Annotation of src/usr.bin/mandoc/mansearch.c, Revision 1.46
1.46 ! schwarze 1: /* $OpenBSD: mansearch.c,v 1.45 2015/10/06 18:30:44 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>
26: #include <getopt.h>
1.41 schwarze 27: #include <glob.h>
1.1 schwarze 28: #include <limits.h>
29: #include <regex.h>
30: #include <stdio.h>
31: #include <stdint.h>
32: #include <stddef.h>
33: #include <stdlib.h>
34: #include <string.h>
35: #include <unistd.h>
36:
37: #include <ohash.h>
38: #include <sqlite3.h>
39:
40: #include "mandoc.h"
1.14 schwarze 41: #include "mandoc_aux.h"
1.43 schwarze 42: #include "manconf.h"
1.1 schwarze 43: #include "mansearch.h"
44:
1.11 schwarze 45: extern int mansearch_keymax;
46: extern const char *const mansearch_keynames[];
47:
1.1 schwarze 48: #define SQL_BIND_TEXT(_db, _s, _i, _v) \
49: do { if (SQLITE_OK != sqlite3_bind_text \
50: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.46 ! schwarze 51: warnx("%s", sqlite3_errmsg((_db))); \
1.1 schwarze 52: } while (0)
53: #define SQL_BIND_INT64(_db, _s, _i, _v) \
54: do { if (SQLITE_OK != sqlite3_bind_int64 \
55: ((_s), (_i)++, (_v))) \
1.46 ! schwarze 56: warnx("%s", sqlite3_errmsg((_db))); \
1.1 schwarze 57: } while (0)
58: #define SQL_BIND_BLOB(_db, _s, _i, _v) \
59: do { if (SQLITE_OK != sqlite3_bind_blob \
60: ((_s), (_i)++, (&_v), sizeof(_v), SQLITE_STATIC)) \
1.46 ! schwarze 61: warnx("%s", sqlite3_errmsg((_db))); \
1.1 schwarze 62: } while (0)
63:
64: struct expr {
1.28 schwarze 65: regex_t regexp; /* compiled regexp, if applicable */
66: const char *substr; /* to search for, if applicable */
67: struct expr *next; /* next in sequence */
1.24 schwarze 68: uint64_t bits; /* type-mask */
1.28 schwarze 69: int equal; /* equality, not subsring match */
1.4 schwarze 70: int open; /* opening parentheses before */
71: int and; /* logical AND before */
72: int close; /* closing parentheses after */
1.1 schwarze 73: };
74:
75: struct match {
1.22 schwarze 76: uint64_t pageid; /* identifier in database */
1.37 schwarze 77: uint64_t bits; /* name type mask */
1.17 schwarze 78: char *desc; /* manual page description */
1.35 schwarze 79: int form; /* bit field: formatted, zipped? */
1.1 schwarze 80: };
81:
1.40 schwarze 82: static void buildnames(const struct mansearch *,
83: struct manpage *, sqlite3 *,
1.10 schwarze 84: sqlite3_stmt *, uint64_t,
85: const char *, int form);
1.3 schwarze 86: static char *buildoutput(sqlite3 *, sqlite3_stmt *,
87: uint64_t, uint64_t);
1.1 schwarze 88: static void *hash_alloc(size_t, void *);
1.27 espie 89: static void hash_free(void *, void *);
90: static void *hash_calloc(size_t, size_t, void *);
1.24 schwarze 91: static struct expr *exprcomp(const struct mansearch *,
1.1 schwarze 92: int, char *[]);
93: static void exprfree(struct expr *);
94: static struct expr *exprterm(const struct mansearch *, char *, int);
1.29 schwarze 95: static int manpage_compare(const void *, const void *);
1.4 schwarze 96: static void sql_append(char **sql, size_t *sz,
97: const char *newstr, int count);
1.1 schwarze 98: static void sql_match(sqlite3_context *context,
99: int argc, sqlite3_value **argv);
100: static void sql_regexp(sqlite3_context *context,
101: int argc, sqlite3_value **argv);
1.6 schwarze 102: static char *sql_statement(const struct expr *);
1.19 schwarze 103:
1.24 schwarze 104:
1.19 schwarze 105: int
106: mansearch_setup(int start)
107: {
108: static void *pagecache;
109: int c;
110:
111: #define PC_PAGESIZE 1280
112: #define PC_NUMPAGES 256
113:
114: if (start) {
115: if (NULL != pagecache) {
1.46 ! schwarze 116: warnx("pagecache already enabled");
1.45 schwarze 117: return (int)MANDOCLEVEL_BADARG;
1.19 schwarze 118: }
119:
120: pagecache = mmap(NULL, PC_PAGESIZE * PC_NUMPAGES,
1.31 schwarze 121: PROT_READ | PROT_WRITE,
122: MAP_SHARED | MAP_ANON, -1, 0);
1.19 schwarze 123:
124: if (MAP_FAILED == pagecache) {
125: perror("mmap");
126: pagecache = NULL;
1.45 schwarze 127: return (int)MANDOCLEVEL_SYSERR;
1.19 schwarze 128: }
129:
130: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
131: pagecache, PC_PAGESIZE, PC_NUMPAGES);
132:
133: if (SQLITE_OK == c)
1.45 schwarze 134: return (int)MANDOCLEVEL_OK;
1.19 schwarze 135:
1.46 ! schwarze 136: warnx("pagecache: %s", sqlite3_errstr(c));
1.19 schwarze 137:
138: } else if (NULL == pagecache) {
1.46 ! schwarze 139: warnx("pagecache missing");
1.45 schwarze 140: return (int)MANDOCLEVEL_BADARG;
1.19 schwarze 141: }
142:
143: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
144: perror("munmap");
145: pagecache = NULL;
1.45 schwarze 146: return (int)MANDOCLEVEL_SYSERR;
1.19 schwarze 147: }
148:
149: pagecache = NULL;
1.45 schwarze 150: return (int)MANDOCLEVEL_OK;
1.19 schwarze 151: }
1.1 schwarze 152:
153: int
154: mansearch(const struct mansearch *search,
1.3 schwarze 155: const struct manpaths *paths,
156: int argc, char *argv[],
1.1 schwarze 157: struct manpage **res, size_t *sz)
158: {
1.22 schwarze 159: int64_t pageid;
1.11 schwarze 160: uint64_t outbit, iterbit;
1.1 schwarze 161: char buf[PATH_MAX];
1.2 schwarze 162: char *sql;
1.1 schwarze 163: struct manpage *mpage;
164: struct expr *e, *ep;
165: sqlite3 *db;
1.3 schwarze 166: sqlite3_stmt *s, *s2;
1.1 schwarze 167: struct match *mp;
168: struct ohash_info info;
169: struct ohash htab;
170: unsigned int idx;
171: size_t i, j, cur, maxres;
1.44 schwarze 172: int c, chdir_status, getcwd_status, indexbit;
173:
174: if (argc == 0 || (e = exprcomp(search, argc, argv)) == NULL) {
175: *sz = 0;
1.45 schwarze 176: return 0;
1.44 schwarze 177: }
1.1 schwarze 178:
1.27 espie 179: info.calloc = hash_calloc;
1.1 schwarze 180: info.alloc = hash_alloc;
1.27 espie 181: info.free = hash_free;
1.22 schwarze 182: info.key_offset = offsetof(struct match, pageid);
1.1 schwarze 183:
1.44 schwarze 184: cur = maxres = 0;
1.1 schwarze 185: *res = NULL;
186:
1.33 schwarze 187: if (NULL != search->outkey) {
1.42 schwarze 188: outbit = TYPE_Nd;
1.11 schwarze 189: for (indexbit = 0, iterbit = 1;
190: indexbit < mansearch_keymax;
191: indexbit++, iterbit <<= 1) {
1.33 schwarze 192: if (0 == strcasecmp(search->outkey,
1.11 schwarze 193: mansearch_keynames[indexbit])) {
194: outbit = iterbit;
1.3 schwarze 195: break;
196: }
197: }
1.42 schwarze 198: } else
199: outbit = 0;
1.3 schwarze 200:
1.1 schwarze 201: /*
1.44 schwarze 202: * Remember the original working directory, if possible.
203: * This will be needed if the second or a later directory
204: * is given as a relative path.
205: * Do not error out if the current directory is not
206: * searchable: Maybe it won't be needed after all.
1.1 schwarze 207: */
208:
1.44 schwarze 209: if (getcwd(buf, PATH_MAX) == NULL) {
210: getcwd_status = 0;
211: (void)strlcpy(buf, strerror(errno), sizeof(buf));
212: } else
213: getcwd_status = 1;
1.1 schwarze 214:
1.6 schwarze 215: sql = sql_statement(e);
1.1 schwarze 216:
217: /*
218: * Loop over the directories (containing databases) for us to
219: * search.
220: * Don't let missing/bad databases/directories phase us.
221: * In each, try to open the resident database and, if it opens,
222: * scan it for our match expression.
223: */
224:
1.44 schwarze 225: chdir_status = 0;
1.1 schwarze 226: for (i = 0; i < paths->sz; i++) {
1.44 schwarze 227: if (chdir_status && paths->paths[i][0] != '/') {
228: if ( ! getcwd_status) {
1.46 ! schwarze 229: warnx("%s: getcwd: %s", paths->paths[i], buf);
1.44 schwarze 230: continue;
231: } else if (chdir(buf) == -1) {
232: perror(buf);
233: continue;
234: }
235: }
236: if (chdir(paths->paths[i]) == -1) {
1.1 schwarze 237: perror(paths->paths[i]);
238: continue;
1.24 schwarze 239: }
1.44 schwarze 240: chdir_status = 1;
1.1 schwarze 241:
1.24 schwarze 242: c = sqlite3_open_v2(MANDOC_DB, &db,
243: SQLITE_OPEN_READONLY, NULL);
1.1 schwarze 244:
245: if (SQLITE_OK != c) {
1.46 ! schwarze 246: warn("%s/%s", paths->paths[i], MANDOC_DB);
1.1 schwarze 247: sqlite3_close(db);
248: continue;
249: }
250:
251: /*
252: * Define the SQL functions for substring
253: * and regular expression matching.
254: */
255:
256: c = sqlite3_create_function(db, "match", 2,
1.21 schwarze 257: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
258: NULL, sql_match, NULL, NULL);
1.1 schwarze 259: assert(SQLITE_OK == c);
260: c = sqlite3_create_function(db, "regexp", 2,
1.21 schwarze 261: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
262: NULL, sql_regexp, NULL, NULL);
1.1 schwarze 263: assert(SQLITE_OK == c);
264:
265: j = 1;
266: c = sqlite3_prepare_v2(db, sql, -1, &s, NULL);
267: if (SQLITE_OK != c)
1.46 ! schwarze 268: warnx("%s", sqlite3_errmsg(db));
1.1 schwarze 269:
270: for (ep = e; NULL != ep; ep = ep->next) {
271: if (NULL == ep->substr) {
272: SQL_BIND_BLOB(db, s, j, ep->regexp);
273: } else
274: SQL_BIND_TEXT(db, s, j, ep->substr);
1.18 schwarze 275: if (0 == ((TYPE_Nd | TYPE_Nm) & ep->bits))
1.17 schwarze 276: SQL_BIND_INT64(db, s, j, ep->bits);
1.1 schwarze 277: }
278:
279: memset(&htab, 0, sizeof(struct ohash));
280: ohash_init(&htab, 4, &info);
281:
282: /*
283: * Hash each entry on its [unique] document identifier.
284: * This is a uint64_t.
285: * Instead of using a hash function, simply convert the
286: * uint64_t to a uint32_t, the hash value's type.
287: * This gives good performance and preserves the
288: * distribution of buckets in the table.
289: */
290: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.22 schwarze 291: pageid = sqlite3_column_int64(s, 2);
1.24 schwarze 292: idx = ohash_lookup_memory(&htab,
293: (char *)&pageid, sizeof(uint64_t),
294: (uint32_t)pageid);
1.1 schwarze 295:
296: if (NULL != ohash_find(&htab, idx))
297: continue;
298:
299: mp = mandoc_calloc(1, sizeof(struct match));
1.22 schwarze 300: mp->pageid = pageid;
1.17 schwarze 301: mp->form = sqlite3_column_int(s, 1);
1.37 schwarze 302: mp->bits = sqlite3_column_int64(s, 3);
1.17 schwarze 303: if (TYPE_Nd == outbit)
1.30 schwarze 304: mp->desc = mandoc_strdup((const char *)
1.17 schwarze 305: sqlite3_column_text(s, 0));
1.1 schwarze 306: ohash_insert(&htab, idx, mp);
307: }
308:
309: if (SQLITE_DONE != c)
1.46 ! schwarze 310: warnx("%s", sqlite3_errmsg(db));
1.1 schwarze 311:
312: sqlite3_finalize(s);
313:
1.24 schwarze 314: c = sqlite3_prepare_v2(db,
1.25 schwarze 315: "SELECT sec, arch, name, pageid FROM mlinks "
316: "WHERE pageid=? ORDER BY sec, arch, name",
1.1 schwarze 317: -1, &s, NULL);
318: if (SQLITE_OK != c)
1.46 ! schwarze 319: warnx("%s", sqlite3_errmsg(db));
1.1 schwarze 320:
1.3 schwarze 321: c = sqlite3_prepare_v2(db,
1.25 schwarze 322: "SELECT bits, key, pageid FROM keys "
323: "WHERE pageid=? AND bits & ?",
1.3 schwarze 324: -1, &s2, NULL);
325: if (SQLITE_OK != c)
1.46 ! schwarze 326: warnx("%s", sqlite3_errmsg(db));
1.3 schwarze 327:
1.1 schwarze 328: for (mp = ohash_first(&htab, &idx);
329: NULL != mp;
330: mp = ohash_next(&htab, &idx)) {
331: if (cur + 1 > maxres) {
332: maxres += 1024;
1.26 schwarze 333: *res = mandoc_reallocarray(*res,
334: maxres, sizeof(struct manpage));
1.1 schwarze 335: }
336: mpage = *res + cur;
1.34 schwarze 337: mpage->ipath = i;
1.37 schwarze 338: mpage->bits = mp->bits;
1.29 schwarze 339: mpage->sec = 10;
1.1 schwarze 340: mpage->form = mp->form;
1.40 schwarze 341: buildnames(search, mpage, db, s, mp->pageid,
1.10 schwarze 342: paths->paths[i], mp->form);
1.40 schwarze 343: if (mpage->names != NULL) {
344: mpage->output = TYPE_Nd & outbit ?
345: mp->desc : outbit ?
346: buildoutput(db, s2, mp->pageid, outbit) :
347: NULL;
348: cur++;
349: }
1.1 schwarze 350: free(mp);
351: }
352:
353: sqlite3_finalize(s);
1.3 schwarze 354: sqlite3_finalize(s2);
1.1 schwarze 355: sqlite3_close(db);
356: ohash_delete(&htab);
1.36 schwarze 357:
358: /*
359: * In man(1) mode, prefer matches in earlier trees
360: * over matches in later trees.
361: */
362:
363: if (cur && search->firstmatch)
364: break;
1.1 schwarze 365: }
1.29 schwarze 366: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.44 schwarze 367: if (chdir_status && getcwd_status && chdir(buf) == -1)
368: perror(buf);
1.1 schwarze 369: exprfree(e);
370: free(sql);
371: *sz = cur;
1.45 schwarze 372: return 1;
1.2 schwarze 373: }
374:
1.33 schwarze 375: void
376: mansearch_free(struct manpage *res, size_t sz)
377: {
378: size_t i;
379:
380: for (i = 0; i < sz; i++) {
381: free(res[i].file);
382: free(res[i].names);
383: free(res[i].output);
384: }
385: free(res);
386: }
387:
1.29 schwarze 388: static int
389: manpage_compare(const void *vp1, const void *vp2)
390: {
391: const struct manpage *mp1, *mp2;
392: int diff;
393:
394: mp1 = vp1;
395: mp2 = vp2;
1.45 schwarze 396: return (diff = mp2->bits - mp1->bits) ? diff :
397: (diff = mp1->sec - mp2->sec) ? diff :
398: strcasecmp(mp1->names, mp2->names);
1.29 schwarze 399: }
400:
1.8 schwarze 401: static void
1.40 schwarze 402: buildnames(const struct mansearch *search, struct manpage *mpage,
403: sqlite3 *db, sqlite3_stmt *s,
1.22 schwarze 404: uint64_t pageid, const char *path, int form)
1.2 schwarze 405: {
1.41 schwarze 406: glob_t globinfo;
407: char *firstname, *newnames, *prevsec, *prevarch;
1.10 schwarze 408: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.2 schwarze 409: size_t i;
1.41 schwarze 410: int c, globres;
1.2 schwarze 411:
1.16 schwarze 412: mpage->file = NULL;
1.8 schwarze 413: mpage->names = NULL;
1.41 schwarze 414: firstname = prevsec = prevarch = NULL;
1.2 schwarze 415: i = 1;
1.22 schwarze 416: SQL_BIND_INT64(db, s, i, pageid);
1.2 schwarze 417: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.8 schwarze 418:
1.13 schwarze 419: /* Decide whether we already have some names. */
1.8 schwarze 420:
421: if (NULL == mpage->names) {
1.2 schwarze 422: oldnames = "";
423: sep1 = "";
424: } else {
1.8 schwarze 425: oldnames = mpage->names;
1.2 schwarze 426: sep1 = ", ";
427: }
1.13 schwarze 428:
1.40 schwarze 429: /* Fetch the next name, rejecting sec/arch mismatches. */
1.13 schwarze 430:
1.30 schwarze 431: sec = (const char *)sqlite3_column_text(s, 0);
1.40 schwarze 432: if (search->sec != NULL && strcasecmp(sec, search->sec))
433: continue;
1.30 schwarze 434: arch = (const char *)sqlite3_column_text(s, 1);
1.40 schwarze 435: if (search->arch != NULL && *arch != '\0' &&
436: strcasecmp(arch, search->arch))
437: continue;
1.30 schwarze 438: name = (const char *)sqlite3_column_text(s, 2);
1.29 schwarze 439:
440: /* Remember the first section found. */
441:
442: if (9 < mpage->sec && '1' <= *sec && '9' >= *sec)
443: mpage->sec = (*sec - '1') + 1;
1.13 schwarze 444:
445: /* If the section changed, append the old one. */
446:
447: if (NULL != prevsec &&
448: (strcmp(sec, prevsec) ||
449: strcmp(arch, prevarch))) {
450: sep2 = '\0' == *prevarch ? "" : "/";
1.15 schwarze 451: mandoc_asprintf(&newnames, "%s(%s%s%s)",
452: oldnames, prevsec, sep2, prevarch);
1.13 schwarze 453: free(mpage->names);
454: oldnames = mpage->names = newnames;
455: free(prevsec);
456: free(prevarch);
457: prevsec = prevarch = NULL;
458: }
459:
460: /* Save the new section, to append it later. */
461:
462: if (NULL == prevsec) {
463: prevsec = mandoc_strdup(sec);
464: prevarch = mandoc_strdup(arch);
465: }
466:
467: /* Append the new name. */
468:
1.15 schwarze 469: mandoc_asprintf(&newnames, "%s%s%s",
470: oldnames, sep1, name);
1.8 schwarze 471: free(mpage->names);
472: mpage->names = newnames;
473:
474: /* Also save the first file name encountered. */
475:
1.38 schwarze 476: if (mpage->file != NULL)
1.8 schwarze 477: continue;
478:
1.35 schwarze 479: if (form & FORM_SRC) {
1.10 schwarze 480: sep1 = "man";
481: fsec = sec;
482: } else {
483: sep1 = "cat";
484: fsec = "0";
485: }
1.38 schwarze 486: sep2 = *arch == '\0' ? "" : "/";
487: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
488: path, sep1, sec, sep2, arch, name, fsec);
1.41 schwarze 489: if (access(mpage->file, R_OK) != -1)
490: continue;
491:
492: /* Handle unusual file name extensions. */
493:
494: if (firstname == NULL)
495: firstname = mpage->file;
496: else
497: free(mpage->file);
498: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.*",
499: path, sep1, sec, sep2, arch, name);
500: globres = glob(mpage->file, 0, NULL, &globinfo);
501: free(mpage->file);
502: mpage->file = globres ? NULL :
503: mandoc_strdup(*globinfo.gl_pathv);
504: globfree(&globinfo);
1.2 schwarze 505: }
1.38 schwarze 506: if (c != SQLITE_DONE)
1.46 ! schwarze 507: warnx("%s", sqlite3_errmsg(db));
1.2 schwarze 508: sqlite3_reset(s);
1.41 schwarze 509:
510: /* If none of the files is usable, use the first name. */
511:
512: if (mpage->file == NULL)
513: mpage->file = firstname;
514: else if (mpage->file != firstname)
515: free(firstname);
1.13 schwarze 516:
517: /* Append one final section to the names. */
518:
1.38 schwarze 519: if (prevsec != NULL) {
520: sep2 = *prevarch == '\0' ? "" : "/";
1.15 schwarze 521: mandoc_asprintf(&newnames, "%s(%s%s%s)",
522: mpage->names, prevsec, sep2, prevarch);
1.13 schwarze 523: free(mpage->names);
524: mpage->names = newnames;
525: free(prevsec);
526: free(prevarch);
527: }
1.3 schwarze 528: }
529:
530: static char *
1.22 schwarze 531: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.3 schwarze 532: {
533: char *output, *newoutput;
534: const char *oldoutput, *sep1, *data;
535: size_t i;
536: int c;
537:
538: output = NULL;
539: i = 1;
1.22 schwarze 540: SQL_BIND_INT64(db, s, i, pageid);
1.3 schwarze 541: SQL_BIND_INT64(db, s, i, outbit);
542: while (SQLITE_ROW == (c = sqlite3_step(s))) {
543: if (NULL == output) {
544: oldoutput = "";
545: sep1 = "";
546: } else {
547: oldoutput = output;
548: sep1 = " # ";
549: }
1.30 schwarze 550: data = (const char *)sqlite3_column_text(s, 1);
1.15 schwarze 551: mandoc_asprintf(&newoutput, "%s%s%s",
552: oldoutput, sep1, data);
1.3 schwarze 553: free(output);
554: output = newoutput;
555: }
556: if (SQLITE_DONE != c)
1.46 ! schwarze 557: warnx("%s", sqlite3_errmsg(db));
1.3 schwarze 558: sqlite3_reset(s);
1.45 schwarze 559: return output;
1.1 schwarze 560: }
561:
562: /*
563: * Implement substring match as an application-defined SQL function.
564: * Using the SQL LIKE or GLOB operators instead would be a bad idea
565: * because that would require escaping metacharacters in the string
566: * being searched for.
567: */
568: static void
569: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
570: {
571:
572: assert(2 == argc);
573: sqlite3_result_int(context, NULL != strcasestr(
574: (const char *)sqlite3_value_text(argv[1]),
575: (const char *)sqlite3_value_text(argv[0])));
576: }
577:
578: /*
579: * Implement regular expression match
580: * as an application-defined SQL function.
581: */
582: static void
583: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
584: {
585:
586: assert(2 == argc);
587: sqlite3_result_int(context, !regexec(
588: (regex_t *)sqlite3_value_blob(argv[0]),
589: (const char *)sqlite3_value_text(argv[1]),
590: 0, NULL, 0));
591: }
592:
1.4 schwarze 593: static void
594: sql_append(char **sql, size_t *sz, const char *newstr, int count)
595: {
596: size_t newsz;
597:
598: newsz = 1 < count ? (size_t)count : strlen(newstr);
599: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
600: if (1 < count)
601: memset(*sql + *sz, *newstr, (size_t)count);
602: else
603: memcpy(*sql + *sz, newstr, newsz);
604: *sz += newsz;
605: (*sql)[*sz] = '\0';
606: }
607:
1.1 schwarze 608: /*
609: * Prepare the search SQL statement.
610: */
611: static char *
1.6 schwarze 612: sql_statement(const struct expr *e)
1.1 schwarze 613: {
614: char *sql;
615: size_t sz;
1.4 schwarze 616: int needop;
1.1 schwarze 617:
1.37 schwarze 618: sql = mandoc_strdup(e->equal ?
619: "SELECT desc, form, pageid, bits "
620: "FROM mpages NATURAL JOIN names WHERE " :
621: "SELECT desc, form, pageid, 0 FROM mpages WHERE ");
1.1 schwarze 622: sz = strlen(sql);
623:
1.4 schwarze 624: for (needop = 0; NULL != e; e = e->next) {
625: if (e->and)
626: sql_append(&sql, &sz, " AND ", 1);
627: else if (needop)
628: sql_append(&sql, &sz, " OR ", 1);
629: if (e->open)
630: sql_append(&sql, &sz, "(", e->open);
1.17 schwarze 631: sql_append(&sql, &sz,
632: TYPE_Nd & e->bits
633: ? (NULL == e->substr
634: ? "desc REGEXP ?"
635: : "desc MATCH ?")
1.18 schwarze 636: : TYPE_Nm == e->bits
637: ? (NULL == e->substr
1.22 schwarze 638: ? "pageid IN (SELECT pageid FROM names "
1.18 schwarze 639: "WHERE name REGEXP ?)"
1.28 schwarze 640: : e->equal
1.37 schwarze 641: ? "name = ? "
1.22 schwarze 642: : "pageid IN (SELECT pageid FROM names "
1.18 schwarze 643: "WHERE name MATCH ?)")
1.17 schwarze 644: : (NULL == e->substr
1.22 schwarze 645: ? "pageid IN (SELECT pageid FROM keys "
1.17 schwarze 646: "WHERE key REGEXP ? AND bits & ?)"
1.22 schwarze 647: : "pageid IN (SELECT pageid FROM keys "
1.17 schwarze 648: "WHERE key MATCH ? AND bits & ?)"), 1);
1.4 schwarze 649: if (e->close)
650: sql_append(&sql, &sz, ")", e->close);
651: needop = 1;
1.1 schwarze 652: }
653:
1.45 schwarze 654: return sql;
1.1 schwarze 655: }
656:
657: /*
658: * Compile a set of string tokens into an expression.
659: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
660: * "(", "foo=bar", etc.).
661: */
662: static struct expr *
663: exprcomp(const struct mansearch *search, int argc, char *argv[])
664: {
1.18 schwarze 665: uint64_t mask;
1.4 schwarze 666: int i, toopen, logic, igncase, toclose;
1.18 schwarze 667: struct expr *first, *prev, *cur, *next;
1.1 schwarze 668:
669: first = cur = NULL;
1.40 schwarze 670: logic = igncase = toopen = toclose = 0;
1.1 schwarze 671:
672: for (i = 0; i < argc; i++) {
1.4 schwarze 673: if (0 == strcmp("(", argv[i])) {
674: if (igncase)
675: goto fail;
676: toopen++;
677: toclose++;
678: continue;
679: } else if (0 == strcmp(")", argv[i])) {
680: if (toopen || logic || igncase || NULL == cur)
681: goto fail;
682: cur->close++;
683: if (0 > --toclose)
684: goto fail;
685: continue;
686: } else if (0 == strcmp("-a", argv[i])) {
687: if (toopen || logic || igncase || NULL == cur)
688: goto fail;
689: logic = 1;
690: continue;
691: } else if (0 == strcmp("-o", argv[i])) {
692: if (toopen || logic || igncase || NULL == cur)
693: goto fail;
694: logic = 2;
695: continue;
696: } else if (0 == strcmp("-i", argv[i])) {
697: if (igncase)
698: goto fail;
699: igncase = 1;
700: continue;
1.1 schwarze 701: }
1.4 schwarze 702: next = exprterm(search, argv[i], !igncase);
703: if (NULL == next)
704: goto fail;
1.17 schwarze 705: if (NULL == first)
706: first = next;
707: else
1.1 schwarze 708: cur->next = next;
1.18 schwarze 709: prev = cur = next;
1.17 schwarze 710:
711: /*
712: * Searching for descriptions must be split out
713: * because they are stored in the mpages table,
714: * not in the keys table.
715: */
716:
1.18 schwarze 717: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
718: if (mask & cur->bits && ~mask & cur->bits) {
719: next = mandoc_calloc(1,
720: sizeof(struct expr));
721: memcpy(next, cur, sizeof(struct expr));
722: prev->open = 1;
723: cur->bits = mask;
724: cur->next = next;
725: cur = next;
726: cur->bits &= ~mask;
727: }
728: }
729: prev->and = (1 == logic);
730: prev->open += toopen;
731: if (cur != prev)
1.17 schwarze 732: cur->close = 1;
1.18 schwarze 733:
1.4 schwarze 734: toopen = logic = igncase = 0;
1.1 schwarze 735: }
1.40 schwarze 736: if ( ! (toopen || logic || igncase || toclose))
1.45 schwarze 737: return first;
1.6 schwarze 738:
1.4 schwarze 739: fail:
740: if (NULL != first)
741: exprfree(first);
1.45 schwarze 742: return NULL;
1.6 schwarze 743: }
744:
745: static struct expr *
1.1 schwarze 746: exprterm(const struct mansearch *search, char *buf, int cs)
747: {
1.6 schwarze 748: char errbuf[BUFSIZ];
1.1 schwarze 749: struct expr *e;
1.28 schwarze 750: char *key, *val;
1.11 schwarze 751: uint64_t iterbit;
752: int i, irc;
1.1 schwarze 753:
754: if ('\0' == *buf)
1.45 schwarze 755: return NULL;
1.1 schwarze 756:
757: e = mandoc_calloc(1, sizeof(struct expr));
758:
1.33 schwarze 759: if (search->argmode == ARG_NAME) {
760: e->bits = TYPE_Nm;
1.1 schwarze 761: e->substr = buf;
1.28 schwarze 762: e->equal = 1;
1.45 schwarze 763: return e;
1.1 schwarze 764: }
765:
766: /*
1.33 schwarze 767: * Separate macro keys from search string.
768: * If needed, request regular expression handling
769: * by setting e->substr to NULL.
1.1 schwarze 770: */
771:
1.33 schwarze 772: if (search->argmode == ARG_WORD) {
773: e->bits = TYPE_Nm;
774: e->substr = NULL;
775: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
776: cs = 0;
777: } else if ((val = strpbrk(buf, "=~")) == NULL) {
778: e->bits = TYPE_Nm | TYPE_Nd;
1.1 schwarze 779: e->substr = buf;
1.28 schwarze 780: } else {
781: if (val == buf)
1.33 schwarze 782: e->bits = TYPE_Nm | TYPE_Nd;
1.28 schwarze 783: if ('=' == *val)
784: e->substr = val + 1;
785: *val++ = '\0';
1.12 schwarze 786: if (NULL != strstr(buf, "arch"))
787: cs = 0;
1.28 schwarze 788: }
789:
790: /* Compile regular expressions. */
791:
792: if (NULL == e->substr) {
793: irc = regcomp(&e->regexp, val,
794: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.33 schwarze 795: if (search->argmode == ARG_WORD)
1.28 schwarze 796: free(val);
797: if (irc) {
1.6 schwarze 798: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
1.46 ! schwarze 799: warnx("regcomp: %s", errbuf);
1.1 schwarze 800: free(e);
1.45 schwarze 801: return NULL;
1.1 schwarze 802: }
1.28 schwarze 803: }
804:
805: if (e->bits)
1.45 schwarze 806: return e;
1.1 schwarze 807:
808: /*
809: * Parse out all possible fields.
810: * If the field doesn't resolve, bail.
811: */
812:
813: while (NULL != (key = strsep(&buf, ","))) {
814: if ('\0' == *key)
815: continue;
1.11 schwarze 816: for (i = 0, iterbit = 1;
817: i < mansearch_keymax;
818: i++, iterbit <<= 1) {
819: if (0 == strcasecmp(key,
820: mansearch_keynames[i])) {
821: e->bits |= iterbit;
822: break;
823: }
824: }
825: if (i == mansearch_keymax) {
826: if (strcasecmp(key, "any")) {
827: free(e);
1.45 schwarze 828: return NULL;
1.11 schwarze 829: }
830: e->bits |= ~0ULL;
1.1 schwarze 831: }
832: }
833:
1.45 schwarze 834: return e;
1.1 schwarze 835: }
836:
837: static void
838: exprfree(struct expr *p)
839: {
840: struct expr *pp;
841:
842: while (NULL != p) {
843: pp = p->next;
844: free(p);
845: p = pp;
846: }
847: }
848:
849: static void *
1.27 espie 850: hash_calloc(size_t nmemb, size_t sz, void *arg)
1.1 schwarze 851: {
852:
1.45 schwarze 853: return mandoc_calloc(nmemb, sz);
1.1 schwarze 854: }
855:
856: static void *
857: hash_alloc(size_t sz, void *arg)
858: {
859:
1.45 schwarze 860: return mandoc_malloc(sz);
1.1 schwarze 861: }
862:
863: static void
1.27 espie 864: hash_free(void *p, void *arg)
1.1 schwarze 865: {
866:
867: free(p);
868: }