Annotation of mandoc/mansearch.c, Revision 1.62
1.62 ! schwarze 1: /* $Id: mansearch.c,v 1.61 2015/11/06 21:19:09 schwarze Exp $ */
1.1 kristaps 2: /*
3: * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.53 schwarze 4: * Copyright (c) 2013, 2014, 2015 Ingo Schwarze <schwarze@openbsd.org>
1.1 kristaps 5: *
6: * Permission to use, copy, modify, and distribute this software for any
7: * purpose with or without fee is hereby granted, provided that the above
8: * copyright notice and this permission notice appear in all copies.
9: *
1.56 schwarze 10: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
1.1 kristaps 11: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
1.56 schwarze 12: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
1.1 kristaps 13: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17: */
18: #include "config.h"
19:
1.28 schwarze 20: #include <sys/mman.h>
1.43 schwarze 21: #include <sys/types.h>
22:
1.1 kristaps 23: #include <assert.h>
1.62 ! schwarze 24: #if HAVE_ERR
1.59 schwarze 25: #include <err.h>
1.62 ! schwarze 26: #endif
1.52 schwarze 27: #include <errno.h>
1.1 kristaps 28: #include <fcntl.h>
29: #include <getopt.h>
1.54 schwarze 30: #include <glob.h>
1.6 schwarze 31: #include <limits.h>
1.8 schwarze 32: #include <regex.h>
1.1 kristaps 33: #include <stdio.h>
34: #include <stdint.h>
35: #include <stddef.h>
36: #include <stdlib.h>
37: #include <string.h>
38: #include <unistd.h>
39:
40: #include <sqlite3.h>
1.40 schwarze 41: #ifndef SQLITE_DETERMINISTIC
42: #define SQLITE_DETERMINISTIC 0
43: #endif
1.1 kristaps 44:
45: #include "mandoc.h"
1.23 schwarze 46: #include "mandoc_aux.h"
1.60 schwarze 47: #include "mandoc_ohash.h"
1.56 schwarze 48: #include "manconf.h"
1.1 kristaps 49: #include "mansearch.h"
50:
1.20 schwarze 51: extern int mansearch_keymax;
52: extern const char *const mansearch_keynames[];
53:
1.3 kristaps 54: #define SQL_BIND_TEXT(_db, _s, _i, _v) \
1.8 schwarze 55: do { if (SQLITE_OK != sqlite3_bind_text \
1.2 kristaps 56: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.59 schwarze 57: warnx("%s", sqlite3_errmsg((_db))); \
1.8 schwarze 58: } while (0)
1.3 kristaps 59: #define SQL_BIND_INT64(_db, _s, _i, _v) \
1.8 schwarze 60: do { if (SQLITE_OK != sqlite3_bind_int64 \
1.2 kristaps 61: ((_s), (_i)++, (_v))) \
1.59 schwarze 62: warnx("%s", sqlite3_errmsg((_db))); \
1.8 schwarze 63: } while (0)
64: #define SQL_BIND_BLOB(_db, _s, _i, _v) \
65: do { if (SQLITE_OK != sqlite3_bind_blob \
66: ((_s), (_i)++, (&_v), sizeof(_v), SQLITE_STATIC)) \
1.59 schwarze 67: warnx("%s", sqlite3_errmsg((_db))); \
1.8 schwarze 68: } while (0)
1.2 kristaps 69:
1.1 kristaps 70: struct expr {
1.38 schwarze 71: regex_t regexp; /* compiled regexp, if applicable */
72: const char *substr; /* to search for, if applicable */
73: struct expr *next; /* next in sequence */
1.34 schwarze 74: uint64_t bits; /* type-mask */
1.38 schwarze 75: int equal; /* equality, not subsring match */
1.13 schwarze 76: int open; /* opening parentheses before */
77: int and; /* logical AND before */
78: int close; /* closing parentheses after */
1.1 kristaps 79: };
80:
81: struct match {
1.32 schwarze 82: uint64_t pageid; /* identifier in database */
1.50 schwarze 83: uint64_t bits; /* name type mask */
1.26 schwarze 84: char *desc; /* manual page description */
1.48 schwarze 85: int form; /* bit field: formatted, zipped? */
1.1 kristaps 86: };
87:
1.53 schwarze 88: static void buildnames(const struct mansearch *,
89: struct manpage *, sqlite3 *,
1.19 schwarze 90: sqlite3_stmt *, uint64_t,
91: const char *, int form);
1.12 schwarze 92: static char *buildoutput(sqlite3 *, sqlite3_stmt *,
93: uint64_t, uint64_t);
1.34 schwarze 94: static struct expr *exprcomp(const struct mansearch *,
1.5 kristaps 95: int, char *[]);
1.1 kristaps 96: static void exprfree(struct expr *);
1.8 schwarze 97: static struct expr *exprterm(const struct mansearch *, char *, int);
1.39 schwarze 98: static int manpage_compare(const void *, const void *);
1.13 schwarze 99: static void sql_append(char **sql, size_t *sz,
100: const char *newstr, int count);
1.7 schwarze 101: static void sql_match(sqlite3_context *context,
102: int argc, sqlite3_value **argv);
1.8 schwarze 103: static void sql_regexp(sqlite3_context *context,
104: int argc, sqlite3_value **argv);
1.15 schwarze 105: static char *sql_statement(const struct expr *);
1.28 schwarze 106:
1.34 schwarze 107:
1.28 schwarze 108: int
109: mansearch_setup(int start)
110: {
111: static void *pagecache;
112: int c;
113:
114: #define PC_PAGESIZE 1280
115: #define PC_NUMPAGES 256
116:
117: if (start) {
118: if (NULL != pagecache) {
1.59 schwarze 119: warnx("pagecache already enabled");
1.58 schwarze 120: return (int)MANDOCLEVEL_BADARG;
1.28 schwarze 121: }
122:
123: pagecache = mmap(NULL, PC_PAGESIZE * PC_NUMPAGES,
1.42 schwarze 124: PROT_READ | PROT_WRITE,
125: MAP_SHARED | MAP_ANON, -1, 0);
1.28 schwarze 126:
127: if (MAP_FAILED == pagecache) {
128: perror("mmap");
129: pagecache = NULL;
1.58 schwarze 130: return (int)MANDOCLEVEL_SYSERR;
1.28 schwarze 131: }
132:
133: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
134: pagecache, PC_PAGESIZE, PC_NUMPAGES);
135:
136: if (SQLITE_OK == c)
1.58 schwarze 137: return (int)MANDOCLEVEL_OK;
1.28 schwarze 138:
1.59 schwarze 139: warnx("pagecache: %s", sqlite3_errstr(c));
1.28 schwarze 140:
141: } else if (NULL == pagecache) {
1.59 schwarze 142: warnx("pagecache missing");
1.58 schwarze 143: return (int)MANDOCLEVEL_BADARG;
1.28 schwarze 144: }
145:
146: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
147: perror("munmap");
148: pagecache = NULL;
1.58 schwarze 149: return (int)MANDOCLEVEL_SYSERR;
1.28 schwarze 150: }
151:
152: pagecache = NULL;
1.58 schwarze 153: return (int)MANDOCLEVEL_OK;
1.28 schwarze 154: }
1.1 kristaps 155:
156: int
1.5 kristaps 157: mansearch(const struct mansearch *search,
1.12 schwarze 158: const struct manpaths *paths,
159: int argc, char *argv[],
1.1 kristaps 160: struct manpage **res, size_t *sz)
161: {
1.32 schwarze 162: int64_t pageid;
1.20 schwarze 163: uint64_t outbit, iterbit;
1.6 schwarze 164: char buf[PATH_MAX];
1.11 schwarze 165: char *sql;
1.10 schwarze 166: struct manpage *mpage;
1.1 kristaps 167: struct expr *e, *ep;
168: sqlite3 *db;
1.12 schwarze 169: sqlite3_stmt *s, *s2;
1.1 kristaps 170: struct match *mp;
171: struct ohash htab;
172: unsigned int idx;
173: size_t i, j, cur, maxres;
1.57 schwarze 174: int c, chdir_status, getcwd_status, indexbit;
175:
176: if (argc == 0 || (e = exprcomp(search, argc, argv)) == NULL) {
177: *sz = 0;
1.58 schwarze 178: return 0;
1.57 schwarze 179: }
1.1 kristaps 180:
1.57 schwarze 181: cur = maxres = 0;
1.1 kristaps 182: *res = NULL;
183:
1.45 schwarze 184: if (NULL != search->outkey) {
1.55 schwarze 185: outbit = TYPE_Nd;
1.20 schwarze 186: for (indexbit = 0, iterbit = 1;
187: indexbit < mansearch_keymax;
188: indexbit++, iterbit <<= 1) {
1.45 schwarze 189: if (0 == strcasecmp(search->outkey,
1.20 schwarze 190: mansearch_keynames[indexbit])) {
191: outbit = iterbit;
1.12 schwarze 192: break;
193: }
194: }
1.55 schwarze 195: } else
196: outbit = 0;
1.12 schwarze 197:
1.1 kristaps 198: /*
1.57 schwarze 199: * Remember the original working directory, if possible.
200: * This will be needed if the second or a later directory
201: * is given as a relative path.
202: * Do not error out if the current directory is not
203: * searchable: Maybe it won't be needed after all.
1.1 kristaps 204: */
205:
1.57 schwarze 206: if (getcwd(buf, PATH_MAX) == NULL) {
207: getcwd_status = 0;
208: (void)strlcpy(buf, strerror(errno), sizeof(buf));
209: } else
210: getcwd_status = 1;
1.1 kristaps 211:
1.15 schwarze 212: sql = sql_statement(e);
1.1 kristaps 213:
214: /*
215: * Loop over the directories (containing databases) for us to
216: * search.
217: * Don't let missing/bad databases/directories phase us.
218: * In each, try to open the resident database and, if it opens,
219: * scan it for our match expression.
220: */
221:
1.57 schwarze 222: chdir_status = 0;
1.1 kristaps 223: for (i = 0; i < paths->sz; i++) {
1.57 schwarze 224: if (chdir_status && paths->paths[i][0] != '/') {
225: if ( ! getcwd_status) {
1.59 schwarze 226: warnx("%s: getcwd: %s", paths->paths[i], buf);
1.57 schwarze 227: continue;
228: } else if (chdir(buf) == -1) {
229: perror(buf);
230: continue;
231: }
232: }
233: if (chdir(paths->paths[i]) == -1) {
1.1 kristaps 234: perror(paths->paths[i]);
235: continue;
1.34 schwarze 236: }
1.57 schwarze 237: chdir_status = 1;
1.1 kristaps 238:
1.34 schwarze 239: c = sqlite3_open_v2(MANDOC_DB, &db,
240: SQLITE_OPEN_READONLY, NULL);
1.1 kristaps 241:
1.2 kristaps 242: if (SQLITE_OK != c) {
1.59 schwarze 243: warn("%s/%s", paths->paths[i], MANDOC_DB);
1.1 kristaps 244: sqlite3_close(db);
245: continue;
246: }
247:
1.8 schwarze 248: /*
249: * Define the SQL functions for substring
250: * and regular expression matching.
251: */
1.7 schwarze 252:
253: c = sqlite3_create_function(db, "match", 2,
1.31 schwarze 254: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
255: NULL, sql_match, NULL, NULL);
1.8 schwarze 256: assert(SQLITE_OK == c);
257: c = sqlite3_create_function(db, "regexp", 2,
1.31 schwarze 258: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
259: NULL, sql_regexp, NULL, NULL);
1.8 schwarze 260: assert(SQLITE_OK == c);
1.7 schwarze 261:
1.1 kristaps 262: j = 1;
1.2 kristaps 263: c = sqlite3_prepare_v2(db, sql, -1, &s, NULL);
264: if (SQLITE_OK != c)
1.59 schwarze 265: warnx("%s", sqlite3_errmsg(db));
1.1 kristaps 266:
267: for (ep = e; NULL != ep; ep = ep->next) {
1.8 schwarze 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);
1.27 schwarze 272: if (0 == ((TYPE_Nd | TYPE_Nm) & ep->bits))
1.26 schwarze 273: SQL_BIND_INT64(db, s, j, ep->bits);
1.1 kristaps 274: }
275:
1.60 schwarze 276: mandoc_ohash_init(&htab, 4, offsetof(struct match, pageid));
1.1 kristaps 277:
278: /*
279: * Hash each entry on its [unique] document identifier.
280: * This is a uint64_t.
281: * Instead of using a hash function, simply convert the
282: * uint64_t to a uint32_t, the hash value's type.
283: * This gives good performance and preserves the
284: * distribution of buckets in the table.
285: */
1.2 kristaps 286: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.32 schwarze 287: pageid = sqlite3_column_int64(s, 2);
1.34 schwarze 288: idx = ohash_lookup_memory(&htab,
289: (char *)&pageid, sizeof(uint64_t),
290: (uint32_t)pageid);
1.1 kristaps 291:
292: if (NULL != ohash_find(&htab, idx))
293: continue;
294:
295: mp = mandoc_calloc(1, sizeof(struct match));
1.32 schwarze 296: mp->pageid = pageid;
1.26 schwarze 297: mp->form = sqlite3_column_int(s, 1);
1.50 schwarze 298: mp->bits = sqlite3_column_int64(s, 3);
1.26 schwarze 299: if (TYPE_Nd == outbit)
1.41 schwarze 300: mp->desc = mandoc_strdup((const char *)
1.26 schwarze 301: sqlite3_column_text(s, 0));
1.1 kristaps 302: ohash_insert(&htab, idx, mp);
303: }
304:
1.2 kristaps 305: if (SQLITE_DONE != c)
1.59 schwarze 306: warnx("%s", sqlite3_errmsg(db));
1.2 kristaps 307:
1.1 kristaps 308: sqlite3_finalize(s);
1.10 schwarze 309:
1.34 schwarze 310: c = sqlite3_prepare_v2(db,
1.35 schwarze 311: "SELECT sec, arch, name, pageid FROM mlinks "
312: "WHERE pageid=? ORDER BY sec, arch, name",
1.10 schwarze 313: -1, &s, NULL);
314: if (SQLITE_OK != c)
1.59 schwarze 315: warnx("%s", sqlite3_errmsg(db));
1.1 kristaps 316:
1.12 schwarze 317: c = sqlite3_prepare_v2(db,
1.35 schwarze 318: "SELECT bits, key, pageid FROM keys "
319: "WHERE pageid=? AND bits & ?",
1.12 schwarze 320: -1, &s2, NULL);
321: if (SQLITE_OK != c)
1.59 schwarze 322: warnx("%s", sqlite3_errmsg(db));
1.12 schwarze 323:
1.1 kristaps 324: for (mp = ohash_first(&htab, &idx);
325: NULL != mp;
326: mp = ohash_next(&htab, &idx)) {
327: if (cur + 1 > maxres) {
328: maxres += 1024;
1.36 schwarze 329: *res = mandoc_reallocarray(*res,
330: maxres, sizeof(struct manpage));
1.1 kristaps 331: }
1.10 schwarze 332: mpage = *res + cur;
1.47 schwarze 333: mpage->ipath = i;
1.50 schwarze 334: mpage->bits = mp->bits;
1.39 schwarze 335: mpage->sec = 10;
1.10 schwarze 336: mpage->form = mp->form;
1.53 schwarze 337: buildnames(search, mpage, db, s, mp->pageid,
1.19 schwarze 338: paths->paths[i], mp->form);
1.53 schwarze 339: if (mpage->names != NULL) {
340: mpage->output = TYPE_Nd & outbit ?
341: mp->desc : outbit ?
342: buildoutput(db, s2, mp->pageid, outbit) :
343: NULL;
344: cur++;
345: }
1.1 kristaps 346: free(mp);
347: }
1.10 schwarze 348:
349: sqlite3_finalize(s);
1.12 schwarze 350: sqlite3_finalize(s2);
1.10 schwarze 351: sqlite3_close(db);
1.1 kristaps 352: ohash_delete(&htab);
1.49 schwarze 353:
354: /*
355: * In man(1) mode, prefer matches in earlier trees
356: * over matches in later trees.
357: */
358:
359: if (cur && search->firstmatch)
360: break;
1.1 kristaps 361: }
1.39 schwarze 362: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.57 schwarze 363: if (chdir_status && getcwd_status && chdir(buf) == -1)
364: perror(buf);
1.1 kristaps 365: exprfree(e);
366: free(sql);
367: *sz = cur;
1.58 schwarze 368: return 1;
1.11 schwarze 369: }
370:
1.45 schwarze 371: void
372: mansearch_free(struct manpage *res, size_t sz)
373: {
374: size_t i;
375:
376: for (i = 0; i < sz; i++) {
377: free(res[i].file);
378: free(res[i].names);
379: free(res[i].output);
380: }
381: free(res);
382: }
383:
1.39 schwarze 384: static int
385: manpage_compare(const void *vp1, const void *vp2)
386: {
387: const struct manpage *mp1, *mp2;
388: int diff;
389:
390: mp1 = vp1;
391: mp2 = vp2;
1.58 schwarze 392: return (diff = mp2->bits - mp1->bits) ? diff :
393: (diff = mp1->sec - mp2->sec) ? diff :
394: strcasecmp(mp1->names, mp2->names);
1.39 schwarze 395: }
396:
1.17 schwarze 397: static void
1.53 schwarze 398: buildnames(const struct mansearch *search, struct manpage *mpage,
399: sqlite3 *db, sqlite3_stmt *s,
1.32 schwarze 400: uint64_t pageid, const char *path, int form)
1.11 schwarze 401: {
1.54 schwarze 402: glob_t globinfo;
403: char *firstname, *newnames, *prevsec, *prevarch;
1.19 schwarze 404: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.11 schwarze 405: size_t i;
1.54 schwarze 406: int c, globres;
1.11 schwarze 407:
1.25 schwarze 408: mpage->file = NULL;
1.17 schwarze 409: mpage->names = NULL;
1.54 schwarze 410: firstname = prevsec = prevarch = NULL;
1.11 schwarze 411: i = 1;
1.32 schwarze 412: SQL_BIND_INT64(db, s, i, pageid);
1.11 schwarze 413: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.17 schwarze 414:
1.22 schwarze 415: /* Decide whether we already have some names. */
1.17 schwarze 416:
417: if (NULL == mpage->names) {
1.11 schwarze 418: oldnames = "";
419: sep1 = "";
420: } else {
1.17 schwarze 421: oldnames = mpage->names;
1.11 schwarze 422: sep1 = ", ";
423: }
1.22 schwarze 424:
1.53 schwarze 425: /* Fetch the next name, rejecting sec/arch mismatches. */
1.22 schwarze 426:
1.41 schwarze 427: sec = (const char *)sqlite3_column_text(s, 0);
1.53 schwarze 428: if (search->sec != NULL && strcasecmp(sec, search->sec))
429: continue;
1.41 schwarze 430: arch = (const char *)sqlite3_column_text(s, 1);
1.53 schwarze 431: if (search->arch != NULL && *arch != '\0' &&
432: strcasecmp(arch, search->arch))
433: continue;
1.41 schwarze 434: name = (const char *)sqlite3_column_text(s, 2);
1.39 schwarze 435:
436: /* Remember the first section found. */
437:
438: if (9 < mpage->sec && '1' <= *sec && '9' >= *sec)
439: mpage->sec = (*sec - '1') + 1;
1.22 schwarze 440:
441: /* If the section changed, append the old one. */
442:
443: if (NULL != prevsec &&
444: (strcmp(sec, prevsec) ||
445: strcmp(arch, prevarch))) {
446: sep2 = '\0' == *prevarch ? "" : "/";
1.24 schwarze 447: mandoc_asprintf(&newnames, "%s(%s%s%s)",
448: oldnames, prevsec, sep2, prevarch);
1.22 schwarze 449: free(mpage->names);
450: oldnames = mpage->names = newnames;
451: free(prevsec);
452: free(prevarch);
453: prevsec = prevarch = NULL;
454: }
455:
456: /* Save the new section, to append it later. */
457:
458: if (NULL == prevsec) {
459: prevsec = mandoc_strdup(sec);
460: prevarch = mandoc_strdup(arch);
461: }
462:
463: /* Append the new name. */
464:
1.24 schwarze 465: mandoc_asprintf(&newnames, "%s%s%s",
466: oldnames, sep1, name);
1.17 schwarze 467: free(mpage->names);
468: mpage->names = newnames;
469:
470: /* Also save the first file name encountered. */
471:
1.51 schwarze 472: if (mpage->file != NULL)
1.17 schwarze 473: continue;
474:
1.48 schwarze 475: if (form & FORM_SRC) {
1.19 schwarze 476: sep1 = "man";
477: fsec = sec;
478: } else {
479: sep1 = "cat";
480: fsec = "0";
481: }
1.51 schwarze 482: sep2 = *arch == '\0' ? "" : "/";
483: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
484: path, sep1, sec, sep2, arch, name, fsec);
1.54 schwarze 485: if (access(mpage->file, R_OK) != -1)
486: continue;
487:
488: /* Handle unusual file name extensions. */
489:
490: if (firstname == NULL)
491: firstname = mpage->file;
492: else
493: free(mpage->file);
494: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.*",
495: path, sep1, sec, sep2, arch, name);
496: globres = glob(mpage->file, 0, NULL, &globinfo);
497: free(mpage->file);
498: mpage->file = globres ? NULL :
499: mandoc_strdup(*globinfo.gl_pathv);
500: globfree(&globinfo);
1.11 schwarze 501: }
1.51 schwarze 502: if (c != SQLITE_DONE)
1.59 schwarze 503: warnx("%s", sqlite3_errmsg(db));
1.11 schwarze 504: sqlite3_reset(s);
1.54 schwarze 505:
506: /* If none of the files is usable, use the first name. */
507:
508: if (mpage->file == NULL)
509: mpage->file = firstname;
510: else if (mpage->file != firstname)
511: free(firstname);
1.22 schwarze 512:
513: /* Append one final section to the names. */
514:
1.51 schwarze 515: if (prevsec != NULL) {
516: sep2 = *prevarch == '\0' ? "" : "/";
1.24 schwarze 517: mandoc_asprintf(&newnames, "%s(%s%s%s)",
518: mpage->names, prevsec, sep2, prevarch);
1.22 schwarze 519: free(mpage->names);
520: mpage->names = newnames;
521: free(prevsec);
522: free(prevarch);
523: }
1.12 schwarze 524: }
525:
526: static char *
1.32 schwarze 527: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.12 schwarze 528: {
529: char *output, *newoutput;
530: const char *oldoutput, *sep1, *data;
531: size_t i;
532: int c;
533:
534: output = NULL;
535: i = 1;
1.32 schwarze 536: SQL_BIND_INT64(db, s, i, pageid);
1.12 schwarze 537: SQL_BIND_INT64(db, s, i, outbit);
538: while (SQLITE_ROW == (c = sqlite3_step(s))) {
539: if (NULL == output) {
540: oldoutput = "";
541: sep1 = "";
542: } else {
543: oldoutput = output;
544: sep1 = " # ";
545: }
1.41 schwarze 546: data = (const char *)sqlite3_column_text(s, 1);
1.24 schwarze 547: mandoc_asprintf(&newoutput, "%s%s%s",
548: oldoutput, sep1, data);
1.12 schwarze 549: free(output);
550: output = newoutput;
551: }
552: if (SQLITE_DONE != c)
1.59 schwarze 553: warnx("%s", sqlite3_errmsg(db));
1.12 schwarze 554: sqlite3_reset(s);
1.58 schwarze 555: return output;
1.1 kristaps 556: }
557:
558: /*
1.7 schwarze 559: * Implement substring match as an application-defined SQL function.
560: * Using the SQL LIKE or GLOB operators instead would be a bad idea
561: * because that would require escaping metacharacters in the string
562: * being searched for.
563: */
564: static void
565: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
566: {
567:
568: assert(2 == argc);
569: sqlite3_result_int(context, NULL != strcasestr(
570: (const char *)sqlite3_value_text(argv[1]),
571: (const char *)sqlite3_value_text(argv[0])));
572: }
573:
574: /*
1.8 schwarze 575: * Implement regular expression match
576: * as an application-defined SQL function.
577: */
578: static void
579: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
580: {
581:
582: assert(2 == argc);
583: sqlite3_result_int(context, !regexec(
584: (regex_t *)sqlite3_value_blob(argv[0]),
585: (const char *)sqlite3_value_text(argv[1]),
586: 0, NULL, 0));
587: }
588:
1.13 schwarze 589: static void
590: sql_append(char **sql, size_t *sz, const char *newstr, int count)
591: {
592: size_t newsz;
593:
594: newsz = 1 < count ? (size_t)count : strlen(newstr);
595: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
596: if (1 < count)
597: memset(*sql + *sz, *newstr, (size_t)count);
598: else
599: memcpy(*sql + *sz, newstr, newsz);
600: *sz += newsz;
601: (*sql)[*sz] = '\0';
602: }
603:
1.8 schwarze 604: /*
1.1 kristaps 605: * Prepare the search SQL statement.
606: */
607: static char *
1.15 schwarze 608: sql_statement(const struct expr *e)
1.1 kristaps 609: {
610: char *sql;
611: size_t sz;
1.13 schwarze 612: int needop;
1.1 kristaps 613:
1.50 schwarze 614: sql = mandoc_strdup(e->equal ?
615: "SELECT desc, form, pageid, bits "
616: "FROM mpages NATURAL JOIN names WHERE " :
617: "SELECT desc, form, pageid, 0 FROM mpages WHERE ");
1.1 kristaps 618: sz = strlen(sql);
1.2 kristaps 619:
1.13 schwarze 620: for (needop = 0; NULL != e; e = e->next) {
621: if (e->and)
622: sql_append(&sql, &sz, " AND ", 1);
623: else if (needop)
624: sql_append(&sql, &sz, " OR ", 1);
625: if (e->open)
626: sql_append(&sql, &sz, "(", e->open);
1.26 schwarze 627: sql_append(&sql, &sz,
628: TYPE_Nd & e->bits
629: ? (NULL == e->substr
630: ? "desc REGEXP ?"
631: : "desc MATCH ?")
1.27 schwarze 632: : TYPE_Nm == e->bits
633: ? (NULL == e->substr
1.32 schwarze 634: ? "pageid IN (SELECT pageid FROM names "
1.27 schwarze 635: "WHERE name REGEXP ?)"
1.38 schwarze 636: : e->equal
1.50 schwarze 637: ? "name = ? "
1.32 schwarze 638: : "pageid IN (SELECT pageid FROM names "
1.27 schwarze 639: "WHERE name MATCH ?)")
1.26 schwarze 640: : (NULL == e->substr
1.32 schwarze 641: ? "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 642: "WHERE key REGEXP ? AND bits & ?)"
1.32 schwarze 643: : "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 644: "WHERE key MATCH ? AND bits & ?)"), 1);
1.13 schwarze 645: if (e->close)
646: sql_append(&sql, &sz, ")", e->close);
647: needop = 1;
1.1 kristaps 648: }
649:
1.58 schwarze 650: return sql;
1.1 kristaps 651: }
652:
653: /*
654: * Compile a set of string tokens into an expression.
655: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
656: * "(", "foo=bar", etc.).
657: */
658: static struct expr *
1.5 kristaps 659: exprcomp(const struct mansearch *search, int argc, char *argv[])
1.1 kristaps 660: {
1.27 schwarze 661: uint64_t mask;
1.13 schwarze 662: int i, toopen, logic, igncase, toclose;
1.27 schwarze 663: struct expr *first, *prev, *cur, *next;
1.1 kristaps 664:
665: first = cur = NULL;
1.53 schwarze 666: logic = igncase = toopen = toclose = 0;
1.1 kristaps 667:
668: for (i = 0; i < argc; i++) {
1.13 schwarze 669: if (0 == strcmp("(", argv[i])) {
670: if (igncase)
671: goto fail;
672: toopen++;
673: toclose++;
674: continue;
675: } else if (0 == strcmp(")", argv[i])) {
676: if (toopen || logic || igncase || NULL == cur)
677: goto fail;
678: cur->close++;
679: if (0 > --toclose)
680: goto fail;
681: continue;
682: } else if (0 == strcmp("-a", argv[i])) {
683: if (toopen || logic || igncase || NULL == cur)
684: goto fail;
685: logic = 1;
686: continue;
687: } else if (0 == strcmp("-o", argv[i])) {
688: if (toopen || logic || igncase || NULL == cur)
689: goto fail;
690: logic = 2;
691: continue;
692: } else if (0 == strcmp("-i", argv[i])) {
693: if (igncase)
694: goto fail;
695: igncase = 1;
696: continue;
1.1 kristaps 697: }
1.13 schwarze 698: next = exprterm(search, argv[i], !igncase);
699: if (NULL == next)
700: goto fail;
1.26 schwarze 701: if (NULL == first)
702: first = next;
703: else
1.1 kristaps 704: cur->next = next;
1.27 schwarze 705: prev = cur = next;
1.26 schwarze 706:
707: /*
708: * Searching for descriptions must be split out
709: * because they are stored in the mpages table,
710: * not in the keys table.
711: */
712:
1.27 schwarze 713: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
714: if (mask & cur->bits && ~mask & cur->bits) {
715: next = mandoc_calloc(1,
716: sizeof(struct expr));
717: memcpy(next, cur, sizeof(struct expr));
718: prev->open = 1;
719: cur->bits = mask;
720: cur->next = next;
721: cur = next;
722: cur->bits &= ~mask;
723: }
724: }
725: prev->and = (1 == logic);
726: prev->open += toopen;
727: if (cur != prev)
1.26 schwarze 728: cur->close = 1;
1.27 schwarze 729:
1.13 schwarze 730: toopen = logic = igncase = 0;
1.1 kristaps 731: }
1.53 schwarze 732: if ( ! (toopen || logic || igncase || toclose))
1.58 schwarze 733: return first;
1.15 schwarze 734:
1.13 schwarze 735: fail:
736: if (NULL != first)
737: exprfree(first);
1.58 schwarze 738: return NULL;
1.15 schwarze 739: }
740:
741: static struct expr *
1.8 schwarze 742: exprterm(const struct mansearch *search, char *buf, int cs)
1.1 kristaps 743: {
1.15 schwarze 744: char errbuf[BUFSIZ];
1.1 kristaps 745: struct expr *e;
1.38 schwarze 746: char *key, *val;
1.20 schwarze 747: uint64_t iterbit;
748: int i, irc;
1.1 kristaps 749:
750: if ('\0' == *buf)
1.58 schwarze 751: return NULL;
1.1 kristaps 752:
753: e = mandoc_calloc(1, sizeof(struct expr));
754:
1.45 schwarze 755: if (search->argmode == ARG_NAME) {
756: e->bits = TYPE_Nm;
1.8 schwarze 757: e->substr = buf;
1.38 schwarze 758: e->equal = 1;
1.58 schwarze 759: return e;
1.5 kristaps 760: }
761:
1.1 kristaps 762: /*
1.45 schwarze 763: * Separate macro keys from search string.
764: * If needed, request regular expression handling
765: * by setting e->substr to NULL.
1.1 kristaps 766: */
767:
1.45 schwarze 768: if (search->argmode == ARG_WORD) {
769: e->bits = TYPE_Nm;
770: e->substr = NULL;
1.61 schwarze 771: #if HAVE_REWB_BSD
1.45 schwarze 772: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
1.61 schwarze 773: #elif HAVE_REWB_SYSV
774: mandoc_asprintf(&val, "\\<%s\\>", buf);
775: #else
776: mandoc_asprintf(&val,
777: "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", buf);
778: #endif
1.46 schwarze 779: cs = 0;
1.45 schwarze 780: } else if ((val = strpbrk(buf, "=~")) == NULL) {
781: e->bits = TYPE_Nm | TYPE_Nd;
1.8 schwarze 782: e->substr = buf;
1.38 schwarze 783: } else {
784: if (val == buf)
1.45 schwarze 785: e->bits = TYPE_Nm | TYPE_Nd;
1.38 schwarze 786: if ('=' == *val)
787: e->substr = val + 1;
788: *val++ = '\0';
1.21 schwarze 789: if (NULL != strstr(buf, "arch"))
790: cs = 0;
1.38 schwarze 791: }
792:
793: /* Compile regular expressions. */
794:
795: if (NULL == e->substr) {
796: irc = regcomp(&e->regexp, val,
797: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.45 schwarze 798: if (search->argmode == ARG_WORD)
1.38 schwarze 799: free(val);
800: if (irc) {
1.15 schwarze 801: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
1.59 schwarze 802: warnx("regcomp: %s", errbuf);
1.8 schwarze 803: free(e);
1.58 schwarze 804: return NULL;
1.8 schwarze 805: }
1.38 schwarze 806: }
807:
808: if (e->bits)
1.58 schwarze 809: return e;
1.1 kristaps 810:
811: /*
812: * Parse out all possible fields.
813: * If the field doesn't resolve, bail.
814: */
815:
816: while (NULL != (key = strsep(&buf, ","))) {
817: if ('\0' == *key)
818: continue;
1.20 schwarze 819: for (i = 0, iterbit = 1;
820: i < mansearch_keymax;
821: i++, iterbit <<= 1) {
822: if (0 == strcasecmp(key,
823: mansearch_keynames[i])) {
824: e->bits |= iterbit;
825: break;
826: }
827: }
828: if (i == mansearch_keymax) {
829: if (strcasecmp(key, "any")) {
830: free(e);
1.58 schwarze 831: return NULL;
1.20 schwarze 832: }
833: e->bits |= ~0ULL;
1.1 kristaps 834: }
835: }
836:
1.58 schwarze 837: return e;
1.1 kristaps 838: }
839:
840: static void
841: exprfree(struct expr *p)
842: {
843: struct expr *pp;
844:
845: while (NULL != p) {
846: pp = p->next;
847: free(p);
848: p = pp;
849: }
850: }
CVSweb