Annotation of src/usr.bin/mandoc/mansearch.c, Revision 1.50
1.50 ! schwarze 1: /* $OpenBSD: mansearch.c,v 1.49 2016/01/08 15:01:58 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.40 schwarze 4: * Copyright (c) 2013, 2014, 2015 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: *
1.43 schwarze 10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
1.1 schwarze 11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
1.43 schwarze 12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
1.1 schwarze 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: */
1.33 schwarze 18:
1.19 schwarze 19: #include <sys/mman.h>
1.33 schwarze 20: #include <sys/types.h>
21:
1.1 schwarze 22: #include <assert.h>
1.46 schwarze 23: #include <err.h>
1.39 schwarze 24: #include <errno.h>
1.1 schwarze 25: #include <fcntl.h>
1.41 schwarze 26: #include <glob.h>
1.1 schwarze 27: #include <limits.h>
28: #include <regex.h>
29: #include <stdio.h>
30: #include <stdint.h>
31: #include <stddef.h>
32: #include <stdlib.h>
33: #include <string.h>
34: #include <unistd.h>
35:
36: #include <sqlite3.h>
37:
38: #include "mandoc.h"
1.14 schwarze 39: #include "mandoc_aux.h"
1.47 schwarze 40: #include "mandoc_ohash.h"
1.43 schwarze 41: #include "manconf.h"
1.1 schwarze 42: #include "mansearch.h"
43:
1.11 schwarze 44: extern int mansearch_keymax;
45: extern const char *const mansearch_keynames[];
46:
1.1 schwarze 47: #define SQL_BIND_TEXT(_db, _s, _i, _v) \
48: do { if (SQLITE_OK != sqlite3_bind_text \
49: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.48 schwarze 50: errx((int)MANDOCLEVEL_SYSERR, "%s", sqlite3_errmsg((_db))); \
1.1 schwarze 51: } while (0)
52: #define SQL_BIND_INT64(_db, _s, _i, _v) \
53: do { if (SQLITE_OK != sqlite3_bind_int64 \
54: ((_s), (_i)++, (_v))) \
1.48 schwarze 55: errx((int)MANDOCLEVEL_SYSERR, "%s", sqlite3_errmsg((_db))); \
1.1 schwarze 56: } while (0)
57: #define SQL_BIND_BLOB(_db, _s, _i, _v) \
58: do { if (SQLITE_OK != sqlite3_bind_blob \
59: ((_s), (_i)++, (&_v), sizeof(_v), SQLITE_STATIC)) \
1.48 schwarze 60: errx((int)MANDOCLEVEL_SYSERR, "%s", sqlite3_errmsg((_db))); \
1.1 schwarze 61: } while (0)
62:
63: struct expr {
1.28 schwarze 64: regex_t regexp; /* compiled regexp, if applicable */
65: const char *substr; /* to search for, if applicable */
66: struct expr *next; /* next in sequence */
1.24 schwarze 67: uint64_t bits; /* type-mask */
1.28 schwarze 68: int equal; /* equality, not subsring match */
1.4 schwarze 69: int open; /* opening parentheses before */
70: int and; /* logical AND before */
71: int close; /* closing parentheses after */
1.1 schwarze 72: };
73:
74: struct match {
1.22 schwarze 75: uint64_t pageid; /* identifier in database */
1.37 schwarze 76: uint64_t bits; /* name type mask */
1.17 schwarze 77: char *desc; /* manual page description */
1.35 schwarze 78: int form; /* bit field: formatted, zipped? */
1.1 schwarze 79: };
80:
1.40 schwarze 81: static void buildnames(const struct mansearch *,
82: struct manpage *, sqlite3 *,
1.10 schwarze 83: sqlite3_stmt *, uint64_t,
84: const char *, int form);
1.3 schwarze 85: static char *buildoutput(sqlite3 *, sqlite3_stmt *,
86: uint64_t, uint64_t);
1.24 schwarze 87: static struct expr *exprcomp(const struct mansearch *,
1.1 schwarze 88: int, char *[]);
89: static void exprfree(struct expr *);
90: static struct expr *exprterm(const struct mansearch *, char *, int);
1.29 schwarze 91: static int manpage_compare(const void *, const void *);
1.4 schwarze 92: static void sql_append(char **sql, size_t *sz,
93: const char *newstr, int count);
1.1 schwarze 94: static void sql_match(sqlite3_context *context,
95: int argc, sqlite3_value **argv);
96: static void sql_regexp(sqlite3_context *context,
97: int argc, sqlite3_value **argv);
1.6 schwarze 98: static char *sql_statement(const struct expr *);
1.19 schwarze 99:
1.24 schwarze 100:
1.19 schwarze 101: int
102: mansearch_setup(int start)
103: {
104: static void *pagecache;
105: int c;
106:
107: #define PC_PAGESIZE 1280
108: #define PC_NUMPAGES 256
109:
110: if (start) {
111: if (NULL != pagecache) {
1.46 schwarze 112: warnx("pagecache already enabled");
1.45 schwarze 113: return (int)MANDOCLEVEL_BADARG;
1.19 schwarze 114: }
115:
116: pagecache = mmap(NULL, PC_PAGESIZE * PC_NUMPAGES,
1.31 schwarze 117: PROT_READ | PROT_WRITE,
118: MAP_SHARED | MAP_ANON, -1, 0);
1.19 schwarze 119:
120: if (MAP_FAILED == pagecache) {
1.49 schwarze 121: warn("mmap");
1.19 schwarze 122: pagecache = NULL;
1.45 schwarze 123: return (int)MANDOCLEVEL_SYSERR;
1.19 schwarze 124: }
125:
126: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
127: pagecache, PC_PAGESIZE, PC_NUMPAGES);
128:
129: if (SQLITE_OK == c)
1.45 schwarze 130: return (int)MANDOCLEVEL_OK;
1.19 schwarze 131:
1.46 schwarze 132: warnx("pagecache: %s", sqlite3_errstr(c));
1.19 schwarze 133:
134: } else if (NULL == pagecache) {
1.46 schwarze 135: warnx("pagecache missing");
1.45 schwarze 136: return (int)MANDOCLEVEL_BADARG;
1.19 schwarze 137: }
138:
139: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
1.49 schwarze 140: warn("munmap");
1.19 schwarze 141: pagecache = NULL;
1.45 schwarze 142: return (int)MANDOCLEVEL_SYSERR;
1.19 schwarze 143: }
144:
145: pagecache = NULL;
1.45 schwarze 146: return (int)MANDOCLEVEL_OK;
1.19 schwarze 147: }
1.1 schwarze 148:
149: int
150: mansearch(const struct mansearch *search,
1.3 schwarze 151: const struct manpaths *paths,
152: int argc, char *argv[],
1.1 schwarze 153: struct manpage **res, size_t *sz)
154: {
1.22 schwarze 155: int64_t pageid;
1.11 schwarze 156: uint64_t outbit, iterbit;
1.1 schwarze 157: char buf[PATH_MAX];
1.2 schwarze 158: char *sql;
1.1 schwarze 159: struct manpage *mpage;
160: struct expr *e, *ep;
161: sqlite3 *db;
1.3 schwarze 162: sqlite3_stmt *s, *s2;
1.1 schwarze 163: struct match *mp;
164: struct ohash htab;
165: unsigned int idx;
166: size_t i, j, cur, maxres;
1.44 schwarze 167: int c, chdir_status, getcwd_status, indexbit;
168:
169: if (argc == 0 || (e = exprcomp(search, argc, argv)) == NULL) {
170: *sz = 0;
1.45 schwarze 171: return 0;
1.44 schwarze 172: }
1.1 schwarze 173:
1.44 schwarze 174: cur = maxres = 0;
1.1 schwarze 175: *res = NULL;
176:
1.33 schwarze 177: if (NULL != search->outkey) {
1.42 schwarze 178: outbit = TYPE_Nd;
1.11 schwarze 179: for (indexbit = 0, iterbit = 1;
180: indexbit < mansearch_keymax;
181: indexbit++, iterbit <<= 1) {
1.33 schwarze 182: if (0 == strcasecmp(search->outkey,
1.11 schwarze 183: mansearch_keynames[indexbit])) {
184: outbit = iterbit;
1.3 schwarze 185: break;
186: }
187: }
1.42 schwarze 188: } else
189: outbit = 0;
1.3 schwarze 190:
1.1 schwarze 191: /*
1.44 schwarze 192: * Remember the original working directory, if possible.
193: * This will be needed if the second or a later directory
194: * is given as a relative path.
195: * Do not error out if the current directory is not
196: * searchable: Maybe it won't be needed after all.
1.1 schwarze 197: */
198:
1.44 schwarze 199: if (getcwd(buf, PATH_MAX) == NULL) {
200: getcwd_status = 0;
201: (void)strlcpy(buf, strerror(errno), sizeof(buf));
202: } else
203: getcwd_status = 1;
1.1 schwarze 204:
1.6 schwarze 205: sql = sql_statement(e);
1.1 schwarze 206:
207: /*
208: * Loop over the directories (containing databases) for us to
209: * search.
210: * Don't let missing/bad databases/directories phase us.
211: * In each, try to open the resident database and, if it opens,
212: * scan it for our match expression.
213: */
214:
1.44 schwarze 215: chdir_status = 0;
1.1 schwarze 216: for (i = 0; i < paths->sz; i++) {
1.44 schwarze 217: if (chdir_status && paths->paths[i][0] != '/') {
218: if ( ! getcwd_status) {
1.46 schwarze 219: warnx("%s: getcwd: %s", paths->paths[i], buf);
1.44 schwarze 220: continue;
221: } else if (chdir(buf) == -1) {
1.49 schwarze 222: warn("%s", buf);
1.44 schwarze 223: continue;
224: }
225: }
226: if (chdir(paths->paths[i]) == -1) {
1.49 schwarze 227: warn("%s", paths->paths[i]);
1.1 schwarze 228: continue;
1.24 schwarze 229: }
1.44 schwarze 230: chdir_status = 1;
1.1 schwarze 231:
1.24 schwarze 232: c = sqlite3_open_v2(MANDOC_DB, &db,
233: SQLITE_OPEN_READONLY, NULL);
1.1 schwarze 234:
235: if (SQLITE_OK != c) {
1.46 schwarze 236: warn("%s/%s", paths->paths[i], MANDOC_DB);
1.1 schwarze 237: sqlite3_close(db);
238: continue;
239: }
240:
241: /*
242: * Define the SQL functions for substring
243: * and regular expression matching.
244: */
245:
246: c = sqlite3_create_function(db, "match", 2,
1.21 schwarze 247: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
248: NULL, sql_match, NULL, NULL);
1.1 schwarze 249: assert(SQLITE_OK == c);
250: c = sqlite3_create_function(db, "regexp", 2,
1.21 schwarze 251: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
252: NULL, sql_regexp, NULL, NULL);
1.1 schwarze 253: assert(SQLITE_OK == c);
254:
255: j = 1;
256: c = sqlite3_prepare_v2(db, sql, -1, &s, NULL);
257: if (SQLITE_OK != c)
1.48 schwarze 258: errx((int)MANDOCLEVEL_SYSERR,
259: "%s", sqlite3_errmsg(db));
1.1 schwarze 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:
1.47 schwarze 270: mandoc_ohash_init(&htab, 4, offsetof(struct match, pageid));
1.1 schwarze 271:
272: /*
273: * Hash each entry on its [unique] document identifier.
274: * This is a uint64_t.
275: * Instead of using a hash function, simply convert the
276: * uint64_t to a uint32_t, the hash value's type.
277: * This gives good performance and preserves the
278: * distribution of buckets in the table.
279: */
280: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.22 schwarze 281: pageid = sqlite3_column_int64(s, 2);
1.24 schwarze 282: idx = ohash_lookup_memory(&htab,
283: (char *)&pageid, sizeof(uint64_t),
284: (uint32_t)pageid);
1.1 schwarze 285:
286: if (NULL != ohash_find(&htab, idx))
287: continue;
288:
289: mp = mandoc_calloc(1, sizeof(struct match));
1.22 schwarze 290: mp->pageid = pageid;
1.17 schwarze 291: mp->form = sqlite3_column_int(s, 1);
1.37 schwarze 292: mp->bits = sqlite3_column_int64(s, 3);
1.17 schwarze 293: if (TYPE_Nd == outbit)
1.30 schwarze 294: mp->desc = mandoc_strdup((const char *)
1.17 schwarze 295: sqlite3_column_text(s, 0));
1.1 schwarze 296: ohash_insert(&htab, idx, mp);
297: }
298:
299: if (SQLITE_DONE != c)
1.46 schwarze 300: warnx("%s", sqlite3_errmsg(db));
1.1 schwarze 301:
302: sqlite3_finalize(s);
303:
1.24 schwarze 304: c = sqlite3_prepare_v2(db,
1.25 schwarze 305: "SELECT sec, arch, name, pageid FROM mlinks "
306: "WHERE pageid=? ORDER BY sec, arch, name",
1.1 schwarze 307: -1, &s, NULL);
308: if (SQLITE_OK != c)
1.48 schwarze 309: errx((int)MANDOCLEVEL_SYSERR,
310: "%s", sqlite3_errmsg(db));
1.1 schwarze 311:
1.3 schwarze 312: c = sqlite3_prepare_v2(db,
1.25 schwarze 313: "SELECT bits, key, pageid FROM keys "
314: "WHERE pageid=? AND bits & ?",
1.3 schwarze 315: -1, &s2, NULL);
316: if (SQLITE_OK != c)
1.48 schwarze 317: errx((int)MANDOCLEVEL_SYSERR,
318: "%s", sqlite3_errmsg(db));
1.3 schwarze 319:
1.1 schwarze 320: for (mp = ohash_first(&htab, &idx);
321: NULL != mp;
322: mp = ohash_next(&htab, &idx)) {
323: if (cur + 1 > maxres) {
324: maxres += 1024;
1.26 schwarze 325: *res = mandoc_reallocarray(*res,
326: maxres, sizeof(struct manpage));
1.1 schwarze 327: }
328: mpage = *res + cur;
1.34 schwarze 329: mpage->ipath = i;
1.37 schwarze 330: mpage->bits = mp->bits;
1.29 schwarze 331: mpage->sec = 10;
1.1 schwarze 332: mpage->form = mp->form;
1.40 schwarze 333: buildnames(search, mpage, db, s, mp->pageid,
1.10 schwarze 334: paths->paths[i], mp->form);
1.40 schwarze 335: if (mpage->names != NULL) {
336: mpage->output = TYPE_Nd & outbit ?
337: mp->desc : outbit ?
338: buildoutput(db, s2, mp->pageid, outbit) :
339: NULL;
340: cur++;
341: }
1.1 schwarze 342: free(mp);
343: }
344:
345: sqlite3_finalize(s);
1.3 schwarze 346: sqlite3_finalize(s2);
1.1 schwarze 347: sqlite3_close(db);
348: ohash_delete(&htab);
1.36 schwarze 349:
350: /*
351: * In man(1) mode, prefer matches in earlier trees
352: * over matches in later trees.
353: */
354:
355: if (cur && search->firstmatch)
356: break;
1.1 schwarze 357: }
1.29 schwarze 358: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.44 schwarze 359: if (chdir_status && getcwd_status && chdir(buf) == -1)
1.49 schwarze 360: warn("%s", buf);
1.1 schwarze 361: exprfree(e);
362: free(sql);
363: *sz = cur;
1.45 schwarze 364: return 1;
1.2 schwarze 365: }
366:
1.33 schwarze 367: void
368: mansearch_free(struct manpage *res, size_t sz)
369: {
370: size_t i;
371:
372: for (i = 0; i < sz; i++) {
373: free(res[i].file);
374: free(res[i].names);
375: free(res[i].output);
376: }
377: free(res);
378: }
379:
1.29 schwarze 380: static int
381: manpage_compare(const void *vp1, const void *vp2)
382: {
383: const struct manpage *mp1, *mp2;
384: int diff;
385:
386: mp1 = vp1;
387: mp2 = vp2;
1.45 schwarze 388: return (diff = mp2->bits - mp1->bits) ? diff :
389: (diff = mp1->sec - mp2->sec) ? diff :
390: strcasecmp(mp1->names, mp2->names);
1.29 schwarze 391: }
392:
1.8 schwarze 393: static void
1.40 schwarze 394: buildnames(const struct mansearch *search, struct manpage *mpage,
395: sqlite3 *db, sqlite3_stmt *s,
1.22 schwarze 396: uint64_t pageid, const char *path, int form)
1.2 schwarze 397: {
1.41 schwarze 398: glob_t globinfo;
399: char *firstname, *newnames, *prevsec, *prevarch;
1.10 schwarze 400: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.2 schwarze 401: size_t i;
1.41 schwarze 402: int c, globres;
1.2 schwarze 403:
1.16 schwarze 404: mpage->file = NULL;
1.8 schwarze 405: mpage->names = NULL;
1.41 schwarze 406: firstname = prevsec = prevarch = NULL;
1.2 schwarze 407: i = 1;
1.22 schwarze 408: SQL_BIND_INT64(db, s, i, pageid);
1.2 schwarze 409: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.8 schwarze 410:
1.13 schwarze 411: /* Decide whether we already have some names. */
1.8 schwarze 412:
413: if (NULL == mpage->names) {
1.2 schwarze 414: oldnames = "";
415: sep1 = "";
416: } else {
1.8 schwarze 417: oldnames = mpage->names;
1.2 schwarze 418: sep1 = ", ";
419: }
1.13 schwarze 420:
1.40 schwarze 421: /* Fetch the next name, rejecting sec/arch mismatches. */
1.13 schwarze 422:
1.30 schwarze 423: sec = (const char *)sqlite3_column_text(s, 0);
1.40 schwarze 424: if (search->sec != NULL && strcasecmp(sec, search->sec))
425: continue;
1.30 schwarze 426: arch = (const char *)sqlite3_column_text(s, 1);
1.40 schwarze 427: if (search->arch != NULL && *arch != '\0' &&
428: strcasecmp(arch, search->arch))
429: continue;
1.30 schwarze 430: name = (const char *)sqlite3_column_text(s, 2);
1.29 schwarze 431:
432: /* Remember the first section found. */
433:
434: if (9 < mpage->sec && '1' <= *sec && '9' >= *sec)
435: mpage->sec = (*sec - '1') + 1;
1.13 schwarze 436:
437: /* If the section changed, append the old one. */
438:
439: if (NULL != prevsec &&
440: (strcmp(sec, prevsec) ||
441: strcmp(arch, prevarch))) {
442: sep2 = '\0' == *prevarch ? "" : "/";
1.15 schwarze 443: mandoc_asprintf(&newnames, "%s(%s%s%s)",
444: oldnames, prevsec, sep2, prevarch);
1.13 schwarze 445: free(mpage->names);
446: oldnames = mpage->names = newnames;
447: free(prevsec);
448: free(prevarch);
449: prevsec = prevarch = NULL;
450: }
451:
452: /* Save the new section, to append it later. */
453:
454: if (NULL == prevsec) {
455: prevsec = mandoc_strdup(sec);
456: prevarch = mandoc_strdup(arch);
457: }
458:
459: /* Append the new name. */
460:
1.15 schwarze 461: mandoc_asprintf(&newnames, "%s%s%s",
462: oldnames, sep1, name);
1.8 schwarze 463: free(mpage->names);
464: mpage->names = newnames;
465:
466: /* Also save the first file name encountered. */
467:
1.38 schwarze 468: if (mpage->file != NULL)
1.8 schwarze 469: continue;
470:
1.35 schwarze 471: if (form & FORM_SRC) {
1.10 schwarze 472: sep1 = "man";
473: fsec = sec;
474: } else {
475: sep1 = "cat";
476: fsec = "0";
477: }
1.38 schwarze 478: sep2 = *arch == '\0' ? "" : "/";
479: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
480: path, sep1, sec, sep2, arch, name, fsec);
1.41 schwarze 481: if (access(mpage->file, R_OK) != -1)
482: continue;
483:
484: /* Handle unusual file name extensions. */
485:
486: if (firstname == NULL)
487: firstname = mpage->file;
488: else
489: free(mpage->file);
490: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.*",
491: path, sep1, sec, sep2, arch, name);
492: globres = glob(mpage->file, 0, NULL, &globinfo);
493: free(mpage->file);
494: mpage->file = globres ? NULL :
495: mandoc_strdup(*globinfo.gl_pathv);
496: globfree(&globinfo);
1.2 schwarze 497: }
1.38 schwarze 498: if (c != SQLITE_DONE)
1.46 schwarze 499: warnx("%s", sqlite3_errmsg(db));
1.2 schwarze 500: sqlite3_reset(s);
1.41 schwarze 501:
502: /* If none of the files is usable, use the first name. */
503:
504: if (mpage->file == NULL)
505: mpage->file = firstname;
506: else if (mpage->file != firstname)
507: free(firstname);
1.13 schwarze 508:
509: /* Append one final section to the names. */
510:
1.38 schwarze 511: if (prevsec != NULL) {
512: sep2 = *prevarch == '\0' ? "" : "/";
1.15 schwarze 513: mandoc_asprintf(&newnames, "%s(%s%s%s)",
514: mpage->names, prevsec, sep2, prevarch);
1.13 schwarze 515: free(mpage->names);
516: mpage->names = newnames;
517: free(prevsec);
518: free(prevarch);
519: }
1.3 schwarze 520: }
521:
522: static char *
1.22 schwarze 523: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.3 schwarze 524: {
525: char *output, *newoutput;
526: const char *oldoutput, *sep1, *data;
527: size_t i;
528: int c;
529:
530: output = NULL;
531: i = 1;
1.22 schwarze 532: SQL_BIND_INT64(db, s, i, pageid);
1.3 schwarze 533: SQL_BIND_INT64(db, s, i, outbit);
534: while (SQLITE_ROW == (c = sqlite3_step(s))) {
535: if (NULL == output) {
536: oldoutput = "";
537: sep1 = "";
538: } else {
539: oldoutput = output;
540: sep1 = " # ";
541: }
1.30 schwarze 542: data = (const char *)sqlite3_column_text(s, 1);
1.15 schwarze 543: mandoc_asprintf(&newoutput, "%s%s%s",
544: oldoutput, sep1, data);
1.3 schwarze 545: free(output);
546: output = newoutput;
547: }
548: if (SQLITE_DONE != c)
1.46 schwarze 549: warnx("%s", sqlite3_errmsg(db));
1.3 schwarze 550: sqlite3_reset(s);
1.45 schwarze 551: return output;
1.1 schwarze 552: }
553:
554: /*
555: * Implement substring match as an application-defined SQL function.
556: * Using the SQL LIKE or GLOB operators instead would be a bad idea
557: * because that would require escaping metacharacters in the string
558: * being searched for.
559: */
560: static void
561: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
562: {
563:
564: assert(2 == argc);
565: sqlite3_result_int(context, NULL != strcasestr(
566: (const char *)sqlite3_value_text(argv[1]),
567: (const char *)sqlite3_value_text(argv[0])));
568: }
569:
570: /*
571: * Implement regular expression match
572: * as an application-defined SQL function.
573: */
574: static void
575: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
576: {
577:
578: assert(2 == argc);
579: sqlite3_result_int(context, !regexec(
580: (regex_t *)sqlite3_value_blob(argv[0]),
581: (const char *)sqlite3_value_text(argv[1]),
582: 0, NULL, 0));
583: }
584:
1.4 schwarze 585: static void
586: sql_append(char **sql, size_t *sz, const char *newstr, int count)
587: {
588: size_t newsz;
589:
590: newsz = 1 < count ? (size_t)count : strlen(newstr);
591: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
592: if (1 < count)
593: memset(*sql + *sz, *newstr, (size_t)count);
594: else
595: memcpy(*sql + *sz, newstr, newsz);
596: *sz += newsz;
597: (*sql)[*sz] = '\0';
598: }
599:
1.1 schwarze 600: /*
601: * Prepare the search SQL statement.
602: */
603: static char *
1.6 schwarze 604: sql_statement(const struct expr *e)
1.1 schwarze 605: {
606: char *sql;
607: size_t sz;
1.4 schwarze 608: int needop;
1.1 schwarze 609:
1.37 schwarze 610: sql = mandoc_strdup(e->equal ?
611: "SELECT desc, form, pageid, bits "
612: "FROM mpages NATURAL JOIN names WHERE " :
613: "SELECT desc, form, pageid, 0 FROM mpages WHERE ");
1.1 schwarze 614: sz = strlen(sql);
615:
1.4 schwarze 616: for (needop = 0; NULL != e; e = e->next) {
617: if (e->and)
618: sql_append(&sql, &sz, " AND ", 1);
619: else if (needop)
620: sql_append(&sql, &sz, " OR ", 1);
621: if (e->open)
622: sql_append(&sql, &sz, "(", e->open);
1.17 schwarze 623: sql_append(&sql, &sz,
624: TYPE_Nd & e->bits
625: ? (NULL == e->substr
626: ? "desc REGEXP ?"
627: : "desc MATCH ?")
1.18 schwarze 628: : TYPE_Nm == e->bits
629: ? (NULL == e->substr
1.22 schwarze 630: ? "pageid IN (SELECT pageid FROM names "
1.18 schwarze 631: "WHERE name REGEXP ?)"
1.28 schwarze 632: : e->equal
1.37 schwarze 633: ? "name = ? "
1.22 schwarze 634: : "pageid IN (SELECT pageid FROM names "
1.18 schwarze 635: "WHERE name MATCH ?)")
1.17 schwarze 636: : (NULL == e->substr
1.22 schwarze 637: ? "pageid IN (SELECT pageid FROM keys "
1.17 schwarze 638: "WHERE key REGEXP ? AND bits & ?)"
1.22 schwarze 639: : "pageid IN (SELECT pageid FROM keys "
1.17 schwarze 640: "WHERE key MATCH ? AND bits & ?)"), 1);
1.4 schwarze 641: if (e->close)
642: sql_append(&sql, &sz, ")", e->close);
643: needop = 1;
1.1 schwarze 644: }
645:
1.45 schwarze 646: return sql;
1.1 schwarze 647: }
648:
649: /*
650: * Compile a set of string tokens into an expression.
651: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
652: * "(", "foo=bar", etc.).
653: */
654: static struct expr *
655: exprcomp(const struct mansearch *search, int argc, char *argv[])
656: {
1.18 schwarze 657: uint64_t mask;
1.4 schwarze 658: int i, toopen, logic, igncase, toclose;
1.18 schwarze 659: struct expr *first, *prev, *cur, *next;
1.1 schwarze 660:
661: first = cur = NULL;
1.40 schwarze 662: logic = igncase = toopen = toclose = 0;
1.1 schwarze 663:
664: for (i = 0; i < argc; i++) {
1.4 schwarze 665: if (0 == strcmp("(", argv[i])) {
666: if (igncase)
667: goto fail;
668: toopen++;
669: toclose++;
670: continue;
671: } else if (0 == strcmp(")", argv[i])) {
672: if (toopen || logic || igncase || NULL == cur)
673: goto fail;
674: cur->close++;
675: if (0 > --toclose)
676: goto fail;
677: continue;
678: } else if (0 == strcmp("-a", argv[i])) {
679: if (toopen || logic || igncase || NULL == cur)
680: goto fail;
681: logic = 1;
682: continue;
683: } else if (0 == strcmp("-o", argv[i])) {
684: if (toopen || logic || igncase || NULL == cur)
685: goto fail;
686: logic = 2;
687: continue;
688: } else if (0 == strcmp("-i", argv[i])) {
689: if (igncase)
690: goto fail;
691: igncase = 1;
692: continue;
1.1 schwarze 693: }
1.4 schwarze 694: next = exprterm(search, argv[i], !igncase);
695: if (NULL == next)
696: goto fail;
1.17 schwarze 697: if (NULL == first)
698: first = next;
699: else
1.1 schwarze 700: cur->next = next;
1.18 schwarze 701: prev = cur = next;
1.17 schwarze 702:
703: /*
704: * Searching for descriptions must be split out
705: * because they are stored in the mpages table,
706: * not in the keys table.
707: */
708:
1.18 schwarze 709: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
710: if (mask & cur->bits && ~mask & cur->bits) {
711: next = mandoc_calloc(1,
712: sizeof(struct expr));
713: memcpy(next, cur, sizeof(struct expr));
714: prev->open = 1;
715: cur->bits = mask;
716: cur->next = next;
717: cur = next;
718: cur->bits &= ~mask;
719: }
720: }
721: prev->and = (1 == logic);
722: prev->open += toopen;
723: if (cur != prev)
1.17 schwarze 724: cur->close = 1;
1.18 schwarze 725:
1.4 schwarze 726: toopen = logic = igncase = 0;
1.1 schwarze 727: }
1.40 schwarze 728: if ( ! (toopen || logic || igncase || toclose))
1.45 schwarze 729: return first;
1.6 schwarze 730:
1.4 schwarze 731: fail:
732: if (NULL != first)
733: exprfree(first);
1.45 schwarze 734: return NULL;
1.6 schwarze 735: }
736:
737: static struct expr *
1.1 schwarze 738: exprterm(const struct mansearch *search, char *buf, int cs)
739: {
1.6 schwarze 740: char errbuf[BUFSIZ];
1.1 schwarze 741: struct expr *e;
1.28 schwarze 742: char *key, *val;
1.11 schwarze 743: uint64_t iterbit;
744: int i, irc;
1.1 schwarze 745:
746: if ('\0' == *buf)
1.45 schwarze 747: return NULL;
1.1 schwarze 748:
749: e = mandoc_calloc(1, sizeof(struct expr));
750:
1.33 schwarze 751: if (search->argmode == ARG_NAME) {
752: e->bits = TYPE_Nm;
1.1 schwarze 753: e->substr = buf;
1.28 schwarze 754: e->equal = 1;
1.45 schwarze 755: return e;
1.1 schwarze 756: }
757:
758: /*
1.33 schwarze 759: * Separate macro keys from search string.
760: * If needed, request regular expression handling
761: * by setting e->substr to NULL.
1.1 schwarze 762: */
763:
1.33 schwarze 764: if (search->argmode == ARG_WORD) {
765: e->bits = TYPE_Nm;
766: e->substr = NULL;
767: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
768: cs = 0;
769: } else if ((val = strpbrk(buf, "=~")) == NULL) {
770: e->bits = TYPE_Nm | TYPE_Nd;
1.1 schwarze 771: e->substr = buf;
1.28 schwarze 772: } else {
773: if (val == buf)
1.33 schwarze 774: e->bits = TYPE_Nm | TYPE_Nd;
1.28 schwarze 775: if ('=' == *val)
776: e->substr = val + 1;
777: *val++ = '\0';
1.12 schwarze 778: if (NULL != strstr(buf, "arch"))
779: cs = 0;
1.28 schwarze 780: }
781:
782: /* Compile regular expressions. */
783:
784: if (NULL == e->substr) {
785: irc = regcomp(&e->regexp, val,
786: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.33 schwarze 787: if (search->argmode == ARG_WORD)
1.28 schwarze 788: free(val);
789: if (irc) {
1.6 schwarze 790: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
1.46 schwarze 791: warnx("regcomp: %s", errbuf);
1.1 schwarze 792: free(e);
1.45 schwarze 793: return NULL;
1.1 schwarze 794: }
1.28 schwarze 795: }
796:
797: if (e->bits)
1.45 schwarze 798: return e;
1.1 schwarze 799:
800: /*
801: * Parse out all possible fields.
802: * If the field doesn't resolve, bail.
803: */
804:
805: while (NULL != (key = strsep(&buf, ","))) {
806: if ('\0' == *key)
807: continue;
1.11 schwarze 808: for (i = 0, iterbit = 1;
809: i < mansearch_keymax;
810: i++, iterbit <<= 1) {
811: if (0 == strcasecmp(key,
812: mansearch_keynames[i])) {
813: e->bits |= iterbit;
814: break;
815: }
816: }
817: if (i == mansearch_keymax) {
818: if (strcasecmp(key, "any")) {
819: free(e);
1.45 schwarze 820: return NULL;
1.11 schwarze 821: }
822: e->bits |= ~0ULL;
1.1 schwarze 823: }
824: }
825:
1.45 schwarze 826: return e;
1.1 schwarze 827: }
828:
829: static void
830: exprfree(struct expr *p)
831: {
832: struct expr *pp;
833:
834: while (NULL != p) {
835: pp = p->next;
836: free(p);
837: p = pp;
838: }
839: }