Annotation of src/usr.bin/mandoc/mansearch.c, Revision 1.57
1.57 ! schwarze 1: /* $OpenBSD: mansearch.c,v 1.56 2017/05/17 21:18:41 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.53 schwarze 4: * Copyright (c) 2013-2017 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:
36: #include "mandoc.h"
1.14 schwarze 37: #include "mandoc_aux.h"
1.47 schwarze 38: #include "mandoc_ohash.h"
1.43 schwarze 39: #include "manconf.h"
1.1 schwarze 40: #include "mansearch.h"
1.51 schwarze 41: #include "dbm.h"
1.1 schwarze 42:
43: struct expr {
1.51 schwarze 44: /* Used for terms: */
45: struct dbm_match match; /* Match type and expression. */
46: uint64_t bits; /* Type mask. */
47: /* Used for OR and AND groups: */
48: struct expr *next; /* Next child in the parent group. */
49: struct expr *child; /* First child in this group. */
50: enum { EXPR_TERM, EXPR_OR, EXPR_AND } type;
1.1 schwarze 51: };
52:
1.51 schwarze 53: const char *const mansearch_keynames[KEY_MAX] = {
54: "arch", "sec", "Xr", "Ar", "Fa", "Fl", "Dv", "Fn",
55: "Ic", "Pa", "Cm", "Li", "Em", "Cd", "Va", "Ft",
56: "Tn", "Er", "Ev", "Sy", "Sh", "In", "Ss", "Ox",
57: "An", "Mt", "St", "Bx", "At", "Nx", "Fx", "Lk",
58: "Ms", "Bsx", "Dx", "Rs", "Vt", "Lb", "Nm", "Nd"
1.1 schwarze 59: };
60:
1.51 schwarze 61:
62: static struct ohash *manmerge(struct expr *, struct ohash *);
63: static struct ohash *manmerge_term(struct expr *, struct ohash *);
64: static struct ohash *manmerge_or(struct expr *, struct ohash *);
65: static struct ohash *manmerge_and(struct expr *, struct ohash *);
66: static char *buildnames(const struct dbm_page *);
1.53 schwarze 67: static char *buildoutput(size_t, struct dbm_page *);
68: static size_t lstlen(const char *, size_t);
69: static void lstcat(char *, size_t *, const char *, const char *);
1.51 schwarze 70: static int lstmatch(const char *, const char *);
1.24 schwarze 71: static struct expr *exprcomp(const struct mansearch *,
1.51 schwarze 72: int, char *[], int *);
73: static struct expr *expr_and(const struct mansearch *,
74: int, char *[], int *);
75: static struct expr *exprterm(const struct mansearch *,
76: int, char *[], int *);
1.1 schwarze 77: static void exprfree(struct expr *);
1.29 schwarze 78: static int manpage_compare(const void *, const void *);
1.19 schwarze 79:
1.24 schwarze 80:
1.19 schwarze 81: int
1.1 schwarze 82: mansearch(const struct mansearch *search,
1.3 schwarze 83: const struct manpaths *paths,
84: int argc, char *argv[],
1.1 schwarze 85: struct manpage **res, size_t *sz)
86: {
87: char buf[PATH_MAX];
1.51 schwarze 88: struct dbm_res *rp;
89: struct expr *e;
90: struct dbm_page *page;
1.1 schwarze 91: struct manpage *mpage;
1.51 schwarze 92: struct ohash *htab;
93: size_t cur, i, maxres, outkey;
94: unsigned int slot;
95: int argi, chdir_status, getcwd_status, im;
1.44 schwarze 96:
1.51 schwarze 97: argi = 0;
98: if ((e = exprcomp(search, argc, argv, &argi)) == NULL) {
1.44 schwarze 99: *sz = 0;
1.45 schwarze 100: return 0;
1.44 schwarze 101: }
1.1 schwarze 102:
1.44 schwarze 103: cur = maxres = 0;
1.57 ! schwarze 104: if (res != NULL)
! 105: *res = NULL;
1.1 schwarze 106:
1.51 schwarze 107: outkey = KEY_Nd;
108: if (search->outkey != NULL)
109: for (im = 0; im < KEY_MAX; im++)
1.33 schwarze 110: if (0 == strcasecmp(search->outkey,
1.51 schwarze 111: mansearch_keynames[im])) {
112: outkey = im;
1.3 schwarze 113: break;
114: }
115:
1.1 schwarze 116: /*
1.44 schwarze 117: * Remember the original working directory, if possible.
118: * This will be needed if the second or a later directory
119: * is given as a relative path.
120: * Do not error out if the current directory is not
121: * searchable: Maybe it won't be needed after all.
1.1 schwarze 122: */
123:
1.44 schwarze 124: if (getcwd(buf, PATH_MAX) == NULL) {
125: getcwd_status = 0;
126: (void)strlcpy(buf, strerror(errno), sizeof(buf));
127: } else
128: getcwd_status = 1;
1.1 schwarze 129:
130: /*
131: * Loop over the directories (containing databases) for us to
132: * search.
133: * Don't let missing/bad databases/directories phase us.
134: * In each, try to open the resident database and, if it opens,
135: * scan it for our match expression.
136: */
137:
1.44 schwarze 138: chdir_status = 0;
1.1 schwarze 139: for (i = 0; i < paths->sz; i++) {
1.44 schwarze 140: if (chdir_status && paths->paths[i][0] != '/') {
141: if ( ! getcwd_status) {
1.46 schwarze 142: warnx("%s: getcwd: %s", paths->paths[i], buf);
1.44 schwarze 143: continue;
144: } else if (chdir(buf) == -1) {
1.49 schwarze 145: warn("%s", buf);
1.44 schwarze 146: continue;
147: }
148: }
149: if (chdir(paths->paths[i]) == -1) {
1.49 schwarze 150: warn("%s", paths->paths[i]);
1.1 schwarze 151: continue;
1.24 schwarze 152: }
1.44 schwarze 153: chdir_status = 1;
1.1 schwarze 154:
1.51 schwarze 155: if (dbm_open(MANDOC_DB) == -1) {
1.56 schwarze 156: if (errno != ENOENT)
157: warn("%s/%s", paths->paths[i], MANDOC_DB);
1.1 schwarze 158: continue;
159: }
160:
1.51 schwarze 161: if ((htab = manmerge(e, NULL)) == NULL) {
162: dbm_close();
163: continue;
1.1 schwarze 164: }
165:
1.51 schwarze 166: for (rp = ohash_first(htab, &slot); rp != NULL;
167: rp = ohash_next(htab, &slot)) {
168: page = dbm_page_get(rp->page);
1.1 schwarze 169:
1.51 schwarze 170: if (lstmatch(search->sec, page->sect) == 0 ||
171: lstmatch(search->arch, page->arch) == 0)
1.1 schwarze 172: continue;
173:
1.57 ! schwarze 174: if (res == NULL) {
! 175: cur = 1;
! 176: break;
! 177: }
1.1 schwarze 178: if (cur + 1 > maxres) {
179: maxres += 1024;
1.26 schwarze 180: *res = mandoc_reallocarray(*res,
1.51 schwarze 181: maxres, sizeof(**res));
1.1 schwarze 182: }
183: mpage = *res + cur;
1.51 schwarze 184: mandoc_asprintf(&mpage->file, "%s/%s",
185: paths->paths[i], page->file + 1);
186: mpage->names = buildnames(page);
1.53 schwarze 187: mpage->output = buildoutput(outkey, page);
1.34 schwarze 188: mpage->ipath = i;
1.51 schwarze 189: mpage->bits = rp->bits;
190: mpage->sec = *page->sect - '0';
191: if (mpage->sec < 0 || mpage->sec > 9)
192: mpage->sec = 10;
193: mpage->form = *page->file;
194: free(rp);
195: cur++;
196: }
197: ohash_delete(htab);
198: free(htab);
199: dbm_close();
1.36 schwarze 200:
201: /*
202: * In man(1) mode, prefer matches in earlier trees
203: * over matches in later trees.
204: */
205:
206: if (cur && search->firstmatch)
207: break;
1.1 schwarze 208: }
1.57 ! schwarze 209: if (res != NULL)
! 210: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.44 schwarze 211: if (chdir_status && getcwd_status && chdir(buf) == -1)
1.49 schwarze 212: warn("%s", buf);
1.1 schwarze 213: exprfree(e);
214: *sz = cur;
1.57 ! schwarze 215: return res != NULL || cur;
1.2 schwarze 216: }
217:
1.51 schwarze 218: /*
219: * Merge the results for the expression tree rooted at e
220: * into the the result list htab.
221: */
222: static struct ohash *
223: manmerge(struct expr *e, struct ohash *htab)
1.33 schwarze 224: {
1.51 schwarze 225: switch (e->type) {
226: case EXPR_TERM:
227: return manmerge_term(e, htab);
228: case EXPR_OR:
229: return manmerge_or(e->child, htab);
230: case EXPR_AND:
231: return manmerge_and(e->child, htab);
232: default:
233: abort();
1.33 schwarze 234: }
235: }
236:
1.51 schwarze 237: static struct ohash *
238: manmerge_term(struct expr *e, struct ohash *htab)
1.29 schwarze 239: {
1.51 schwarze 240: struct dbm_res res, *rp;
241: uint64_t ib;
242: unsigned int slot;
243: int im;
244:
245: if (htab == NULL) {
246: htab = mandoc_malloc(sizeof(*htab));
247: mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
248: }
249:
250: for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
251: if ((e->bits & ib) == 0)
252: continue;
253:
254: switch (ib) {
255: case TYPE_arch:
256: dbm_page_byarch(&e->match);
257: break;
258: case TYPE_sec:
259: dbm_page_bysect(&e->match);
260: break;
261: case TYPE_Nm:
262: dbm_page_byname(&e->match);
263: break;
264: case TYPE_Nd:
265: dbm_page_bydesc(&e->match);
266: break;
267: default:
268: dbm_page_bymacro(im - 2, &e->match);
269: break;
270: }
271:
272: /*
273: * When hashing for deduplication, use the unique
274: * page ID itself instead of a hash function;
275: * that is quite efficient.
276: */
1.29 schwarze 277:
1.51 schwarze 278: for (;;) {
279: res = dbm_page_next();
280: if (res.page == -1)
281: break;
282: slot = ohash_lookup_memory(htab,
283: (char *)&res, sizeof(res.page), res.page);
284: if ((rp = ohash_find(htab, slot)) != NULL) {
285: rp->bits |= res.bits;
286: continue;
287: }
288: rp = mandoc_malloc(sizeof(*rp));
289: *rp = res;
290: ohash_insert(htab, slot, rp);
291: }
292: }
293: return htab;
1.29 schwarze 294: }
295:
1.51 schwarze 296: static struct ohash *
297: manmerge_or(struct expr *e, struct ohash *htab)
298: {
299: while (e != NULL) {
300: htab = manmerge(e, htab);
301: e = e->next;
302: }
303: return htab;
304: }
1.13 schwarze 305:
1.51 schwarze 306: static struct ohash *
307: manmerge_and(struct expr *e, struct ohash *htab)
308: {
309: struct ohash *hand, *h1, *h2;
310: struct dbm_res *res;
311: unsigned int slot1, slot2;
1.13 schwarze 312:
1.51 schwarze 313: /* Evaluate the first term of the AND clause. */
1.29 schwarze 314:
1.51 schwarze 315: hand = manmerge(e, NULL);
1.29 schwarze 316:
1.51 schwarze 317: while ((e = e->next) != NULL) {
1.13 schwarze 318:
1.51 schwarze 319: /* Evaluate the next term and prepare for ANDing. */
1.13 schwarze 320:
1.51 schwarze 321: h2 = manmerge(e, NULL);
322: if (ohash_entries(h2) < ohash_entries(hand)) {
323: h1 = h2;
324: h2 = hand;
325: } else
326: h1 = hand;
327: hand = mandoc_malloc(sizeof(*hand));
328: mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
1.13 schwarze 329:
1.51 schwarze 330: /* Keep all pages that are in both result sets. */
1.13 schwarze 331:
1.51 schwarze 332: for (res = ohash_first(h1, &slot1); res != NULL;
333: res = ohash_next(h1, &slot1)) {
334: if (ohash_find(h2, ohash_lookup_memory(h2,
335: (char *)res, sizeof(res->page),
336: res->page)) == NULL)
337: free(res);
338: else
339: ohash_insert(hand, ohash_lookup_memory(hand,
340: (char *)res, sizeof(res->page),
341: res->page), res);
1.13 schwarze 342: }
343:
1.51 schwarze 344: /* Discard the merged results. */
345:
346: for (res = ohash_first(h2, &slot2); res != NULL;
347: res = ohash_next(h2, &slot2))
348: free(res);
349: ohash_delete(h2);
350: free(h2);
351: ohash_delete(h1);
352: free(h1);
353: }
1.13 schwarze 354:
1.51 schwarze 355: /* Merge the result of the AND into htab. */
1.8 schwarze 356:
1.51 schwarze 357: if (htab == NULL)
358: return hand;
1.8 schwarze 359:
1.51 schwarze 360: for (res = ohash_first(hand, &slot1); res != NULL;
361: res = ohash_next(hand, &slot1)) {
362: slot2 = ohash_lookup_memory(htab,
363: (char *)res, sizeof(res->page), res->page);
364: if (ohash_find(htab, slot2) == NULL)
365: ohash_insert(htab, slot2, res);
366: else
367: free(res);
368: }
369:
370: /* Discard the merged result. */
1.8 schwarze 371:
1.51 schwarze 372: ohash_delete(hand);
373: free(hand);
374: return htab;
375: }
1.41 schwarze 376:
1.51 schwarze 377: void
378: mansearch_free(struct manpage *res, size_t sz)
379: {
380: size_t i;
1.41 schwarze 381:
1.51 schwarze 382: for (i = 0; i < sz; i++) {
383: free(res[i].file);
384: free(res[i].names);
385: free(res[i].output);
1.13 schwarze 386: }
1.51 schwarze 387: free(res);
388: }
389:
390: static int
391: manpage_compare(const void *vp1, const void *vp2)
392: {
393: const struct manpage *mp1, *mp2;
394: int diff;
395:
396: mp1 = vp1;
397: mp2 = vp2;
398: return (diff = mp2->bits - mp1->bits) ? diff :
399: (diff = mp1->sec - mp2->sec) ? diff :
400: strcasecmp(mp1->names, mp2->names);
1.3 schwarze 401: }
402:
403: static char *
1.51 schwarze 404: buildnames(const struct dbm_page *page)
1.3 schwarze 405: {
1.51 schwarze 406: char *buf;
407: size_t i, sz;
1.3 schwarze 408:
1.53 schwarze 409: sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
410: (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
1.51 schwarze 411: buf = mandoc_malloc(sz);
412: i = 0;
1.53 schwarze 413: lstcat(buf, &i, page->name, ", ");
1.51 schwarze 414: buf[i++] = '(';
1.53 schwarze 415: lstcat(buf, &i, page->sect, ", ");
1.51 schwarze 416: if (page->arch != NULL) {
417: buf[i++] = '/';
1.53 schwarze 418: lstcat(buf, &i, page->arch, ", ");
1.51 schwarze 419: }
420: buf[i++] = ')';
421: buf[i++] = '\0';
422: assert(i == sz);
423: return buf;
1.1 schwarze 424: }
425:
426: /*
1.51 schwarze 427: * Count the buffer space needed to print the NUL-terminated
1.53 schwarze 428: * list of NUL-terminated strings, when printing sep separator
1.51 schwarze 429: * characters between strings.
1.1 schwarze 430: */
1.51 schwarze 431: static size_t
1.53 schwarze 432: lstlen(const char *cp, size_t sep)
1.1 schwarze 433: {
1.51 schwarze 434: size_t sz;
1.1 schwarze 435:
1.51 schwarze 436: for (sz = 0;; sz++) {
437: if (cp[0] == '\0') {
438: if (cp[1] == '\0')
439: break;
1.53 schwarze 440: sz += sep - 1;
1.51 schwarze 441: } else if (cp[0] < ' ')
442: sz--;
443: cp++;
444: }
445: return sz;
1.1 schwarze 446: }
447:
448: /*
1.51 schwarze 449: * Print the NUL-terminated list of NUL-terminated strings
1.53 schwarze 450: * into the buffer, seperating strings with sep.
1.1 schwarze 451: */
452: static void
1.53 schwarze 453: lstcat(char *buf, size_t *i, const char *cp, const char *sep)
1.1 schwarze 454: {
1.53 schwarze 455: const char *s;
456:
1.51 schwarze 457: for (;;) {
458: if (cp[0] == '\0') {
459: if (cp[1] == '\0')
460: break;
1.53 schwarze 461: s = sep;
462: while (*s != '\0')
463: buf[(*i)++] = *s++;
1.51 schwarze 464: } else if (cp[0] >= ' ')
465: buf[(*i)++] = cp[0];
466: cp++;
467: }
1.1 schwarze 468: }
469:
1.51 schwarze 470: /*
471: * Return 1 if the string *want occurs in any of the strings
472: * in the NUL-terminated string list *have, or 0 otherwise.
473: * If either argument is NULL or empty, assume no filtering
474: * is desired and return 1.
475: */
476: static int
477: lstmatch(const char *want, const char *have)
1.4 schwarze 478: {
1.51 schwarze 479: if (want == NULL || have == NULL || *have == '\0')
480: return 1;
481: while (*have != '\0') {
482: if (strcasestr(have, want) != NULL)
483: return 1;
484: have = strchr(have, '\0') + 1;
485: }
486: return 0;
1.4 schwarze 487: }
488:
1.1 schwarze 489: /*
1.53 schwarze 490: * Build a list of values taken by the macro im in the manual page.
1.1 schwarze 491: */
492: static char *
1.53 schwarze 493: buildoutput(size_t im, struct dbm_page *page)
1.1 schwarze 494: {
1.53 schwarze 495: const char *oldoutput, *sep, *input;
1.51 schwarze 496: char *output, *newoutput, *value;
1.53 schwarze 497: size_t sz, i;
498:
499: switch (im) {
500: case KEY_Nd:
501: return mandoc_strdup(page->desc);
502: case KEY_Nm:
503: input = page->name;
504: break;
505: case KEY_sec:
506: input = page->sect;
507: break;
508: case KEY_arch:
509: input = page->arch;
510: if (input == NULL)
511: input = "all\0";
512: break;
513: default:
514: input = NULL;
515: break;
516: }
517:
518: if (input != NULL) {
519: sz = lstlen(input, 3) + 1;
520: output = mandoc_malloc(sz);
521: i = 0;
522: lstcat(output, &i, input, " # ");
1.54 schwarze 523: output[i++] = '\0';
524: assert(i == sz);
1.53 schwarze 525: return output;
526: }
1.51 schwarze 527:
528: output = NULL;
1.53 schwarze 529: dbm_macro_bypage(im - 2, page->addr);
1.51 schwarze 530: while ((value = dbm_macro_next()) != NULL) {
531: if (output == NULL) {
532: oldoutput = "";
533: sep = "";
534: } else {
535: oldoutput = output;
536: sep = " # ";
537: }
538: mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
539: free(output);
540: output = newoutput;
1.1 schwarze 541: }
1.51 schwarze 542: return output;
1.1 schwarze 543: }
544:
545: /*
546: * Compile a set of string tokens into an expression.
547: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
548: * "(", "foo=bar", etc.).
549: */
550: static struct expr *
1.51 schwarze 551: exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 schwarze 552: {
1.51 schwarze 553: struct expr *parent, *child;
554: int needterm, nested;
555:
556: if ((nested = *argi) == argc)
557: return NULL;
558: needterm = 1;
559: parent = child = NULL;
560: while (*argi < argc) {
561: if (strcmp(")", argv[*argi]) == 0) {
562: if (needterm)
563: warnx("missing term "
564: "before closing parenthesis");
565: needterm = 0;
566: if (nested)
567: break;
568: warnx("ignoring unmatched right parenthesis");
569: ++*argi;
1.4 schwarze 570: continue;
1.51 schwarze 571: }
572: if (strcmp("-o", argv[*argi]) == 0) {
573: if (needterm) {
574: if (*argi > 0)
575: warnx("ignoring -o after %s",
576: argv[*argi - 1]);
577: else
578: warnx("ignoring initial -o");
579: }
580: needterm = 1;
581: ++*argi;
1.4 schwarze 582: continue;
1.51 schwarze 583: }
584: needterm = 0;
585: if (child == NULL) {
586: child = expr_and(search, argc, argv, argi);
1.4 schwarze 587: continue;
1.1 schwarze 588: }
1.51 schwarze 589: if (parent == NULL) {
590: parent = mandoc_calloc(1, sizeof(*parent));
591: parent->type = EXPR_OR;
592: parent->next = NULL;
593: parent->child = child;
594: }
595: child->next = expr_and(search, argc, argv, argi);
596: child = child->next;
597: }
598: if (needterm && *argi)
599: warnx("ignoring trailing %s", argv[*argi - 1]);
600: return parent == NULL ? child : parent;
601: }
1.17 schwarze 602:
1.51 schwarze 603: static struct expr *
604: expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
605: {
606: struct expr *parent, *child;
607: int needterm;
1.17 schwarze 608:
1.51 schwarze 609: needterm = 1;
610: parent = child = NULL;
611: while (*argi < argc) {
612: if (strcmp(")", argv[*argi]) == 0) {
613: if (needterm)
614: warnx("missing term "
615: "before closing parenthesis");
616: needterm = 0;
617: break;
618: }
619: if (strcmp("-o", argv[*argi]) == 0)
620: break;
621: if (strcmp("-a", argv[*argi]) == 0) {
622: if (needterm) {
623: if (*argi > 0)
624: warnx("ignoring -a after %s",
625: argv[*argi - 1]);
626: else
627: warnx("ignoring initial -a");
1.18 schwarze 628: }
1.51 schwarze 629: needterm = 1;
630: ++*argi;
631: continue;
1.18 schwarze 632: }
1.51 schwarze 633: if (needterm == 0)
634: break;
635: if (child == NULL) {
636: child = exprterm(search, argc, argv, argi);
637: if (child != NULL)
638: needterm = 0;
639: continue;
640: }
641: needterm = 0;
642: if (parent == NULL) {
643: parent = mandoc_calloc(1, sizeof(*parent));
644: parent->type = EXPR_AND;
645: parent->next = NULL;
646: parent->child = child;
647: }
648: child->next = exprterm(search, argc, argv, argi);
649: if (child->next != NULL) {
650: child = child->next;
651: needterm = 0;
652: }
653: }
654: if (needterm && *argi)
655: warnx("ignoring trailing %s", argv[*argi - 1]);
656: return parent == NULL ? child : parent;
1.6 schwarze 657: }
658:
659: static struct expr *
1.51 schwarze 660: exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 schwarze 661: {
1.6 schwarze 662: char errbuf[BUFSIZ];
1.1 schwarze 663: struct expr *e;
1.28 schwarze 664: char *key, *val;
1.11 schwarze 665: uint64_t iterbit;
1.51 schwarze 666: int cs, i, irc;
1.1 schwarze 667:
1.51 schwarze 668: if (strcmp("(", argv[*argi]) == 0) {
669: ++*argi;
670: e = exprcomp(search, argc, argv, argi);
671: if (*argi < argc) {
672: assert(strcmp(")", argv[*argi]) == 0);
673: ++*argi;
674: } else
675: warnx("unclosed parenthesis");
676: return e;
677: }
1.1 schwarze 678:
1.55 schwarze 679: if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
680: cs = 0;
681: ++*argi;
682: } else
683: cs = 1;
684:
1.51 schwarze 685: e = mandoc_calloc(1, sizeof(*e));
686: e->type = EXPR_TERM;
687: e->bits = 0;
688: e->next = NULL;
689: e->child = NULL;
1.1 schwarze 690:
1.33 schwarze 691: if (search->argmode == ARG_NAME) {
692: e->bits = TYPE_Nm;
1.51 schwarze 693: e->match.type = DBM_EXACT;
694: e->match.str = argv[(*argi)++];
1.45 schwarze 695: return e;
1.1 schwarze 696: }
697:
698: /*
1.33 schwarze 699: * Separate macro keys from search string.
1.51 schwarze 700: * If needed, request regular expression handling.
1.1 schwarze 701: */
702:
1.33 schwarze 703: if (search->argmode == ARG_WORD) {
704: e->bits = TYPE_Nm;
1.51 schwarze 705: e->match.type = DBM_REGEX;
706: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
1.33 schwarze 707: cs = 0;
1.51 schwarze 708: } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
1.33 schwarze 709: e->bits = TYPE_Nm | TYPE_Nd;
1.51 schwarze 710: e->match.type = DBM_SUB;
711: e->match.str = argv[*argi];
1.28 schwarze 712: } else {
1.51 schwarze 713: if (val == argv[*argi])
1.33 schwarze 714: e->bits = TYPE_Nm | TYPE_Nd;
1.51 schwarze 715: if (*val == '=') {
716: e->match.type = DBM_SUB;
717: e->match.str = val + 1;
718: } else
719: e->match.type = DBM_REGEX;
1.28 schwarze 720: *val++ = '\0';
1.51 schwarze 721: if (strstr(argv[*argi], "arch") != NULL)
1.12 schwarze 722: cs = 0;
1.28 schwarze 723: }
724:
725: /* Compile regular expressions. */
726:
1.51 schwarze 727: if (e->match.type == DBM_REGEX) {
728: e->match.re = mandoc_malloc(sizeof(*e->match.re));
729: irc = regcomp(e->match.re, val,
1.28 schwarze 730: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.51 schwarze 731: if (irc) {
732: regerror(irc, e->match.re, errbuf, sizeof(errbuf));
733: warnx("regcomp /%s/: %s", val, errbuf);
734: }
1.33 schwarze 735: if (search->argmode == ARG_WORD)
1.28 schwarze 736: free(val);
737: if (irc) {
1.51 schwarze 738: free(e->match.re);
1.1 schwarze 739: free(e);
1.51 schwarze 740: ++*argi;
1.45 schwarze 741: return NULL;
1.1 schwarze 742: }
1.28 schwarze 743: }
744:
1.51 schwarze 745: if (e->bits) {
746: ++*argi;
1.45 schwarze 747: return e;
1.51 schwarze 748: }
1.1 schwarze 749:
750: /*
751: * Parse out all possible fields.
752: * If the field doesn't resolve, bail.
753: */
754:
1.51 schwarze 755: while (NULL != (key = strsep(&argv[*argi], ","))) {
1.1 schwarze 756: if ('\0' == *key)
757: continue;
1.51 schwarze 758: for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
759: if (0 == strcasecmp(key, mansearch_keynames[i])) {
1.11 schwarze 760: e->bits |= iterbit;
761: break;
762: }
763: }
1.51 schwarze 764: if (i == KEY_MAX) {
765: if (strcasecmp(key, "any"))
766: warnx("treating unknown key "
767: "\"%s\" as \"any\"", key);
1.11 schwarze 768: e->bits |= ~0ULL;
1.1 schwarze 769: }
770: }
771:
1.51 schwarze 772: ++*argi;
1.45 schwarze 773: return e;
1.1 schwarze 774: }
775:
776: static void
1.51 schwarze 777: exprfree(struct expr *e)
1.1 schwarze 778: {
1.51 schwarze 779: if (e->next != NULL)
780: exprfree(e->next);
781: if (e->child != NULL)
782: exprfree(e->child);
783: free(e);
1.1 schwarze 784: }