Annotation of src/usr.bin/mandoc/mansearch.c, Revision 1.63
1.63 ! schwarze 1: /* $OpenBSD: mansearch.c,v 1.62 2018/11/22 12:01:42 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 "
191: "bogus %s entry, run makewhatis %s",
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);
1.34 schwarze 199: mpage->ipath = i;
1.51 schwarze 200: mpage->sec = *page->sect - '0';
201: if (mpage->sec < 0 || mpage->sec > 9)
202: mpage->sec = 10;
203: mpage->form = *page->file;
204: free(rp);
205: cur++;
206: }
207: ohash_delete(htab);
208: free(htab);
209: dbm_close();
1.36 schwarze 210:
211: /*
212: * In man(1) mode, prefer matches in earlier trees
213: * over matches in later trees.
214: */
215:
216: if (cur && search->firstmatch)
217: break;
1.1 schwarze 218: }
1.57 schwarze 219: if (res != NULL)
220: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.44 schwarze 221: if (chdir_status && getcwd_status && chdir(buf) == -1)
1.49 schwarze 222: warn("%s", buf);
1.1 schwarze 223: exprfree(e);
224: *sz = cur;
1.57 schwarze 225: return res != NULL || cur;
1.2 schwarze 226: }
227:
1.51 schwarze 228: /*
229: * Merge the results for the expression tree rooted at e
230: * into the the result list htab.
231: */
232: static struct ohash *
233: manmerge(struct expr *e, struct ohash *htab)
1.33 schwarze 234: {
1.51 schwarze 235: switch (e->type) {
236: case EXPR_TERM:
237: return manmerge_term(e, htab);
238: case EXPR_OR:
239: return manmerge_or(e->child, htab);
240: case EXPR_AND:
241: return manmerge_and(e->child, htab);
242: default:
243: abort();
1.33 schwarze 244: }
245: }
246:
1.51 schwarze 247: static struct ohash *
248: manmerge_term(struct expr *e, struct ohash *htab)
1.29 schwarze 249: {
1.51 schwarze 250: struct dbm_res res, *rp;
251: uint64_t ib;
252: unsigned int slot;
253: int im;
254:
255: if (htab == NULL) {
256: htab = mandoc_malloc(sizeof(*htab));
257: mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
258: }
259:
260: for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
261: if ((e->bits & ib) == 0)
262: continue;
263:
264: switch (ib) {
265: case TYPE_arch:
266: dbm_page_byarch(&e->match);
267: break;
268: case TYPE_sec:
269: dbm_page_bysect(&e->match);
270: break;
271: case TYPE_Nm:
272: dbm_page_byname(&e->match);
273: break;
274: case TYPE_Nd:
275: dbm_page_bydesc(&e->match);
276: break;
277: default:
278: dbm_page_bymacro(im - 2, &e->match);
279: break;
280: }
281:
282: /*
283: * When hashing for deduplication, use the unique
284: * page ID itself instead of a hash function;
285: * that is quite efficient.
286: */
1.29 schwarze 287:
1.51 schwarze 288: for (;;) {
289: res = dbm_page_next();
290: if (res.page == -1)
291: break;
292: slot = ohash_lookup_memory(htab,
293: (char *)&res, sizeof(res.page), res.page);
1.62 schwarze 294: if ((rp = ohash_find(htab, slot)) != NULL)
1.51 schwarze 295: continue;
296: rp = mandoc_malloc(sizeof(*rp));
297: *rp = res;
298: ohash_insert(htab, slot, rp);
299: }
300: }
301: return htab;
1.29 schwarze 302: }
303:
1.51 schwarze 304: static struct ohash *
305: manmerge_or(struct expr *e, struct ohash *htab)
306: {
307: while (e != NULL) {
308: htab = manmerge(e, htab);
309: e = e->next;
310: }
311: return htab;
312: }
1.13 schwarze 313:
1.51 schwarze 314: static struct ohash *
315: manmerge_and(struct expr *e, struct ohash *htab)
316: {
317: struct ohash *hand, *h1, *h2;
318: struct dbm_res *res;
319: unsigned int slot1, slot2;
1.13 schwarze 320:
1.51 schwarze 321: /* Evaluate the first term of the AND clause. */
1.29 schwarze 322:
1.51 schwarze 323: hand = manmerge(e, NULL);
1.29 schwarze 324:
1.51 schwarze 325: while ((e = e->next) != NULL) {
1.13 schwarze 326:
1.51 schwarze 327: /* Evaluate the next term and prepare for ANDing. */
1.13 schwarze 328:
1.51 schwarze 329: h2 = manmerge(e, NULL);
330: if (ohash_entries(h2) < ohash_entries(hand)) {
331: h1 = h2;
332: h2 = hand;
333: } else
334: h1 = hand;
335: hand = mandoc_malloc(sizeof(*hand));
336: mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
1.13 schwarze 337:
1.51 schwarze 338: /* Keep all pages that are in both result sets. */
1.13 schwarze 339:
1.51 schwarze 340: for (res = ohash_first(h1, &slot1); res != NULL;
341: res = ohash_next(h1, &slot1)) {
342: if (ohash_find(h2, ohash_lookup_memory(h2,
343: (char *)res, sizeof(res->page),
344: res->page)) == NULL)
345: free(res);
346: else
347: ohash_insert(hand, ohash_lookup_memory(hand,
348: (char *)res, sizeof(res->page),
349: res->page), res);
1.13 schwarze 350: }
351:
1.51 schwarze 352: /* Discard the merged results. */
353:
354: for (res = ohash_first(h2, &slot2); res != NULL;
355: res = ohash_next(h2, &slot2))
356: free(res);
357: ohash_delete(h2);
358: free(h2);
359: ohash_delete(h1);
360: free(h1);
361: }
1.13 schwarze 362:
1.51 schwarze 363: /* Merge the result of the AND into htab. */
1.8 schwarze 364:
1.51 schwarze 365: if (htab == NULL)
366: return hand;
1.8 schwarze 367:
1.51 schwarze 368: for (res = ohash_first(hand, &slot1); res != NULL;
369: res = ohash_next(hand, &slot1)) {
370: slot2 = ohash_lookup_memory(htab,
371: (char *)res, sizeof(res->page), res->page);
372: if (ohash_find(htab, slot2) == NULL)
373: ohash_insert(htab, slot2, res);
374: else
375: free(res);
376: }
377:
378: /* Discard the merged result. */
1.8 schwarze 379:
1.51 schwarze 380: ohash_delete(hand);
381: free(hand);
382: return htab;
383: }
1.41 schwarze 384:
1.51 schwarze 385: void
386: mansearch_free(struct manpage *res, size_t sz)
387: {
388: size_t i;
1.41 schwarze 389:
1.51 schwarze 390: for (i = 0; i < sz; i++) {
391: free(res[i].file);
392: free(res[i].names);
393: free(res[i].output);
1.13 schwarze 394: }
1.51 schwarze 395: free(res);
396: }
397:
398: static int
399: manpage_compare(const void *vp1, const void *vp2)
400: {
401: const struct manpage *mp1, *mp2;
1.58 schwarze 402: const char *cp1, *cp2;
403: size_t sz1, sz2;
1.51 schwarze 404: int diff;
405:
406: mp1 = vp1;
407: mp2 = vp2;
1.62 schwarze 408: if ((diff = mp1->sec - mp2->sec))
1.58 schwarze 409: return diff;
410:
411: /* Fall back to alphabetic ordering of names. */
412: sz1 = strcspn(mp1->names, "(");
413: sz2 = strcspn(mp2->names, "(");
414: if (sz1 < sz2)
415: sz1 = sz2;
416: if ((diff = strncasecmp(mp1->names, mp2->names, sz1)))
417: return diff;
418:
419: /* For identical names and sections, prefer arch-dependent. */
420: cp1 = strchr(mp1->names + sz1, '/');
421: cp2 = strchr(mp2->names + sz2, '/');
422: return cp1 != NULL && cp2 != NULL ? strcasecmp(cp1, cp2) :
423: cp1 != NULL ? -1 : cp2 != NULL ? 1 : 0;
1.3 schwarze 424: }
425:
426: static char *
1.51 schwarze 427: buildnames(const struct dbm_page *page)
1.3 schwarze 428: {
1.51 schwarze 429: char *buf;
430: size_t i, sz;
1.3 schwarze 431:
1.53 schwarze 432: sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
433: (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
1.51 schwarze 434: buf = mandoc_malloc(sz);
435: i = 0;
1.53 schwarze 436: lstcat(buf, &i, page->name, ", ");
1.51 schwarze 437: buf[i++] = '(';
1.53 schwarze 438: lstcat(buf, &i, page->sect, ", ");
1.51 schwarze 439: if (page->arch != NULL) {
440: buf[i++] = '/';
1.53 schwarze 441: lstcat(buf, &i, page->arch, ", ");
1.51 schwarze 442: }
443: buf[i++] = ')';
444: buf[i++] = '\0';
445: assert(i == sz);
446: return buf;
1.1 schwarze 447: }
448:
449: /*
1.51 schwarze 450: * Count the buffer space needed to print the NUL-terminated
1.53 schwarze 451: * list of NUL-terminated strings, when printing sep separator
1.51 schwarze 452: * characters between strings.
1.1 schwarze 453: */
1.51 schwarze 454: static size_t
1.53 schwarze 455: lstlen(const char *cp, size_t sep)
1.1 schwarze 456: {
1.51 schwarze 457: size_t sz;
1.1 schwarze 458:
1.59 schwarze 459: for (sz = 0; *cp != '\0'; cp++) {
460:
461: /* Skip names appearing only in the SYNOPSIS. */
462: if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
463: while (*cp != '\0')
464: cp++;
465: continue;
466: }
467:
468: /* Skip name class markers. */
469: if (*cp < ' ')
470: cp++;
471:
472: /* Print a separator before each but the first string. */
473: if (sz)
474: sz += sep;
475:
476: /* Copy one string. */
477: while (*cp != '\0') {
478: sz++;
479: cp++;
480: }
1.51 schwarze 481: }
482: return sz;
1.1 schwarze 483: }
484:
485: /*
1.51 schwarze 486: * Print the NUL-terminated list of NUL-terminated strings
1.53 schwarze 487: * into the buffer, seperating strings with sep.
1.1 schwarze 488: */
489: static void
1.53 schwarze 490: lstcat(char *buf, size_t *i, const char *cp, const char *sep)
1.1 schwarze 491: {
1.59 schwarze 492: const char *s;
493: size_t i_start;
1.53 schwarze 494:
1.59 schwarze 495: for (i_start = *i; *cp != '\0'; cp++) {
496:
497: /* Skip names appearing only in the SYNOPSIS. */
498: if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
499: while (*cp != '\0')
500: cp++;
501: continue;
502: }
503:
504: /* Skip name class markers. */
505: if (*cp < ' ')
506: cp++;
507:
508: /* Print a separator before each but the first string. */
509: if (*i > i_start) {
1.53 schwarze 510: s = sep;
511: while (*s != '\0')
512: buf[(*i)++] = *s++;
1.59 schwarze 513: }
514:
515: /* Copy one string. */
516: while (*cp != '\0')
517: buf[(*i)++] = *cp++;
1.51 schwarze 518: }
1.59 schwarze 519:
1.1 schwarze 520: }
521:
1.51 schwarze 522: /*
523: * Return 1 if the string *want occurs in any of the strings
524: * in the NUL-terminated string list *have, or 0 otherwise.
525: * If either argument is NULL or empty, assume no filtering
526: * is desired and return 1.
527: */
528: static int
529: lstmatch(const char *want, const char *have)
1.4 schwarze 530: {
1.51 schwarze 531: if (want == NULL || have == NULL || *have == '\0')
532: return 1;
533: while (*have != '\0') {
534: if (strcasestr(have, want) != NULL)
535: return 1;
536: have = strchr(have, '\0') + 1;
537: }
538: return 0;
1.4 schwarze 539: }
540:
1.1 schwarze 541: /*
1.53 schwarze 542: * Build a list of values taken by the macro im in the manual page.
1.1 schwarze 543: */
544: static char *
1.53 schwarze 545: buildoutput(size_t im, struct dbm_page *page)
1.1 schwarze 546: {
1.53 schwarze 547: const char *oldoutput, *sep, *input;
1.51 schwarze 548: char *output, *newoutput, *value;
1.53 schwarze 549: size_t sz, i;
550:
551: switch (im) {
552: case KEY_Nd:
553: return mandoc_strdup(page->desc);
554: case KEY_Nm:
555: input = page->name;
556: break;
557: case KEY_sec:
558: input = page->sect;
559: break;
560: case KEY_arch:
561: input = page->arch;
562: if (input == NULL)
563: input = "all\0";
564: break;
565: default:
566: input = NULL;
567: break;
568: }
569:
570: if (input != NULL) {
571: sz = lstlen(input, 3) + 1;
572: output = mandoc_malloc(sz);
573: i = 0;
574: lstcat(output, &i, input, " # ");
1.54 schwarze 575: output[i++] = '\0';
576: assert(i == sz);
1.53 schwarze 577: return output;
578: }
1.51 schwarze 579:
580: output = NULL;
1.53 schwarze 581: dbm_macro_bypage(im - 2, page->addr);
1.51 schwarze 582: while ((value = dbm_macro_next()) != NULL) {
583: if (output == NULL) {
584: oldoutput = "";
585: sep = "";
586: } else {
587: oldoutput = output;
588: sep = " # ";
589: }
590: mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
591: free(output);
592: output = newoutput;
1.1 schwarze 593: }
1.51 schwarze 594: return output;
1.1 schwarze 595: }
596:
597: /*
598: * Compile a set of string tokens into an expression.
599: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
600: * "(", "foo=bar", etc.).
601: */
602: static struct expr *
1.51 schwarze 603: exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 schwarze 604: {
1.51 schwarze 605: struct expr *parent, *child;
606: int needterm, nested;
607:
608: if ((nested = *argi) == argc)
609: return NULL;
610: needterm = 1;
611: parent = child = NULL;
612: while (*argi < argc) {
613: if (strcmp(")", argv[*argi]) == 0) {
614: if (needterm)
615: warnx("missing term "
616: "before closing parenthesis");
617: needterm = 0;
618: if (nested)
619: break;
620: warnx("ignoring unmatched right parenthesis");
621: ++*argi;
1.4 schwarze 622: continue;
1.51 schwarze 623: }
624: if (strcmp("-o", argv[*argi]) == 0) {
625: if (needterm) {
626: if (*argi > 0)
627: warnx("ignoring -o after %s",
628: argv[*argi - 1]);
629: else
630: warnx("ignoring initial -o");
631: }
632: needterm = 1;
633: ++*argi;
1.4 schwarze 634: continue;
1.51 schwarze 635: }
636: needterm = 0;
637: if (child == NULL) {
638: child = expr_and(search, argc, argv, argi);
1.4 schwarze 639: continue;
1.1 schwarze 640: }
1.51 schwarze 641: if (parent == NULL) {
642: parent = mandoc_calloc(1, sizeof(*parent));
643: parent->type = EXPR_OR;
644: parent->next = NULL;
645: parent->child = child;
646: }
647: child->next = expr_and(search, argc, argv, argi);
648: child = child->next;
649: }
650: if (needterm && *argi)
651: warnx("ignoring trailing %s", argv[*argi - 1]);
652: return parent == NULL ? child : parent;
653: }
1.17 schwarze 654:
1.51 schwarze 655: static struct expr *
656: expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
657: {
658: struct expr *parent, *child;
659: int needterm;
1.17 schwarze 660:
1.51 schwarze 661: needterm = 1;
662: parent = child = NULL;
663: while (*argi < argc) {
664: if (strcmp(")", argv[*argi]) == 0) {
665: if (needterm)
666: warnx("missing term "
667: "before closing parenthesis");
668: needterm = 0;
669: break;
670: }
671: if (strcmp("-o", argv[*argi]) == 0)
672: break;
673: if (strcmp("-a", argv[*argi]) == 0) {
674: if (needterm) {
675: if (*argi > 0)
676: warnx("ignoring -a after %s",
677: argv[*argi - 1]);
678: else
679: warnx("ignoring initial -a");
1.18 schwarze 680: }
1.51 schwarze 681: needterm = 1;
682: ++*argi;
683: continue;
1.18 schwarze 684: }
1.51 schwarze 685: if (needterm == 0)
686: break;
687: if (child == NULL) {
688: child = exprterm(search, argc, argv, argi);
689: if (child != NULL)
690: needterm = 0;
691: continue;
692: }
693: needterm = 0;
694: if (parent == NULL) {
695: parent = mandoc_calloc(1, sizeof(*parent));
696: parent->type = EXPR_AND;
697: parent->next = NULL;
698: parent->child = child;
699: }
700: child->next = exprterm(search, argc, argv, argi);
701: if (child->next != NULL) {
702: child = child->next;
703: needterm = 0;
704: }
705: }
706: if (needterm && *argi)
707: warnx("ignoring trailing %s", argv[*argi - 1]);
708: return parent == NULL ? child : parent;
1.6 schwarze 709: }
710:
711: static struct expr *
1.51 schwarze 712: exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 schwarze 713: {
1.6 schwarze 714: char errbuf[BUFSIZ];
1.1 schwarze 715: struct expr *e;
1.28 schwarze 716: char *key, *val;
1.11 schwarze 717: uint64_t iterbit;
1.51 schwarze 718: int cs, i, irc;
1.1 schwarze 719:
1.51 schwarze 720: if (strcmp("(", argv[*argi]) == 0) {
721: ++*argi;
722: e = exprcomp(search, argc, argv, argi);
723: if (*argi < argc) {
724: assert(strcmp(")", argv[*argi]) == 0);
725: ++*argi;
726: } else
727: warnx("unclosed parenthesis");
728: return e;
729: }
1.1 schwarze 730:
1.55 schwarze 731: if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
732: cs = 0;
733: ++*argi;
734: } else
735: cs = 1;
736:
1.51 schwarze 737: e = mandoc_calloc(1, sizeof(*e));
738: e->type = EXPR_TERM;
739: e->bits = 0;
740: e->next = NULL;
741: e->child = NULL;
1.1 schwarze 742:
1.33 schwarze 743: if (search->argmode == ARG_NAME) {
744: e->bits = TYPE_Nm;
1.51 schwarze 745: e->match.type = DBM_EXACT;
746: e->match.str = argv[(*argi)++];
1.45 schwarze 747: return e;
1.1 schwarze 748: }
749:
750: /*
1.33 schwarze 751: * Separate macro keys from search string.
1.51 schwarze 752: * If needed, request regular expression handling.
1.1 schwarze 753: */
754:
1.33 schwarze 755: if (search->argmode == ARG_WORD) {
756: e->bits = TYPE_Nm;
1.51 schwarze 757: e->match.type = DBM_REGEX;
758: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
1.33 schwarze 759: cs = 0;
1.51 schwarze 760: } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
1.33 schwarze 761: e->bits = TYPE_Nm | TYPE_Nd;
1.61 schwarze 762: e->match.type = DBM_REGEX;
763: val = argv[*argi];
764: cs = 0;
1.28 schwarze 765: } else {
1.51 schwarze 766: if (val == argv[*argi])
1.33 schwarze 767: e->bits = TYPE_Nm | TYPE_Nd;
1.51 schwarze 768: if (*val == '=') {
769: e->match.type = DBM_SUB;
770: e->match.str = val + 1;
771: } else
772: e->match.type = DBM_REGEX;
1.28 schwarze 773: *val++ = '\0';
1.51 schwarze 774: if (strstr(argv[*argi], "arch") != NULL)
1.12 schwarze 775: cs = 0;
1.28 schwarze 776: }
777:
778: /* Compile regular expressions. */
779:
1.51 schwarze 780: if (e->match.type == DBM_REGEX) {
781: e->match.re = mandoc_malloc(sizeof(*e->match.re));
782: irc = regcomp(e->match.re, val,
1.28 schwarze 783: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.51 schwarze 784: if (irc) {
785: regerror(irc, e->match.re, errbuf, sizeof(errbuf));
786: warnx("regcomp /%s/: %s", val, errbuf);
787: }
1.33 schwarze 788: if (search->argmode == ARG_WORD)
1.28 schwarze 789: free(val);
790: if (irc) {
1.51 schwarze 791: free(e->match.re);
1.1 schwarze 792: free(e);
1.51 schwarze 793: ++*argi;
1.45 schwarze 794: return NULL;
1.1 schwarze 795: }
1.28 schwarze 796: }
797:
1.51 schwarze 798: if (e->bits) {
799: ++*argi;
1.45 schwarze 800: return e;
1.51 schwarze 801: }
1.1 schwarze 802:
803: /*
804: * Parse out all possible fields.
805: * If the field doesn't resolve, bail.
806: */
807:
1.51 schwarze 808: while (NULL != (key = strsep(&argv[*argi], ","))) {
1.1 schwarze 809: if ('\0' == *key)
810: continue;
1.51 schwarze 811: for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
812: if (0 == strcasecmp(key, mansearch_keynames[i])) {
1.11 schwarze 813: e->bits |= iterbit;
814: break;
815: }
816: }
1.51 schwarze 817: if (i == KEY_MAX) {
818: if (strcasecmp(key, "any"))
819: warnx("treating unknown key "
820: "\"%s\" as \"any\"", key);
1.11 schwarze 821: e->bits |= ~0ULL;
1.1 schwarze 822: }
823: }
824:
1.51 schwarze 825: ++*argi;
1.45 schwarze 826: return e;
1.1 schwarze 827: }
828:
829: static void
1.51 schwarze 830: exprfree(struct expr *e)
1.1 schwarze 831: {
1.51 schwarze 832: if (e->next != NULL)
833: exprfree(e->next);
834: if (e->child != NULL)
835: exprfree(e->child);
836: free(e);
1.1 schwarze 837: }