Annotation of mandoc/mansearch.c, Revision 1.74
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;
1.74 ! 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 ||
174: lstmatch(search->arch, page->arch) == 0)
1.1 kristaps 175: continue;
176:
1.74 ! schwarze 177: if (res == NULL) {
! 178: cur = 1;
! 179: break;
! 180: }
1.1 kristaps 181: if (cur + 1 > maxres) {
182: maxres += 1024;
1.36 schwarze 183: *res = mandoc_reallocarray(*res,
1.66 schwarze 184: maxres, sizeof(**res));
1.1 kristaps 185: }
1.10 schwarze 186: mpage = *res + cur;
1.66 schwarze 187: mandoc_asprintf(&mpage->file, "%s/%s",
188: paths->paths[i], page->file + 1);
189: mpage->names = buildnames(page);
1.70 schwarze 190: mpage->output = buildoutput(outkey, page);
1.47 schwarze 191: mpage->ipath = i;
1.66 schwarze 192: mpage->bits = rp->bits;
193: mpage->sec = *page->sect - '0';
194: if (mpage->sec < 0 || mpage->sec > 9)
195: mpage->sec = 10;
196: mpage->form = *page->file;
197: free(rp);
198: cur++;
199: }
200: ohash_delete(htab);
201: free(htab);
202: dbm_close();
1.49 schwarze 203:
204: /*
205: * In man(1) mode, prefer matches in earlier trees
206: * over matches in later trees.
207: */
208:
209: if (cur && search->firstmatch)
210: break;
1.1 kristaps 211: }
1.74 ! schwarze 212: if (res != NULL)
! 213: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.57 schwarze 214: if (chdir_status && getcwd_status && chdir(buf) == -1)
1.64 schwarze 215: warn("%s", buf);
1.1 kristaps 216: exprfree(e);
217: *sz = cur;
1.74 ! schwarze 218: return res != NULL || cur;
1.11 schwarze 219: }
220:
1.66 schwarze 221: /*
222: * Merge the results for the expression tree rooted at e
223: * into the the result list htab.
224: */
225: static struct ohash *
226: manmerge(struct expr *e, struct ohash *htab)
1.45 schwarze 227: {
1.66 schwarze 228: switch (e->type) {
229: case EXPR_TERM:
230: return manmerge_term(e, htab);
231: case EXPR_OR:
232: return manmerge_or(e->child, htab);
233: case EXPR_AND:
234: return manmerge_and(e->child, htab);
235: default:
236: abort();
1.45 schwarze 237: }
238: }
239:
1.66 schwarze 240: static struct ohash *
241: manmerge_term(struct expr *e, struct ohash *htab)
1.39 schwarze 242: {
1.66 schwarze 243: struct dbm_res res, *rp;
244: uint64_t ib;
245: unsigned int slot;
246: int im;
247:
248: if (htab == NULL) {
249: htab = mandoc_malloc(sizeof(*htab));
250: mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
251: }
252:
253: for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
254: if ((e->bits & ib) == 0)
255: continue;
256:
257: switch (ib) {
258: case TYPE_arch:
259: dbm_page_byarch(&e->match);
260: break;
261: case TYPE_sec:
262: dbm_page_bysect(&e->match);
263: break;
264: case TYPE_Nm:
265: dbm_page_byname(&e->match);
266: break;
267: case TYPE_Nd:
268: dbm_page_bydesc(&e->match);
269: break;
270: default:
271: dbm_page_bymacro(im - 2, &e->match);
272: break;
273: }
274:
275: /*
276: * When hashing for deduplication, use the unique
277: * page ID itself instead of a hash function;
278: * that is quite efficient.
279: */
1.39 schwarze 280:
1.66 schwarze 281: for (;;) {
282: res = dbm_page_next();
283: if (res.page == -1)
284: break;
285: slot = ohash_lookup_memory(htab,
286: (char *)&res, sizeof(res.page), res.page);
287: if ((rp = ohash_find(htab, slot)) != NULL) {
288: rp->bits |= res.bits;
289: continue;
290: }
291: rp = mandoc_malloc(sizeof(*rp));
292: *rp = res;
293: ohash_insert(htab, slot, rp);
294: }
295: }
296: return htab;
1.39 schwarze 297: }
298:
1.66 schwarze 299: static struct ohash *
300: manmerge_or(struct expr *e, struct ohash *htab)
301: {
302: while (e != NULL) {
303: htab = manmerge(e, htab);
304: e = e->next;
305: }
306: return htab;
307: }
1.22 schwarze 308:
1.66 schwarze 309: static struct ohash *
310: manmerge_and(struct expr *e, struct ohash *htab)
311: {
312: struct ohash *hand, *h1, *h2;
313: struct dbm_res *res;
314: unsigned int slot1, slot2;
1.22 schwarze 315:
1.66 schwarze 316: /* Evaluate the first term of the AND clause. */
1.39 schwarze 317:
1.66 schwarze 318: hand = manmerge(e, NULL);
1.39 schwarze 319:
1.66 schwarze 320: while ((e = e->next) != NULL) {
1.22 schwarze 321:
1.66 schwarze 322: /* Evaluate the next term and prepare for ANDing. */
1.22 schwarze 323:
1.66 schwarze 324: h2 = manmerge(e, NULL);
325: if (ohash_entries(h2) < ohash_entries(hand)) {
326: h1 = h2;
327: h2 = hand;
328: } else
329: h1 = hand;
330: hand = mandoc_malloc(sizeof(*hand));
331: mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
1.22 schwarze 332:
1.66 schwarze 333: /* Keep all pages that are in both result sets. */
1.22 schwarze 334:
1.66 schwarze 335: for (res = ohash_first(h1, &slot1); res != NULL;
336: res = ohash_next(h1, &slot1)) {
337: if (ohash_find(h2, ohash_lookup_memory(h2,
338: (char *)res, sizeof(res->page),
339: res->page)) == NULL)
340: free(res);
341: else
342: ohash_insert(hand, ohash_lookup_memory(hand,
343: (char *)res, sizeof(res->page),
344: res->page), res);
1.22 schwarze 345: }
346:
1.66 schwarze 347: /* Discard the merged results. */
1.22 schwarze 348:
1.66 schwarze 349: for (res = ohash_first(h2, &slot2); res != NULL;
350: res = ohash_next(h2, &slot2))
351: free(res);
352: ohash_delete(h2);
353: free(h2);
354: ohash_delete(h1);
355: free(h1);
356: }
1.17 schwarze 357:
1.66 schwarze 358: /* Merge the result of the AND into htab. */
1.17 schwarze 359:
1.66 schwarze 360: if (htab == NULL)
361: return hand;
362:
363: for (res = ohash_first(hand, &slot1); res != NULL;
364: res = ohash_next(hand, &slot1)) {
365: slot2 = ohash_lookup_memory(htab,
366: (char *)res, sizeof(res->page), res->page);
367: if (ohash_find(htab, slot2) == NULL)
368: ohash_insert(htab, slot2, res);
369: else
370: free(res);
371: }
372:
373: /* Discard the merged result. */
1.17 schwarze 374:
1.66 schwarze 375: ohash_delete(hand);
376: free(hand);
377: return htab;
378: }
1.54 schwarze 379:
1.66 schwarze 380: void
381: mansearch_free(struct manpage *res, size_t sz)
382: {
383: size_t i;
1.54 schwarze 384:
1.66 schwarze 385: for (i = 0; i < sz; i++) {
386: free(res[i].file);
387: free(res[i].names);
388: free(res[i].output);
1.22 schwarze 389: }
1.66 schwarze 390: free(res);
391: }
392:
393: static int
394: manpage_compare(const void *vp1, const void *vp2)
395: {
396: const struct manpage *mp1, *mp2;
397: int diff;
398:
399: mp1 = vp1;
400: mp2 = vp2;
401: return (diff = mp2->bits - mp1->bits) ? diff :
402: (diff = mp1->sec - mp2->sec) ? diff :
403: strcasecmp(mp1->names, mp2->names);
1.12 schwarze 404: }
405:
406: static char *
1.66 schwarze 407: buildnames(const struct dbm_page *page)
1.12 schwarze 408: {
1.66 schwarze 409: char *buf;
410: size_t i, sz;
1.12 schwarze 411:
1.70 schwarze 412: sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
413: (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
1.66 schwarze 414: buf = mandoc_malloc(sz);
415: i = 0;
1.70 schwarze 416: lstcat(buf, &i, page->name, ", ");
1.66 schwarze 417: buf[i++] = '(';
1.70 schwarze 418: lstcat(buf, &i, page->sect, ", ");
1.66 schwarze 419: if (page->arch != NULL) {
420: buf[i++] = '/';
1.70 schwarze 421: lstcat(buf, &i, page->arch, ", ");
1.66 schwarze 422: }
423: buf[i++] = ')';
424: buf[i++] = '\0';
425: assert(i == sz);
426: return buf;
1.1 kristaps 427: }
428:
429: /*
1.66 schwarze 430: * Count the buffer space needed to print the NUL-terminated
1.70 schwarze 431: * list of NUL-terminated strings, when printing sep separator
1.66 schwarze 432: * characters between strings.
1.7 schwarze 433: */
1.66 schwarze 434: static size_t
1.70 schwarze 435: lstlen(const char *cp, size_t sep)
1.7 schwarze 436: {
1.66 schwarze 437: size_t sz;
1.7 schwarze 438:
1.66 schwarze 439: for (sz = 0;; sz++) {
440: if (cp[0] == '\0') {
441: if (cp[1] == '\0')
442: break;
1.70 schwarze 443: sz += sep - 1;
1.66 schwarze 444: } else if (cp[0] < ' ')
445: sz--;
446: cp++;
447: }
448: return sz;
1.7 schwarze 449: }
450:
451: /*
1.66 schwarze 452: * Print the NUL-terminated list of NUL-terminated strings
1.70 schwarze 453: * into the buffer, seperating strings with sep.
1.8 schwarze 454: */
455: static void
1.70 schwarze 456: lstcat(char *buf, size_t *i, const char *cp, const char *sep)
1.8 schwarze 457: {
1.70 schwarze 458: const char *s;
459:
1.66 schwarze 460: for (;;) {
461: if (cp[0] == '\0') {
462: if (cp[1] == '\0')
463: break;
1.70 schwarze 464: s = sep;
465: while (*s != '\0')
466: buf[(*i)++] = *s++;
1.66 schwarze 467: } else if (cp[0] >= ' ')
468: buf[(*i)++] = cp[0];
469: cp++;
470: }
1.8 schwarze 471: }
472:
1.66 schwarze 473: /*
474: * Return 1 if the string *want occurs in any of the strings
475: * in the NUL-terminated string list *have, or 0 otherwise.
476: * If either argument is NULL or empty, assume no filtering
477: * is desired and return 1.
478: */
479: static int
480: lstmatch(const char *want, const char *have)
1.13 schwarze 481: {
1.66 schwarze 482: if (want == NULL || have == NULL || *have == '\0')
483: return 1;
484: while (*have != '\0') {
485: if (strcasestr(have, want) != NULL)
486: return 1;
487: have = strchr(have, '\0') + 1;
488: }
489: return 0;
1.13 schwarze 490: }
491:
1.8 schwarze 492: /*
1.70 schwarze 493: * Build a list of values taken by the macro im in the manual page.
1.1 kristaps 494: */
495: static char *
1.70 schwarze 496: buildoutput(size_t im, struct dbm_page *page)
1.1 kristaps 497: {
1.70 schwarze 498: const char *oldoutput, *sep, *input;
1.66 schwarze 499: char *output, *newoutput, *value;
1.70 schwarze 500: size_t sz, i;
501:
502: switch (im) {
503: case KEY_Nd:
504: return mandoc_strdup(page->desc);
505: case KEY_Nm:
506: input = page->name;
507: break;
508: case KEY_sec:
509: input = page->sect;
510: break;
511: case KEY_arch:
512: input = page->arch;
513: if (input == NULL)
514: input = "all\0";
515: break;
516: default:
517: input = NULL;
518: break;
519: }
520:
521: if (input != NULL) {
522: sz = lstlen(input, 3) + 1;
523: output = mandoc_malloc(sz);
524: i = 0;
525: lstcat(output, &i, input, " # ");
1.71 schwarze 526: output[i++] = '\0';
527: assert(i == sz);
1.70 schwarze 528: return output;
529: }
1.66 schwarze 530:
531: output = NULL;
1.70 schwarze 532: dbm_macro_bypage(im - 2, page->addr);
1.66 schwarze 533: while ((value = dbm_macro_next()) != NULL) {
534: if (output == NULL) {
535: oldoutput = "";
536: sep = "";
537: } else {
538: oldoutput = output;
539: sep = " # ";
540: }
541: mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
542: free(output);
543: output = newoutput;
1.1 kristaps 544: }
1.66 schwarze 545: return output;
1.1 kristaps 546: }
547:
548: /*
549: * Compile a set of string tokens into an expression.
550: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
551: * "(", "foo=bar", etc.).
552: */
553: static struct expr *
1.66 schwarze 554: exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 555: {
1.66 schwarze 556: struct expr *parent, *child;
557: int needterm, nested;
558:
559: if ((nested = *argi) == argc)
560: return NULL;
561: needterm = 1;
562: parent = child = NULL;
563: while (*argi < argc) {
564: if (strcmp(")", argv[*argi]) == 0) {
565: if (needterm)
566: warnx("missing term "
567: "before closing parenthesis");
568: needterm = 0;
569: if (nested)
570: break;
571: warnx("ignoring unmatched right parenthesis");
572: ++*argi;
1.13 schwarze 573: continue;
1.66 schwarze 574: }
575: if (strcmp("-o", argv[*argi]) == 0) {
576: if (needterm) {
577: if (*argi > 0)
578: warnx("ignoring -o after %s",
579: argv[*argi - 1]);
580: else
581: warnx("ignoring initial -o");
582: }
583: needterm = 1;
584: ++*argi;
1.13 schwarze 585: continue;
1.66 schwarze 586: }
587: needterm = 0;
588: if (child == NULL) {
589: child = expr_and(search, argc, argv, argi);
1.13 schwarze 590: continue;
1.1 kristaps 591: }
1.66 schwarze 592: if (parent == NULL) {
593: parent = mandoc_calloc(1, sizeof(*parent));
594: parent->type = EXPR_OR;
595: parent->next = NULL;
596: parent->child = child;
597: }
598: child->next = expr_and(search, argc, argv, argi);
599: child = child->next;
600: }
601: if (needterm && *argi)
602: warnx("ignoring trailing %s", argv[*argi - 1]);
603: return parent == NULL ? child : parent;
604: }
1.26 schwarze 605:
1.66 schwarze 606: static struct expr *
607: expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
608: {
609: struct expr *parent, *child;
610: int needterm;
1.26 schwarze 611:
1.66 schwarze 612: needterm = 1;
613: parent = child = NULL;
614: while (*argi < argc) {
615: if (strcmp(")", argv[*argi]) == 0) {
616: if (needterm)
617: warnx("missing term "
618: "before closing parenthesis");
619: needterm = 0;
620: break;
621: }
622: if (strcmp("-o", argv[*argi]) == 0)
623: break;
624: if (strcmp("-a", argv[*argi]) == 0) {
625: if (needterm) {
626: if (*argi > 0)
627: warnx("ignoring -a after %s",
628: argv[*argi - 1]);
629: else
630: warnx("ignoring initial -a");
1.27 schwarze 631: }
1.66 schwarze 632: needterm = 1;
633: ++*argi;
634: continue;
1.27 schwarze 635: }
1.66 schwarze 636: if (needterm == 0)
637: break;
638: if (child == NULL) {
639: child = exprterm(search, argc, argv, argi);
640: if (child != NULL)
641: needterm = 0;
642: continue;
643: }
644: needterm = 0;
645: if (parent == NULL) {
646: parent = mandoc_calloc(1, sizeof(*parent));
647: parent->type = EXPR_AND;
648: parent->next = NULL;
649: parent->child = child;
650: }
651: child->next = exprterm(search, argc, argv, argi);
652: if (child->next != NULL) {
653: child = child->next;
654: needterm = 0;
655: }
656: }
657: if (needterm && *argi)
658: warnx("ignoring trailing %s", argv[*argi - 1]);
659: return parent == NULL ? child : parent;
1.15 schwarze 660: }
661:
662: static struct expr *
1.66 schwarze 663: exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 664: {
1.15 schwarze 665: char errbuf[BUFSIZ];
1.1 kristaps 666: struct expr *e;
1.38 schwarze 667: char *key, *val;
1.20 schwarze 668: uint64_t iterbit;
1.66 schwarze 669: int cs, i, irc;
1.1 kristaps 670:
1.66 schwarze 671: if (strcmp("(", argv[*argi]) == 0) {
672: ++*argi;
673: e = exprcomp(search, argc, argv, argi);
674: if (*argi < argc) {
675: assert(strcmp(")", argv[*argi]) == 0);
676: ++*argi;
677: } else
678: warnx("unclosed parenthesis");
679: return e;
680: }
1.1 kristaps 681:
1.72 schwarze 682: if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
683: cs = 0;
684: ++*argi;
685: } else
686: cs = 1;
687:
1.66 schwarze 688: e = mandoc_calloc(1, sizeof(*e));
689: e->type = EXPR_TERM;
690: e->bits = 0;
691: e->next = NULL;
692: e->child = NULL;
1.1 kristaps 693:
1.45 schwarze 694: if (search->argmode == ARG_NAME) {
695: e->bits = TYPE_Nm;
1.66 schwarze 696: e->match.type = DBM_EXACT;
697: e->match.str = argv[(*argi)++];
1.58 schwarze 698: return e;
1.5 kristaps 699: }
700:
1.1 kristaps 701: /*
1.45 schwarze 702: * Separate macro keys from search string.
1.66 schwarze 703: * If needed, request regular expression handling.
1.1 kristaps 704: */
705:
1.45 schwarze 706: if (search->argmode == ARG_WORD) {
707: e->bits = TYPE_Nm;
1.66 schwarze 708: e->match.type = DBM_REGEX;
1.61 schwarze 709: #if HAVE_REWB_BSD
1.66 schwarze 710: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
1.61 schwarze 711: #elif HAVE_REWB_SYSV
1.66 schwarze 712: mandoc_asprintf(&val, "\\<%s\\>", argv[*argi]);
1.61 schwarze 713: #else
714: mandoc_asprintf(&val,
1.66 schwarze 715: "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", argv[*argi]);
1.61 schwarze 716: #endif
1.46 schwarze 717: cs = 0;
1.66 schwarze 718: } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
1.45 schwarze 719: e->bits = TYPE_Nm | TYPE_Nd;
1.66 schwarze 720: e->match.type = DBM_SUB;
721: e->match.str = argv[*argi];
1.38 schwarze 722: } else {
1.66 schwarze 723: if (val == argv[*argi])
1.45 schwarze 724: e->bits = TYPE_Nm | TYPE_Nd;
1.66 schwarze 725: if (*val == '=') {
726: e->match.type = DBM_SUB;
727: e->match.str = val + 1;
728: } else
729: e->match.type = DBM_REGEX;
1.38 schwarze 730: *val++ = '\0';
1.66 schwarze 731: if (strstr(argv[*argi], "arch") != NULL)
1.21 schwarze 732: cs = 0;
1.38 schwarze 733: }
734:
735: /* Compile regular expressions. */
736:
1.66 schwarze 737: if (e->match.type == DBM_REGEX) {
738: e->match.re = mandoc_malloc(sizeof(*e->match.re));
739: irc = regcomp(e->match.re, val,
1.38 schwarze 740: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.66 schwarze 741: if (irc) {
742: regerror(irc, e->match.re, errbuf, sizeof(errbuf));
743: warnx("regcomp /%s/: %s", val, errbuf);
744: }
1.45 schwarze 745: if (search->argmode == ARG_WORD)
1.38 schwarze 746: free(val);
747: if (irc) {
1.66 schwarze 748: free(e->match.re);
1.8 schwarze 749: free(e);
1.66 schwarze 750: ++*argi;
1.58 schwarze 751: return NULL;
1.8 schwarze 752: }
1.38 schwarze 753: }
754:
1.66 schwarze 755: if (e->bits) {
756: ++*argi;
1.58 schwarze 757: return e;
1.66 schwarze 758: }
1.1 kristaps 759:
760: /*
761: * Parse out all possible fields.
762: * If the field doesn't resolve, bail.
763: */
764:
1.66 schwarze 765: while (NULL != (key = strsep(&argv[*argi], ","))) {
1.1 kristaps 766: if ('\0' == *key)
767: continue;
1.66 schwarze 768: for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
769: if (0 == strcasecmp(key, mansearch_keynames[i])) {
1.20 schwarze 770: e->bits |= iterbit;
771: break;
772: }
773: }
1.66 schwarze 774: if (i == KEY_MAX) {
775: if (strcasecmp(key, "any"))
776: warnx("treating unknown key "
777: "\"%s\" as \"any\"", key);
1.20 schwarze 778: e->bits |= ~0ULL;
1.1 kristaps 779: }
780: }
781:
1.66 schwarze 782: ++*argi;
1.58 schwarze 783: return e;
1.1 kristaps 784: }
785:
786: static void
1.66 schwarze 787: exprfree(struct expr *e)
1.1 kristaps 788: {
1.66 schwarze 789: if (e->next != NULL)
790: exprfree(e->next);
791: if (e->child != NULL)
792: exprfree(e->child);
793: free(e);
1.1 kristaps 794: }
CVSweb