Annotation of mandoc/mansearch.c, Revision 1.76
1.76 ! schwarze 1: /* $Id$ */
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 ||
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);
191: mpage->names = buildnames(page);
1.70 schwarze 192: mpage->output = buildoutput(outkey, page);
1.47 schwarze 193: mpage->ipath = i;
1.66 schwarze 194: mpage->bits = rp->bits;
195: mpage->sec = *page->sect - '0';
196: if (mpage->sec < 0 || mpage->sec > 9)
197: mpage->sec = 10;
198: mpage->form = *page->file;
199: free(rp);
200: cur++;
201: }
202: ohash_delete(htab);
203: free(htab);
204: dbm_close();
1.49 schwarze 205:
206: /*
207: * In man(1) mode, prefer matches in earlier trees
208: * over matches in later trees.
209: */
210:
211: if (cur && search->firstmatch)
212: break;
1.1 kristaps 213: }
1.74 schwarze 214: if (res != NULL)
215: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.57 schwarze 216: if (chdir_status && getcwd_status && chdir(buf) == -1)
1.64 schwarze 217: warn("%s", buf);
1.1 kristaps 218: exprfree(e);
219: *sz = cur;
1.74 schwarze 220: return res != NULL || cur;
1.11 schwarze 221: }
222:
1.66 schwarze 223: /*
224: * Merge the results for the expression tree rooted at e
225: * into the the result list htab.
226: */
227: static struct ohash *
228: manmerge(struct expr *e, struct ohash *htab)
1.45 schwarze 229: {
1.66 schwarze 230: switch (e->type) {
231: case EXPR_TERM:
232: return manmerge_term(e, htab);
233: case EXPR_OR:
234: return manmerge_or(e->child, htab);
235: case EXPR_AND:
236: return manmerge_and(e->child, htab);
237: default:
238: abort();
1.45 schwarze 239: }
240: }
241:
1.66 schwarze 242: static struct ohash *
243: manmerge_term(struct expr *e, struct ohash *htab)
1.39 schwarze 244: {
1.66 schwarze 245: struct dbm_res res, *rp;
246: uint64_t ib;
247: unsigned int slot;
248: int im;
249:
250: if (htab == NULL) {
251: htab = mandoc_malloc(sizeof(*htab));
252: mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page));
253: }
254:
255: for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) {
256: if ((e->bits & ib) == 0)
257: continue;
258:
259: switch (ib) {
260: case TYPE_arch:
261: dbm_page_byarch(&e->match);
262: break;
263: case TYPE_sec:
264: dbm_page_bysect(&e->match);
265: break;
266: case TYPE_Nm:
267: dbm_page_byname(&e->match);
268: break;
269: case TYPE_Nd:
270: dbm_page_bydesc(&e->match);
271: break;
272: default:
273: dbm_page_bymacro(im - 2, &e->match);
274: break;
275: }
276:
277: /*
278: * When hashing for deduplication, use the unique
279: * page ID itself instead of a hash function;
280: * that is quite efficient.
281: */
1.39 schwarze 282:
1.66 schwarze 283: for (;;) {
284: res = dbm_page_next();
285: if (res.page == -1)
286: break;
287: slot = ohash_lookup_memory(htab,
288: (char *)&res, sizeof(res.page), res.page);
289: if ((rp = ohash_find(htab, slot)) != NULL) {
290: rp->bits |= res.bits;
291: continue;
292: }
293: rp = mandoc_malloc(sizeof(*rp));
294: *rp = res;
295: ohash_insert(htab, slot, rp);
296: }
297: }
298: return htab;
1.39 schwarze 299: }
300:
1.66 schwarze 301: static struct ohash *
302: manmerge_or(struct expr *e, struct ohash *htab)
303: {
304: while (e != NULL) {
305: htab = manmerge(e, htab);
306: e = e->next;
307: }
308: return htab;
309: }
1.22 schwarze 310:
1.66 schwarze 311: static struct ohash *
312: manmerge_and(struct expr *e, struct ohash *htab)
313: {
314: struct ohash *hand, *h1, *h2;
315: struct dbm_res *res;
316: unsigned int slot1, slot2;
1.22 schwarze 317:
1.66 schwarze 318: /* Evaluate the first term of the AND clause. */
1.39 schwarze 319:
1.66 schwarze 320: hand = manmerge(e, NULL);
1.39 schwarze 321:
1.66 schwarze 322: while ((e = e->next) != NULL) {
1.22 schwarze 323:
1.66 schwarze 324: /* Evaluate the next term and prepare for ANDing. */
1.22 schwarze 325:
1.66 schwarze 326: h2 = manmerge(e, NULL);
327: if (ohash_entries(h2) < ohash_entries(hand)) {
328: h1 = h2;
329: h2 = hand;
330: } else
331: h1 = hand;
332: hand = mandoc_malloc(sizeof(*hand));
333: mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page));
1.22 schwarze 334:
1.66 schwarze 335: /* Keep all pages that are in both result sets. */
1.22 schwarze 336:
1.66 schwarze 337: for (res = ohash_first(h1, &slot1); res != NULL;
338: res = ohash_next(h1, &slot1)) {
339: if (ohash_find(h2, ohash_lookup_memory(h2,
340: (char *)res, sizeof(res->page),
341: res->page)) == NULL)
342: free(res);
343: else
344: ohash_insert(hand, ohash_lookup_memory(hand,
345: (char *)res, sizeof(res->page),
346: res->page), res);
1.22 schwarze 347: }
348:
1.66 schwarze 349: /* Discard the merged results. */
1.22 schwarze 350:
1.66 schwarze 351: for (res = ohash_first(h2, &slot2); res != NULL;
352: res = ohash_next(h2, &slot2))
353: free(res);
354: ohash_delete(h2);
355: free(h2);
356: ohash_delete(h1);
357: free(h1);
358: }
1.17 schwarze 359:
1.66 schwarze 360: /* Merge the result of the AND into htab. */
1.17 schwarze 361:
1.66 schwarze 362: if (htab == NULL)
363: return hand;
364:
365: for (res = ohash_first(hand, &slot1); res != NULL;
366: res = ohash_next(hand, &slot1)) {
367: slot2 = ohash_lookup_memory(htab,
368: (char *)res, sizeof(res->page), res->page);
369: if (ohash_find(htab, slot2) == NULL)
370: ohash_insert(htab, slot2, res);
371: else
372: free(res);
373: }
374:
375: /* Discard the merged result. */
1.17 schwarze 376:
1.66 schwarze 377: ohash_delete(hand);
378: free(hand);
379: return htab;
380: }
1.54 schwarze 381:
1.66 schwarze 382: void
383: mansearch_free(struct manpage *res, size_t sz)
384: {
385: size_t i;
1.54 schwarze 386:
1.66 schwarze 387: for (i = 0; i < sz; i++) {
388: free(res[i].file);
389: free(res[i].names);
390: free(res[i].output);
1.22 schwarze 391: }
1.66 schwarze 392: free(res);
393: }
394:
395: static int
396: manpage_compare(const void *vp1, const void *vp2)
397: {
398: const struct manpage *mp1, *mp2;
1.75 schwarze 399: const char *cp1, *cp2;
400: size_t sz1, sz2;
1.66 schwarze 401: int diff;
402:
403: mp1 = vp1;
404: mp2 = vp2;
1.75 schwarze 405: if ((diff = mp2->bits - mp1->bits) ||
406: (diff = mp1->sec - mp2->sec))
407: return diff;
408:
409: /* Fall back to alphabetic ordering of names. */
410: sz1 = strcspn(mp1->names, "(");
411: sz2 = strcspn(mp2->names, "(");
412: if (sz1 < sz2)
413: sz1 = sz2;
414: if ((diff = strncasecmp(mp1->names, mp2->names, sz1)))
415: return diff;
416:
417: /* For identical names and sections, prefer arch-dependent. */
418: cp1 = strchr(mp1->names + sz1, '/');
419: cp2 = strchr(mp2->names + sz2, '/');
420: return cp1 != NULL && cp2 != NULL ? strcasecmp(cp1, cp2) :
421: cp1 != NULL ? -1 : cp2 != NULL ? 1 : 0;
1.12 schwarze 422: }
423:
424: static char *
1.66 schwarze 425: buildnames(const struct dbm_page *page)
1.12 schwarze 426: {
1.66 schwarze 427: char *buf;
428: size_t i, sz;
1.12 schwarze 429:
1.70 schwarze 430: sz = lstlen(page->name, 2) + 1 + lstlen(page->sect, 2) +
431: (page->arch == NULL ? 0 : 1 + lstlen(page->arch, 2)) + 2;
1.66 schwarze 432: buf = mandoc_malloc(sz);
433: i = 0;
1.70 schwarze 434: lstcat(buf, &i, page->name, ", ");
1.66 schwarze 435: buf[i++] = '(';
1.70 schwarze 436: lstcat(buf, &i, page->sect, ", ");
1.66 schwarze 437: if (page->arch != NULL) {
438: buf[i++] = '/';
1.70 schwarze 439: lstcat(buf, &i, page->arch, ", ");
1.66 schwarze 440: }
441: buf[i++] = ')';
442: buf[i++] = '\0';
443: assert(i == sz);
444: return buf;
1.1 kristaps 445: }
446:
447: /*
1.66 schwarze 448: * Count the buffer space needed to print the NUL-terminated
1.70 schwarze 449: * list of NUL-terminated strings, when printing sep separator
1.66 schwarze 450: * characters between strings.
1.7 schwarze 451: */
1.66 schwarze 452: static size_t
1.70 schwarze 453: lstlen(const char *cp, size_t sep)
1.7 schwarze 454: {
1.66 schwarze 455: size_t sz;
1.7 schwarze 456:
1.76 ! schwarze 457: for (sz = 0; *cp != '\0'; cp++) {
! 458:
! 459: /* Skip names appearing only in the SYNOPSIS. */
! 460: if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
! 461: while (*cp != '\0')
! 462: cp++;
! 463: continue;
! 464: }
! 465:
! 466: /* Skip name class markers. */
! 467: if (*cp < ' ')
! 468: cp++;
! 469:
! 470: /* Print a separator before each but the first string. */
! 471: if (sz)
! 472: sz += sep;
! 473:
! 474: /* Copy one string. */
! 475: while (*cp != '\0') {
! 476: sz++;
! 477: cp++;
! 478: }
1.66 schwarze 479: }
480: return sz;
1.7 schwarze 481: }
482:
483: /*
1.66 schwarze 484: * Print the NUL-terminated list of NUL-terminated strings
1.70 schwarze 485: * into the buffer, seperating strings with sep.
1.8 schwarze 486: */
487: static void
1.70 schwarze 488: lstcat(char *buf, size_t *i, const char *cp, const char *sep)
1.8 schwarze 489: {
1.76 ! schwarze 490: const char *s;
! 491: size_t i_start;
1.70 schwarze 492:
1.76 ! schwarze 493: for (i_start = *i; *cp != '\0'; cp++) {
! 494:
! 495: /* Skip names appearing only in the SYNOPSIS. */
! 496: if (*cp <= (char)(NAME_SYN & NAME_MASK)) {
! 497: while (*cp != '\0')
! 498: cp++;
! 499: continue;
! 500: }
! 501:
! 502: /* Skip name class markers. */
! 503: if (*cp < ' ')
! 504: cp++;
! 505:
! 506: /* Print a separator before each but the first string. */
! 507: if (*i > i_start) {
1.70 schwarze 508: s = sep;
509: while (*s != '\0')
510: buf[(*i)++] = *s++;
1.76 ! schwarze 511: }
! 512:
! 513: /* Copy one string. */
! 514: while (*cp != '\0')
! 515: buf[(*i)++] = *cp++;
1.66 schwarze 516: }
1.76 ! schwarze 517:
1.8 schwarze 518: }
519:
1.66 schwarze 520: /*
521: * Return 1 if the string *want occurs in any of the strings
522: * in the NUL-terminated string list *have, or 0 otherwise.
523: * If either argument is NULL or empty, assume no filtering
524: * is desired and return 1.
525: */
526: static int
527: lstmatch(const char *want, const char *have)
1.13 schwarze 528: {
1.66 schwarze 529: if (want == NULL || have == NULL || *have == '\0')
530: return 1;
531: while (*have != '\0') {
532: if (strcasestr(have, want) != NULL)
533: return 1;
534: have = strchr(have, '\0') + 1;
535: }
536: return 0;
1.13 schwarze 537: }
538:
1.8 schwarze 539: /*
1.70 schwarze 540: * Build a list of values taken by the macro im in the manual page.
1.1 kristaps 541: */
542: static char *
1.70 schwarze 543: buildoutput(size_t im, struct dbm_page *page)
1.1 kristaps 544: {
1.70 schwarze 545: const char *oldoutput, *sep, *input;
1.66 schwarze 546: char *output, *newoutput, *value;
1.70 schwarze 547: size_t sz, i;
548:
549: switch (im) {
550: case KEY_Nd:
551: return mandoc_strdup(page->desc);
552: case KEY_Nm:
553: input = page->name;
554: break;
555: case KEY_sec:
556: input = page->sect;
557: break;
558: case KEY_arch:
559: input = page->arch;
560: if (input == NULL)
561: input = "all\0";
562: break;
563: default:
564: input = NULL;
565: break;
566: }
567:
568: if (input != NULL) {
569: sz = lstlen(input, 3) + 1;
570: output = mandoc_malloc(sz);
571: i = 0;
572: lstcat(output, &i, input, " # ");
1.71 schwarze 573: output[i++] = '\0';
574: assert(i == sz);
1.70 schwarze 575: return output;
576: }
1.66 schwarze 577:
578: output = NULL;
1.70 schwarze 579: dbm_macro_bypage(im - 2, page->addr);
1.66 schwarze 580: while ((value = dbm_macro_next()) != NULL) {
581: if (output == NULL) {
582: oldoutput = "";
583: sep = "";
584: } else {
585: oldoutput = output;
586: sep = " # ";
587: }
588: mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value);
589: free(output);
590: output = newoutput;
1.1 kristaps 591: }
1.66 schwarze 592: return output;
1.1 kristaps 593: }
594:
595: /*
596: * Compile a set of string tokens into an expression.
597: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
598: * "(", "foo=bar", etc.).
599: */
600: static struct expr *
1.66 schwarze 601: exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 602: {
1.66 schwarze 603: struct expr *parent, *child;
604: int needterm, nested;
605:
606: if ((nested = *argi) == argc)
607: return NULL;
608: needterm = 1;
609: parent = child = NULL;
610: while (*argi < argc) {
611: if (strcmp(")", argv[*argi]) == 0) {
612: if (needterm)
613: warnx("missing term "
614: "before closing parenthesis");
615: needterm = 0;
616: if (nested)
617: break;
618: warnx("ignoring unmatched right parenthesis");
619: ++*argi;
1.13 schwarze 620: continue;
1.66 schwarze 621: }
622: if (strcmp("-o", argv[*argi]) == 0) {
623: if (needterm) {
624: if (*argi > 0)
625: warnx("ignoring -o after %s",
626: argv[*argi - 1]);
627: else
628: warnx("ignoring initial -o");
629: }
630: needterm = 1;
631: ++*argi;
1.13 schwarze 632: continue;
1.66 schwarze 633: }
634: needterm = 0;
635: if (child == NULL) {
636: child = expr_and(search, argc, argv, argi);
1.13 schwarze 637: continue;
1.1 kristaps 638: }
1.66 schwarze 639: if (parent == NULL) {
640: parent = mandoc_calloc(1, sizeof(*parent));
641: parent->type = EXPR_OR;
642: parent->next = NULL;
643: parent->child = child;
644: }
645: child->next = expr_and(search, argc, argv, argi);
646: child = child->next;
647: }
648: if (needterm && *argi)
649: warnx("ignoring trailing %s", argv[*argi - 1]);
650: return parent == NULL ? child : parent;
651: }
1.26 schwarze 652:
1.66 schwarze 653: static struct expr *
654: expr_and(const struct mansearch *search, int argc, char *argv[], int *argi)
655: {
656: struct expr *parent, *child;
657: int needterm;
1.26 schwarze 658:
1.66 schwarze 659: needterm = 1;
660: parent = child = NULL;
661: while (*argi < argc) {
662: if (strcmp(")", argv[*argi]) == 0) {
663: if (needterm)
664: warnx("missing term "
665: "before closing parenthesis");
666: needterm = 0;
667: break;
668: }
669: if (strcmp("-o", argv[*argi]) == 0)
670: break;
671: if (strcmp("-a", argv[*argi]) == 0) {
672: if (needterm) {
673: if (*argi > 0)
674: warnx("ignoring -a after %s",
675: argv[*argi - 1]);
676: else
677: warnx("ignoring initial -a");
1.27 schwarze 678: }
1.66 schwarze 679: needterm = 1;
680: ++*argi;
681: continue;
1.27 schwarze 682: }
1.66 schwarze 683: if (needterm == 0)
684: break;
685: if (child == NULL) {
686: child = exprterm(search, argc, argv, argi);
687: if (child != NULL)
688: needterm = 0;
689: continue;
690: }
691: needterm = 0;
692: if (parent == NULL) {
693: parent = mandoc_calloc(1, sizeof(*parent));
694: parent->type = EXPR_AND;
695: parent->next = NULL;
696: parent->child = child;
697: }
698: child->next = exprterm(search, argc, argv, argi);
699: if (child->next != NULL) {
700: child = child->next;
701: needterm = 0;
702: }
703: }
704: if (needterm && *argi)
705: warnx("ignoring trailing %s", argv[*argi - 1]);
706: return parent == NULL ? child : parent;
1.15 schwarze 707: }
708:
709: static struct expr *
1.66 schwarze 710: exprterm(const struct mansearch *search, int argc, char *argv[], int *argi)
1.1 kristaps 711: {
1.15 schwarze 712: char errbuf[BUFSIZ];
1.1 kristaps 713: struct expr *e;
1.38 schwarze 714: char *key, *val;
1.20 schwarze 715: uint64_t iterbit;
1.66 schwarze 716: int cs, i, irc;
1.1 kristaps 717:
1.66 schwarze 718: if (strcmp("(", argv[*argi]) == 0) {
719: ++*argi;
720: e = exprcomp(search, argc, argv, argi);
721: if (*argi < argc) {
722: assert(strcmp(")", argv[*argi]) == 0);
723: ++*argi;
724: } else
725: warnx("unclosed parenthesis");
726: return e;
727: }
1.1 kristaps 728:
1.72 schwarze 729: if (strcmp("-i", argv[*argi]) == 0 && *argi + 1 < argc) {
730: cs = 0;
731: ++*argi;
732: } else
733: cs = 1;
734:
1.66 schwarze 735: e = mandoc_calloc(1, sizeof(*e));
736: e->type = EXPR_TERM;
737: e->bits = 0;
738: e->next = NULL;
739: e->child = NULL;
1.1 kristaps 740:
1.45 schwarze 741: if (search->argmode == ARG_NAME) {
742: e->bits = TYPE_Nm;
1.66 schwarze 743: e->match.type = DBM_EXACT;
744: e->match.str = argv[(*argi)++];
1.58 schwarze 745: return e;
1.5 kristaps 746: }
747:
1.1 kristaps 748: /*
1.45 schwarze 749: * Separate macro keys from search string.
1.66 schwarze 750: * If needed, request regular expression handling.
1.1 kristaps 751: */
752:
1.45 schwarze 753: if (search->argmode == ARG_WORD) {
754: e->bits = TYPE_Nm;
1.66 schwarze 755: e->match.type = DBM_REGEX;
1.61 schwarze 756: #if HAVE_REWB_BSD
1.66 schwarze 757: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]);
1.61 schwarze 758: #elif HAVE_REWB_SYSV
1.66 schwarze 759: mandoc_asprintf(&val, "\\<%s\\>", argv[*argi]);
1.61 schwarze 760: #else
761: mandoc_asprintf(&val,
1.66 schwarze 762: "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", argv[*argi]);
1.61 schwarze 763: #endif
1.46 schwarze 764: cs = 0;
1.66 schwarze 765: } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) {
1.45 schwarze 766: e->bits = TYPE_Nm | TYPE_Nd;
1.66 schwarze 767: e->match.type = DBM_SUB;
768: e->match.str = argv[*argi];
1.38 schwarze 769: } else {
1.66 schwarze 770: if (val == argv[*argi])
1.45 schwarze 771: e->bits = TYPE_Nm | TYPE_Nd;
1.66 schwarze 772: if (*val == '=') {
773: e->match.type = DBM_SUB;
774: e->match.str = val + 1;
775: } else
776: e->match.type = DBM_REGEX;
1.38 schwarze 777: *val++ = '\0';
1.66 schwarze 778: if (strstr(argv[*argi], "arch") != NULL)
1.21 schwarze 779: cs = 0;
1.38 schwarze 780: }
781:
782: /* Compile regular expressions. */
783:
1.66 schwarze 784: if (e->match.type == DBM_REGEX) {
785: e->match.re = mandoc_malloc(sizeof(*e->match.re));
786: irc = regcomp(e->match.re, val,
1.38 schwarze 787: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.66 schwarze 788: if (irc) {
789: regerror(irc, e->match.re, errbuf, sizeof(errbuf));
790: warnx("regcomp /%s/: %s", val, errbuf);
791: }
1.45 schwarze 792: if (search->argmode == ARG_WORD)
1.38 schwarze 793: free(val);
794: if (irc) {
1.66 schwarze 795: free(e->match.re);
1.8 schwarze 796: free(e);
1.66 schwarze 797: ++*argi;
1.58 schwarze 798: return NULL;
1.8 schwarze 799: }
1.38 schwarze 800: }
801:
1.66 schwarze 802: if (e->bits) {
803: ++*argi;
1.58 schwarze 804: return e;
1.66 schwarze 805: }
1.1 kristaps 806:
807: /*
808: * Parse out all possible fields.
809: * If the field doesn't resolve, bail.
810: */
811:
1.66 schwarze 812: while (NULL != (key = strsep(&argv[*argi], ","))) {
1.1 kristaps 813: if ('\0' == *key)
814: continue;
1.66 schwarze 815: for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) {
816: if (0 == strcasecmp(key, mansearch_keynames[i])) {
1.20 schwarze 817: e->bits |= iterbit;
818: break;
819: }
820: }
1.66 schwarze 821: if (i == KEY_MAX) {
822: if (strcasecmp(key, "any"))
823: warnx("treating unknown key "
824: "\"%s\" as \"any\"", key);
1.20 schwarze 825: e->bits |= ~0ULL;
1.1 kristaps 826: }
827: }
828:
1.66 schwarze 829: ++*argi;
1.58 schwarze 830: return e;
1.1 kristaps 831: }
832:
833: static void
1.66 schwarze 834: exprfree(struct expr *e)
1.1 kristaps 835: {
1.66 schwarze 836: if (e->next != NULL)
837: exprfree(e->next);
838: if (e->child != NULL)
839: exprfree(e->child);
840: free(e);
1.1 kristaps 841: }
CVSweb