Annotation of mandoc/mansearch.c, Revision 1.79
1.79 ! schwarze 1: /* $Id: mansearch.c,v 1.78 2018/11/19 19:27:37 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:
39: #include "mandoc.h"
1.23 schwarze 40: #include "mandoc_aux.h"
1.60 schwarze 41: #include "mandoc_ohash.h"
1.56 schwarze 42: #include "manconf.h"
1.1 kristaps 43: #include "mansearch.h"
1.66 schwarze 44: #include "dbm.h"
1.2 kristaps 45:
1.1 kristaps 46: struct expr {
1.66 schwarze 47: /* Used for terms: */
48: struct dbm_match match; /* Match type and expression. */
49: uint64_t bits; /* Type mask. */
50: /* Used for OR and AND groups: */
51: struct expr *next; /* Next child in the parent group. */
52: struct expr *child; /* First child in this group. */
53: enum { EXPR_TERM, EXPR_OR, EXPR_AND } type;
1.1 kristaps 54: };
55:
1.66 schwarze 56: const char *const mansearch_keynames[KEY_MAX] = {
57: "arch", "sec", "Xr", "Ar", "Fa", "Fl", "Dv", "Fn",
58: "Ic", "Pa", "Cm", "Li", "Em", "Cd", "Va", "Ft",
59: "Tn", "Er", "Ev", "Sy", "Sh", "In", "Ss", "Ox",
60: "An", "Mt", "St", "Bx", "At", "Nx", "Fx", "Lk",
61: "Ms", "Bsx", "Dx", "Rs", "Vt", "Lb", "Nm", "Nd"
1.1 kristaps 62: };
63:
1.66 schwarze 64:
65: static struct ohash *manmerge(struct expr *, struct ohash *);
66: static struct ohash *manmerge_term(struct expr *, struct ohash *);
67: static struct ohash *manmerge_or(struct expr *, struct ohash *);
68: static struct ohash *manmerge_and(struct expr *, struct ohash *);
69: static char *buildnames(const struct dbm_page *);
1.70 schwarze 70: static char *buildoutput(size_t, struct dbm_page *);
71: static size_t lstlen(const char *, size_t);
72: static void lstcat(char *, size_t *, const char *, const char *);
1.66 schwarze 73: static int lstmatch(const char *, const char *);
1.34 schwarze 74: static struct expr *exprcomp(const struct mansearch *,
1.66 schwarze 75: int, char *[], int *);
76: static struct expr *expr_and(const struct mansearch *,
77: int, char *[], int *);
78: static struct expr *exprterm(const struct mansearch *,
79: int, char *[], int *);
1.1 kristaps 80: static void exprfree(struct expr *);
1.39 schwarze 81: static int manpage_compare(const void *, const void *);
1.28 schwarze 82:
1.1 kristaps 83:
84: int
1.5 kristaps 85: mansearch(const struct mansearch *search,
1.12 schwarze 86: const struct manpaths *paths,
87: int argc, char *argv[],
1.1 kristaps 88: struct manpage **res, size_t *sz)
89: {
1.6 schwarze 90: char buf[PATH_MAX];
1.66 schwarze 91: struct dbm_res *rp;
92: struct expr *e;
93: struct dbm_page *page;
1.10 schwarze 94: struct manpage *mpage;
1.66 schwarze 95: struct ohash *htab;
96: size_t cur, i, maxres, outkey;
97: unsigned int slot;
98: int argi, chdir_status, getcwd_status, im;
1.57 schwarze 99:
1.66 schwarze 100: argi = 0;
101: if ((e = exprcomp(search, argc, argv, &argi)) == NULL) {
1.57 schwarze 102: *sz = 0;
1.58 schwarze 103: return 0;
1.57 schwarze 104: }
1.1 kristaps 105:
1.57 schwarze 106: cur = maxres = 0;
1.74 schwarze 107: if (res != NULL)
108: *res = NULL;
1.1 kristaps 109:
1.66 schwarze 110: outkey = KEY_Nd;
111: if (search->outkey != NULL)
112: for (im = 0; im < KEY_MAX; im++)
1.45 schwarze 113: if (0 == strcasecmp(search->outkey,
1.66 schwarze 114: mansearch_keynames[im])) {
115: outkey = im;
1.12 schwarze 116: break;
117: }
118:
1.1 kristaps 119: /*
1.57 schwarze 120: * Remember the original working directory, if possible.
121: * This will be needed if the second or a later directory
122: * is given as a relative path.
123: * Do not error out if the current directory is not
124: * searchable: Maybe it won't be needed after all.
1.1 kristaps 125: */
126:
1.57 schwarze 127: if (getcwd(buf, PATH_MAX) == NULL) {
128: getcwd_status = 0;
129: (void)strlcpy(buf, strerror(errno), sizeof(buf));
130: } else
131: getcwd_status = 1;
1.1 kristaps 132:
133: /*
134: * Loop over the directories (containing databases) for us to
135: * search.
136: * Don't let missing/bad databases/directories phase us.
137: * In each, try to open the resident database and, if it opens,
138: * scan it for our match expression.
139: */
140:
1.57 schwarze 141: chdir_status = 0;
1.1 kristaps 142: for (i = 0; i < paths->sz; i++) {
1.57 schwarze 143: if (chdir_status && paths->paths[i][0] != '/') {
144: if ( ! getcwd_status) {
1.59 schwarze 145: warnx("%s: getcwd: %s", paths->paths[i], buf);
1.57 schwarze 146: continue;
147: } else if (chdir(buf) == -1) {
1.64 schwarze 148: warn("%s", buf);
1.57 schwarze 149: continue;
150: }
151: }
152: if (chdir(paths->paths[i]) == -1) {
1.64 schwarze 153: warn("%s", paths->paths[i]);
1.1 kristaps 154: continue;
1.34 schwarze 155: }
1.57 schwarze 156: chdir_status = 1;
1.1 kristaps 157:
1.66 schwarze 158: if (dbm_open(MANDOC_DB) == -1) {
1.73 schwarze 159: if (errno != ENOENT)
160: warn("%s/%s", paths->paths[i], MANDOC_DB);
1.1 kristaps 161: continue;
162: }
163:
1.66 schwarze 164: if ((htab = manmerge(e, NULL)) == NULL) {
165: dbm_close();
166: continue;
1.1 kristaps 167: }
168:
1.66 schwarze 169: for (rp = ohash_first(htab, &slot); rp != NULL;
170: rp = ohash_next(htab, &slot)) {
171: page = dbm_page_get(rp->page);
1.1 kristaps 172:
1.66 schwarze 173: if (lstmatch(search->sec, page->sect) == 0 ||
1.76 schwarze 174: lstmatch(search->arch, page->arch) == 0 ||
175: (search->argmode == ARG_NAME &&
176: rp->bits <= (int32_t)(NAME_SYN & NAME_MASK)))
1.1 kristaps 177: continue;
178:
1.74 schwarze 179: if (res == NULL) {
180: cur = 1;
181: break;
182: }
1.1 kristaps 183: if (cur + 1 > maxres) {
184: maxres += 1024;
1.36 schwarze 185: *res = mandoc_reallocarray(*res,
1.66 schwarze 186: maxres, sizeof(**res));
1.1 kristaps 187: }
1.10 schwarze 188: mpage = *res + cur;
1.66 schwarze 189: mandoc_asprintf(&mpage->file, "%s/%s",
190: paths->paths[i], page->file + 1);
1.77 schwarze 191: if (access(chdir_status ? page->file + 1 :
192: mpage->file, R_OK) == -1) {
193: warn("%s", mpage->file);
194: warnx("outdated mandoc.db contains "
195: "bogus %s entry, run makewhatis %s",
196: page->file + 1, paths->paths[i]);
197: free(mpage->file);
198: free(rp);
199: continue;
200: }
1.66 schwarze 201: mpage->names = buildnames(page);
1.70 schwarze 202: mpage->output = buildoutput(outkey, page);
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);
1.79 ! schwarze 298: if ((rp = ohash_find(htab, slot)) != NULL)
1.66 schwarze 299: continue;
300: rp = mandoc_malloc(sizeof(*rp));
301: *rp = res;
302: ohash_insert(htab, slot, rp);
303: }
304: }
305: return htab;
1.39 schwarze 306: }
307:
1.66 schwarze 308: static struct ohash *
309: manmerge_or(struct expr *e, struct ohash *htab)
310: {
311: while (e != NULL) {
312: htab = manmerge(e, htab);
313: e = e->next;
314: }
315: return htab;
316: }
1.22 schwarze 317:
1.66 schwarze 318: static struct ohash *
319: manmerge_and(struct expr *e, struct ohash *htab)
320: {
321: struct ohash *hand, *h1, *h2;
322: struct dbm_res *res;
323: unsigned int slot1, slot2;
1.22 schwarze 324:
1.66 schwarze 325: /* Evaluate the first term of the AND clause. */
1.39 schwarze 326:
1.66 schwarze 327: hand = manmerge(e, NULL);
1.39 schwarze 328:
1.66 schwarze 329: while ((e = e->next) != NULL) {
1.22 schwarze 330:
1.66 schwarze 331: /* Evaluate the next term and prepare for ANDing. */
1.22 schwarze 332:
1.66 schwarze 333: h2 = manmerge(e, NULL);
334: if (ohash_entries(h2) < ohash_entries(hand)) {
335: h1 = h2;
336: h2 = hand;
337: } else
338: h1 = hand;
339: hand = mandoc_malloc(sizeof(*hand));
340: mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
1.22 schwarze 341:
1.66 schwarze 342: /* Keep all pages that are in both result sets. */
1.22 schwarze 343:
1.66 schwarze 344: for (res = ohash_first(h1, &slot1); res != NULL;
345: res = ohash_next(h1, &slot1)) {
346: if (ohash_find(h2, ohash_lookup_memory(h2,
347: (char *)res, sizeof(res->page),
348: res->page)) == NULL)
349: free(res);
350: else
351: ohash_insert(hand, ohash_lookup_memory(hand,
352: (char *)res, sizeof(res->page),
353: res->page), res);
1.22 schwarze 354: }
355:
1.66 schwarze 356: /* Discard the merged results. */
1.22 schwarze 357:
1.66 schwarze 358: for (res = ohash_first(h2, &slot2); res != NULL;
359: res = ohash_next(h2, &slot2))
360: free(res);
361: ohash_delete(h2);
362: free(h2);
363: ohash_delete(h1);
364: free(h1);
365: }
1.17 schwarze 366:
1.66 schwarze 367: /* Merge the result of the AND into htab. */
1.17 schwarze 368:
1.66 schwarze 369: if (htab == NULL)
370: return hand;
371:
372: for (res = ohash_first(hand, &slot1); res != NULL;
373: res = ohash_next(hand, &slot1)) {
374: slot2 = ohash_lookup_memory(htab,
375: (char *)res, sizeof(res->page), res->page);
376: if (ohash_find(htab, slot2) == NULL)
377: ohash_insert(htab, slot2, res);
378: else
379: free(res);
380: }
381:
382: /* Discard the merged result. */
1.17 schwarze 383:
1.66 schwarze 384: ohash_delete(hand);
385: free(hand);
386: return htab;
387: }
1.54 schwarze 388:
1.66 schwarze 389: void
390: mansearch_free(struct manpage *res, size_t sz)
391: {
392: size_t i;
1.54 schwarze 393:
1.66 schwarze 394: for (i = 0; i < sz; i++) {
395: free(res[i].file);
396: free(res[i].names);
397: free(res[i].output);
1.22 schwarze 398: }
1.66 schwarze 399: free(res);
400: }
401:
402: static int
403: manpage_compare(const void *vp1, const void *vp2)
404: {
405: const struct manpage *mp1, *mp2;
1.75 schwarze 406: const char *cp1, *cp2;
407: size_t sz1, sz2;
1.66 schwarze 408: int diff;
409:
410: mp1 = vp1;
411: mp2 = vp2;
1.79 ! schwarze 412: if ((diff = mp1->sec - mp2->sec))
1.75 schwarze 413: return diff;
414:
415: /* Fall back to alphabetic ordering of names. */
416: sz1 = strcspn(mp1->names, "(");
417: sz2 = strcspn(mp2->names, "(");
418: if (sz1 < sz2)
419: sz1 = sz2;
420: if ((diff = strncasecmp(mp1->names, mp2->names, sz1)))
421: return diff;
422:
423: /* For identical names and sections, prefer arch-dependent. */
424: cp1 = strchr(mp1->names + sz1, '/');
425: cp2 = strchr(mp2->names + sz2, '/');
426: return cp1 != NULL && cp2 != NULL ? strcasecmp(cp1, cp2) :
427: cp1 != NULL ? -1 : cp2 != NULL ? 1 : 0;
1.12 schwarze 428: }
429:
430: static char *
1.66 schwarze 431: buildnames(const struct dbm_page *page)
1.12 schwarze 432: {
1.66 schwarze 433: char *buf;
434: size_t i, sz;
1.12 schwarze 435:
1.70 schwarze 436: sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
437: (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
1.66 schwarze 438: buf = mandoc_malloc(sz);
439: i = 0;
1.70 schwarze 440: lstcat(buf, &i, page->name, ", ");
1.66 schwarze 441: buf[i++] = '(';
1.70 schwarze 442: lstcat(buf, &i, page->sect, ", ");
1.66 schwarze 443: if (page->arch != NULL) {
444: buf[i++] = '/';
1.70 schwarze 445: lstcat(buf, &i, page->arch, ", ");
1.66 schwarze 446: }
447: buf[i++] = ')';
448: buf[i++] = '\0';
449: assert(i == sz);
450: return buf;
1.1 kristaps 451: }
452:
453: /*
1.66 schwarze 454: * Count the buffer space needed to print the NUL-terminated
1.70 schwarze 455: * list of NUL-terminated strings, when printing sep separator
1.66 schwarze 456: * characters between strings.
1.7 schwarze 457: */
1.66 schwarze 458: static size_t
1.70 schwarze 459: lstlen(const char *cp, size_t sep)
1.7 schwarze 460: {
1.66 schwarze 461: size_t sz;
1.7 schwarze 462:
1.76 schwarze 463: for (sz = 0; *cp != '\0'; cp++) {
464:
465: /* Skip names appearing only in the SYNOPSIS. */
466: if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
467: while (*cp != '\0')
468: cp++;
469: continue;
470: }
471:
472: /* Skip name class markers. */
473: if (*cp < ' ')
474: cp++;
475:
476: /* Print a separator before each but the first string. */
477: if (sz)
478: sz += sep;
479:
480: /* Copy one string. */
481: while (*cp != '\0') {
482: sz++;
483: cp++;
484: }
1.66 schwarze 485: }
486: return sz;
1.7 schwarze 487: }
488:
489: /*
1.66 schwarze 490: * Print the NUL-terminated list of NUL-terminated strings
1.70 schwarze 491: * into the buffer, seperating strings with sep.
1.8 schwarze 492: */
493: static void
1.70 schwarze 494: lstcat(char *buf, size_t *i, const char *cp, const char *sep)
1.8 schwarze 495: {
1.76 schwarze 496: const char *s;
497: size_t i_start;
1.70 schwarze 498:
1.76 schwarze 499: for (i_start = *i; *cp != '\0'; cp++) {
500:
501: /* Skip names appearing only in the SYNOPSIS. */
502: if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
503: while (*cp != '\0')
504: cp++;
505: continue;
506: }
507:
508: /* Skip name class markers. */
509: if (*cp < ' ')
510: cp++;
511:
512: /* Print a separator before each but the first string. */
513: if (*i > i_start) {
1.70 schwarze 514: s = sep;
515: while (*s != '\0')
516: buf[(*i)++] = *s++;
1.76 schwarze 517: }
518:
519: /* Copy one string. */
520: while (*cp != '\0')
521: buf[(*i)++] = *cp++;
1.66 schwarze 522: }
1.76 schwarze 523:
1.8 schwarze 524: }
525:
1.66 schwarze 526: /*
527: * Return 1 if the string *want occurs in any of the strings
528: * in the NUL-terminated string list *have, or 0 otherwise.
529: * If either argument is NULL or empty, assume no filtering
530: * is desired and return 1.
531: */
532: static int
533: lstmatch(const char *want, const char *have)
1.13 schwarze 534: {
1.66 schwarze 535: if (want == NULL || have == NULL || *have == '\0')
536: return 1;
537: while (*have != '\0') {
538: if (strcasestr(have, want) != NULL)
539: return 1;
540: have = strchr(have, '\0') + 1;
541: }
542: return 0;
1.13 schwarze 543: }
544:
1.8 schwarze 545: /*
1.70 schwarze 546: * Build a list of values taken by the macro im in the manual page.
1.1 kristaps 547: */
548: static char *
1.70 schwarze 549: buildoutput(size_t im, struct dbm_page *page)
1.1 kristaps 550: {
1.70 schwarze 551: const char *oldoutput, *sep, *input;
1.66 schwarze 552: char *output, *newoutput, *value;
1.70 schwarze 553: size_t sz, i;
554:
555: switch (im) {
556: case KEY_Nd:
557: return mandoc_strdup(page->desc);
558: case KEY_Nm:
559: input = page->name;
560: break;
561: case KEY_sec:
562: input = page->sect;
563: break;
564: case KEY_arch:
565: input = page->arch;
566: if (input == NULL)
567: input = "all\0";
568: break;
569: default:
570: input = NULL;
571: break;
572: }
573:
574: if (input != NULL) {
575: sz = lstlen(input, 3) + 1;
576: output = mandoc_malloc(sz);
577: i = 0;
578: lstcat(output, &i, input, " # ");
1.71 schwarze 579: output[i++] = '\0';
580: assert(i == sz);
1.70 schwarze 581: return output;
582: }
1.66 schwarze 583:
584: output = NULL;
1.70 schwarze 585: dbm_macro_bypage(im - 2, page->addr);
1.66 schwarze 586: while ((value = dbm_macro_next()) != NULL) {
587: if (output == NULL) {
588: oldoutput = "";
589: sep = "";
590: } else {
591: oldoutput = output;
592: sep = " # ";
593: }
594: mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
595: free(output);
596: output = newoutput;
1.1 kristaps 597: }
1.66 schwarze 598: return output;
1.1 kristaps 599: }
600:
601: /*
602: * Compile a set of string tokens into an expression.
603: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
604: * "(", "foo=bar", etc.).
605: */
606: static struct expr *
1.66 schwarze 607: exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 608: {
1.66 schwarze 609: struct expr *parent, *child;
610: int needterm, nested;
611:
612: if ((nested = *argi) == argc)
613: return NULL;
614: needterm = 1;
615: parent = child = NULL;
616: while (*argi < argc) {
617: if (strcmp(")", argv[*argi]) == 0) {
618: if (needterm)
619: warnx("missing term "
620: "before closing parenthesis");
621: needterm = 0;
622: if (nested)
623: break;
624: warnx("ignoring unmatched right parenthesis");
625: ++*argi;
1.13 schwarze 626: continue;
1.66 schwarze 627: }
628: if (strcmp("-o", argv[*argi]) == 0) {
629: if (needterm) {
630: if (*argi > 0)
631: warnx("ignoring -o after %s",
632: argv[*argi - 1]);
633: else
634: warnx("ignoring initial -o");
635: }
636: needterm = 1;
637: ++*argi;
1.13 schwarze 638: continue;
1.66 schwarze 639: }
640: needterm = 0;
641: if (child == NULL) {
642: child = expr_and(search, argc, argv, argi);
1.13 schwarze 643: continue;
1.1 kristaps 644: }
1.66 schwarze 645: if (parent == NULL) {
646: parent = mandoc_calloc(1, sizeof(*parent));
647: parent->type = EXPR_OR;
648: parent->next = NULL;
649: parent->child = child;
650: }
651: child->next = expr_and(search, argc, argv, argi);
652: child = child->next;
653: }
654: if (needterm && *argi)
655: warnx("ignoring trailing %s", argv[*argi - 1]);
656: return parent == NULL ? child : parent;
657: }
1.26 schwarze 658:
1.66 schwarze 659: static struct expr *
660: expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
661: {
662: struct expr *parent, *child;
663: int needterm;
1.26 schwarze 664:
1.66 schwarze 665: needterm = 1;
666: parent = child = NULL;
667: while (*argi < argc) {
668: if (strcmp(")", argv[*argi]) == 0) {
669: if (needterm)
670: warnx("missing term "
671: "before closing parenthesis");
672: needterm = 0;
673: break;
674: }
675: if (strcmp("-o", argv[*argi]) == 0)
676: break;
677: if (strcmp("-a", argv[*argi]) == 0) {
678: if (needterm) {
679: if (*argi > 0)
680: warnx("ignoring -a after %s",
681: argv[*argi - 1]);
682: else
683: warnx("ignoring initial -a");
1.27 schwarze 684: }
1.66 schwarze 685: needterm = 1;
686: ++*argi;
687: continue;
1.27 schwarze 688: }
1.66 schwarze 689: if (needterm == 0)
690: break;
691: if (child == NULL) {
692: child = exprterm(search, argc, argv, argi);
693: if (child != NULL)
694: needterm = 0;
695: continue;
696: }
697: needterm = 0;
698: if (parent == NULL) {
699: parent = mandoc_calloc(1, sizeof(*parent));
700: parent->type = EXPR_AND;
701: parent->next = NULL;
702: parent->child = child;
703: }
704: child->next = exprterm(search, argc, argv, argi);
705: if (child->next != NULL) {
706: child = child->next;
707: needterm = 0;
708: }
709: }
710: if (needterm && *argi)
711: warnx("ignoring trailing %s", argv[*argi - 1]);
712: return parent == NULL ? child : parent;
1.15 schwarze 713: }
714:
715: static struct expr *
1.66 schwarze 716: exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 717: {
1.15 schwarze 718: char errbuf[BUFSIZ];
1.1 kristaps 719: struct expr *e;
1.38 schwarze 720: char *key, *val;
1.20 schwarze 721: uint64_t iterbit;
1.66 schwarze 722: int cs, i, irc;
1.1 kristaps 723:
1.66 schwarze 724: if (strcmp("(", argv[*argi]) == 0) {
725: ++*argi;
726: e = exprcomp(search, argc, argv, argi);
727: if (*argi < argc) {
728: assert(strcmp(")", argv[*argi]) == 0);
729: ++*argi;
730: } else
731: warnx("unclosed parenthesis");
732: return e;
733: }
1.1 kristaps 734:
1.72 schwarze 735: if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
736: cs = 0;
737: ++*argi;
738: } else
739: cs = 1;
740:
1.66 schwarze 741: e = mandoc_calloc(1, sizeof(*e));
742: e->type = EXPR_TERM;
743: e->bits = 0;
744: e->next = NULL;
745: e->child = NULL;
1.1 kristaps 746:
1.45 schwarze 747: if (search->argmode == ARG_NAME) {
748: e->bits = TYPE_Nm;
1.66 schwarze 749: e->match.type = DBM_EXACT;
750: e->match.str = argv[(*argi)++];
1.58 schwarze 751: return e;
1.5 kristaps 752: }
753:
1.1 kristaps 754: /*
1.45 schwarze 755: * Separate macro keys from search string.
1.66 schwarze 756: * If needed, request regular expression handling.
1.1 kristaps 757: */
758:
1.45 schwarze 759: if (search->argmode == ARG_WORD) {
760: e->bits = TYPE_Nm;
1.66 schwarze 761: e->match.type = DBM_REGEX;
1.61 schwarze 762: #if HAVE_REWB_BSD
1.66 schwarze 763: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
1.61 schwarze 764: #elif HAVE_REWB_SYSV
1.66 schwarze 765: mandoc_asprintf(&val, "\\<%s\\>", argv[*argi]);
1.61 schwarze 766: #else
767: mandoc_asprintf(&val,
1.66 schwarze 768: "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", argv[*argi]);
1.61 schwarze 769: #endif
1.46 schwarze 770: cs = 0;
1.66 schwarze 771: } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
1.45 schwarze 772: e->bits = TYPE_Nm | TYPE_Nd;
1.78 schwarze 773: e->match.type = DBM_REGEX;
774: val = argv[*argi];
775: cs = 0;
1.38 schwarze 776: } else {
1.66 schwarze 777: if (val == argv[*argi])
1.45 schwarze 778: e->bits = TYPE_Nm | TYPE_Nd;
1.66 schwarze 779: if (*val == '=') {
780: e->match.type = DBM_SUB;
781: e->match.str = val + 1;
782: } else
783: e->match.type = DBM_REGEX;
1.38 schwarze 784: *val++ = '\0';
1.66 schwarze 785: if (strstr(argv[*argi], "arch") != NULL)
1.21 schwarze 786: cs = 0;
1.38 schwarze 787: }
788:
789: /* Compile regular expressions. */
790:
1.66 schwarze 791: if (e->match.type == DBM_REGEX) {
792: e->match.re = mandoc_malloc(sizeof(*e->match.re));
793: irc = regcomp(e->match.re, val,
1.38 schwarze 794: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.66 schwarze 795: if (irc) {
796: regerror(irc, e->match.re, errbuf, sizeof(errbuf));
797: warnx("regcomp /%s/: %s", val, errbuf);
798: }
1.45 schwarze 799: if (search->argmode == ARG_WORD)
1.38 schwarze 800: free(val);
801: if (irc) {
1.66 schwarze 802: free(e->match.re);
1.8 schwarze 803: free(e);
1.66 schwarze 804: ++*argi;
1.58 schwarze 805: return NULL;
1.8 schwarze 806: }
1.38 schwarze 807: }
808:
1.66 schwarze 809: if (e->bits) {
810: ++*argi;
1.58 schwarze 811: return e;
1.66 schwarze 812: }
1.1 kristaps 813:
814: /*
815: * Parse out all possible fields.
816: * If the field doesn't resolve, bail.
817: */
818:
1.66 schwarze 819: while (NULL != (key = strsep(&argv[*argi], ","))) {
1.1 kristaps 820: if ('\0' == *key)
821: continue;
1.66 schwarze 822: for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
823: if (0 == strcasecmp(key, mansearch_keynames[i])) {
1.20 schwarze 824: e->bits |= iterbit;
825: break;
826: }
827: }
1.66 schwarze 828: if (i == KEY_MAX) {
829: if (strcasecmp(key, "any"))
830: warnx("treating unknown key "
831: "\"%s\" as \"any\"", key);
1.20 schwarze 832: e->bits |= ~0ULL;
1.1 kristaps 833: }
834: }
835:
1.66 schwarze 836: ++*argi;
1.58 schwarze 837: return e;
1.1 kristaps 838: }
839:
840: static void
1.66 schwarze 841: exprfree(struct expr *e)
1.1 kristaps 842: {
1.66 schwarze 843: if (e->next != NULL)
844: exprfree(e->next);
845: if (e->child != NULL)
846: exprfree(e->child);
847: free(e);
1.1 kristaps 848: }
CVSweb