);
206: cur++;
207: }
208: ohash_delete(htab);
209: free(htab);
210: dbm_close();
1.36 schwarze 211:
212: /*
213: * In man(1) mode, prefer matches in earlier trees
214: * over matches in later trees.
215: */
216:
217: if (cur && search->firstmatch)
218: break;
1.1 schwarze 219: }
1.66 ! schwarze 220: if (res != NULL && cur > 1)
1.57 schwarze 221: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.44 schwarze 222: if (chdir_status && getcwd_status && chdir(buf) == -1)
1.49 schwarze 223: warn("%s", buf);
1.1 schwarze 224: exprfree(e);
225: *sz = cur;
1.57 schwarze 226: return res != NULL || cur;
1.2 schwarze 227: }
228:
1.51 schwarze 229: /*
230: * Merge the results for the expression tree rooted at e
231: * into the the result list htab.
232: */
233: static struct ohash *
234: manmerge(struct expr *e, struct ohash *htab)
1.33 schwarze 235: {
1.51 schwarze 236: switch (e->type) {
237: case EXPR_TERM:
238: return manmerge_term(e, htab);
239: case EXPR_OR:
240: return manmerge_or(e->child, htab);
241: case EXPR_AND:
242: return manmerge_and(e->child, htab);
243: default:
244: abort();
1.33 schwarze 245: }
246: }
247:
1.51 schwarze 248: static struct ohash *
249: manmerge_term(struct expr *e, struct ohash *htab)
1.29 schwarze 250: {
1.51 schwarze 251: struct dbm_res res, *rp;
252: uint64_t ib;
253: unsigned int slot;
254: int im;
255:
256: if (htab == NULL) {
257: htab = mandoc_malloc(sizeof(*htab));
258: mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
259: }
260:
261: for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
262: if ((e->bits & ib) == 0)
263: continue;
264:
265: switch (ib) {
266: case TYPE_arch:
267: dbm_page_byarch(&e->match);
268: break;
269: case TYPE_sec:
270: dbm_page_bysect(&e->match);
271: break;
272: case TYPE_Nm:
273: dbm_page_byname(&e->match);
274: break;
275: case TYPE_Nd:
276: dbm_page_bydesc(&e->match);
277: break;
278: default:
279: dbm_page_bymacro(im - 2, &e->match);
280: break;
281: }
282:
283: /*
284: * When hashing for deduplication, use the unique
285: * page ID itself instead of a hash function;
286: * that is quite efficient.
287: */
1.29 schwarze 288:
1.51 schwarze 289: for (;;) {
290: res = dbm_page_next();
291: if (res.page == -1)
292: break;
293: slot = ohash_lookup_memory(htab,
294: (char *)&res, sizeof(res.page), res.page);
1.64 schwarze 295: if ((rp = ohash_find(htab, slot)) != NULL) {
296: rp->bits |= res.bits;
1.51 schwarze 297: continue;
1.64 schwarze 298: }
1.51 schwarze 299: rp = mandoc_malloc(sizeof(*rp));
300: *rp = res;
301: ohash_insert(htab, slot, rp);
302: }
303: }
304: return htab;
1.29 schwarze 305: }
306:
1.51 schwarze 307: static struct ohash *
308: manmerge_or(struct expr *e, struct ohash *htab)
309: {
310: while (e != NULL) {
311: htab = manmerge(e, htab);
312: e = e->next;
313: }
314: return htab;
315: }
1.13 schwarze 316:
1.51 schwarze 317: static struct ohash *
318: manmerge_and(struct expr *e, struct ohash *htab)
319: {
320: struct ohash *hand, *h1, *h2;
321: struct dbm_res *res;
322: unsigned int slot1, slot2;
1.13 schwarze 323:
1.51 schwarze 324: /* Evaluate the first term of the AND clause. */
1.29 schwarze 325:
1.51 schwarze 326: hand = manmerge(e, NULL);
1.29 schwarze 327:
1.51 schwarze 328: while ((e = e->next) != NULL) {
1.13 schwarze 329:
1.51 schwarze 330: /* Evaluate the next term and prepare for ANDing. */
1.13 schwarze 331:
1.51 schwarze 332: h2 = manmerge(e, NULL);
333: if (ohash_entries(h2) < ohash_entries(hand)) {
334: h1 = h2;
335: h2 = hand;
336: } else
337: h1 = hand;
338: hand = mandoc_malloc(sizeof(*hand));
339: mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
1.13 schwarze 340:
1.51 schwarze 341: /* Keep all pages that are in both result sets. */
1.13 schwarze 342:
1.51 schwarze 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),
347: res->page)) == NULL)
348: free(res);
349: else
350: ohash_insert(hand, ohash_lookup_memory(hand,
351: (char *)res, sizeof(res->page),
352: res->page), res);
1.13 schwarze 353: }
354:
1.51 schwarze 355: /* Discard the merged results. */
356:
357: for (res = ohash_first(h2, &slot2); res != NULL;
358: res = ohash_next(h2, &slot2))
359: free(res);
360: ohash_delete(h2);
361: free(h2);
362: ohash_delete(h1);
363: free(h1);
364: }
1.13 schwarze 365:
1.51 schwarze 366: /* Merge the result of the AND into htab. */
1.8 schwarze 367:
1.51 schwarze 368: if (htab == NULL)
369: return hand;
1.8 schwarze 370:
1.51 schwarze 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);
377: else
378: free(res);
379: }
380:
381: /* Discard the merged result. */
1.8 schwarze 382:
1.51 schwarze 383: ohash_delete(hand);
384: free(hand);
385: return htab;
386: }
1.41 schwarze 387:
1.51 schwarze 388: void
389: mansearch_free(struct manpage *res, size_t sz)
390: {
391: size_t i;
1.41 schwarze 392:
1.51 schwarze 393: for (i = 0; i < sz; i++) {
394: free(res[i].file);
395: free(res[i].names);
396: free(res[i].output);
1.13 schwarze 397: }
1.51 schwarze 398: free(res);
399: }
400:
401: static int
402: manpage_compare(const void *vp1, const void *vp2)
403: {
404: const struct manpage *mp1, *mp2;
1.58 schwarze 405: const char *cp1, *cp2;
406: size_t sz1, sz2;
1.51 schwarze 407: int diff;
408:
409: mp1 = vp1;
410: mp2 = vp2;
1.64 schwarze 411: if ((diff = mp2->bits - mp1->bits) ||
412: (diff = mp1->sec - mp2->sec))
1.58 schwarze 413: return diff;
414:
415: /* Fall back to alphabetic ordering of names. */
416: sz1 = strcspn(mp1->names, "(");
417: sz2 = strcspn(mp2->names, "(");
418: if (sz1 < sz2)
419: sz1 = sz2;
420: if ((diff = strncasecmp(mp1->names, mp2->names, sz1)))
421: return diff;
422:
423: /* For identical names and sections, prefer arch-dependent. */
424: cp1 = strchr(mp1->names + sz1, '/');
425: cp2 = strchr(mp2->names + sz2, '/');
426: return cp1 != NULL && cp2 != NULL ? strcasecmp(cp1, cp2) :
427: cp1 != NULL ? -1 : cp2 != NULL ? 1 : 0;
1.3 schwarze 428: }
429:
430: static char *
1.51 schwarze 431: buildnames(const struct dbm_page *page)
1.3 schwarze 432: {
1.51 schwarze 433: char *buf;
434: size_t i, sz;
1.3 schwarze 435:
1.53 schwarze 436: sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
437: (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
1.51 schwarze 438: buf = mandoc_malloc(sz);
439: i = 0;
1.53 schwarze 440: lstcat(buf, &i, page->name, ", ");
1.51 schwarze 441: buf[i++] = '(';
1.53 schwarze 442: lstcat(buf, &i, page->sect, ", ");
1.51 schwarze 443: if (page->arch != NULL) {
444: buf[i++] = '/';
1.53 schwarze 445: lstcat(buf, &i, page->arch, ", ");
1.51 schwarze 446: }
447: buf[i++] = ')';
448: buf[i++] = '\0';
449: assert(i == sz);
450: return buf;
1.1 schwarze 451: }
452:
453: /*
1.51 schwarze 454: * Count the buffer space needed to print the NUL-terminated
1.53 schwarze 455: * list of NUL-terminated strings, when printing sep separator
1.51 schwarze 456: * characters between strings.
1.1 schwarze 457: */
1.51 schwarze 458: static size_t
1.53 schwarze 459: lstlen(const char *cp, size_t sep)
1.1 schwarze 460: {
1.51 schwarze 461: size_t sz;
1.1 schwarze 462:
1.59 schwarze 463: for (sz = 0; *cp != '\0'; cp++) {
464:
465: /* Skip names appearing only in the SYNOPSIS. */
466: if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
467: while (*cp != '\0')
468: cp++;
469: continue;
470: }
471:
472: /* Skip name class markers. */
473: if (*cp < ' ')
474: cp++;
475:
476: /* Print a separator before each but the first string. */
477: if (sz)
478: sz += sep;
479:
480: /* Copy one string. */
481: while (*cp != '\0') {
482: sz++;
483: cp++;
484: }
1.51 schwarze 485: }
486: return sz;
1.1 schwarze 487: }
488:
489: /*
1.51 schwarze 490: * Print the NUL-terminated list of NUL-terminated strings
1.53 schwarze 491: * into the buffer, seperating strings with sep.
1.1 schwarze 492: */
493: static void
1.53 schwarze 494: lstcat(char *buf, size_t *i, const char *cp, const char *sep)
1.1 schwarze 495: {
1.59 schwarze 496: const char *s;
497: size_t i_start;
1.53 schwarze 498:
1.59 schwarze 499: for (i_start = *i; *cp != '\0'; cp++) {
500:
501: /* Skip names appearing only in the SYNOPSIS. */
502: if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
503: while (*cp != '\0')
504: cp++;
505: continue;
506: }
507:
508: /* Skip name class markers. */
509: if (*cp < ' ')
510: cp++;
511:
512: /* Print a separator before each but the first string. */
513: if (*i > i_start) {
1.53 schwarze 514: s = sep;
515: while (*s != '\0')
516: buf[(*i)++] = *s++;
1.59 schwarze 517: }
518:
519: /* Copy one string. */
520: while (*cp != '\0')
521: buf[(*i)++] = *cp++;
1.51 schwarze 522: }
1.59 schwarze 523:
1.1 schwarze 524: }
525:
1.51 schwarze 526: /*
527: * Return 1 if the string *want occurs in any of the strings
528: * in the NUL-terminated string list *have, or 0 otherwise.
529: * If either argument is NULL or empty, assume no filtering
530: * is desired and return 1.
531: */
532: static int
533: lstmatch(const char *want, const char *have)
1.4 schwarze 534: {
1.51 schwarze 535: if (want == NULL || have == NULL || *have == '\0')
536: return 1;
537: while (*have != '\0') {
538: if (strcasestr(have, want) != NULL)
539: return 1;
540: have = strchr(have, '\0') + 1;
541: }
542: return 0;
1.4 schwarze 543: }
544:
1.1 schwarze 545: /*
1.53 schwarze 546: * Build a list of values taken by the macro im in the manual page.
1.1 schwarze 547: */
548: static char *
1.53 schwarze 549: buildoutput(size_t im, struct dbm_page *page)
1.1 schwarze 550: {
1.53 schwarze 551: const char *oldoutput, *sep, *input;
1.51 schwarze 552: char *output, *newoutput, *value;
1.53 schwarze 553: size_t sz, i;
554:
555: switch (im) {
556: case KEY_Nd:
557: return mandoc_strdup(page->desc);
558: case KEY_Nm:
559: input = page->name;
560: break;
561: case KEY_sec:
562: input = page->sect;
563: break;
564: case KEY_arch:
565: input = page->arch;
566: if (input == NULL)
567: input = "all\0";
568: break;
569: default:
570: input = NULL;
571: break;
572: }
573:
574: if (input != NULL) {
575: sz = lstlen(input, 3) + 1;
576: output = mandoc_malloc(sz);
577: i = 0;
578: lstcat(output, &i, input, " # ");
1.54 schwarze 579: output[i++] = '\0';
580: assert(i == sz);
1.53 schwarze 581: return output;
582: }
1.51 schwarze 583:
584: output = NULL;
1.53 schwarze 585: dbm_macro_bypage(im - 2, page->addr);
1.51 schwarze 586: while ((value = dbm_macro_next()) != NULL) {
587: if (output == NULL) {
588: oldoutput = "";
589: sep = "";
590: } else {
591: oldoutput = output;
592: sep = " # ";
593: }
594: mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
595: free(output);
596: output = newoutput;
1.1 schwarze 597: }
1.51 schwarze 598: return output;
1.1 schwarze 599: }
600:
601: /*
602: * Compile a set of string tokens into an expression.
603: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
604: * "(", "foo=bar", etc.).
605: */
606: static struct expr *
1.51 schwarze 607: exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 schwarze 608: {
1.51 schwarze 609: struct expr *parent, *child;
610: int needterm, nested;
611:
612: if ((nested = *argi) == argc)
613: return NULL;
614: needterm = 1;
615: parent = child = NULL;
616: while (*argi < argc) {
617: if (strcmp(")", argv[*argi]) == 0) {
618: if (needterm)
619: warnx("missing term "
620: "before closing parenthesis");
621: needterm = 0;
622: if (nested)
623: break;
624: warnx("ignoring unmatched right parenthesis");
625: ++*argi;
1.4 schwarze 626: continue;
1.51 schwarze 627: }
628: if (strcmp("-o", argv[*argi]) == 0) {
629: if (needterm) {
630: if (*argi > 0)
631: warnx("ignoring -o after %s",
632: argv[*argi - 1]);
633: else
634: warnx("ignoring initial -o");
635: }
636: needterm = 1;
637: ++*argi;
1.4 schwarze 638: continue;
1.51 schwarze 639: }
640: needterm = 0;
641: if (child == NULL) {
642: child = expr_and(search, argc, argv, argi);
1.4 schwarze 643: continue;
1.1 schwarze 644: }
1.51 schwarze 645: if (parent == NULL) {
646: parent = mandoc_calloc(1, sizeof(*parent));
647: parent->type = EXPR_OR;
648: parent->next = NULL;
649: parent->child = child;
650: }
651: child->next = expr_and(search, argc, argv, argi);
652: child = child->next;
653: }
654: if (needterm && *argi)
655: warnx("ignoring trailing %s", argv[*argi - 1]);
656: return parent == NULL ? child : parent;
657: }
1.17 schwarze 658:
1.51 schwarze 659: static struct expr *
660: expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
661: {
662: struct expr *parent, *child;
663: int needterm;
1.17 schwarze 664:
1.51 schwarze 665: needterm = 1;
666: parent = child = NULL;
667: while (*argi < argc) {
668: if (strcmp(")", argv[*argi]) == 0) {
669: if (needterm)
670: warnx("missing term "
671: "before closing parenthesis");
672: needterm = 0;
673: break;
674: }
675: if (strcmp("-o", argv[*argi]) == 0)
676: break;
677: if (strcmp("-a", argv[*argi]) == 0) {
678: if (needterm) {
679: if (*argi > 0)
680: warnx("ignoring -a after %s",
681: argv[*argi - 1]);
682: else
683: warnx("ignoring initial -a");
1.18 schwarze 684: }
1.51 schwarze 685: needterm = 1;
686: ++*argi;
687: continue;
1.18 schwarze 688: }
1.51 schwarze 689: if (needterm == 0)
690: break;
691: if (child == NULL) {
692: child = exprterm(search, argc, argv, argi);
693: if (child != NULL)
694: needterm = 0;
695: continue;
696: }
697: needterm = 0;
698: if (parent == NULL) {
699: parent = mandoc_calloc(1, sizeof(*parent));
700: parent->type = EXPR_AND;
701: parent->next = NULL;
702: parent->child = child;
703: }
704: child->next = exprterm(search, argc, argv, argi);
705: if (child->next != NULL) {
706: child = child->next;
707: needterm = 0;
708: }
709: }
710: if (needterm && *argi)
711: warnx("ignoring trailing %s", argv[*argi - 1]);
712: return parent == NULL ? child : parent;
1.6 schwarze 713: }
714:
715: static struct expr *
1.51 schwarze 716: exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 schwarze 717: {
1.6 schwarze 718: char errbuf[BUFSIZ];
1.1 schwarze 719: struct expr *e;
1.28 schwarze 720: char *key, *val;
1.11 schwarze 721: uint64_t iterbit;
1.51 schwarze 722: int cs, i, irc;
1.1 schwarze 723:
1.51 schwarze 724: if (strcmp("(", argv[*argi]) == 0) {
725: ++*argi;
726: e = exprcomp(search, argc, argv, argi);
727: if (*argi < argc) {
728: assert(strcmp(")", argv[*argi]) == 0);
729: ++*argi;
730: } else
731: warnx("unclosed parenthesis");
732: return e;
733: }
1.1 schwarze 734:
1.55 schwarze 735: if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
736: cs = 0;
737: ++*argi;
738: } else
739: cs = 1;
740:
1.51 schwarze 741: e = mandoc_calloc(1, sizeof(*e));
742: e->type = EXPR_TERM;
743: e->bits = 0;
744: e->next = NULL;
745: e->child = NULL;
1.1 schwarze 746:
1.33 schwarze 747: if (search->argmode == ARG_NAME) {
748: e->bits = TYPE_Nm;
1.51 schwarze 749: e->match.type = DBM_EXACT;
750: e->match.str = argv[(*argi)++];
1.45 schwarze 751: return e;
1.1 schwarze 752: }
753:
754: /*
1.33 schwarze 755: * Separate macro keys from search string.
1.51 schwarze 756: * If needed, request regular expression handling.
1.1 schwarze 757: */
758:
1.33 schwarze 759: if (search->argmode == ARG_WORD) {
760: e->bits = TYPE_Nm;
1.51 schwarze 761: e->match.type = DBM_REGEX;
762: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
1.33 schwarze 763: cs = 0;
1.51 schwarze 764: } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
1.33 schwarze 765: e->bits = TYPE_Nm | TYPE_Nd;
1.61 schwarze 766: e->match.type = DBM_REGEX;
767: val = argv[*argi];
768: cs = 0;
1.28 schwarze 769: } else {
1.51 schwarze 770: if (val == argv[*argi])
1.33 schwarze 771: e->bits = TYPE_Nm | TYPE_Nd;
1.51 schwarze 772: if (*val == '=') {
773: e->match.type = DBM_SUB;
774: e->match.str = val + 1;
775: } else
776: e->match.type = DBM_REGEX;
1.28 schwarze 777: *val++ = '\0';
1.51 schwarze 778: if (strstr(argv[*argi], "arch") != NULL)
1.12 schwarze 779: cs = 0;
1.28 schwarze 780: }
781:
782: /* Compile regular expressions. */
783:
1.51 schwarze 784: if (e->match.type == DBM_REGEX) {
785: e->match.re = mandoc_malloc(sizeof(*e->match.re));
786: irc = regcomp(e->match.re, val,
1.28 schwarze 787: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.51 schwarze 788: if (irc) {
789: regerror(irc, e->match.re, errbuf, sizeof(errbuf));
790: warnx("regcomp /%s/: %s", val, errbuf);
791: }
1.33 schwarze 792: if (search->argmode == ARG_WORD)
1.28 schwarze 793: free(val);
794: if (irc) {
1.51 schwarze 795: free(e->match.re);
1.1 schwarze 796: free(e);
1.51 schwarze 797: ++*argi;
1.45 schwarze 798: return NULL;
1.1 schwarze 799: }
1.28 schwarze 800: }
801:
1.51 schwarze 802: if (e->bits) {
803: ++*argi;
1.45 schwarze 804: return e;
1.51 schwarze 805: }
1.1 schwarze 806:
807: /*
808: * Parse out all possible fields.
809: * If the field doesn't resolve, bail.
810: */
811:
1.51 schwarze 812: while (NULL != (key = strsep(&argv[*argi], ","))) {
1.1 schwarze 813: if ('\0' == *key)
814: continue;
1.51 schwarze 815: for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
816: if (0 == strcasecmp(key, mansearch_keynames[i])) {
1.11 schwarze 817: e->bits |= iterbit;
818: break;
819: }
820: }
1.51 schwarze 821: if (i == KEY_MAX) {
822: if (strcasecmp(key, "any"))
823: warnx("treating unknown key "
824: "\"%s\" as \"any\"", key);
1.11 schwarze 825: e->bits |= ~0ULL;
1.1 schwarze 826: }
827: }
828:
1.51 schwarze 829: ++*argi;
1.45 schwarze 830: return e;
1.1 schwarze 831: }
832:
833: static void
1.51 schwarze 834: exprfree(struct expr *e)
1.1 schwarze 835: {
1.51 schwarze 836: if (e->next != NULL)
837: exprfree(e->next);
838: if (e->child != NULL)
839: exprfree(e->child);
840: free(e);
1.1 schwarze 841: }