Annotation of src/usr.bin/mandoc/mansearch.c, Revision 1.24
1.24 ! schwarze 1: /* $Id: mansearch.c,v 1.23 2014/04/17 19:19:54 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.4 schwarze 4: * Copyright (c) 2013, 2014 Ingo Schwarze <schwarze@openbsd.org>
1.1 schwarze 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: *
10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
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: */
1.19 schwarze 18: #include <sys/mman.h>
1.1 schwarze 19: #include <assert.h>
20: #include <fcntl.h>
21: #include <getopt.h>
22: #include <limits.h>
23: #include <regex.h>
24: #include <stdio.h>
25: #include <stdint.h>
26: #include <stddef.h>
27: #include <stdlib.h>
28: #include <string.h>
29: #include <unistd.h>
30:
31: #include <ohash.h>
32: #include <sqlite3.h>
33:
34: #include "mandoc.h"
1.14 schwarze 35: #include "mandoc_aux.h"
1.1 schwarze 36: #include "manpath.h"
37: #include "mansearch.h"
38:
1.11 schwarze 39: extern int mansearch_keymax;
40: extern const char *const mansearch_keynames[];
41:
1.1 schwarze 42: #define SQL_BIND_TEXT(_db, _s, _i, _v) \
43: do { if (SQLITE_OK != sqlite3_bind_text \
44: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
45: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
46: } while (0)
47: #define SQL_BIND_INT64(_db, _s, _i, _v) \
48: do { if (SQLITE_OK != sqlite3_bind_int64 \
49: ((_s), (_i)++, (_v))) \
50: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
51: } while (0)
52: #define SQL_BIND_BLOB(_db, _s, _i, _v) \
53: do { if (SQLITE_OK != sqlite3_bind_blob \
54: ((_s), (_i)++, (&_v), sizeof(_v), SQLITE_STATIC)) \
55: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
56: } while (0)
57:
58: struct expr {
1.24 ! schwarze 59: uint64_t bits; /* type-mask */
1.1 schwarze 60: const char *substr; /* to search for, if applicable */
61: regex_t regexp; /* compiled regexp, if applicable */
1.4 schwarze 62: int open; /* opening parentheses before */
63: int and; /* logical AND before */
64: int close; /* closing parentheses after */
1.1 schwarze 65: struct expr *next; /* next in sequence */
66: };
67:
68: struct match {
1.22 schwarze 69: uint64_t pageid; /* identifier in database */
1.17 schwarze 70: char *desc; /* manual page description */
1.1 schwarze 71: int form; /* 0 == catpage */
72: };
73:
1.8 schwarze 74: static void buildnames(struct manpage *, sqlite3 *,
1.10 schwarze 75: sqlite3_stmt *, uint64_t,
76: const char *, int form);
1.3 schwarze 77: static char *buildoutput(sqlite3 *, sqlite3_stmt *,
78: uint64_t, uint64_t);
1.1 schwarze 79: static void *hash_alloc(size_t, void *);
80: static void hash_free(void *, size_t, void *);
81: static void *hash_halloc(size_t, void *);
1.24 ! schwarze 82: static struct expr *exprcomp(const struct mansearch *,
1.1 schwarze 83: int, char *[]);
84: static void exprfree(struct expr *);
1.6 schwarze 85: static struct expr *exprspec(struct expr *, uint64_t,
86: const char *, const char *);
1.1 schwarze 87: static struct expr *exprterm(const struct mansearch *, char *, int);
1.4 schwarze 88: static void sql_append(char **sql, size_t *sz,
89: const char *newstr, int count);
1.1 schwarze 90: static void sql_match(sqlite3_context *context,
91: int argc, sqlite3_value **argv);
92: static void sql_regexp(sqlite3_context *context,
93: int argc, sqlite3_value **argv);
1.6 schwarze 94: static char *sql_statement(const struct expr *);
1.19 schwarze 95:
1.24 ! schwarze 96:
1.19 schwarze 97: int
98: mansearch_setup(int start)
99: {
100: static void *pagecache;
101: int c;
102:
103: #define PC_PAGESIZE 1280
104: #define PC_NUMPAGES 256
105:
106: if (start) {
107: if (NULL != pagecache) {
108: fprintf(stderr, "pagecache already enabled\n");
109: return((int)MANDOCLEVEL_BADARG);
110: }
111:
112: pagecache = mmap(NULL, PC_PAGESIZE * PC_NUMPAGES,
113: PROT_READ | PROT_WRITE, MAP_ANON, -1, 0);
114:
115: if (MAP_FAILED == pagecache) {
116: perror("mmap");
117: pagecache = NULL;
118: return((int)MANDOCLEVEL_SYSERR);
119: }
120:
121: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
122: pagecache, PC_PAGESIZE, PC_NUMPAGES);
123:
124: if (SQLITE_OK == c)
125: return((int)MANDOCLEVEL_OK);
126:
127: fprintf(stderr, "pagecache: %s\n", sqlite3_errstr(c));
128:
129: } else if (NULL == pagecache) {
130: fprintf(stderr, "pagecache missing\n");
131: return((int)MANDOCLEVEL_BADARG);
132: }
133:
134: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
135: perror("munmap");
136: pagecache = NULL;
137: return((int)MANDOCLEVEL_SYSERR);
138: }
139:
140: pagecache = NULL;
141: return((int)MANDOCLEVEL_OK);
142: }
1.1 schwarze 143:
144: int
145: mansearch(const struct mansearch *search,
1.3 schwarze 146: const struct manpaths *paths,
147: int argc, char *argv[],
148: const char *outkey,
1.1 schwarze 149: struct manpage **res, size_t *sz)
150: {
1.11 schwarze 151: int fd, rc, c, indexbit;
1.22 schwarze 152: int64_t pageid;
1.11 schwarze 153: uint64_t outbit, iterbit;
1.1 schwarze 154: char buf[PATH_MAX];
1.2 schwarze 155: char *sql;
1.1 schwarze 156: struct manpage *mpage;
157: struct expr *e, *ep;
158: sqlite3 *db;
1.3 schwarze 159: sqlite3_stmt *s, *s2;
1.1 schwarze 160: struct match *mp;
161: struct ohash_info info;
162: struct ohash htab;
163: unsigned int idx;
164: size_t i, j, cur, maxres;
165:
166: memset(&info, 0, sizeof(struct ohash_info));
167:
168: info.halloc = hash_halloc;
169: info.alloc = hash_alloc;
170: info.hfree = hash_free;
1.22 schwarze 171: info.key_offset = offsetof(struct match, pageid);
1.1 schwarze 172:
173: *sz = cur = maxres = 0;
174: sql = NULL;
175: *res = NULL;
176: fd = -1;
177: e = NULL;
178: rc = 0;
179:
180: if (0 == argc)
181: goto out;
182: if (NULL == (e = exprcomp(search, argc, argv)))
183: goto out;
184:
1.3 schwarze 185: outbit = 0;
186: if (NULL != outkey) {
1.11 schwarze 187: for (indexbit = 0, iterbit = 1;
188: indexbit < mansearch_keymax;
189: indexbit++, iterbit <<= 1) {
190: if (0 == strcasecmp(outkey,
191: mansearch_keynames[indexbit])) {
192: outbit = iterbit;
1.3 schwarze 193: break;
194: }
195: }
196: }
197:
1.1 schwarze 198: /*
199: * Save a descriptor to the current working directory.
200: * Since pathnames in the "paths" variable might be relative,
201: * and we'll be chdir()ing into them, we need to keep a handle
202: * on our current directory from which to start the chdir().
203: */
204:
205: if (NULL == getcwd(buf, PATH_MAX)) {
1.20 schwarze 206: perror("getcwd");
1.1 schwarze 207: goto out;
208: } else if (-1 == (fd = open(buf, O_RDONLY, 0))) {
209: perror(buf);
210: goto out;
211: }
212:
1.6 schwarze 213: sql = sql_statement(e);
1.1 schwarze 214:
215: /*
216: * Loop over the directories (containing databases) for us to
217: * search.
218: * Don't let missing/bad databases/directories phase us.
219: * In each, try to open the resident database and, if it opens,
220: * scan it for our match expression.
221: */
222:
223: for (i = 0; i < paths->sz; i++) {
224: if (-1 == fchdir(fd)) {
225: perror(buf);
226: free(*res);
227: break;
228: } else if (-1 == chdir(paths->paths[i])) {
229: perror(paths->paths[i]);
230: continue;
1.24 ! schwarze 231: }
1.1 schwarze 232:
1.24 ! schwarze 233: c = sqlite3_open_v2(MANDOC_DB, &db,
! 234: SQLITE_OPEN_READONLY, NULL);
1.1 schwarze 235:
236: if (SQLITE_OK != c) {
237: perror(MANDOC_DB);
238: sqlite3_close(db);
239: continue;
240: }
241:
242: /*
243: * Define the SQL functions for substring
244: * and regular expression matching.
245: */
246:
247: c = sqlite3_create_function(db, "match", 2,
1.21 schwarze 248: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
249: NULL, sql_match, NULL, NULL);
1.1 schwarze 250: assert(SQLITE_OK == c);
251: c = sqlite3_create_function(db, "regexp", 2,
1.21 schwarze 252: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
253: NULL, sql_regexp, NULL, NULL);
1.1 schwarze 254: assert(SQLITE_OK == c);
255:
256: j = 1;
257: c = sqlite3_prepare_v2(db, sql, -1, &s, NULL);
258: if (SQLITE_OK != c)
259: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
260:
261: for (ep = e; NULL != ep; ep = ep->next) {
262: if (NULL == ep->substr) {
263: SQL_BIND_BLOB(db, s, j, ep->regexp);
264: } else
265: SQL_BIND_TEXT(db, s, j, ep->substr);
1.18 schwarze 266: if (0 == ((TYPE_Nd | TYPE_Nm) & ep->bits))
1.17 schwarze 267: SQL_BIND_INT64(db, s, j, ep->bits);
1.1 schwarze 268: }
269:
270: memset(&htab, 0, sizeof(struct ohash));
271: ohash_init(&htab, 4, &info);
272:
273: /*
274: * Hash each entry on its [unique] document identifier.
275: * This is a uint64_t.
276: * Instead of using a hash function, simply convert the
277: * uint64_t to a uint32_t, the hash value's type.
278: * This gives good performance and preserves the
279: * distribution of buckets in the table.
280: */
281: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.22 schwarze 282: pageid = sqlite3_column_int64(s, 2);
1.24 ! schwarze 283: idx = ohash_lookup_memory(&htab,
! 284: (char *)&pageid, sizeof(uint64_t),
! 285: (uint32_t)pageid);
1.1 schwarze 286:
287: if (NULL != ohash_find(&htab, idx))
288: continue;
289:
290: mp = mandoc_calloc(1, sizeof(struct match));
1.22 schwarze 291: mp->pageid = pageid;
1.17 schwarze 292: mp->form = sqlite3_column_int(s, 1);
293: if (TYPE_Nd == outbit)
294: mp->desc = mandoc_strdup(
295: sqlite3_column_text(s, 0));
1.1 schwarze 296: ohash_insert(&htab, idx, mp);
297: }
298:
299: if (SQLITE_DONE != c)
300: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
301:
302: sqlite3_finalize(s);
303:
1.24 ! schwarze 304: c = sqlite3_prepare_v2(db,
1.13 schwarze 305: "SELECT * FROM mlinks WHERE pageid=?"
306: " ORDER BY sec, arch, name",
1.1 schwarze 307: -1, &s, NULL);
308: if (SQLITE_OK != c)
309: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
310:
1.3 schwarze 311: c = sqlite3_prepare_v2(db,
312: "SELECT * FROM keys WHERE pageid=? AND bits & ?",
313: -1, &s2, NULL);
314: if (SQLITE_OK != c)
315: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
316:
1.1 schwarze 317: for (mp = ohash_first(&htab, &idx);
318: NULL != mp;
319: mp = ohash_next(&htab, &idx)) {
320: if (cur + 1 > maxres) {
321: maxres += 1024;
1.24 ! schwarze 322: *res = mandoc_realloc(*res,
! 323: maxres * sizeof(struct manpage));
1.1 schwarze 324: }
325: mpage = *res + cur;
326: mpage->form = mp->form;
1.22 schwarze 327: buildnames(mpage, db, s, mp->pageid,
1.10 schwarze 328: paths->paths[i], mp->form);
1.17 schwarze 329: mpage->output = TYPE_Nd & outbit ?
330: mp->desc : outbit ?
1.22 schwarze 331: buildoutput(db, s2, mp->pageid, outbit) : NULL;
1.1 schwarze 332:
333: free(mp);
334: cur++;
335: }
336:
337: sqlite3_finalize(s);
1.3 schwarze 338: sqlite3_finalize(s2);
1.1 schwarze 339: sqlite3_close(db);
340: ohash_delete(&htab);
341: }
342: rc = 1;
343: out:
1.20 schwarze 344: if (-1 != fd) {
345: if (-1 == fchdir(fd))
346: perror(buf);
347: close(fd);
348: }
1.1 schwarze 349: exprfree(e);
350: free(sql);
351: *sz = cur;
352: return(rc);
1.2 schwarze 353: }
354:
1.8 schwarze 355: static void
356: buildnames(struct manpage *mpage, sqlite3 *db, sqlite3_stmt *s,
1.22 schwarze 357: uint64_t pageid, const char *path, int form)
1.2 schwarze 358: {
1.13 schwarze 359: char *newnames, *prevsec, *prevarch;
1.10 schwarze 360: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.2 schwarze 361: size_t i;
362: int c;
363:
1.16 schwarze 364: mpage->file = NULL;
1.8 schwarze 365: mpage->names = NULL;
1.13 schwarze 366: prevsec = prevarch = NULL;
1.2 schwarze 367: i = 1;
1.22 schwarze 368: SQL_BIND_INT64(db, s, i, pageid);
1.2 schwarze 369: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.8 schwarze 370:
1.13 schwarze 371: /* Decide whether we already have some names. */
1.8 schwarze 372:
373: if (NULL == mpage->names) {
1.2 schwarze 374: oldnames = "";
375: sep1 = "";
376: } else {
1.8 schwarze 377: oldnames = mpage->names;
1.2 schwarze 378: sep1 = ", ";
379: }
1.13 schwarze 380:
381: /* Fetch the next name. */
382:
1.10 schwarze 383: sec = sqlite3_column_text(s, 0);
384: arch = sqlite3_column_text(s, 1);
385: name = sqlite3_column_text(s, 2);
1.13 schwarze 386:
387: /* If the section changed, append the old one. */
388:
389: if (NULL != prevsec &&
390: (strcmp(sec, prevsec) ||
391: strcmp(arch, prevarch))) {
392: sep2 = '\0' == *prevarch ? "" : "/";
1.15 schwarze 393: mandoc_asprintf(&newnames, "%s(%s%s%s)",
394: oldnames, prevsec, sep2, prevarch);
1.13 schwarze 395: free(mpage->names);
396: oldnames = mpage->names = newnames;
397: free(prevsec);
398: free(prevarch);
399: prevsec = prevarch = NULL;
400: }
401:
402: /* Save the new section, to append it later. */
403:
404: if (NULL == prevsec) {
405: prevsec = mandoc_strdup(sec);
406: prevarch = mandoc_strdup(arch);
407: }
408:
409: /* Append the new name. */
410:
1.15 schwarze 411: mandoc_asprintf(&newnames, "%s%s%s",
412: oldnames, sep1, name);
1.8 schwarze 413: free(mpage->names);
414: mpage->names = newnames;
415:
416: /* Also save the first file name encountered. */
417:
418: if (NULL != mpage->file)
419: continue;
420:
1.10 schwarze 421: if (form) {
422: sep1 = "man";
423: fsec = sec;
424: } else {
425: sep1 = "cat";
426: fsec = "0";
427: }
1.13 schwarze 428: sep2 = '\0' == *arch ? "" : "/";
1.15 schwarze 429: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
430: path, sep1, sec, sep2, arch, name, fsec);
1.2 schwarze 431: }
432: if (SQLITE_DONE != c)
433: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
434: sqlite3_reset(s);
1.13 schwarze 435:
436: /* Append one final section to the names. */
437:
438: if (NULL != prevsec) {
439: sep2 = '\0' == *prevarch ? "" : "/";
1.15 schwarze 440: mandoc_asprintf(&newnames, "%s(%s%s%s)",
441: mpage->names, prevsec, sep2, prevarch);
1.13 schwarze 442: free(mpage->names);
443: mpage->names = newnames;
444: free(prevsec);
445: free(prevarch);
446: }
1.3 schwarze 447: }
448:
449: static char *
1.22 schwarze 450: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.3 schwarze 451: {
452: char *output, *newoutput;
453: const char *oldoutput, *sep1, *data;
454: size_t i;
455: int c;
456:
457: output = NULL;
458: i = 1;
1.22 schwarze 459: SQL_BIND_INT64(db, s, i, pageid);
1.3 schwarze 460: SQL_BIND_INT64(db, s, i, outbit);
461: while (SQLITE_ROW == (c = sqlite3_step(s))) {
462: if (NULL == output) {
463: oldoutput = "";
464: sep1 = "";
465: } else {
466: oldoutput = output;
467: sep1 = " # ";
468: }
469: data = sqlite3_column_text(s, 1);
1.15 schwarze 470: mandoc_asprintf(&newoutput, "%s%s%s",
471: oldoutput, sep1, data);
1.3 schwarze 472: free(output);
473: output = newoutput;
474: }
475: if (SQLITE_DONE != c)
476: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
477: sqlite3_reset(s);
478: return(output);
1.1 schwarze 479: }
480:
481: /*
482: * Implement substring match as an application-defined SQL function.
483: * Using the SQL LIKE or GLOB operators instead would be a bad idea
484: * because that would require escaping metacharacters in the string
485: * being searched for.
486: */
487: static void
488: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
489: {
490:
491: assert(2 == argc);
492: sqlite3_result_int(context, NULL != strcasestr(
493: (const char *)sqlite3_value_text(argv[1]),
494: (const char *)sqlite3_value_text(argv[0])));
495: }
496:
497: /*
498: * Implement regular expression match
499: * as an application-defined SQL function.
500: */
501: static void
502: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
503: {
504:
505: assert(2 == argc);
506: sqlite3_result_int(context, !regexec(
507: (regex_t *)sqlite3_value_blob(argv[0]),
508: (const char *)sqlite3_value_text(argv[1]),
509: 0, NULL, 0));
510: }
511:
1.4 schwarze 512: static void
513: sql_append(char **sql, size_t *sz, const char *newstr, int count)
514: {
515: size_t newsz;
516:
517: newsz = 1 < count ? (size_t)count : strlen(newstr);
518: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
519: if (1 < count)
520: memset(*sql + *sz, *newstr, (size_t)count);
521: else
522: memcpy(*sql + *sz, newstr, newsz);
523: *sz += newsz;
524: (*sql)[*sz] = '\0';
525: }
526:
1.1 schwarze 527: /*
528: * Prepare the search SQL statement.
529: */
530: static char *
1.6 schwarze 531: sql_statement(const struct expr *e)
1.1 schwarze 532: {
533: char *sql;
534: size_t sz;
1.4 schwarze 535: int needop;
1.1 schwarze 536:
1.4 schwarze 537: sql = mandoc_strdup("SELECT * FROM mpages WHERE ");
1.1 schwarze 538: sz = strlen(sql);
539:
1.4 schwarze 540: for (needop = 0; NULL != e; e = e->next) {
541: if (e->and)
542: sql_append(&sql, &sz, " AND ", 1);
543: else if (needop)
544: sql_append(&sql, &sz, " OR ", 1);
545: if (e->open)
546: sql_append(&sql, &sz, "(", e->open);
1.17 schwarze 547: sql_append(&sql, &sz,
548: TYPE_Nd & e->bits
549: ? (NULL == e->substr
550: ? "desc REGEXP ?"
551: : "desc MATCH ?")
1.18 schwarze 552: : TYPE_Nm == e->bits
553: ? (NULL == e->substr
1.22 schwarze 554: ? "pageid IN (SELECT pageid FROM names "
1.18 schwarze 555: "WHERE name REGEXP ?)"
1.22 schwarze 556: : "pageid IN (SELECT pageid FROM names "
1.18 schwarze 557: "WHERE name MATCH ?)")
1.17 schwarze 558: : (NULL == e->substr
1.22 schwarze 559: ? "pageid IN (SELECT pageid FROM keys "
1.17 schwarze 560: "WHERE key REGEXP ? AND bits & ?)"
1.22 schwarze 561: : "pageid IN (SELECT pageid FROM keys "
1.17 schwarze 562: "WHERE key MATCH ? AND bits & ?)"), 1);
1.4 schwarze 563: if (e->close)
564: sql_append(&sql, &sz, ")", e->close);
565: needop = 1;
1.1 schwarze 566: }
567:
568: return(sql);
569: }
570:
571: /*
572: * Compile a set of string tokens into an expression.
573: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
574: * "(", "foo=bar", etc.).
575: */
576: static struct expr *
577: exprcomp(const struct mansearch *search, int argc, char *argv[])
578: {
1.18 schwarze 579: uint64_t mask;
1.4 schwarze 580: int i, toopen, logic, igncase, toclose;
1.18 schwarze 581: struct expr *first, *prev, *cur, *next;
1.1 schwarze 582:
583: first = cur = NULL;
1.6 schwarze 584: logic = igncase = toclose = 0;
1.23 schwarze 585: toopen = NULL != search->sec || NULL != search->arch;
1.1 schwarze 586:
587: for (i = 0; i < argc; i++) {
1.4 schwarze 588: if (0 == strcmp("(", argv[i])) {
589: if (igncase)
590: goto fail;
591: toopen++;
592: toclose++;
593: continue;
594: } else if (0 == strcmp(")", argv[i])) {
595: if (toopen || logic || igncase || NULL == cur)
596: goto fail;
597: cur->close++;
598: if (0 > --toclose)
599: goto fail;
600: continue;
601: } else if (0 == strcmp("-a", argv[i])) {
602: if (toopen || logic || igncase || NULL == cur)
603: goto fail;
604: logic = 1;
605: continue;
606: } else if (0 == strcmp("-o", argv[i])) {
607: if (toopen || logic || igncase || NULL == cur)
608: goto fail;
609: logic = 2;
610: continue;
611: } else if (0 == strcmp("-i", argv[i])) {
612: if (igncase)
613: goto fail;
614: igncase = 1;
615: continue;
1.1 schwarze 616: }
1.4 schwarze 617: next = exprterm(search, argv[i], !igncase);
618: if (NULL == next)
619: goto fail;
1.17 schwarze 620: if (NULL == first)
621: first = next;
622: else
1.1 schwarze 623: cur->next = next;
1.18 schwarze 624: prev = cur = next;
1.17 schwarze 625:
626: /*
627: * Searching for descriptions must be split out
628: * because they are stored in the mpages table,
629: * not in the keys table.
630: */
631:
1.18 schwarze 632: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
633: if (mask & cur->bits && ~mask & cur->bits) {
634: next = mandoc_calloc(1,
635: sizeof(struct expr));
636: memcpy(next, cur, sizeof(struct expr));
637: prev->open = 1;
638: cur->bits = mask;
639: cur->next = next;
640: cur = next;
641: cur->bits &= ~mask;
642: }
643: }
644: prev->and = (1 == logic);
645: prev->open += toopen;
646: if (cur != prev)
1.17 schwarze 647: cur->close = 1;
1.18 schwarze 648:
1.4 schwarze 649: toopen = logic = igncase = 0;
1.1 schwarze 650: }
1.6 schwarze 651: if (toopen || logic || igncase || toclose)
652: goto fail;
653:
1.23 schwarze 654: if (NULL != search->sec || NULL != search->arch)
655: cur->close++;
656: if (NULL != search->arch)
657: cur = exprspec(cur, TYPE_arch, search->arch, "^(%s|any)$");
658: if (NULL != search->sec)
659: exprspec(cur, TYPE_sec, search->sec, "^%s$");
1.6 schwarze 660:
661: return(first);
662:
1.4 schwarze 663: fail:
664: if (NULL != first)
665: exprfree(first);
666: return(NULL);
1.1 schwarze 667: }
668:
669: static struct expr *
1.6 schwarze 670: exprspec(struct expr *cur, uint64_t key, const char *value,
671: const char *format)
672: {
673: char errbuf[BUFSIZ];
674: char *cp;
675: int irc;
676:
1.15 schwarze 677: mandoc_asprintf(&cp, format, value);
1.6 schwarze 678: cur->next = mandoc_calloc(1, sizeof(struct expr));
679: cur = cur->next;
680: cur->and = 1;
681: cur->bits = key;
682: if (0 != (irc = regcomp(&cur->regexp, cp,
683: REG_EXTENDED | REG_NOSUB | REG_ICASE))) {
684: regerror(irc, &cur->regexp, errbuf, sizeof(errbuf));
685: fprintf(stderr, "regcomp: %s\n", errbuf);
686: cur->substr = value;
687: }
688: free(cp);
689: return(cur);
690: }
691:
692: static struct expr *
1.1 schwarze 693: exprterm(const struct mansearch *search, char *buf, int cs)
694: {
1.6 schwarze 695: char errbuf[BUFSIZ];
1.1 schwarze 696: struct expr *e;
697: char *key, *v;
1.11 schwarze 698: uint64_t iterbit;
699: int i, irc;
1.1 schwarze 700:
701: if ('\0' == *buf)
702: return(NULL);
703:
704: e = mandoc_calloc(1, sizeof(struct expr));
705:
706: /*"whatis" mode uses an opaque string and default fields. */
707:
708: if (MANSEARCH_WHATIS & search->flags) {
709: e->substr = buf;
710: e->bits = search->deftype;
711: return(e);
712: }
713:
714: /*
715: * If no =~ is specified, search with equality over names and
716: * descriptions.
717: * If =~ begins the phrase, use name and description fields.
718: */
719:
720: if (NULL == (v = strpbrk(buf, "=~"))) {
721: e->substr = buf;
722: e->bits = search->deftype;
723: return(e);
724: } else if (v == buf)
725: e->bits = search->deftype;
726:
727: if ('~' == *v++) {
1.12 schwarze 728: if (NULL != strstr(buf, "arch"))
729: cs = 0;
1.6 schwarze 730: if (0 != (irc = regcomp(&e->regexp, v,
731: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE)))) {
732: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
733: fprintf(stderr, "regcomp: %s\n", errbuf);
1.1 schwarze 734: free(e);
735: return(NULL);
736: }
737: } else
738: e->substr = v;
739: v[-1] = '\0';
740:
741: /*
742: * Parse out all possible fields.
743: * If the field doesn't resolve, bail.
744: */
745:
746: while (NULL != (key = strsep(&buf, ","))) {
747: if ('\0' == *key)
748: continue;
1.11 schwarze 749: for (i = 0, iterbit = 1;
750: i < mansearch_keymax;
751: i++, iterbit <<= 1) {
752: if (0 == strcasecmp(key,
753: mansearch_keynames[i])) {
754: e->bits |= iterbit;
755: break;
756: }
757: }
758: if (i == mansearch_keymax) {
759: if (strcasecmp(key, "any")) {
760: free(e);
761: return(NULL);
762: }
763: e->bits |= ~0ULL;
1.1 schwarze 764: }
765: }
766:
767: return(e);
768: }
769:
770: static void
771: exprfree(struct expr *p)
772: {
773: struct expr *pp;
774:
775: while (NULL != p) {
776: pp = p->next;
777: free(p);
778: p = pp;
779: }
780: }
781:
782: static void *
783: hash_halloc(size_t sz, void *arg)
784: {
785:
786: return(mandoc_calloc(sz, 1));
787: }
788:
789: static void *
790: hash_alloc(size_t sz, void *arg)
791: {
792:
793: return(mandoc_malloc(sz));
794: }
795:
796: static void
797: hash_free(void *p, size_t sz, void *arg)
798: {
799:
800: free(p);
801: }