Annotation of src/usr.bin/mandoc/mansearch.c, Revision 1.56
1.56 ! schwarze 1: /* $OpenBSD: mansearch.c,v 1.55 2017/04/18 15:04:35 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;
1.1 schwarze 104: *res = NULL;
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 ||
170: lstmatch(search->arch, page->arch) == 0)
1.1 schwarze 171: continue;
172:
173: if (cur + 1 > maxres) {
174: maxres += 1024;
1.26 schwarze 175: *res = mandoc_reallocarray(*res,
1.51 schwarze 176: maxres, sizeof(**res));
1.1 schwarze 177: }
178: mpage = *res + cur;
1.51 schwarze 179: mandoc_asprintf(&mpage->file, "%s/%s",
180: paths->paths[i], page->file + 1);
181: mpage->names = buildnames(page);
1.53 schwarze 182: mpage->output = buildoutput(outkey, page);
1.34 schwarze 183: mpage->ipath = i;
1.51 schwarze 184: mpage->bits = rp->bits;
185: mpage->sec = *page->sect - '0';
186: if (mpage->sec < 0 || mpage->sec > 9)
187: mpage->sec = 10;
188: mpage->form = *page->file;
189: free(rp);
190: cur++;
191: }
192: ohash_delete(htab);
193: free(htab);
194: dbm_close();
1.36 schwarze 195:
196: /*
197: * In man(1) mode, prefer matches in earlier trees
198: * over matches in later trees.
199: */
200:
201: if (cur && search->firstmatch)
202: break;
1.1 schwarze 203: }
1.29 schwarze 204: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.44 schwarze 205: if (chdir_status && getcwd_status && chdir(buf) == -1)
1.49 schwarze 206: warn("%s", buf);
1.1 schwarze 207: exprfree(e);
208: *sz = cur;
1.45 schwarze 209: return 1;
1.2 schwarze 210: }
211:
1.51 schwarze 212: /*
213: * Merge the results for the expression tree rooted at e
214: * into the the result list htab.
215: */
216: static struct ohash *
217: manmerge(struct expr *e, struct ohash *htab)
1.33 schwarze 218: {
1.51 schwarze 219: switch (e->type) {
220: case EXPR_TERM:
221: return manmerge_term(e, htab);
222: case EXPR_OR:
223: return manmerge_or(e->child, htab);
224: case EXPR_AND:
225: return manmerge_and(e->child, htab);
226: default:
227: abort();
1.33 schwarze 228: }
229: }
230:
1.51 schwarze 231: static struct ohash *
232: manmerge_term(struct expr *e, struct ohash *htab)
1.29 schwarze 233: {
1.51 schwarze 234: struct dbm_res res, *rp;
235: uint64_t ib;
236: unsigned int slot;
237: int im;
238:
239: if (htab == NULL) {
240: htab = mandoc_malloc(sizeof(*htab));
241: mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
242: }
243:
244: for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
245: if ((e->bits & ib) == 0)
246: continue;
247:
248: switch (ib) {
249: case TYPE_arch:
250: dbm_page_byarch(&e->match);
251: break;
252: case TYPE_sec:
253: dbm_page_bysect(&e->match);
254: break;
255: case TYPE_Nm:
256: dbm_page_byname(&e->match);
257: break;
258: case TYPE_Nd:
259: dbm_page_bydesc(&e->match);
260: break;
261: default:
262: dbm_page_bymacro(im - 2, &e->match);
263: break;
264: }
265:
266: /*
267: * When hashing for deduplication, use the unique
268: * page ID itself instead of a hash function;
269: * that is quite efficient.
270: */
1.29 schwarze 271:
1.51 schwarze 272: for (;;) {
273: res = dbm_page_next();
274: if (res.page == -1)
275: break;
276: slot = ohash_lookup_memory(htab,
277: (char *)&res, sizeof(res.page), res.page);
278: if ((rp = ohash_find(htab, slot)) != NULL) {
279: rp->bits |= res.bits;
280: continue;
281: }
282: rp = mandoc_malloc(sizeof(*rp));
283: *rp = res;
284: ohash_insert(htab, slot, rp);
285: }
286: }
287: return htab;
1.29 schwarze 288: }
289:
1.51 schwarze 290: static struct ohash *
291: manmerge_or(struct expr *e, struct ohash *htab)
292: {
293: while (e != NULL) {
294: htab = manmerge(e, htab);
295: e = e->next;
296: }
297: return htab;
298: }
1.13 schwarze 299:
1.51 schwarze 300: static struct ohash *
301: manmerge_and(struct expr *e, struct ohash *htab)
302: {
303: struct ohash *hand, *h1, *h2;
304: struct dbm_res *res;
305: unsigned int slot1, slot2;
1.13 schwarze 306:
1.51 schwarze 307: /* Evaluate the first term of the AND clause. */
1.29 schwarze 308:
1.51 schwarze 309: hand = manmerge(e, NULL);
1.29 schwarze 310:
1.51 schwarze 311: while ((e = e->next) != NULL) {
1.13 schwarze 312:
1.51 schwarze 313: /* Evaluate the next term and prepare for ANDing. */
1.13 schwarze 314:
1.51 schwarze 315: h2 = manmerge(e, NULL);
316: if (ohash_entries(h2) < ohash_entries(hand)) {
317: h1 = h2;
318: h2 = hand;
319: } else
320: h1 = hand;
321: hand = mandoc_malloc(sizeof(*hand));
322: mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
1.13 schwarze 323:
1.51 schwarze 324: /* Keep all pages that are in both result sets. */
1.13 schwarze 325:
1.51 schwarze 326: for (res = ohash_first(h1, &slot1); res != NULL;
327: res = ohash_next(h1, &slot1)) {
328: if (ohash_find(h2, ohash_lookup_memory(h2,
329: (char *)res, sizeof(res->page),
330: res->page)) == NULL)
331: free(res);
332: else
333: ohash_insert(hand, ohash_lookup_memory(hand,
334: (char *)res, sizeof(res->page),
335: res->page), res);
1.13 schwarze 336: }
337:
1.51 schwarze 338: /* Discard the merged results. */
339:
340: for (res = ohash_first(h2, &slot2); res != NULL;
341: res = ohash_next(h2, &slot2))
342: free(res);
343: ohash_delete(h2);
344: free(h2);
345: ohash_delete(h1);
346: free(h1);
347: }
1.13 schwarze 348:
1.51 schwarze 349: /* Merge the result of the AND into htab. */
1.8 schwarze 350:
1.51 schwarze 351: if (htab == NULL)
352: return hand;
1.8 schwarze 353:
1.51 schwarze 354: for (res = ohash_first(hand, &slot1); res != NULL;
355: res = ohash_next(hand, &slot1)) {
356: slot2 = ohash_lookup_memory(htab,
357: (char *)res, sizeof(res->page), res->page);
358: if (ohash_find(htab, slot2) == NULL)
359: ohash_insert(htab, slot2, res);
360: else
361: free(res);
362: }
363:
364: /* Discard the merged result. */
1.8 schwarze 365:
1.51 schwarze 366: ohash_delete(hand);
367: free(hand);
368: return htab;
369: }
1.41 schwarze 370:
1.51 schwarze 371: void
372: mansearch_free(struct manpage *res, size_t sz)
373: {
374: size_t i;
1.41 schwarze 375:
1.51 schwarze 376: for (i = 0; i < sz; i++) {
377: free(res[i].file);
378: free(res[i].names);
379: free(res[i].output);
1.13 schwarze 380: }
1.51 schwarze 381: free(res);
382: }
383:
384: static int
385: manpage_compare(const void *vp1, const void *vp2)
386: {
387: const struct manpage *mp1, *mp2;
388: int diff;
389:
390: mp1 = vp1;
391: mp2 = vp2;
392: return (diff = mp2->bits - mp1->bits) ? diff :
393: (diff = mp1->sec - mp2->sec) ? diff :
394: strcasecmp(mp1->names, mp2->names);
1.3 schwarze 395: }
396:
397: static char *
1.51 schwarze 398: buildnames(const struct dbm_page *page)
1.3 schwarze 399: {
1.51 schwarze 400: char *buf;
401: size_t i, sz;
1.3 schwarze 402:
1.53 schwarze 403: sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
404: (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
1.51 schwarze 405: buf = mandoc_malloc(sz);
406: i = 0;
1.53 schwarze 407: lstcat(buf, &i, page->name, ", ");
1.51 schwarze 408: buf[i++] = '(';
1.53 schwarze 409: lstcat(buf, &i, page->sect, ", ");
1.51 schwarze 410: if (page->arch != NULL) {
411: buf[i++] = '/';
1.53 schwarze 412: lstcat(buf, &i, page->arch, ", ");
1.51 schwarze 413: }
414: buf[i++] = ')';
415: buf[i++] = '\0';
416: assert(i == sz);
417: return buf;
1.1 schwarze 418: }
419:
420: /*
1.51 schwarze 421: * Count the buffer space needed to print the NUL-terminated
1.53 schwarze 422: * list of NUL-terminated strings, when printing sep separator
1.51 schwarze 423: * characters between strings.
1.1 schwarze 424: */
1.51 schwarze 425: static size_t
1.53 schwarze 426: lstlen(const char *cp, size_t sep)
1.1 schwarze 427: {
1.51 schwarze 428: size_t sz;
1.1 schwarze 429:
1.51 schwarze 430: for (sz = 0;; sz++) {
431: if (cp[0] == '\0') {
432: if (cp[1] == '\0')
433: break;
1.53 schwarze 434: sz += sep - 1;
1.51 schwarze 435: } else if (cp[0] < ' ')
436: sz--;
437: cp++;
438: }
439: return sz;
1.1 schwarze 440: }
441:
442: /*
1.51 schwarze 443: * Print the NUL-terminated list of NUL-terminated strings
1.53 schwarze 444: * into the buffer, seperating strings with sep.
1.1 schwarze 445: */
446: static void
1.53 schwarze 447: lstcat(char *buf, size_t *i, const char *cp, const char *sep)
1.1 schwarze 448: {
1.53 schwarze 449: const char *s;
450:
1.51 schwarze 451: for (;;) {
452: if (cp[0] == '\0') {
453: if (cp[1] == '\0')
454: break;
1.53 schwarze 455: s = sep;
456: while (*s != '\0')
457: buf[(*i)++] = *s++;
1.51 schwarze 458: } else if (cp[0] >= ' ')
459: buf[(*i)++] = cp[0];
460: cp++;
461: }
1.1 schwarze 462: }
463:
1.51 schwarze 464: /*
465: * Return 1 if the string *want occurs in any of the strings
466: * in the NUL-terminated string list *have, or 0 otherwise.
467: * If either argument is NULL or empty, assume no filtering
468: * is desired and return 1.
469: */
470: static int
471: lstmatch(const char *want, const char *have)
1.4 schwarze 472: {
1.51 schwarze 473: if (want == NULL || have == NULL || *have == '\0')
474: return 1;
475: while (*have != '\0') {
476: if (strcasestr(have, want) != NULL)
477: return 1;
478: have = strchr(have, '\0') + 1;
479: }
480: return 0;
1.4 schwarze 481: }
482:
1.1 schwarze 483: /*
1.53 schwarze 484: * Build a list of values taken by the macro im in the manual page.
1.1 schwarze 485: */
486: static char *
1.53 schwarze 487: buildoutput(size_t im, struct dbm_page *page)
1.1 schwarze 488: {
1.53 schwarze 489: const char *oldoutput, *sep, *input;
1.51 schwarze 490: char *output, *newoutput, *value;
1.53 schwarze 491: size_t sz, i;
492:
493: switch (im) {
494: case KEY_Nd:
495: return mandoc_strdup(page->desc);
496: case KEY_Nm:
497: input = page->name;
498: break;
499: case KEY_sec:
500: input = page->sect;
501: break;
502: case KEY_arch:
503: input = page->arch;
504: if (input == NULL)
505: input = "all\0";
506: break;
507: default:
508: input = NULL;
509: break;
510: }
511:
512: if (input != NULL) {
513: sz = lstlen(input, 3) + 1;
514: output = mandoc_malloc(sz);
515: i = 0;
516: lstcat(output, &i, input, " # ");
1.54 schwarze 517: output[i++] = '\0';
518: assert(i == sz);
1.53 schwarze 519: return output;
520: }
1.51 schwarze 521:
522: output = NULL;
1.53 schwarze 523: dbm_macro_bypage(im - 2, page->addr);
1.51 schwarze 524: while ((value = dbm_macro_next()) != NULL) {
525: if (output == NULL) {
526: oldoutput = "";
527: sep = "";
528: } else {
529: oldoutput = output;
530: sep = " # ";
531: }
532: mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
533: free(output);
534: output = newoutput;
1.1 schwarze 535: }
1.51 schwarze 536: return output;
1.1 schwarze 537: }
538:
539: /*
540: * Compile a set of string tokens into an expression.
541: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
542: * "(", "foo=bar", etc.).
543: */
544: static struct expr *
1.51 schwarze 545: exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 schwarze 546: {
1.51 schwarze 547: struct expr *parent, *child;
548: int needterm, nested;
549:
550: if ((nested = *argi) == argc)
551: return NULL;
552: needterm = 1;
553: parent = child = NULL;
554: while (*argi < argc) {
555: if (strcmp(")", argv[*argi]) == 0) {
556: if (needterm)
557: warnx("missing term "
558: "before closing parenthesis");
559: needterm = 0;
560: if (nested)
561: break;
562: warnx("ignoring unmatched right parenthesis");
563: ++*argi;
1.4 schwarze 564: continue;
1.51 schwarze 565: }
566: if (strcmp("-o", argv[*argi]) == 0) {
567: if (needterm) {
568: if (*argi > 0)
569: warnx("ignoring -o after %s",
570: argv[*argi - 1]);
571: else
572: warnx("ignoring initial -o");
573: }
574: needterm = 1;
575: ++*argi;
1.4 schwarze 576: continue;
1.51 schwarze 577: }
578: needterm = 0;
579: if (child == NULL) {
580: child = expr_and(search, argc, argv, argi);
1.4 schwarze 581: continue;
1.1 schwarze 582: }
1.51 schwarze 583: if (parent == NULL) {
584: parent = mandoc_calloc(1, sizeof(*parent));
585: parent->type = EXPR_OR;
586: parent->next = NULL;
587: parent->child = child;
588: }
589: child->next = expr_and(search, argc, argv, argi);
590: child = child->next;
591: }
592: if (needterm && *argi)
593: warnx("ignoring trailing %s", argv[*argi - 1]);
594: return parent == NULL ? child : parent;
595: }
1.17 schwarze 596:
1.51 schwarze 597: static struct expr *
598: expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
599: {
600: struct expr *parent, *child;
601: int needterm;
1.17 schwarze 602:
1.51 schwarze 603: needterm = 1;
604: parent = child = NULL;
605: while (*argi < argc) {
606: if (strcmp(")", argv[*argi]) == 0) {
607: if (needterm)
608: warnx("missing term "
609: "before closing parenthesis");
610: needterm = 0;
611: break;
612: }
613: if (strcmp("-o", argv[*argi]) == 0)
614: break;
615: if (strcmp("-a", argv[*argi]) == 0) {
616: if (needterm) {
617: if (*argi > 0)
618: warnx("ignoring -a after %s",
619: argv[*argi - 1]);
620: else
621: warnx("ignoring initial -a");
1.18 schwarze 622: }
1.51 schwarze 623: needterm = 1;
624: ++*argi;
625: continue;
1.18 schwarze 626: }
1.51 schwarze 627: if (needterm == 0)
628: break;
629: if (child == NULL) {
630: child = exprterm(search, argc, argv, argi);
631: if (child != NULL)
632: needterm = 0;
633: continue;
634: }
635: needterm = 0;
636: if (parent == NULL) {
637: parent = mandoc_calloc(1, sizeof(*parent));
638: parent->type = EXPR_AND;
639: parent->next = NULL;
640: parent->child = child;
641: }
642: child->next = exprterm(search, argc, argv, argi);
643: if (child->next != NULL) {
644: child = child->next;
645: needterm = 0;
646: }
647: }
648: if (needterm && *argi)
649: warnx("ignoring trailing %s", argv[*argi - 1]);
650: return parent == NULL ? child : parent;
1.6 schwarze 651: }
652:
653: static struct expr *
1.51 schwarze 654: exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 schwarze 655: {
1.6 schwarze 656: char errbuf[BUFSIZ];
1.1 schwarze 657: struct expr *e;
1.28 schwarze 658: char *key, *val;
1.11 schwarze 659: uint64_t iterbit;
1.51 schwarze 660: int cs, i, irc;
1.1 schwarze 661:
1.51 schwarze 662: if (strcmp("(", argv[*argi]) == 0) {
663: ++*argi;
664: e = exprcomp(search, argc, argv, argi);
665: if (*argi < argc) {
666: assert(strcmp(")", argv[*argi]) == 0);
667: ++*argi;
668: } else
669: warnx("unclosed parenthesis");
670: return e;
671: }
1.1 schwarze 672:
1.55 schwarze 673: if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
674: cs = 0;
675: ++*argi;
676: } else
677: cs = 1;
678:
1.51 schwarze 679: e = mandoc_calloc(1, sizeof(*e));
680: e->type = EXPR_TERM;
681: e->bits = 0;
682: e->next = NULL;
683: e->child = NULL;
1.1 schwarze 684:
1.33 schwarze 685: if (search->argmode == ARG_NAME) {
686: e->bits = TYPE_Nm;
1.51 schwarze 687: e->match.type = DBM_EXACT;
688: e->match.str = argv[(*argi)++];
1.45 schwarze 689: return e;
1.1 schwarze 690: }
691:
692: /*
1.33 schwarze 693: * Separate macro keys from search string.
1.51 schwarze 694: * If needed, request regular expression handling.
1.1 schwarze 695: */
696:
1.33 schwarze 697: if (search->argmode == ARG_WORD) {
698: e->bits = TYPE_Nm;
1.51 schwarze 699: e->match.type = DBM_REGEX;
700: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
1.33 schwarze 701: cs = 0;
1.51 schwarze 702: } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
1.33 schwarze 703: e->bits = TYPE_Nm | TYPE_Nd;
1.51 schwarze 704: e->match.type = DBM_SUB;
705: e->match.str = argv[*argi];
1.28 schwarze 706: } else {
1.51 schwarze 707: if (val == argv[*argi])
1.33 schwarze 708: e->bits = TYPE_Nm | TYPE_Nd;
1.51 schwarze 709: if (*val == '=') {
710: e->match.type = DBM_SUB;
711: e->match.str = val + 1;
712: } else
713: e->match.type = DBM_REGEX;
1.28 schwarze 714: *val++ = '\0';
1.51 schwarze 715: if (strstr(argv[*argi], "arch") != NULL)
1.12 schwarze 716: cs = 0;
1.28 schwarze 717: }
718:
719: /* Compile regular expressions. */
720:
1.51 schwarze 721: if (e->match.type == DBM_REGEX) {
722: e->match.re = mandoc_malloc(sizeof(*e->match.re));
723: irc = regcomp(e->match.re, val,
1.28 schwarze 724: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.51 schwarze 725: if (irc) {
726: regerror(irc, e->match.re, errbuf, sizeof(errbuf));
727: warnx("regcomp /%s/: %s", val, errbuf);
728: }
1.33 schwarze 729: if (search->argmode == ARG_WORD)
1.28 schwarze 730: free(val);
731: if (irc) {
1.51 schwarze 732: free(e->match.re);
1.1 schwarze 733: free(e);
1.51 schwarze 734: ++*argi;
1.45 schwarze 735: return NULL;
1.1 schwarze 736: }
1.28 schwarze 737: }
738:
1.51 schwarze 739: if (e->bits) {
740: ++*argi;
1.45 schwarze 741: return e;
1.51 schwarze 742: }
1.1 schwarze 743:
744: /*
745: * Parse out all possible fields.
746: * If the field doesn't resolve, bail.
747: */
748:
1.51 schwarze 749: while (NULL != (key = strsep(&argv[*argi], ","))) {
1.1 schwarze 750: if ('\0' == *key)
751: continue;
1.51 schwarze 752: for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
753: if (0 == strcasecmp(key, mansearch_keynames[i])) {
1.11 schwarze 754: e->bits |= iterbit;
755: break;
756: }
757: }
1.51 schwarze 758: if (i == KEY_MAX) {
759: if (strcasecmp(key, "any"))
760: warnx("treating unknown key "
761: "\"%s\" as \"any\"", key);
1.11 schwarze 762: e->bits |= ~0ULL;
1.1 schwarze 763: }
764: }
765:
1.51 schwarze 766: ++*argi;
1.45 schwarze 767: return e;
1.1 schwarze 768: }
769:
770: static void
1.51 schwarze 771: exprfree(struct expr *e)
1.1 schwarze 772: {
1.51 schwarze 773: if (e->next != NULL)
774: exprfree(e->next);
775: if (e->child != NULL)
776: exprfree(e->child);
777: free(e);
1.1 schwarze 778: }