Annotation of src/usr.bin/mandoc/mansearch.c, Revision 1.41
1.41 ! schwarze 1: /* $OpenBSD: mansearch.c,v 1.40 2015/01/20 18:19:39 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: *
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: */
1.33 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>
1.39 schwarze 23: #include <errno.h>
1.1 schwarze 24: #include <fcntl.h>
25: #include <getopt.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 <ohash.h>
37: #include <sqlite3.h>
38:
39: #include "mandoc.h"
1.14 schwarze 40: #include "mandoc_aux.h"
1.1 schwarze 41: #include "manpath.h"
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)) \
50: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
51: } while (0)
52: #define SQL_BIND_INT64(_db, _s, _i, _v) \
53: do { if (SQLITE_OK != sqlite3_bind_int64 \
54: ((_s), (_i)++, (_v))) \
55: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
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)) \
60: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
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);
1.1 schwarze 87: static void *hash_alloc(size_t, void *);
1.27 espie 88: static void hash_free(void *, void *);
89: static void *hash_calloc(size_t, size_t, void *);
1.24 schwarze 90: static struct expr *exprcomp(const struct mansearch *,
1.1 schwarze 91: int, char *[]);
92: static void exprfree(struct expr *);
93: static struct expr *exprterm(const struct mansearch *, char *, int);
1.29 schwarze 94: static int manpage_compare(const void *, const void *);
1.4 schwarze 95: static void sql_append(char **sql, size_t *sz,
96: const char *newstr, int count);
1.1 schwarze 97: static void sql_match(sqlite3_context *context,
98: int argc, sqlite3_value **argv);
99: static void sql_regexp(sqlite3_context *context,
100: int argc, sqlite3_value **argv);
1.6 schwarze 101: static char *sql_statement(const struct expr *);
1.19 schwarze 102:
1.24 schwarze 103:
1.19 schwarze 104: int
105: mansearch_setup(int start)
106: {
107: static void *pagecache;
108: int c;
109:
110: #define PC_PAGESIZE 1280
111: #define PC_NUMPAGES 256
112:
113: if (start) {
114: if (NULL != pagecache) {
115: fprintf(stderr, "pagecache already enabled\n");
116: return((int)MANDOCLEVEL_BADARG);
117: }
118:
119: pagecache = mmap(NULL, PC_PAGESIZE * PC_NUMPAGES,
1.31 schwarze 120: PROT_READ | PROT_WRITE,
121: MAP_SHARED | MAP_ANON, -1, 0);
1.19 schwarze 122:
123: if (MAP_FAILED == pagecache) {
124: perror("mmap");
125: pagecache = NULL;
126: return((int)MANDOCLEVEL_SYSERR);
127: }
128:
129: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
130: pagecache, PC_PAGESIZE, PC_NUMPAGES);
131:
132: if (SQLITE_OK == c)
133: return((int)MANDOCLEVEL_OK);
134:
135: fprintf(stderr, "pagecache: %s\n", sqlite3_errstr(c));
136:
137: } else if (NULL == pagecache) {
138: fprintf(stderr, "pagecache missing\n");
139: return((int)MANDOCLEVEL_BADARG);
140: }
141:
142: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
143: perror("munmap");
144: pagecache = NULL;
145: return((int)MANDOCLEVEL_SYSERR);
146: }
147:
148: pagecache = NULL;
149: return((int)MANDOCLEVEL_OK);
150: }
1.1 schwarze 151:
152: int
153: mansearch(const struct mansearch *search,
1.3 schwarze 154: const struct manpaths *paths,
155: int argc, char *argv[],
1.1 schwarze 156: struct manpage **res, size_t *sz)
157: {
1.11 schwarze 158: int fd, rc, c, indexbit;
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;
172:
1.27 espie 173: info.calloc = hash_calloc;
1.1 schwarze 174: info.alloc = hash_alloc;
1.27 espie 175: info.free = hash_free;
1.22 schwarze 176: info.key_offset = offsetof(struct match, pageid);
1.1 schwarze 177:
178: *sz = cur = maxres = 0;
179: sql = NULL;
180: *res = NULL;
181: fd = -1;
182: e = NULL;
183: rc = 0;
184:
185: if (0 == argc)
186: goto out;
187: if (NULL == (e = exprcomp(search, argc, argv)))
188: goto out;
189:
1.3 schwarze 190: outbit = 0;
1.33 schwarze 191: if (NULL != search->outkey) {
1.11 schwarze 192: for (indexbit = 0, iterbit = 1;
193: indexbit < mansearch_keymax;
194: indexbit++, iterbit <<= 1) {
1.33 schwarze 195: if (0 == strcasecmp(search->outkey,
1.11 schwarze 196: mansearch_keynames[indexbit])) {
197: outbit = iterbit;
1.3 schwarze 198: break;
199: }
200: }
201: }
202:
1.1 schwarze 203: /*
204: * Save a descriptor to the current working directory.
205: * Since pathnames in the "paths" variable might be relative,
206: * and we'll be chdir()ing into them, we need to keep a handle
207: * on our current directory from which to start the chdir().
208: */
209:
210: if (NULL == getcwd(buf, PATH_MAX)) {
1.20 schwarze 211: perror("getcwd");
1.1 schwarze 212: goto out;
213: } else if (-1 == (fd = open(buf, O_RDONLY, 0))) {
214: perror(buf);
215: goto out;
216: }
217:
1.6 schwarze 218: sql = sql_statement(e);
1.1 schwarze 219:
220: /*
221: * Loop over the directories (containing databases) for us to
222: * search.
223: * Don't let missing/bad databases/directories phase us.
224: * In each, try to open the resident database and, if it opens,
225: * scan it for our match expression.
226: */
227:
228: for (i = 0; i < paths->sz; i++) {
229: if (-1 == fchdir(fd)) {
230: perror(buf);
231: free(*res);
232: break;
233: } else if (-1 == chdir(paths->paths[i])) {
234: perror(paths->paths[i]);
235: continue;
1.24 schwarze 236: }
1.1 schwarze 237:
1.24 schwarze 238: c = sqlite3_open_v2(MANDOC_DB, &db,
239: SQLITE_OPEN_READONLY, NULL);
1.1 schwarze 240:
241: if (SQLITE_OK != c) {
1.39 schwarze 242: fprintf(stderr, "%s/%s: %s\n",
243: paths->paths[i], MANDOC_DB, strerror(errno));
1.1 schwarze 244: sqlite3_close(db);
245: continue;
246: }
247:
248: /*
249: * Define the SQL functions for substring
250: * and regular expression matching.
251: */
252:
253: c = sqlite3_create_function(db, "match", 2,
1.21 schwarze 254: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
255: NULL, sql_match, NULL, NULL);
1.1 schwarze 256: assert(SQLITE_OK == c);
257: c = sqlite3_create_function(db, "regexp", 2,
1.21 schwarze 258: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
259: NULL, sql_regexp, NULL, NULL);
1.1 schwarze 260: assert(SQLITE_OK == c);
261:
262: j = 1;
263: c = sqlite3_prepare_v2(db, sql, -1, &s, NULL);
264: if (SQLITE_OK != c)
265: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
266:
267: for (ep = e; NULL != ep; ep = ep->next) {
268: if (NULL == ep->substr) {
269: SQL_BIND_BLOB(db, s, j, ep->regexp);
270: } else
271: SQL_BIND_TEXT(db, s, j, ep->substr);
1.18 schwarze 272: if (0 == ((TYPE_Nd | TYPE_Nm) & ep->bits))
1.17 schwarze 273: SQL_BIND_INT64(db, s, j, ep->bits);
1.1 schwarze 274: }
275:
276: memset(&htab, 0, sizeof(struct ohash));
277: ohash_init(&htab, 4, &info);
278:
279: /*
280: * Hash each entry on its [unique] document identifier.
281: * This is a uint64_t.
282: * Instead of using a hash function, simply convert the
283: * uint64_t to a uint32_t, the hash value's type.
284: * This gives good performance and preserves the
285: * distribution of buckets in the table.
286: */
287: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.22 schwarze 288: pageid = sqlite3_column_int64(s, 2);
1.24 schwarze 289: idx = ohash_lookup_memory(&htab,
290: (char *)&pageid, sizeof(uint64_t),
291: (uint32_t)pageid);
1.1 schwarze 292:
293: if (NULL != ohash_find(&htab, idx))
294: continue;
295:
296: mp = mandoc_calloc(1, sizeof(struct match));
1.22 schwarze 297: mp->pageid = pageid;
1.17 schwarze 298: mp->form = sqlite3_column_int(s, 1);
1.37 schwarze 299: mp->bits = sqlite3_column_int64(s, 3);
1.17 schwarze 300: if (TYPE_Nd == outbit)
1.30 schwarze 301: mp->desc = mandoc_strdup((const char *)
1.17 schwarze 302: sqlite3_column_text(s, 0));
1.1 schwarze 303: ohash_insert(&htab, idx, mp);
304: }
305:
306: if (SQLITE_DONE != c)
307: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
308:
309: sqlite3_finalize(s);
310:
1.24 schwarze 311: c = sqlite3_prepare_v2(db,
1.25 schwarze 312: "SELECT sec, arch, name, pageid FROM mlinks "
313: "WHERE pageid=? ORDER BY sec, arch, name",
1.1 schwarze 314: -1, &s, NULL);
315: if (SQLITE_OK != c)
316: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
317:
1.3 schwarze 318: c = sqlite3_prepare_v2(db,
1.25 schwarze 319: "SELECT bits, key, pageid FROM keys "
320: "WHERE pageid=? AND bits & ?",
1.3 schwarze 321: -1, &s2, NULL);
322: if (SQLITE_OK != c)
323: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
324:
1.1 schwarze 325: for (mp = ohash_first(&htab, &idx);
326: NULL != mp;
327: mp = ohash_next(&htab, &idx)) {
328: if (cur + 1 > maxres) {
329: maxres += 1024;
1.26 schwarze 330: *res = mandoc_reallocarray(*res,
331: maxres, sizeof(struct manpage));
1.1 schwarze 332: }
333: mpage = *res + cur;
1.34 schwarze 334: mpage->ipath = i;
1.37 schwarze 335: mpage->bits = mp->bits;
1.29 schwarze 336: mpage->sec = 10;
1.1 schwarze 337: mpage->form = mp->form;
1.40 schwarze 338: buildnames(search, mpage, db, s, mp->pageid,
1.10 schwarze 339: paths->paths[i], mp->form);
1.40 schwarze 340: if (mpage->names != NULL) {
341: mpage->output = TYPE_Nd & outbit ?
342: mp->desc : outbit ?
343: buildoutput(db, s2, mp->pageid, outbit) :
344: NULL;
345: cur++;
346: }
1.1 schwarze 347: free(mp);
348: }
349:
350: sqlite3_finalize(s);
1.3 schwarze 351: sqlite3_finalize(s2);
1.1 schwarze 352: sqlite3_close(db);
353: ohash_delete(&htab);
1.36 schwarze 354:
355: /*
356: * In man(1) mode, prefer matches in earlier trees
357: * over matches in later trees.
358: */
359:
360: if (cur && search->firstmatch)
361: break;
1.1 schwarze 362: }
1.29 schwarze 363: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.1 schwarze 364: rc = 1;
365: out:
1.20 schwarze 366: if (-1 != fd) {
367: if (-1 == fchdir(fd))
368: perror(buf);
369: close(fd);
370: }
1.1 schwarze 371: exprfree(e);
372: free(sql);
373: *sz = cur;
374: return(rc);
1.2 schwarze 375: }
376:
1.33 schwarze 377: void
378: mansearch_free(struct manpage *res, size_t sz)
379: {
380: size_t i;
381:
382: for (i = 0; i < sz; i++) {
383: free(res[i].file);
384: free(res[i].names);
385: free(res[i].output);
386: }
387: free(res);
388: }
389:
1.29 schwarze 390: static int
391: manpage_compare(const void *vp1, const void *vp2)
392: {
393: const struct manpage *mp1, *mp2;
394: int diff;
395:
396: mp1 = vp1;
397: mp2 = vp2;
1.37 schwarze 398: return( (diff = mp2->bits - mp1->bits) ? diff :
399: (diff = mp1->sec - mp2->sec) ? diff :
400: strcasecmp(mp1->names, mp2->names));
1.29 schwarze 401: }
402:
1.8 schwarze 403: static void
1.40 schwarze 404: buildnames(const struct mansearch *search, struct manpage *mpage,
405: sqlite3 *db, sqlite3_stmt *s,
1.22 schwarze 406: uint64_t pageid, const char *path, int form)
1.2 schwarze 407: {
1.41 ! schwarze 408: glob_t globinfo;
! 409: char *firstname, *newnames, *prevsec, *prevarch;
1.10 schwarze 410: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.2 schwarze 411: size_t i;
1.41 ! schwarze 412: int c, globres;
1.2 schwarze 413:
1.16 schwarze 414: mpage->file = NULL;
1.8 schwarze 415: mpage->names = NULL;
1.41 ! schwarze 416: firstname = prevsec = prevarch = NULL;
1.2 schwarze 417: i = 1;
1.22 schwarze 418: SQL_BIND_INT64(db, s, i, pageid);
1.2 schwarze 419: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.8 schwarze 420:
1.13 schwarze 421: /* Decide whether we already have some names. */
1.8 schwarze 422:
423: if (NULL == mpage->names) {
1.2 schwarze 424: oldnames = "";
425: sep1 = "";
426: } else {
1.8 schwarze 427: oldnames = mpage->names;
1.2 schwarze 428: sep1 = ", ";
429: }
1.13 schwarze 430:
1.40 schwarze 431: /* Fetch the next name, rejecting sec/arch mismatches. */
1.13 schwarze 432:
1.30 schwarze 433: sec = (const char *)sqlite3_column_text(s, 0);
1.40 schwarze 434: if (search->sec != NULL && strcasecmp(sec, search->sec))
435: continue;
1.30 schwarze 436: arch = (const char *)sqlite3_column_text(s, 1);
1.40 schwarze 437: if (search->arch != NULL && *arch != '\0' &&
438: strcasecmp(arch, search->arch))
439: continue;
1.30 schwarze 440: name = (const char *)sqlite3_column_text(s, 2);
1.29 schwarze 441:
442: /* Remember the first section found. */
443:
444: if (9 < mpage->sec && '1' <= *sec && '9' >= *sec)
445: mpage->sec = (*sec - '1') + 1;
1.13 schwarze 446:
447: /* If the section changed, append the old one. */
448:
449: if (NULL != prevsec &&
450: (strcmp(sec, prevsec) ||
451: strcmp(arch, prevarch))) {
452: sep2 = '\0' == *prevarch ? "" : "/";
1.15 schwarze 453: mandoc_asprintf(&newnames, "%s(%s%s%s)",
454: oldnames, prevsec, sep2, prevarch);
1.13 schwarze 455: free(mpage->names);
456: oldnames = mpage->names = newnames;
457: free(prevsec);
458: free(prevarch);
459: prevsec = prevarch = NULL;
460: }
461:
462: /* Save the new section, to append it later. */
463:
464: if (NULL == prevsec) {
465: prevsec = mandoc_strdup(sec);
466: prevarch = mandoc_strdup(arch);
467: }
468:
469: /* Append the new name. */
470:
1.15 schwarze 471: mandoc_asprintf(&newnames, "%s%s%s",
472: oldnames, sep1, name);
1.8 schwarze 473: free(mpage->names);
474: mpage->names = newnames;
475:
476: /* Also save the first file name encountered. */
477:
1.38 schwarze 478: if (mpage->file != NULL)
1.8 schwarze 479: continue;
480:
1.35 schwarze 481: if (form & FORM_SRC) {
1.10 schwarze 482: sep1 = "man";
483: fsec = sec;
484: } else {
485: sep1 = "cat";
486: fsec = "0";
487: }
1.38 schwarze 488: sep2 = *arch == '\0' ? "" : "/";
489: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
490: path, sep1, sec, sep2, arch, name, fsec);
1.41 ! schwarze 491: if (access(mpage->file, R_OK) != -1)
! 492: continue;
! 493:
! 494: /* Handle unusual file name extensions. */
! 495:
! 496: if (firstname == NULL)
! 497: firstname = mpage->file;
! 498: else
! 499: free(mpage->file);
! 500: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.*",
! 501: path, sep1, sec, sep2, arch, name);
! 502: globres = glob(mpage->file, 0, NULL, &globinfo);
! 503: free(mpage->file);
! 504: mpage->file = globres ? NULL :
! 505: mandoc_strdup(*globinfo.gl_pathv);
! 506: globfree(&globinfo);
1.2 schwarze 507: }
1.38 schwarze 508: if (c != SQLITE_DONE)
1.2 schwarze 509: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
510: sqlite3_reset(s);
1.41 ! schwarze 511:
! 512: /* If none of the files is usable, use the first name. */
! 513:
! 514: if (mpage->file == NULL)
! 515: mpage->file = firstname;
! 516: else if (mpage->file != firstname)
! 517: free(firstname);
1.13 schwarze 518:
519: /* Append one final section to the names. */
520:
1.38 schwarze 521: if (prevsec != NULL) {
522: sep2 = *prevarch == '\0' ? "" : "/";
1.15 schwarze 523: mandoc_asprintf(&newnames, "%s(%s%s%s)",
524: mpage->names, prevsec, sep2, prevarch);
1.13 schwarze 525: free(mpage->names);
526: mpage->names = newnames;
527: free(prevsec);
528: free(prevarch);
529: }
1.3 schwarze 530: }
531:
532: static char *
1.22 schwarze 533: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.3 schwarze 534: {
535: char *output, *newoutput;
536: const char *oldoutput, *sep1, *data;
537: size_t i;
538: int c;
539:
540: output = NULL;
541: i = 1;
1.22 schwarze 542: SQL_BIND_INT64(db, s, i, pageid);
1.3 schwarze 543: SQL_BIND_INT64(db, s, i, outbit);
544: while (SQLITE_ROW == (c = sqlite3_step(s))) {
545: if (NULL == output) {
546: oldoutput = "";
547: sep1 = "";
548: } else {
549: oldoutput = output;
550: sep1 = " # ";
551: }
1.30 schwarze 552: data = (const char *)sqlite3_column_text(s, 1);
1.15 schwarze 553: mandoc_asprintf(&newoutput, "%s%s%s",
554: oldoutput, sep1, data);
1.3 schwarze 555: free(output);
556: output = newoutput;
557: }
558: if (SQLITE_DONE != c)
559: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
560: sqlite3_reset(s);
561: return(output);
1.1 schwarze 562: }
563:
564: /*
565: * Implement substring match as an application-defined SQL function.
566: * Using the SQL LIKE or GLOB operators instead would be a bad idea
567: * because that would require escaping metacharacters in the string
568: * being searched for.
569: */
570: static void
571: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
572: {
573:
574: assert(2 == argc);
575: sqlite3_result_int(context, NULL != strcasestr(
576: (const char *)sqlite3_value_text(argv[1]),
577: (const char *)sqlite3_value_text(argv[0])));
578: }
579:
580: /*
581: * Implement regular expression match
582: * as an application-defined SQL function.
583: */
584: static void
585: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
586: {
587:
588: assert(2 == argc);
589: sqlite3_result_int(context, !regexec(
590: (regex_t *)sqlite3_value_blob(argv[0]),
591: (const char *)sqlite3_value_text(argv[1]),
592: 0, NULL, 0));
593: }
594:
1.4 schwarze 595: static void
596: sql_append(char **sql, size_t *sz, const char *newstr, int count)
597: {
598: size_t newsz;
599:
600: newsz = 1 < count ? (size_t)count : strlen(newstr);
601: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
602: if (1 < count)
603: memset(*sql + *sz, *newstr, (size_t)count);
604: else
605: memcpy(*sql + *sz, newstr, newsz);
606: *sz += newsz;
607: (*sql)[*sz] = '\0';
608: }
609:
1.1 schwarze 610: /*
611: * Prepare the search SQL statement.
612: */
613: static char *
1.6 schwarze 614: sql_statement(const struct expr *e)
1.1 schwarze 615: {
616: char *sql;
617: size_t sz;
1.4 schwarze 618: int needop;
1.1 schwarze 619:
1.37 schwarze 620: sql = mandoc_strdup(e->equal ?
621: "SELECT desc, form, pageid, bits "
622: "FROM mpages NATURAL JOIN names WHERE " :
623: "SELECT desc, form, pageid, 0 FROM mpages WHERE ");
1.1 schwarze 624: sz = strlen(sql);
625:
1.4 schwarze 626: for (needop = 0; NULL != e; e = e->next) {
627: if (e->and)
628: sql_append(&sql, &sz, " AND ", 1);
629: else if (needop)
630: sql_append(&sql, &sz, " OR ", 1);
631: if (e->open)
632: sql_append(&sql, &sz, "(", e->open);
1.17 schwarze 633: sql_append(&sql, &sz,
634: TYPE_Nd & e->bits
635: ? (NULL == e->substr
636: ? "desc REGEXP ?"
637: : "desc MATCH ?")
1.18 schwarze 638: : TYPE_Nm == e->bits
639: ? (NULL == e->substr
1.22 schwarze 640: ? "pageid IN (SELECT pageid FROM names "
1.18 schwarze 641: "WHERE name REGEXP ?)"
1.28 schwarze 642: : e->equal
1.37 schwarze 643: ? "name = ? "
1.22 schwarze 644: : "pageid IN (SELECT pageid FROM names "
1.18 schwarze 645: "WHERE name MATCH ?)")
1.17 schwarze 646: : (NULL == e->substr
1.22 schwarze 647: ? "pageid IN (SELECT pageid FROM keys "
1.17 schwarze 648: "WHERE key REGEXP ? AND bits & ?)"
1.22 schwarze 649: : "pageid IN (SELECT pageid FROM keys "
1.17 schwarze 650: "WHERE key MATCH ? AND bits & ?)"), 1);
1.4 schwarze 651: if (e->close)
652: sql_append(&sql, &sz, ")", e->close);
653: needop = 1;
1.1 schwarze 654: }
655:
656: return(sql);
657: }
658:
659: /*
660: * Compile a set of string tokens into an expression.
661: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
662: * "(", "foo=bar", etc.).
663: */
664: static struct expr *
665: exprcomp(const struct mansearch *search, int argc, char *argv[])
666: {
1.18 schwarze 667: uint64_t mask;
1.4 schwarze 668: int i, toopen, logic, igncase, toclose;
1.18 schwarze 669: struct expr *first, *prev, *cur, *next;
1.1 schwarze 670:
671: first = cur = NULL;
1.40 schwarze 672: logic = igncase = toopen = toclose = 0;
1.1 schwarze 673:
674: for (i = 0; i < argc; i++) {
1.4 schwarze 675: if (0 == strcmp("(", argv[i])) {
676: if (igncase)
677: goto fail;
678: toopen++;
679: toclose++;
680: continue;
681: } else if (0 == strcmp(")", argv[i])) {
682: if (toopen || logic || igncase || NULL == cur)
683: goto fail;
684: cur->close++;
685: if (0 > --toclose)
686: goto fail;
687: continue;
688: } else if (0 == strcmp("-a", argv[i])) {
689: if (toopen || logic || igncase || NULL == cur)
690: goto fail;
691: logic = 1;
692: continue;
693: } else if (0 == strcmp("-o", argv[i])) {
694: if (toopen || logic || igncase || NULL == cur)
695: goto fail;
696: logic = 2;
697: continue;
698: } else if (0 == strcmp("-i", argv[i])) {
699: if (igncase)
700: goto fail;
701: igncase = 1;
702: continue;
1.1 schwarze 703: }
1.4 schwarze 704: next = exprterm(search, argv[i], !igncase);
705: if (NULL == next)
706: goto fail;
1.17 schwarze 707: if (NULL == first)
708: first = next;
709: else
1.1 schwarze 710: cur->next = next;
1.18 schwarze 711: prev = cur = next;
1.17 schwarze 712:
713: /*
714: * Searching for descriptions must be split out
715: * because they are stored in the mpages table,
716: * not in the keys table.
717: */
718:
1.18 schwarze 719: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
720: if (mask & cur->bits && ~mask & cur->bits) {
721: next = mandoc_calloc(1,
722: sizeof(struct expr));
723: memcpy(next, cur, sizeof(struct expr));
724: prev->open = 1;
725: cur->bits = mask;
726: cur->next = next;
727: cur = next;
728: cur->bits &= ~mask;
729: }
730: }
731: prev->and = (1 == logic);
732: prev->open += toopen;
733: if (cur != prev)
1.17 schwarze 734: cur->close = 1;
1.18 schwarze 735:
1.4 schwarze 736: toopen = logic = igncase = 0;
1.1 schwarze 737: }
1.40 schwarze 738: if ( ! (toopen || logic || igncase || toclose))
739: return(first);
1.6 schwarze 740:
1.4 schwarze 741: fail:
742: if (NULL != first)
743: exprfree(first);
744: return(NULL);
1.6 schwarze 745: }
746:
747: static struct expr *
1.1 schwarze 748: exprterm(const struct mansearch *search, char *buf, int cs)
749: {
1.6 schwarze 750: char errbuf[BUFSIZ];
1.1 schwarze 751: struct expr *e;
1.28 schwarze 752: char *key, *val;
1.11 schwarze 753: uint64_t iterbit;
754: int i, irc;
1.1 schwarze 755:
756: if ('\0' == *buf)
757: return(NULL);
758:
759: e = mandoc_calloc(1, sizeof(struct expr));
760:
1.33 schwarze 761: if (search->argmode == ARG_NAME) {
762: e->bits = TYPE_Nm;
1.1 schwarze 763: e->substr = buf;
1.28 schwarze 764: e->equal = 1;
1.1 schwarze 765: return(e);
766: }
767:
768: /*
1.33 schwarze 769: * Separate macro keys from search string.
770: * If needed, request regular expression handling
771: * by setting e->substr to NULL.
1.1 schwarze 772: */
773:
1.33 schwarze 774: if (search->argmode == ARG_WORD) {
775: e->bits = TYPE_Nm;
776: e->substr = NULL;
777: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
778: cs = 0;
779: } else if ((val = strpbrk(buf, "=~")) == NULL) {
780: e->bits = TYPE_Nm | TYPE_Nd;
1.1 schwarze 781: e->substr = buf;
1.28 schwarze 782: } else {
783: if (val == buf)
1.33 schwarze 784: e->bits = TYPE_Nm | TYPE_Nd;
1.28 schwarze 785: if ('=' == *val)
786: e->substr = val + 1;
787: *val++ = '\0';
1.12 schwarze 788: if (NULL != strstr(buf, "arch"))
789: cs = 0;
1.28 schwarze 790: }
791:
792: /* Compile regular expressions. */
793:
794: if (NULL == e->substr) {
795: irc = regcomp(&e->regexp, val,
796: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.33 schwarze 797: if (search->argmode == ARG_WORD)
1.28 schwarze 798: free(val);
799: if (irc) {
1.6 schwarze 800: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
801: fprintf(stderr, "regcomp: %s\n", errbuf);
1.1 schwarze 802: free(e);
803: return(NULL);
804: }
1.28 schwarze 805: }
806:
807: if (e->bits)
808: return(e);
1.1 schwarze 809:
810: /*
811: * Parse out all possible fields.
812: * If the field doesn't resolve, bail.
813: */
814:
815: while (NULL != (key = strsep(&buf, ","))) {
816: if ('\0' == *key)
817: continue;
1.11 schwarze 818: for (i = 0, iterbit = 1;
819: i < mansearch_keymax;
820: i++, iterbit <<= 1) {
821: if (0 == strcasecmp(key,
822: mansearch_keynames[i])) {
823: e->bits |= iterbit;
824: break;
825: }
826: }
827: if (i == mansearch_keymax) {
828: if (strcasecmp(key, "any")) {
829: free(e);
830: return(NULL);
831: }
832: e->bits |= ~0ULL;
1.1 schwarze 833: }
834: }
835:
836: return(e);
837: }
838:
839: static void
840: exprfree(struct expr *p)
841: {
842: struct expr *pp;
843:
844: while (NULL != p) {
845: pp = p->next;
846: free(p);
847: p = pp;
848: }
849: }
850:
851: static void *
1.27 espie 852: hash_calloc(size_t nmemb, size_t sz, void *arg)
1.1 schwarze 853: {
854:
1.27 espie 855: return(mandoc_calloc(nmemb, sz));
1.1 schwarze 856: }
857:
858: static void *
859: hash_alloc(size_t sz, void *arg)
860: {
861:
862: return(mandoc_malloc(sz));
863: }
864:
865: static void
1.27 espie 866: hash_free(void *p, void *arg)
1.1 schwarze 867: {
868:
869: free(p);
870: }