Annotation of mandoc/mansearch.c, Revision 1.73
1.66 schwarze 1: /* $OpenBSD: mansearch.c,v 1.50 2016/07/09 15:23:36 schwarze Exp $ */
1.1 kristaps 2: /*
3: * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.70 schwarze 4: * Copyright (c) 2013-2017 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;
1.1 kristaps 107: *res = NULL;
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 ||
173: lstmatch(search->arch, page->arch) == 0)
1.1 kristaps 174: continue;
175:
176: if (cur + 1 > maxres) {
177: maxres += 1024;
1.36 schwarze 178: *res = mandoc_reallocarray(*res,
1.66 schwarze 179: maxres, sizeof(**res));
1.1 kristaps 180: }
1.10 schwarze 181: mpage = *res + cur;
1.66 schwarze 182: mandoc_asprintf(&mpage->file, "%s/%s",
183: paths->paths[i], page->file + 1);
184: mpage->names = buildnames(page);
1.70 schwarze 185: mpage->output = buildoutput(outkey, page);
1.47 schwarze 186: mpage->ipath = i;
1.66 schwarze 187: mpage->bits = rp->bits;
188: mpage->sec = *page->sect - '0';
189: if (mpage->sec < 0 || mpage->sec > 9)
190: mpage->sec = 10;
191: mpage->form = *page->file;
192: free(rp);
193: cur++;
194: }
195: ohash_delete(htab);
196: free(htab);
197: dbm_close();
1.49 schwarze 198:
199: /*
200: * In man(1) mode, prefer matches in earlier trees
201: * over matches in later trees.
202: */
203:
204: if (cur && search->firstmatch)
205: break;
1.1 kristaps 206: }
1.39 schwarze 207: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.57 schwarze 208: if (chdir_status && getcwd_status && chdir(buf) == -1)
1.64 schwarze 209: warn("%s", buf);
1.1 kristaps 210: exprfree(e);
211: *sz = cur;
1.58 schwarze 212: return 1;
1.11 schwarze 213: }
214:
1.66 schwarze 215: /*
216: * Merge the results for the expression tree rooted at e
217: * into the the result list htab.
218: */
219: static struct ohash *
220: manmerge(struct expr *e, struct ohash *htab)
1.45 schwarze 221: {
1.66 schwarze 222: switch (e->type) {
223: case EXPR_TERM:
224: return manmerge_term(e, htab);
225: case EXPR_OR:
226: return manmerge_or(e->child, htab);
227: case EXPR_AND:
228: return manmerge_and(e->child, htab);
229: default:
230: abort();
1.45 schwarze 231: }
232: }
233:
1.66 schwarze 234: static struct ohash *
235: manmerge_term(struct expr *e, struct ohash *htab)
1.39 schwarze 236: {
1.66 schwarze 237: struct dbm_res res, *rp;
238: uint64_t ib;
239: unsigned int slot;
240: int im;
241:
242: if (htab == NULL) {
243: htab = mandoc_malloc(sizeof(*htab));
244: mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
245: }
246:
247: for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
248: if ((e->bits & ib) == 0)
249: continue;
250:
251: switch (ib) {
252: case TYPE_arch:
253: dbm_page_byarch(&e->match);
254: break;
255: case TYPE_sec:
256: dbm_page_bysect(&e->match);
257: break;
258: case TYPE_Nm:
259: dbm_page_byname(&e->match);
260: break;
261: case TYPE_Nd:
262: dbm_page_bydesc(&e->match);
263: break;
264: default:
265: dbm_page_bymacro(im - 2, &e->match);
266: break;
267: }
268:
269: /*
270: * When hashing for deduplication, use the unique
271: * page ID itself instead of a hash function;
272: * that is quite efficient.
273: */
1.39 schwarze 274:
1.66 schwarze 275: for (;;) {
276: res = dbm_page_next();
277: if (res.page == -1)
278: break;
279: slot = ohash_lookup_memory(htab,
280: (char *)&res, sizeof(res.page), res.page);
281: if ((rp = ohash_find(htab, slot)) != NULL) {
282: rp->bits |= res.bits;
283: continue;
284: }
285: rp = mandoc_malloc(sizeof(*rp));
286: *rp = res;
287: ohash_insert(htab, slot, rp);
288: }
289: }
290: return htab;
1.39 schwarze 291: }
292:
1.66 schwarze 293: static struct ohash *
294: manmerge_or(struct expr *e, struct ohash *htab)
295: {
296: while (e != NULL) {
297: htab = manmerge(e, htab);
298: e = e->next;
299: }
300: return htab;
301: }
1.22 schwarze 302:
1.66 schwarze 303: static struct ohash *
304: manmerge_and(struct expr *e, struct ohash *htab)
305: {
306: struct ohash *hand, *h1, *h2;
307: struct dbm_res *res;
308: unsigned int slot1, slot2;
1.22 schwarze 309:
1.66 schwarze 310: /* Evaluate the first term of the AND clause. */
1.39 schwarze 311:
1.66 schwarze 312: hand = manmerge(e, NULL);
1.39 schwarze 313:
1.66 schwarze 314: while ((e = e->next) != NULL) {
1.22 schwarze 315:
1.66 schwarze 316: /* Evaluate the next term and prepare for ANDing. */
1.22 schwarze 317:
1.66 schwarze 318: h2 = manmerge(e, NULL);
319: if (ohash_entries(h2) < ohash_entries(hand)) {
320: h1 = h2;
321: h2 = hand;
322: } else
323: h1 = hand;
324: hand = mandoc_malloc(sizeof(*hand));
325: mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
1.22 schwarze 326:
1.66 schwarze 327: /* Keep all pages that are in both result sets. */
1.22 schwarze 328:
1.66 schwarze 329: for (res = ohash_first(h1, &slot1); res != NULL;
330: res = ohash_next(h1, &slot1)) {
331: if (ohash_find(h2, ohash_lookup_memory(h2,
332: (char *)res, sizeof(res->page),
333: res->page)) == NULL)
334: free(res);
335: else
336: ohash_insert(hand, ohash_lookup_memory(hand,
337: (char *)res, sizeof(res->page),
338: res->page), res);
1.22 schwarze 339: }
340:
1.66 schwarze 341: /* Discard the merged results. */
1.22 schwarze 342:
1.66 schwarze 343: for (res = ohash_first(h2, &slot2); res != NULL;
344: res = ohash_next(h2, &slot2))
345: free(res);
346: ohash_delete(h2);
347: free(h2);
348: ohash_delete(h1);
349: free(h1);
350: }
1.17 schwarze 351:
1.66 schwarze 352: /* Merge the result of the AND into htab. */
1.17 schwarze 353:
1.66 schwarze 354: if (htab == NULL)
355: return hand;
356:
357: for (res = ohash_first(hand, &slot1); res != NULL;
358: res = ohash_next(hand, &slot1)) {
359: slot2 = ohash_lookup_memory(htab,
360: (char *)res, sizeof(res->page), res->page);
361: if (ohash_find(htab, slot2) == NULL)
362: ohash_insert(htab, slot2, res);
363: else
364: free(res);
365: }
366:
367: /* Discard the merged result. */
1.17 schwarze 368:
1.66 schwarze 369: ohash_delete(hand);
370: free(hand);
371: return htab;
372: }
1.54 schwarze 373:
1.66 schwarze 374: void
375: mansearch_free(struct manpage *res, size_t sz)
376: {
377: size_t i;
1.54 schwarze 378:
1.66 schwarze 379: for (i = 0; i < sz; i++) {
380: free(res[i].file);
381: free(res[i].names);
382: free(res[i].output);
1.22 schwarze 383: }
1.66 schwarze 384: free(res);
385: }
386:
387: static int
388: manpage_compare(const void *vp1, const void *vp2)
389: {
390: const struct manpage *mp1, *mp2;
391: int diff;
392:
393: mp1 = vp1;
394: mp2 = vp2;
395: return (diff = mp2->bits - mp1->bits) ? diff :
396: (diff = mp1->sec - mp2->sec) ? diff :
397: strcasecmp(mp1->names, mp2->names);
1.12 schwarze 398: }
399:
400: static char *
1.66 schwarze 401: buildnames(const struct dbm_page *page)
1.12 schwarze 402: {
1.66 schwarze 403: char *buf;
404: size_t i, sz;
1.12 schwarze 405:
1.70 schwarze 406: sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
407: (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
1.66 schwarze 408: buf = mandoc_malloc(sz);
409: i = 0;
1.70 schwarze 410: lstcat(buf, &i, page->name, ", ");
1.66 schwarze 411: buf[i++] = '(';
1.70 schwarze 412: lstcat(buf, &i, page->sect, ", ");
1.66 schwarze 413: if (page->arch != NULL) {
414: buf[i++] = '/';
1.70 schwarze 415: lstcat(buf, &i, page->arch, ", ");
1.66 schwarze 416: }
417: buf[i++] = ')';
418: buf[i++] = '\0';
419: assert(i == sz);
420: return buf;
1.1 kristaps 421: }
422:
423: /*
1.66 schwarze 424: * Count the buffer space needed to print the NUL-terminated
1.70 schwarze 425: * list of NUL-terminated strings, when printing sep separator
1.66 schwarze 426: * characters between strings.
1.7 schwarze 427: */
1.66 schwarze 428: static size_t
1.70 schwarze 429: lstlen(const char *cp, size_t sep)
1.7 schwarze 430: {
1.66 schwarze 431: size_t sz;
1.7 schwarze 432:
1.66 schwarze 433: for (sz = 0;; sz++) {
434: if (cp[0] == '\0') {
435: if (cp[1] == '\0')
436: break;
1.70 schwarze 437: sz += sep - 1;
1.66 schwarze 438: } else if (cp[0] < ' ')
439: sz--;
440: cp++;
441: }
442: return sz;
1.7 schwarze 443: }
444:
445: /*
1.66 schwarze 446: * Print the NUL-terminated list of NUL-terminated strings
1.70 schwarze 447: * into the buffer, seperating strings with sep.
1.8 schwarze 448: */
449: static void
1.70 schwarze 450: lstcat(char *buf, size_t *i, const char *cp, const char *sep)
1.8 schwarze 451: {
1.70 schwarze 452: const char *s;
453:
1.66 schwarze 454: for (;;) {
455: if (cp[0] == '\0') {
456: if (cp[1] == '\0')
457: break;
1.70 schwarze 458: s = sep;
459: while (*s != '\0')
460: buf[(*i)++] = *s++;
1.66 schwarze 461: } else if (cp[0] >= ' ')
462: buf[(*i)++] = cp[0];
463: cp++;
464: }
1.8 schwarze 465: }
466:
1.66 schwarze 467: /*
468: * Return 1 if the string *want occurs in any of the strings
469: * in the NUL-terminated string list *have, or 0 otherwise.
470: * If either argument is NULL or empty, assume no filtering
471: * is desired and return 1.
472: */
473: static int
474: lstmatch(const char *want, const char *have)
1.13 schwarze 475: {
1.66 schwarze 476: if (want == NULL || have == NULL || *have == '\0')
477: return 1;
478: while (*have != '\0') {
479: if (strcasestr(have, want) != NULL)
480: return 1;
481: have = strchr(have, '\0') + 1;
482: }
483: return 0;
1.13 schwarze 484: }
485:
1.8 schwarze 486: /*
1.70 schwarze 487: * Build a list of values taken by the macro im in the manual page.
1.1 kristaps 488: */
489: static char *
1.70 schwarze 490: buildoutput(size_t im, struct dbm_page *page)
1.1 kristaps 491: {
1.70 schwarze 492: const char *oldoutput, *sep, *input;
1.66 schwarze 493: char *output, *newoutput, *value;
1.70 schwarze 494: size_t sz, i;
495:
496: switch (im) {
497: case KEY_Nd:
498: return mandoc_strdup(page->desc);
499: case KEY_Nm:
500: input = page->name;
501: break;
502: case KEY_sec:
503: input = page->sect;
504: break;
505: case KEY_arch:
506: input = page->arch;
507: if (input == NULL)
508: input = "all\0";
509: break;
510: default:
511: input = NULL;
512: break;
513: }
514:
515: if (input != NULL) {
516: sz = lstlen(input, 3) + 1;
517: output = mandoc_malloc(sz);
518: i = 0;
519: lstcat(output, &i, input, " # ");
1.71 schwarze 520: output[i++] = '\0';
521: assert(i == sz);
1.70 schwarze 522: return output;
523: }
1.66 schwarze 524:
525: output = NULL;
1.70 schwarze 526: dbm_macro_bypage(im - 2, page->addr);
1.66 schwarze 527: while ((value = dbm_macro_next()) != NULL) {
528: if (output == NULL) {
529: oldoutput = "";
530: sep = "";
531: } else {
532: oldoutput = output;
533: sep = " # ";
534: }
535: mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
536: free(output);
537: output = newoutput;
1.1 kristaps 538: }
1.66 schwarze 539: return output;
1.1 kristaps 540: }
541:
542: /*
543: * Compile a set of string tokens into an expression.
544: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
545: * "(", "foo=bar", etc.).
546: */
547: static struct expr *
1.66 schwarze 548: exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 549: {
1.66 schwarze 550: struct expr *parent, *child;
551: int needterm, nested;
552:
553: if ((nested = *argi) == argc)
554: return NULL;
555: needterm = 1;
556: parent = child = NULL;
557: while (*argi < argc) {
558: if (strcmp(")", argv[*argi]) == 0) {
559: if (needterm)
560: warnx("missing term "
561: "before closing parenthesis");
562: needterm = 0;
563: if (nested)
564: break;
565: warnx("ignoring unmatched right parenthesis");
566: ++*argi;
1.13 schwarze 567: continue;
1.66 schwarze 568: }
569: if (strcmp("-o", argv[*argi]) == 0) {
570: if (needterm) {
571: if (*argi > 0)
572: warnx("ignoring -o after %s",
573: argv[*argi - 1]);
574: else
575: warnx("ignoring initial -o");
576: }
577: needterm = 1;
578: ++*argi;
1.13 schwarze 579: continue;
1.66 schwarze 580: }
581: needterm = 0;
582: if (child == NULL) {
583: child = expr_and(search, argc, argv, argi);
1.13 schwarze 584: continue;
1.1 kristaps 585: }
1.66 schwarze 586: if (parent == NULL) {
587: parent = mandoc_calloc(1, sizeof(*parent));
588: parent->type = EXPR_OR;
589: parent->next = NULL;
590: parent->child = child;
591: }
592: child->next = expr_and(search, argc, argv, argi);
593: child = child->next;
594: }
595: if (needterm && *argi)
596: warnx("ignoring trailing %s", argv[*argi - 1]);
597: return parent == NULL ? child : parent;
598: }
1.26 schwarze 599:
1.66 schwarze 600: static struct expr *
601: expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
602: {
603: struct expr *parent, *child;
604: int needterm;
1.26 schwarze 605:
1.66 schwarze 606: needterm = 1;
607: parent = child = NULL;
608: while (*argi < argc) {
609: if (strcmp(")", argv[*argi]) == 0) {
610: if (needterm)
611: warnx("missing term "
612: "before closing parenthesis");
613: needterm = 0;
614: break;
615: }
616: if (strcmp("-o", argv[*argi]) == 0)
617: break;
618: if (strcmp("-a", argv[*argi]) == 0) {
619: if (needterm) {
620: if (*argi > 0)
621: warnx("ignoring -a after %s",
622: argv[*argi - 1]);
623: else
624: warnx("ignoring initial -a");
1.27 schwarze 625: }
1.66 schwarze 626: needterm = 1;
627: ++*argi;
628: continue;
1.27 schwarze 629: }
1.66 schwarze 630: if (needterm == 0)
631: break;
632: if (child == NULL) {
633: child = exprterm(search, argc, argv, argi);
634: if (child != NULL)
635: needterm = 0;
636: continue;
637: }
638: needterm = 0;
639: if (parent == NULL) {
640: parent = mandoc_calloc(1, sizeof(*parent));
641: parent->type = EXPR_AND;
642: parent->next = NULL;
643: parent->child = child;
644: }
645: child->next = exprterm(search, argc, argv, argi);
646: if (child->next != NULL) {
647: child = child->next;
648: needterm = 0;
649: }
650: }
651: if (needterm && *argi)
652: warnx("ignoring trailing %s", argv[*argi - 1]);
653: return parent == NULL ? child : parent;
1.15 schwarze 654: }
655:
656: static struct expr *
1.66 schwarze 657: exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 658: {
1.15 schwarze 659: char errbuf[BUFSIZ];
1.1 kristaps 660: struct expr *e;
1.38 schwarze 661: char *key, *val;
1.20 schwarze 662: uint64_t iterbit;
1.66 schwarze 663: int cs, i, irc;
1.1 kristaps 664:
1.66 schwarze 665: if (strcmp("(", argv[*argi]) == 0) {
666: ++*argi;
667: e = exprcomp(search, argc, argv, argi);
668: if (*argi < argc) {
669: assert(strcmp(")", argv[*argi]) == 0);
670: ++*argi;
671: } else
672: warnx("unclosed parenthesis");
673: return e;
674: }
1.1 kristaps 675:
1.72 schwarze 676: if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
677: cs = 0;
678: ++*argi;
679: } else
680: cs = 1;
681:
1.66 schwarze 682: e = mandoc_calloc(1, sizeof(*e));
683: e->type = EXPR_TERM;
684: e->bits = 0;
685: e->next = NULL;
686: e->child = NULL;
1.1 kristaps 687:
1.45 schwarze 688: if (search->argmode == ARG_NAME) {
689: e->bits = TYPE_Nm;
1.66 schwarze 690: e->match.type = DBM_EXACT;
691: e->match.str = argv[(*argi)++];
1.58 schwarze 692: return e;
1.5 kristaps 693: }
694:
1.1 kristaps 695: /*
1.45 schwarze 696: * Separate macro keys from search string.
1.66 schwarze 697: * If needed, request regular expression handling.
1.1 kristaps 698: */
699:
1.45 schwarze 700: if (search->argmode == ARG_WORD) {
701: e->bits = TYPE_Nm;
1.66 schwarze 702: e->match.type = DBM_REGEX;
1.61 schwarze 703: #if HAVE_REWB_BSD
1.66 schwarze 704: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
1.61 schwarze 705: #elif HAVE_REWB_SYSV
1.66 schwarze 706: mandoc_asprintf(&val, "\\<%s\\>", argv[*argi]);
1.61 schwarze 707: #else
708: mandoc_asprintf(&val,
1.66 schwarze 709: "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", argv[*argi]);
1.61 schwarze 710: #endif
1.46 schwarze 711: cs = 0;
1.66 schwarze 712: } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
1.45 schwarze 713: e->bits = TYPE_Nm | TYPE_Nd;
1.66 schwarze 714: e->match.type = DBM_SUB;
715: e->match.str = argv[*argi];
1.38 schwarze 716: } else {
1.66 schwarze 717: if (val == argv[*argi])
1.45 schwarze 718: e->bits = TYPE_Nm | TYPE_Nd;
1.66 schwarze 719: if (*val == '=') {
720: e->match.type = DBM_SUB;
721: e->match.str = val + 1;
722: } else
723: e->match.type = DBM_REGEX;
1.38 schwarze 724: *val++ = '\0';
1.66 schwarze 725: if (strstr(argv[*argi], "arch") != NULL)
1.21 schwarze 726: cs = 0;
1.38 schwarze 727: }
728:
729: /* Compile regular expressions. */
730:
1.66 schwarze 731: if (e->match.type == DBM_REGEX) {
732: e->match.re = mandoc_malloc(sizeof(*e->match.re));
733: irc = regcomp(e->match.re, val,
1.38 schwarze 734: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.66 schwarze 735: if (irc) {
736: regerror(irc, e->match.re, errbuf, sizeof(errbuf));
737: warnx("regcomp /%s/: %s", val, errbuf);
738: }
1.45 schwarze 739: if (search->argmode == ARG_WORD)
1.38 schwarze 740: free(val);
741: if (irc) {
1.66 schwarze 742: free(e->match.re);
1.8 schwarze 743: free(e);
1.66 schwarze 744: ++*argi;
1.58 schwarze 745: return NULL;
1.8 schwarze 746: }
1.38 schwarze 747: }
748:
1.66 schwarze 749: if (e->bits) {
750: ++*argi;
1.58 schwarze 751: return e;
1.66 schwarze 752: }
1.1 kristaps 753:
754: /*
755: * Parse out all possible fields.
756: * If the field doesn't resolve, bail.
757: */
758:
1.66 schwarze 759: while (NULL != (key = strsep(&argv[*argi], ","))) {
1.1 kristaps 760: if ('\0' == *key)
761: continue;
1.66 schwarze 762: for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
763: if (0 == strcasecmp(key, mansearch_keynames[i])) {
1.20 schwarze 764: e->bits |= iterbit;
765: break;
766: }
767: }
1.66 schwarze 768: if (i == KEY_MAX) {
769: if (strcasecmp(key, "any"))
770: warnx("treating unknown key "
771: "\"%s\" as \"any\"", key);
1.20 schwarze 772: e->bits |= ~0ULL;
1.1 kristaps 773: }
774: }
775:
1.66 schwarze 776: ++*argi;
1.58 schwarze 777: return e;
1.1 kristaps 778: }
779:
780: static void
1.66 schwarze 781: exprfree(struct expr *e)
1.1 kristaps 782: {
1.66 schwarze 783: if (e->next != NULL)
784: exprfree(e->next);
785: if (e->child != NULL)
786: exprfree(e->child);
787: free(e);
1.1 kristaps 788: }
CVSweb