Annotation of mandoc/mansearch.c, Revision 1.70
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) {
1.59 schwarze 158: warn("%s/%s", paths->paths[i], MANDOC_DB);
1.1 kristaps 159: continue;
160: }
161:
1.66 schwarze 162: if ((htab = manmerge(e, NULL)) == NULL) {
163: dbm_close();
164: continue;
1.1 kristaps 165: }
166:
1.66 schwarze 167: for (rp = ohash_first(htab, &slot); rp != NULL;
168: rp = ohash_next(htab, &slot)) {
169: page = dbm_page_get(rp->page);
1.1 kristaps 170:
1.66 schwarze 171: if (lstmatch(search->sec, page->sect) == 0 ||
172: lstmatch(search->arch, page->arch) == 0)
1.1 kristaps 173: continue;
174:
175: if (cur + 1 > maxres) {
176: maxres += 1024;
1.36 schwarze 177: *res = mandoc_reallocarray(*res,
1.66 schwarze 178: maxres, sizeof(**res));
1.1 kristaps 179: }
1.10 schwarze 180: mpage = *res + cur;
1.66 schwarze 181: mandoc_asprintf(&mpage->file, "%s/%s",
182: paths->paths[i], page->file + 1);
183: mpage->names = buildnames(page);
1.70 ! schwarze 184: mpage->output = buildoutput(outkey, page);
1.47 schwarze 185: mpage->ipath = i;
1.66 schwarze 186: mpage->bits = rp->bits;
187: mpage->sec = *page->sect - '0';
188: if (mpage->sec < 0 || mpage->sec > 9)
189: mpage->sec = 10;
190: mpage->form = *page->file;
191: free(rp);
192: cur++;
193: }
194: ohash_delete(htab);
195: free(htab);
196: dbm_close();
1.49 schwarze 197:
198: /*
199: * In man(1) mode, prefer matches in earlier trees
200: * over matches in later trees.
201: */
202:
203: if (cur && search->firstmatch)
204: break;
1.1 kristaps 205: }
1.39 schwarze 206: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.57 schwarze 207: if (chdir_status && getcwd_status && chdir(buf) == -1)
1.64 schwarze 208: warn("%s", buf);
1.1 kristaps 209: exprfree(e);
210: *sz = cur;
1.58 schwarze 211: return 1;
1.11 schwarze 212: }
213:
1.66 schwarze 214: /*
215: * Merge the results for the expression tree rooted at e
216: * into the the result list htab.
217: */
218: static struct ohash *
219: manmerge(struct expr *e, struct ohash *htab)
1.45 schwarze 220: {
1.66 schwarze 221: switch (e->type) {
222: case EXPR_TERM:
223: return manmerge_term(e, htab);
224: case EXPR_OR:
225: return manmerge_or(e->child, htab);
226: case EXPR_AND:
227: return manmerge_and(e->child, htab);
228: default:
229: abort();
1.45 schwarze 230: }
231: }
232:
1.66 schwarze 233: static struct ohash *
234: manmerge_term(struct expr *e, struct ohash *htab)
1.39 schwarze 235: {
1.66 schwarze 236: struct dbm_res res, *rp;
237: uint64_t ib;
238: unsigned int slot;
239: int im;
240:
241: if (htab == NULL) {
242: htab = mandoc_malloc(sizeof(*htab));
243: mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
244: }
245:
246: for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
247: if ((e->bits & ib) == 0)
248: continue;
249:
250: switch (ib) {
251: case TYPE_arch:
252: dbm_page_byarch(&e->match);
253: break;
254: case TYPE_sec:
255: dbm_page_bysect(&e->match);
256: break;
257: case TYPE_Nm:
258: dbm_page_byname(&e->match);
259: break;
260: case TYPE_Nd:
261: dbm_page_bydesc(&e->match);
262: break;
263: default:
264: dbm_page_bymacro(im - 2, &e->match);
265: break;
266: }
267:
268: /*
269: * When hashing for deduplication, use the unique
270: * page ID itself instead of a hash function;
271: * that is quite efficient.
272: */
1.39 schwarze 273:
1.66 schwarze 274: for (;;) {
275: res = dbm_page_next();
276: if (res.page == -1)
277: break;
278: slot = ohash_lookup_memory(htab,
279: (char *)&res, sizeof(res.page), res.page);
280: if ((rp = ohash_find(htab, slot)) != NULL) {
281: rp->bits |= res.bits;
282: continue;
283: }
284: rp = mandoc_malloc(sizeof(*rp));
285: *rp = res;
286: ohash_insert(htab, slot, rp);
287: }
288: }
289: return htab;
1.39 schwarze 290: }
291:
1.66 schwarze 292: static struct ohash *
293: manmerge_or(struct expr *e, struct ohash *htab)
294: {
295: while (e != NULL) {
296: htab = manmerge(e, htab);
297: e = e->next;
298: }
299: return htab;
300: }
1.22 schwarze 301:
1.66 schwarze 302: static struct ohash *
303: manmerge_and(struct expr *e, struct ohash *htab)
304: {
305: struct ohash *hand, *h1, *h2;
306: struct dbm_res *res;
307: unsigned int slot1, slot2;
1.22 schwarze 308:
1.66 schwarze 309: /* Evaluate the first term of the AND clause. */
1.39 schwarze 310:
1.66 schwarze 311: hand = manmerge(e, NULL);
1.39 schwarze 312:
1.66 schwarze 313: while ((e = e->next) != NULL) {
1.22 schwarze 314:
1.66 schwarze 315: /* Evaluate the next term and prepare for ANDing. */
1.22 schwarze 316:
1.66 schwarze 317: h2 = manmerge(e, NULL);
318: if (ohash_entries(h2) < ohash_entries(hand)) {
319: h1 = h2;
320: h2 = hand;
321: } else
322: h1 = hand;
323: hand = mandoc_malloc(sizeof(*hand));
324: mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
1.22 schwarze 325:
1.66 schwarze 326: /* Keep all pages that are in both result sets. */
1.22 schwarze 327:
1.66 schwarze 328: for (res = ohash_first(h1, &slot1); res != NULL;
329: res = ohash_next(h1, &slot1)) {
330: if (ohash_find(h2, ohash_lookup_memory(h2,
331: (char *)res, sizeof(res->page),
332: res->page)) == NULL)
333: free(res);
334: else
335: ohash_insert(hand, ohash_lookup_memory(hand,
336: (char *)res, sizeof(res->page),
337: res->page), res);
1.22 schwarze 338: }
339:
1.66 schwarze 340: /* Discard the merged results. */
1.22 schwarze 341:
1.66 schwarze 342: for (res = ohash_first(h2, &slot2); res != NULL;
343: res = ohash_next(h2, &slot2))
344: free(res);
345: ohash_delete(h2);
346: free(h2);
347: ohash_delete(h1);
348: free(h1);
349: }
1.17 schwarze 350:
1.66 schwarze 351: /* Merge the result of the AND into htab. */
1.17 schwarze 352:
1.66 schwarze 353: if (htab == NULL)
354: return hand;
355:
356: for (res = ohash_first(hand, &slot1); res != NULL;
357: res = ohash_next(hand, &slot1)) {
358: slot2 = ohash_lookup_memory(htab,
359: (char *)res, sizeof(res->page), res->page);
360: if (ohash_find(htab, slot2) == NULL)
361: ohash_insert(htab, slot2, res);
362: else
363: free(res);
364: }
365:
366: /* Discard the merged result. */
1.17 schwarze 367:
1.66 schwarze 368: ohash_delete(hand);
369: free(hand);
370: return htab;
371: }
1.54 schwarze 372:
1.66 schwarze 373: void
374: mansearch_free(struct manpage *res, size_t sz)
375: {
376: size_t i;
1.54 schwarze 377:
1.66 schwarze 378: for (i = 0; i < sz; i++) {
379: free(res[i].file);
380: free(res[i].names);
381: free(res[i].output);
1.22 schwarze 382: }
1.66 schwarze 383: free(res);
384: }
385:
386: static int
387: manpage_compare(const void *vp1, const void *vp2)
388: {
389: const struct manpage *mp1, *mp2;
390: int diff;
391:
392: mp1 = vp1;
393: mp2 = vp2;
394: return (diff = mp2->bits - mp1->bits) ? diff :
395: (diff = mp1->sec - mp2->sec) ? diff :
396: strcasecmp(mp1->names, mp2->names);
1.12 schwarze 397: }
398:
399: static char *
1.66 schwarze 400: buildnames(const struct dbm_page *page)
1.12 schwarze 401: {
1.66 schwarze 402: char *buf;
403: size_t i, sz;
1.12 schwarze 404:
1.70 ! schwarze 405: sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
! 406: (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
1.66 schwarze 407: buf = mandoc_malloc(sz);
408: i = 0;
1.70 ! schwarze 409: lstcat(buf, &i, page->name, ", ");
1.66 schwarze 410: buf[i++] = '(';
1.70 ! schwarze 411: lstcat(buf, &i, page->sect, ", ");
1.66 schwarze 412: if (page->arch != NULL) {
413: buf[i++] = '/';
1.70 ! schwarze 414: lstcat(buf, &i, page->arch, ", ");
1.66 schwarze 415: }
416: buf[i++] = ')';
417: buf[i++] = '\0';
418: assert(i == sz);
419: return buf;
1.1 kristaps 420: }
421:
422: /*
1.66 schwarze 423: * Count the buffer space needed to print the NUL-terminated
1.70 ! schwarze 424: * list of NUL-terminated strings, when printing sep separator
1.66 schwarze 425: * characters between strings.
1.7 schwarze 426: */
1.66 schwarze 427: static size_t
1.70 ! schwarze 428: lstlen(const char *cp, size_t sep)
1.7 schwarze 429: {
1.66 schwarze 430: size_t sz;
1.7 schwarze 431:
1.66 schwarze 432: for (sz = 0;; sz++) {
433: if (cp[0] == '\0') {
434: if (cp[1] == '\0')
435: break;
1.70 ! schwarze 436: sz += sep - 1;
1.66 schwarze 437: } else if (cp[0] < ' ')
438: sz--;
439: cp++;
440: }
441: return sz;
1.7 schwarze 442: }
443:
444: /*
1.66 schwarze 445: * Print the NUL-terminated list of NUL-terminated strings
1.70 ! schwarze 446: * into the buffer, seperating strings with sep.
1.8 schwarze 447: */
448: static void
1.70 ! schwarze 449: lstcat(char *buf, size_t *i, const char *cp, const char *sep)
1.8 schwarze 450: {
1.70 ! schwarze 451: const char *s;
! 452:
1.66 schwarze 453: for (;;) {
454: if (cp[0] == '\0') {
455: if (cp[1] == '\0')
456: break;
1.70 ! schwarze 457: s = sep;
! 458: while (*s != '\0')
! 459: buf[(*i)++] = *s++;
1.66 schwarze 460: } else if (cp[0] >= ' ')
461: buf[(*i)++] = cp[0];
462: cp++;
463: }
1.8 schwarze 464: }
465:
1.66 schwarze 466: /*
467: * Return 1 if the string *want occurs in any of the strings
468: * in the NUL-terminated string list *have, or 0 otherwise.
469: * If either argument is NULL or empty, assume no filtering
470: * is desired and return 1.
471: */
472: static int
473: lstmatch(const char *want, const char *have)
1.13 schwarze 474: {
1.66 schwarze 475: if (want == NULL || have == NULL || *have == '\0')
476: return 1;
477: while (*have != '\0') {
478: if (strcasestr(have, want) != NULL)
479: return 1;
480: have = strchr(have, '\0') + 1;
481: }
482: return 0;
1.13 schwarze 483: }
484:
1.8 schwarze 485: /*
1.70 ! schwarze 486: * Build a list of values taken by the macro im in the manual page.
1.1 kristaps 487: */
488: static char *
1.70 ! schwarze 489: buildoutput(size_t im, struct dbm_page *page)
1.1 kristaps 490: {
1.70 ! schwarze 491: const char *oldoutput, *sep, *input;
1.66 schwarze 492: char *output, *newoutput, *value;
1.70 ! schwarze 493: size_t sz, i;
! 494:
! 495: switch (im) {
! 496: case KEY_Nd:
! 497: return mandoc_strdup(page->desc);
! 498: case KEY_Nm:
! 499: input = page->name;
! 500: break;
! 501: case KEY_sec:
! 502: input = page->sect;
! 503: break;
! 504: case KEY_arch:
! 505: input = page->arch;
! 506: if (input == NULL)
! 507: input = "all\0";
! 508: break;
! 509: default:
! 510: input = NULL;
! 511: break;
! 512: }
! 513:
! 514: if (input != NULL) {
! 515: sz = lstlen(input, 3) + 1;
! 516: output = mandoc_malloc(sz);
! 517: i = 0;
! 518: lstcat(output, &i, input, " # ");
! 519: return output;
! 520: }
1.66 schwarze 521:
522: output = NULL;
1.70 ! schwarze 523: dbm_macro_bypage(im - 2, page->addr);
1.66 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 kristaps 535: }
1.66 schwarze 536: return output;
1.1 kristaps 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 *
1.66 schwarze 545: exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 546: {
1.66 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;
1.13 schwarze 564: continue;
1.66 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;
1.13 schwarze 576: continue;
1.66 schwarze 577: }
578: needterm = 0;
579: if (child == NULL) {
580: child = expr_and(search, argc, argv, argi);
1.13 schwarze 581: continue;
1.1 kristaps 582: }
1.66 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: }
1.26 schwarze 596:
1.66 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;
1.26 schwarze 602:
1.66 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");
1.27 schwarze 622: }
1.66 schwarze 623: needterm = 1;
624: ++*argi;
625: continue;
1.27 schwarze 626: }
1.66 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;
1.15 schwarze 651: }
652:
653: static struct expr *
1.66 schwarze 654: exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 655: {
1.15 schwarze 656: char errbuf[BUFSIZ];
1.1 kristaps 657: struct expr *e;
1.38 schwarze 658: char *key, *val;
1.20 schwarze 659: uint64_t iterbit;
1.66 schwarze 660: int cs, i, irc;
1.1 kristaps 661:
1.66 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 kristaps 672:
1.66 schwarze 673: e = mandoc_calloc(1, sizeof(*e));
674: e->type = EXPR_TERM;
675: e->bits = 0;
676: e->next = NULL;
677: e->child = NULL;
1.1 kristaps 678:
1.45 schwarze 679: if (search->argmode == ARG_NAME) {
680: e->bits = TYPE_Nm;
1.66 schwarze 681: e->match.type = DBM_EXACT;
682: e->match.str = argv[(*argi)++];
1.58 schwarze 683: return e;
1.5 kristaps 684: }
685:
1.1 kristaps 686: /*
1.45 schwarze 687: * Separate macro keys from search string.
1.66 schwarze 688: * If needed, request regular expression handling.
1.1 kristaps 689: */
690:
1.69 schwarze 691: cs = 1;
1.45 schwarze 692: if (search->argmode == ARG_WORD) {
693: e->bits = TYPE_Nm;
1.66 schwarze 694: e->match.type = DBM_REGEX;
1.61 schwarze 695: #if HAVE_REWB_BSD
1.66 schwarze 696: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
1.61 schwarze 697: #elif HAVE_REWB_SYSV
1.66 schwarze 698: mandoc_asprintf(&val, "\\<%s\\>", argv[*argi]);
1.61 schwarze 699: #else
700: mandoc_asprintf(&val,
1.66 schwarze 701: "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", argv[*argi]);
1.61 schwarze 702: #endif
1.46 schwarze 703: cs = 0;
1.66 schwarze 704: } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
1.45 schwarze 705: e->bits = TYPE_Nm | TYPE_Nd;
1.66 schwarze 706: e->match.type = DBM_SUB;
707: e->match.str = argv[*argi];
1.38 schwarze 708: } else {
1.66 schwarze 709: if (val == argv[*argi])
1.45 schwarze 710: e->bits = TYPE_Nm | TYPE_Nd;
1.66 schwarze 711: if (*val == '=') {
712: e->match.type = DBM_SUB;
713: e->match.str = val + 1;
714: } else
715: e->match.type = DBM_REGEX;
1.38 schwarze 716: *val++ = '\0';
1.66 schwarze 717: if (strstr(argv[*argi], "arch") != NULL)
1.21 schwarze 718: cs = 0;
1.38 schwarze 719: }
720:
721: /* Compile regular expressions. */
722:
1.66 schwarze 723: if (e->match.type == DBM_REGEX) {
724: e->match.re = mandoc_malloc(sizeof(*e->match.re));
725: irc = regcomp(e->match.re, val,
1.38 schwarze 726: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.66 schwarze 727: if (irc) {
728: regerror(irc, e->match.re, errbuf, sizeof(errbuf));
729: warnx("regcomp /%s/: %s", val, errbuf);
730: }
1.45 schwarze 731: if (search->argmode == ARG_WORD)
1.38 schwarze 732: free(val);
733: if (irc) {
1.66 schwarze 734: free(e->match.re);
1.8 schwarze 735: free(e);
1.66 schwarze 736: ++*argi;
1.58 schwarze 737: return NULL;
1.8 schwarze 738: }
1.38 schwarze 739: }
740:
1.66 schwarze 741: if (e->bits) {
742: ++*argi;
1.58 schwarze 743: return e;
1.66 schwarze 744: }
1.1 kristaps 745:
746: /*
747: * Parse out all possible fields.
748: * If the field doesn't resolve, bail.
749: */
750:
1.66 schwarze 751: while (NULL != (key = strsep(&argv[*argi], ","))) {
1.1 kristaps 752: if ('\0' == *key)
753: continue;
1.66 schwarze 754: for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
755: if (0 == strcasecmp(key, mansearch_keynames[i])) {
1.20 schwarze 756: e->bits |= iterbit;
757: break;
758: }
759: }
1.66 schwarze 760: if (i == KEY_MAX) {
761: if (strcasecmp(key, "any"))
762: warnx("treating unknown key "
763: "\"%s\" as \"any\"", key);
1.20 schwarze 764: e->bits |= ~0ULL;
1.1 kristaps 765: }
766: }
767:
1.66 schwarze 768: ++*argi;
1.58 schwarze 769: return e;
1.1 kristaps 770: }
771:
772: static void
1.66 schwarze 773: exprfree(struct expr *e)
1.1 kristaps 774: {
1.66 schwarze 775: if (e->next != NULL)
776: exprfree(e->next);
777: if (e->child != NULL)
778: exprfree(e->child);
779: free(e);
1.1 kristaps 780: }
CVSweb