Annotation of mandoc/mansearch.c, Revision 1.82
1.82 ! schwarze 1: /* $Id: mansearch.c,v 1.81 2019/04/30 18:51:57 schwarze Exp $ */
1.1 kristaps 2: /*
3: * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.78 schwarze 4: * Copyright (c) 2013-2018 Ingo Schwarze <schwarze@openbsd.org>
1.1 kristaps 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: *
1.56 schwarze 10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
1.1 kristaps 11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
1.56 schwarze 12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
1.1 kristaps 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: */
1.67 schwarze 18: #include "config.h"
1.1 kristaps 19:
1.28 schwarze 20: #include <sys/mman.h>
1.43 schwarze 21: #include <sys/types.h>
22:
1.1 kristaps 23: #include <assert.h>
1.68 schwarze 24: #if HAVE_ERR
1.59 schwarze 25: #include <err.h>
1.68 schwarze 26: #endif
1.52 schwarze 27: #include <errno.h>
1.1 kristaps 28: #include <fcntl.h>
1.54 schwarze 29: #include <glob.h>
1.6 schwarze 30: #include <limits.h>
1.8 schwarze 31: #include <regex.h>
1.1 kristaps 32: #include <stdio.h>
33: #include <stdint.h>
34: #include <stddef.h>
35: #include <stdlib.h>
36: #include <string.h>
37: #include <unistd.h>
38:
1.23 schwarze 39: #include "mandoc_aux.h"
1.60 schwarze 40: #include "mandoc_ohash.h"
1.56 schwarze 41: #include "manconf.h"
1.1 kristaps 42: #include "mansearch.h"
1.66 schwarze 43: #include "dbm.h"
1.2 kristaps 44:
1.1 kristaps 45: struct expr {
1.66 schwarze 46: /* Used for terms: */
47: struct dbm_match match; /* Match type and expression. */
48: uint64_t bits; /* Type mask. */
49: /* Used for OR and AND groups: */
50: struct expr *next; /* Next child in the parent group. */
51: struct expr *child; /* First child in this group. */
52: enum { EXPR_TERM, EXPR_OR, EXPR_AND } type;
1.1 kristaps 53: };
54:
1.66 schwarze 55: const char *const mansearch_keynames[KEY_MAX] = {
56: "arch", "sec", "Xr", "Ar", "Fa", "Fl", "Dv", "Fn",
57: "Ic", "Pa", "Cm", "Li", "Em", "Cd", "Va", "Ft",
58: "Tn", "Er", "Ev", "Sy", "Sh", "In", "Ss", "Ox",
59: "An", "Mt", "St", "Bx", "At", "Nx", "Fx", "Lk",
60: "Ms", "Bsx", "Dx", "Rs", "Vt", "Lb", "Nm", "Nd"
1.1 kristaps 61: };
62:
1.66 schwarze 63:
64: static struct ohash *manmerge(struct expr *, struct ohash *);
65: static struct ohash *manmerge_term(struct expr *, struct ohash *);
66: static struct ohash *manmerge_or(struct expr *, struct ohash *);
67: static struct ohash *manmerge_and(struct expr *, struct ohash *);
68: static char *buildnames(const struct dbm_page *);
1.70 schwarze 69: static char *buildoutput(size_t, struct dbm_page *);
70: static size_t lstlen(const char *, size_t);
71: static void lstcat(char *, size_t *, const char *, const char *);
1.66 schwarze 72: static int lstmatch(const char *, const char *);
1.34 schwarze 73: static struct expr *exprcomp(const struct mansearch *,
1.66 schwarze 74: int, char *[], int *);
75: static struct expr *expr_and(const struct mansearch *,
76: int, char *[], int *);
77: static struct expr *exprterm(const struct mansearch *,
78: int, char *[], int *);
1.1 kristaps 79: static void exprfree(struct expr *);
1.39 schwarze 80: static int manpage_compare(const void *, const void *);
1.28 schwarze 81:
1.1 kristaps 82:
83: int
1.5 kristaps 84: mansearch(const struct mansearch *search,
1.12 schwarze 85: const struct manpaths *paths,
86: int argc, char *argv[],
1.1 kristaps 87: struct manpage **res, size_t *sz)
88: {
1.6 schwarze 89: char buf[PATH_MAX];
1.66 schwarze 90: struct dbm_res *rp;
91: struct expr *e;
92: struct dbm_page *page;
1.10 schwarze 93: struct manpage *mpage;
1.66 schwarze 94: struct ohash *htab;
95: size_t cur, i, maxres, outkey;
96: unsigned int slot;
97: int argi, chdir_status, getcwd_status, im;
1.57 schwarze 98:
1.66 schwarze 99: argi = 0;
100: if ((e = exprcomp(search, argc, argv, &argi)) == NULL) {
1.57 schwarze 101: *sz = 0;
1.58 schwarze 102: return 0;
1.57 schwarze 103: }
1.1 kristaps 104:
1.57 schwarze 105: cur = maxres = 0;
1.74 schwarze 106: if (res != NULL)
107: *res = NULL;
1.1 kristaps 108:
1.66 schwarze 109: outkey = KEY_Nd;
110: if (search->outkey != NULL)
111: for (im = 0; im < KEY_MAX; im++)
1.45 schwarze 112: if (0 == strcasecmp(search->outkey,
1.66 schwarze 113: mansearch_keynames[im])) {
114: outkey = im;
1.12 schwarze 115: break;
116: }
117:
1.1 kristaps 118: /*
1.57 schwarze 119: * Remember the original working directory, if possible.
120: * This will be needed if the second or a later directory
121: * is given as a relative path.
122: * Do not error out if the current directory is not
123: * searchable: Maybe it won't be needed after all.
1.1 kristaps 124: */
125:
1.57 schwarze 126: if (getcwd(buf, PATH_MAX) == NULL) {
127: getcwd_status = 0;
128: (void)strlcpy(buf, strerror(errno), sizeof(buf));
129: } else
130: getcwd_status = 1;
1.1 kristaps 131:
132: /*
133: * Loop over the directories (containing databases) for us to
134: * search.
135: * Don't let missing/bad databases/directories phase us.
136: * In each, try to open the resident database and, if it opens,
137: * scan it for our match expression.
138: */
139:
1.57 schwarze 140: chdir_status = 0;
1.1 kristaps 141: for (i = 0; i < paths->sz; i++) {
1.57 schwarze 142: if (chdir_status && paths->paths[i][0] != '/') {
143: if ( ! getcwd_status) {
1.59 schwarze 144: warnx("%s: getcwd: %s", paths->paths[i], buf);
1.57 schwarze 145: continue;
146: } else if (chdir(buf) == -1) {
1.64 schwarze 147: warn("%s", buf);
1.57 schwarze 148: continue;
149: }
150: }
151: if (chdir(paths->paths[i]) == -1) {
1.64 schwarze 152: warn("%s", paths->paths[i]);
1.1 kristaps 153: continue;
1.34 schwarze 154: }
1.57 schwarze 155: chdir_status = 1;
1.1 kristaps 156:
1.66 schwarze 157: if (dbm_open(MANDOC_DB) == -1) {
1.73 schwarze 158: if (errno != ENOENT)
159: warn("%s/%s", paths->paths[i], MANDOC_DB);
1.1 kristaps 160: continue;
161: }
162:
1.66 schwarze 163: if ((htab = manmerge(e, NULL)) == NULL) {
164: dbm_close();
165: continue;
1.1 kristaps 166: }
167:
1.66 schwarze 168: for (rp = ohash_first(htab, &slot); rp != NULL;
169: rp = ohash_next(htab, &slot)) {
170: page = dbm_page_get(rp->page);
1.1 kristaps 171:
1.66 schwarze 172: if (lstmatch(search->sec, page->sect) == 0 ||
1.76 schwarze 173: lstmatch(search->arch, page->arch) == 0 ||
174: (search->argmode == ARG_NAME &&
175: rp->bits <= (int32_t)(NAME_SYN & NAME_MASK)))
1.1 kristaps 176: continue;
177:
1.74 schwarze 178: if (res == NULL) {
179: cur = 1;
180: break;
181: }
1.1 kristaps 182: if (cur + 1 > maxres) {
183: maxres += 1024;
1.36 schwarze 184: *res = mandoc_reallocarray(*res,
1.66 schwarze 185: maxres, sizeof(**res));
1.1 kristaps 186: }
1.10 schwarze 187: mpage = *res + cur;
1.66 schwarze 188: mandoc_asprintf(&mpage->file, "%s/%s",
189: paths->paths[i], page->file + 1);
1.77 schwarze 190: if (access(chdir_status ? page->file + 1 :
191: mpage->file, R_OK) == -1) {
192: warn("%s", mpage->file);
193: warnx("outdated mandoc.db contains "
1.82 ! schwarze 194: "bogus %s entry, run makewhatis %s",
1.77 schwarze 195: page->file + 1, paths->paths[i]);
196: free(mpage->file);
197: free(rp);
198: continue;
199: }
1.66 schwarze 200: mpage->names = buildnames(page);
1.70 schwarze 201: mpage->output = buildoutput(outkey, page);
1.81 schwarze 202: mpage->bits = search->firstmatch ? rp->bits : 0;
1.47 schwarze 203: mpage->ipath = i;
1.66 schwarze 204: mpage->sec = *page->sect - '0';
205: if (mpage->sec < 0 || mpage->sec > 9)
206: mpage->sec = 10;
207: mpage->form = *page->file;
208: free(rp);
209: cur++;
210: }
211: ohash_delete(htab);
212: free(htab);
213: dbm_close();
1.49 schwarze 214:
215: /*
216: * In man(1) mode, prefer matches in earlier trees
217: * over matches in later trees.
218: */
219:
220: if (cur && search->firstmatch)
221: break;
1.1 kristaps 222: }
1.74 schwarze 223: if (res != NULL)
224: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.57 schwarze 225: if (chdir_status && getcwd_status && chdir(buf) == -1)
1.64 schwarze 226: warn("%s", buf);
1.1 kristaps 227: exprfree(e);
228: *sz = cur;
1.74 schwarze 229: return res != NULL || cur;
1.11 schwarze 230: }
231:
1.66 schwarze 232: /*
233: * Merge the results for the expression tree rooted at e
234: * into the the result list htab.
235: */
236: static struct ohash *
237: manmerge(struct expr *e, struct ohash *htab)
1.45 schwarze 238: {
1.66 schwarze 239: switch (e->type) {
240: case EXPR_TERM:
241: return manmerge_term(e, htab);
242: case EXPR_OR:
243: return manmerge_or(e->child, htab);
244: case EXPR_AND:
245: return manmerge_and(e->child, htab);
246: default:
247: abort();
1.45 schwarze 248: }
249: }
250:
1.66 schwarze 251: static struct ohash *
252: manmerge_term(struct expr *e, struct ohash *htab)
1.39 schwarze 253: {
1.66 schwarze 254: struct dbm_res res, *rp;
255: uint64_t ib;
256: unsigned int slot;
257: int im;
258:
259: if (htab == NULL) {
260: htab = mandoc_malloc(sizeof(*htab));
261: mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
262: }
263:
264: for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
265: if ((e->bits & ib) == 0)
266: continue;
267:
268: switch (ib) {
269: case TYPE_arch:
270: dbm_page_byarch(&e->match);
271: break;
272: case TYPE_sec:
273: dbm_page_bysect(&e->match);
274: break;
275: case TYPE_Nm:
276: dbm_page_byname(&e->match);
277: break;
278: case TYPE_Nd:
279: dbm_page_bydesc(&e->match);
280: break;
281: default:
282: dbm_page_bymacro(im - 2, &e->match);
283: break;
284: }
285:
286: /*
287: * When hashing for deduplication, use the unique
288: * page ID itself instead of a hash function;
289: * that is quite efficient.
290: */
1.39 schwarze 291:
1.66 schwarze 292: for (;;) {
293: res = dbm_page_next();
294: if (res.page == -1)
295: break;
296: slot = ohash_lookup_memory(htab,
297: (char *)&res, sizeof(res.page), res.page);
1.81 schwarze 298: if ((rp = ohash_find(htab, slot)) != NULL) {
299: rp->bits |= res.bits;
1.66 schwarze 300: continue;
1.81 schwarze 301: }
1.66 schwarze 302: rp = mandoc_malloc(sizeof(*rp));
303: *rp = res;
304: ohash_insert(htab, slot, rp);
305: }
306: }
307: return htab;
1.39 schwarze 308: }
309:
1.66 schwarze 310: static struct ohash *
311: manmerge_or(struct expr *e, struct ohash *htab)
312: {
313: while (e != NULL) {
314: htab = manmerge(e, htab);
315: e = e->next;
316: }
317: return htab;
318: }
1.22 schwarze 319:
1.66 schwarze 320: static struct ohash *
321: manmerge_and(struct expr *e, struct ohash *htab)
322: {
323: struct ohash *hand, *h1, *h2;
324: struct dbm_res *res;
325: unsigned int slot1, slot2;
1.22 schwarze 326:
1.66 schwarze 327: /* Evaluate the first term of the AND clause. */
1.39 schwarze 328:
1.66 schwarze 329: hand = manmerge(e, NULL);
1.39 schwarze 330:
1.66 schwarze 331: while ((e = e->next) != NULL) {
1.22 schwarze 332:
1.66 schwarze 333: /* Evaluate the next term and prepare for ANDing. */
1.22 schwarze 334:
1.66 schwarze 335: h2 = manmerge(e, NULL);
336: if (ohash_entries(h2) < ohash_entries(hand)) {
337: h1 = h2;
338: h2 = hand;
339: } else
340: h1 = hand;
341: hand = mandoc_malloc(sizeof(*hand));
342: mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
1.22 schwarze 343:
1.66 schwarze 344: /* Keep all pages that are in both result sets. */
1.22 schwarze 345:
1.66 schwarze 346: for (res = ohash_first(h1, &slot1); res != NULL;
347: res = ohash_next(h1, &slot1)) {
348: if (ohash_find(h2, ohash_lookup_memory(h2,
349: (char *)res, sizeof(res->page),
350: res->page)) == NULL)
351: free(res);
352: else
353: ohash_insert(hand, ohash_lookup_memory(hand,
354: (char *)res, sizeof(res->page),
355: res->page), res);
1.22 schwarze 356: }
357:
1.66 schwarze 358: /* Discard the merged results. */
1.22 schwarze 359:
1.66 schwarze 360: for (res = ohash_first(h2, &slot2); res != NULL;
361: res = ohash_next(h2, &slot2))
362: free(res);
363: ohash_delete(h2);
364: free(h2);
365: ohash_delete(h1);
366: free(h1);
367: }
1.17 schwarze 368:
1.66 schwarze 369: /* Merge the result of the AND into htab. */
1.17 schwarze 370:
1.66 schwarze 371: if (htab == NULL)
372: return hand;
373:
374: for (res = ohash_first(hand, &slot1); res != NULL;
375: res = ohash_next(hand, &slot1)) {
376: slot2 = ohash_lookup_memory(htab,
377: (char *)res, sizeof(res->page), res->page);
378: if (ohash_find(htab, slot2) == NULL)
379: ohash_insert(htab, slot2, res);
380: else
381: free(res);
382: }
383:
384: /* Discard the merged result. */
1.17 schwarze 385:
1.66 schwarze 386: ohash_delete(hand);
387: free(hand);
388: return htab;
389: }
1.54 schwarze 390:
1.66 schwarze 391: void
392: mansearch_free(struct manpage *res, size_t sz)
393: {
394: size_t i;
1.54 schwarze 395:
1.66 schwarze 396: for (i = 0; i < sz; i++) {
397: free(res[i].file);
398: free(res[i].names);
399: free(res[i].output);
1.22 schwarze 400: }
1.66 schwarze 401: free(res);
402: }
403:
404: static int
405: manpage_compare(const void *vp1, const void *vp2)
406: {
407: const struct manpage *mp1, *mp2;
1.75 schwarze 408: const char *cp1, *cp2;
409: size_t sz1, sz2;
1.66 schwarze 410: int diff;
411:
412: mp1 = vp1;
413: mp2 = vp2;
1.81 schwarze 414: if ((diff = mp2->bits - mp1->bits) ||
415: (diff = mp1->sec - mp2->sec))
1.75 schwarze 416: return diff;
417:
418: /* Fall back to alphabetic ordering of names. */
419: sz1 = strcspn(mp1->names, "(");
420: sz2 = strcspn(mp2->names, "(");
421: if (sz1 < sz2)
422: sz1 = sz2;
423: if ((diff = strncasecmp(mp1->names, mp2->names, sz1)))
424: return diff;
425:
426: /* For identical names and sections, prefer arch-dependent. */
427: cp1 = strchr(mp1->names + sz1, '/');
428: cp2 = strchr(mp2->names + sz2, '/');
429: return cp1 != NULL && cp2 != NULL ? strcasecmp(cp1, cp2) :
430: cp1 != NULL ? -1 : cp2 != NULL ? 1 : 0;
1.12 schwarze 431: }
432:
433: static char *
1.66 schwarze 434: buildnames(const struct dbm_page *page)
1.12 schwarze 435: {
1.66 schwarze 436: char *buf;
437: size_t i, sz;
1.12 schwarze 438:
1.70 schwarze 439: sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
440: (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
1.66 schwarze 441: buf = mandoc_malloc(sz);
442: i = 0;
1.70 schwarze 443: lstcat(buf, &i, page->name, ", ");
1.66 schwarze 444: buf[i++] = '(';
1.70 schwarze 445: lstcat(buf, &i, page->sect, ", ");
1.66 schwarze 446: if (page->arch != NULL) {
447: buf[i++] = '/';
1.70 schwarze 448: lstcat(buf, &i, page->arch, ", ");
1.66 schwarze 449: }
450: buf[i++] = ')';
451: buf[i++] = '\0';
452: assert(i == sz);
453: return buf;
1.1 kristaps 454: }
455:
456: /*
1.66 schwarze 457: * Count the buffer space needed to print the NUL-terminated
1.70 schwarze 458: * list of NUL-terminated strings, when printing sep separator
1.66 schwarze 459: * characters between strings.
1.7 schwarze 460: */
1.66 schwarze 461: static size_t
1.70 schwarze 462: lstlen(const char *cp, size_t sep)
1.7 schwarze 463: {
1.66 schwarze 464: size_t sz;
1.7 schwarze 465:
1.76 schwarze 466: for (sz = 0; *cp != '\0'; cp++) {
467:
468: /* Skip names appearing only in the SYNOPSIS. */
469: if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
470: while (*cp != '\0')
471: cp++;
472: continue;
473: }
474:
475: /* Skip name class markers. */
476: if (*cp < ' ')
477: cp++;
478:
479: /* Print a separator before each but the first string. */
480: if (sz)
481: sz += sep;
482:
483: /* Copy one string. */
484: while (*cp != '\0') {
485: sz++;
486: cp++;
487: }
1.66 schwarze 488: }
489: return sz;
1.7 schwarze 490: }
491:
492: /*
1.66 schwarze 493: * Print the NUL-terminated list of NUL-terminated strings
1.70 schwarze 494: * into the buffer, seperating strings with sep.
1.8 schwarze 495: */
496: static void
1.70 schwarze 497: lstcat(char *buf, size_t *i, const char *cp, const char *sep)
1.8 schwarze 498: {
1.76 schwarze 499: const char *s;
500: size_t i_start;
1.70 schwarze 501:
1.76 schwarze 502: for (i_start = *i; *cp != '\0'; cp++) {
503:
504: /* Skip names appearing only in the SYNOPSIS. */
505: if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
506: while (*cp != '\0')
507: cp++;
508: continue;
509: }
510:
511: /* Skip name class markers. */
512: if (*cp < ' ')
513: cp++;
514:
515: /* Print a separator before each but the first string. */
516: if (*i > i_start) {
1.70 schwarze 517: s = sep;
518: while (*s != '\0')
519: buf[(*i)++] = *s++;
1.76 schwarze 520: }
521:
522: /* Copy one string. */
523: while (*cp != '\0')
524: buf[(*i)++] = *cp++;
1.66 schwarze 525: }
1.76 schwarze 526:
1.8 schwarze 527: }
528:
1.66 schwarze 529: /*
530: * Return 1 if the string *want occurs in any of the strings
531: * in the NUL-terminated string list *have, or 0 otherwise.
532: * If either argument is NULL or empty, assume no filtering
533: * is desired and return 1.
534: */
535: static int
536: lstmatch(const char *want, const char *have)
1.13 schwarze 537: {
1.66 schwarze 538: if (want == NULL || have == NULL || *have == '\0')
539: return 1;
540: while (*have != '\0') {
541: if (strcasestr(have, want) != NULL)
542: return 1;
543: have = strchr(have, '\0') + 1;
544: }
545: return 0;
1.13 schwarze 546: }
547:
1.8 schwarze 548: /*
1.70 schwarze 549: * Build a list of values taken by the macro im in the manual page.
1.1 kristaps 550: */
551: static char *
1.70 schwarze 552: buildoutput(size_t im, struct dbm_page *page)
1.1 kristaps 553: {
1.70 schwarze 554: const char *oldoutput, *sep, *input;
1.66 schwarze 555: char *output, *newoutput, *value;
1.70 schwarze 556: size_t sz, i;
557:
558: switch (im) {
559: case KEY_Nd:
560: return mandoc_strdup(page->desc);
561: case KEY_Nm:
562: input = page->name;
563: break;
564: case KEY_sec:
565: input = page->sect;
566: break;
567: case KEY_arch:
568: input = page->arch;
569: if (input == NULL)
570: input = "all\0";
571: break;
572: default:
573: input = NULL;
574: break;
575: }
576:
577: if (input != NULL) {
578: sz = lstlen(input, 3) + 1;
579: output = mandoc_malloc(sz);
580: i = 0;
581: lstcat(output, &i, input, " # ");
1.71 schwarze 582: output[i++] = '\0';
583: assert(i == sz);
1.70 schwarze 584: return output;
585: }
1.66 schwarze 586:
587: output = NULL;
1.70 schwarze 588: dbm_macro_bypage(im - 2, page->addr);
1.66 schwarze 589: while ((value = dbm_macro_next()) != NULL) {
590: if (output == NULL) {
591: oldoutput = "";
592: sep = "";
593: } else {
594: oldoutput = output;
595: sep = " # ";
596: }
597: mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
598: free(output);
599: output = newoutput;
1.1 kristaps 600: }
1.66 schwarze 601: return output;
1.1 kristaps 602: }
603:
604: /*
605: * Compile a set of string tokens into an expression.
606: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
607: * "(", "foo=bar", etc.).
608: */
609: static struct expr *
1.66 schwarze 610: exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 611: {
1.66 schwarze 612: struct expr *parent, *child;
613: int needterm, nested;
614:
615: if ((nested = *argi) == argc)
616: return NULL;
617: needterm = 1;
618: parent = child = NULL;
619: while (*argi < argc) {
620: if (strcmp(")", argv[*argi]) == 0) {
621: if (needterm)
622: warnx("missing term "
623: "before closing parenthesis");
624: needterm = 0;
625: if (nested)
626: break;
627: warnx("ignoring unmatched right parenthesis");
628: ++*argi;
1.13 schwarze 629: continue;
1.66 schwarze 630: }
631: if (strcmp("-o", argv[*argi]) == 0) {
632: if (needterm) {
633: if (*argi > 0)
634: warnx("ignoring -o after %s",
635: argv[*argi - 1]);
636: else
637: warnx("ignoring initial -o");
638: }
639: needterm = 1;
640: ++*argi;
1.13 schwarze 641: continue;
1.66 schwarze 642: }
643: needterm = 0;
644: if (child == NULL) {
645: child = expr_and(search, argc, argv, argi);
1.13 schwarze 646: continue;
1.1 kristaps 647: }
1.66 schwarze 648: if (parent == NULL) {
649: parent = mandoc_calloc(1, sizeof(*parent));
650: parent->type = EXPR_OR;
651: parent->next = NULL;
652: parent->child = child;
653: }
654: child->next = expr_and(search, argc, argv, argi);
655: child = child->next;
656: }
657: if (needterm && *argi)
658: warnx("ignoring trailing %s", argv[*argi - 1]);
659: return parent == NULL ? child : parent;
660: }
1.26 schwarze 661:
1.66 schwarze 662: static struct expr *
663: expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
664: {
665: struct expr *parent, *child;
666: int needterm;
1.26 schwarze 667:
1.66 schwarze 668: needterm = 1;
669: parent = child = NULL;
670: while (*argi < argc) {
671: if (strcmp(")", argv[*argi]) == 0) {
672: if (needterm)
673: warnx("missing term "
674: "before closing parenthesis");
675: needterm = 0;
676: break;
677: }
678: if (strcmp("-o", argv[*argi]) == 0)
679: break;
680: if (strcmp("-a", argv[*argi]) == 0) {
681: if (needterm) {
682: if (*argi > 0)
683: warnx("ignoring -a after %s",
684: argv[*argi - 1]);
685: else
686: warnx("ignoring initial -a");
1.27 schwarze 687: }
1.66 schwarze 688: needterm = 1;
689: ++*argi;
690: continue;
1.27 schwarze 691: }
1.66 schwarze 692: if (needterm == 0)
693: break;
694: if (child == NULL) {
695: child = exprterm(search, argc, argv, argi);
696: if (child != NULL)
697: needterm = 0;
698: continue;
699: }
700: needterm = 0;
701: if (parent == NULL) {
702: parent = mandoc_calloc(1, sizeof(*parent));
703: parent->type = EXPR_AND;
704: parent->next = NULL;
705: parent->child = child;
706: }
707: child->next = exprterm(search, argc, argv, argi);
708: if (child->next != NULL) {
709: child = child->next;
710: needterm = 0;
711: }
712: }
713: if (needterm && *argi)
714: warnx("ignoring trailing %s", argv[*argi - 1]);
715: return parent == NULL ? child : parent;
1.15 schwarze 716: }
717:
718: static struct expr *
1.66 schwarze 719: exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 720: {
1.15 schwarze 721: char errbuf[BUFSIZ];
1.1 kristaps 722: struct expr *e;
1.38 schwarze 723: char *key, *val;
1.20 schwarze 724: uint64_t iterbit;
1.66 schwarze 725: int cs, i, irc;
1.1 kristaps 726:
1.66 schwarze 727: if (strcmp("(", argv[*argi]) == 0) {
728: ++*argi;
729: e = exprcomp(search, argc, argv, argi);
730: if (*argi < argc) {
731: assert(strcmp(")", argv[*argi]) == 0);
732: ++*argi;
733: } else
734: warnx("unclosed parenthesis");
735: return e;
736: }
1.1 kristaps 737:
1.72 schwarze 738: if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
739: cs = 0;
740: ++*argi;
741: } else
742: cs = 1;
743:
1.66 schwarze 744: e = mandoc_calloc(1, sizeof(*e));
745: e->type = EXPR_TERM;
746: e->bits = 0;
747: e->next = NULL;
748: e->child = NULL;
1.1 kristaps 749:
1.45 schwarze 750: if (search->argmode == ARG_NAME) {
751: e->bits = TYPE_Nm;
1.66 schwarze 752: e->match.type = DBM_EXACT;
753: e->match.str = argv[(*argi)++];
1.58 schwarze 754: return e;
1.5 kristaps 755: }
756:
1.1 kristaps 757: /*
1.45 schwarze 758: * Separate macro keys from search string.
1.66 schwarze 759: * If needed, request regular expression handling.
1.1 kristaps 760: */
761:
1.45 schwarze 762: if (search->argmode == ARG_WORD) {
763: e->bits = TYPE_Nm;
1.66 schwarze 764: e->match.type = DBM_REGEX;
1.61 schwarze 765: #if HAVE_REWB_BSD
1.66 schwarze 766: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
1.61 schwarze 767: #elif HAVE_REWB_SYSV
1.66 schwarze 768: mandoc_asprintf(&val, "\\<%s\\>", argv[*argi]);
1.61 schwarze 769: #else
770: mandoc_asprintf(&val,
1.66 schwarze 771: "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", argv[*argi]);
1.61 schwarze 772: #endif
1.46 schwarze 773: cs = 0;
1.66 schwarze 774: } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
1.45 schwarze 775: e->bits = TYPE_Nm | TYPE_Nd;
1.78 schwarze 776: e->match.type = DBM_REGEX;
777: val = argv[*argi];
778: cs = 0;
1.38 schwarze 779: } else {
1.66 schwarze 780: if (val == argv[*argi])
1.45 schwarze 781: e->bits = TYPE_Nm | TYPE_Nd;
1.66 schwarze 782: if (*val == '=') {
783: e->match.type = DBM_SUB;
784: e->match.str = val + 1;
785: } else
786: e->match.type = DBM_REGEX;
1.38 schwarze 787: *val++ = '\0';
1.66 schwarze 788: if (strstr(argv[*argi], "arch") != NULL)
1.21 schwarze 789: cs = 0;
1.38 schwarze 790: }
791:
792: /* Compile regular expressions. */
793:
1.66 schwarze 794: if (e->match.type == DBM_REGEX) {
795: e->match.re = mandoc_malloc(sizeof(*e->match.re));
796: irc = regcomp(e->match.re, val,
1.38 schwarze 797: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.66 schwarze 798: if (irc) {
799: regerror(irc, e->match.re, errbuf, sizeof(errbuf));
800: warnx("regcomp /%s/: %s", val, errbuf);
801: }
1.45 schwarze 802: if (search->argmode == ARG_WORD)
1.38 schwarze 803: free(val);
804: if (irc) {
1.66 schwarze 805: free(e->match.re);
1.8 schwarze 806: free(e);
1.66 schwarze 807: ++*argi;
1.58 schwarze 808: return NULL;
1.8 schwarze 809: }
1.38 schwarze 810: }
811:
1.66 schwarze 812: if (e->bits) {
813: ++*argi;
1.58 schwarze 814: return e;
1.66 schwarze 815: }
1.1 kristaps 816:
817: /*
818: * Parse out all possible fields.
819: * If the field doesn't resolve, bail.
820: */
821:
1.66 schwarze 822: while (NULL != (key = strsep(&argv[*argi], ","))) {
1.1 kristaps 823: if ('\0' == *key)
824: continue;
1.66 schwarze 825: for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
826: if (0 == strcasecmp(key, mansearch_keynames[i])) {
1.20 schwarze 827: e->bits |= iterbit;
828: break;
829: }
830: }
1.66 schwarze 831: if (i == KEY_MAX) {
832: if (strcasecmp(key, "any"))
833: warnx("treating unknown key "
834: "\"%s\" as \"any\"", key);
1.20 schwarze 835: e->bits |= ~0ULL;
1.1 kristaps 836: }
837: }
838:
1.66 schwarze 839: ++*argi;
1.58 schwarze 840: return e;
1.1 kristaps 841: }
842:
843: static void
1.66 schwarze 844: exprfree(struct expr *e)
1.1 kristaps 845: {
1.66 schwarze 846: if (e->next != NULL)
847: exprfree(e->next);
848: if (e->child != NULL)
849: exprfree(e->child);
850: free(e);
1.1 kristaps 851: }
CVSweb