Annotation of mandoc/mansearch.c, Revision 1.64
1.64 ! schwarze 1: /* $Id: mansearch.c,v 1.63 2015/11/26 07:42:11 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)) \
1.63 schwarze 57: errx((int)MANDOCLEVEL_SYSERR, "%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))) \
1.63 schwarze 62: errx((int)MANDOCLEVEL_SYSERR, "%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)) \
1.63 schwarze 67: errx((int)MANDOCLEVEL_SYSERR, "%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) {
1.64 ! schwarze 128: warn("mmap");
1.28 schwarze 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)) {
1.64 ! schwarze 147: warn("munmap");
1.28 schwarze 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) {
1.64 ! schwarze 229: warn("%s", buf);
1.57 schwarze 230: continue;
231: }
232: }
233: if (chdir(paths->paths[i]) == -1) {
1.64 ! schwarze 234: warn("%s", paths->paths[i]);
1.1 kristaps 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)
1.63 schwarze 265: errx((int)MANDOCLEVEL_SYSERR,
266: "%s", sqlite3_errmsg(db));
1.1 kristaps 267:
268: for (ep = e; NULL != ep; ep = ep->next) {
1.8 schwarze 269: if (NULL == ep->substr) {
270: SQL_BIND_BLOB(db, s, j, ep->regexp);
271: } else
272: SQL_BIND_TEXT(db, s, j, ep->substr);
1.27 schwarze 273: if (0 == ((TYPE_Nd | TYPE_Nm) & ep->bits))
1.26 schwarze 274: SQL_BIND_INT64(db, s, j, ep->bits);
1.1 kristaps 275: }
276:
1.60 schwarze 277: mandoc_ohash_init(&htab, 4, offsetof(struct match, pageid));
1.1 kristaps 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: */
1.2 kristaps 287: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.32 schwarze 288: pageid = sqlite3_column_int64(s, 2);
1.34 schwarze 289: idx = ohash_lookup_memory(&htab,
290: (char *)&pageid, sizeof(uint64_t),
291: (uint32_t)pageid);
1.1 kristaps 292:
293: if (NULL != ohash_find(&htab, idx))
294: continue;
295:
296: mp = mandoc_calloc(1, sizeof(struct match));
1.32 schwarze 297: mp->pageid = pageid;
1.26 schwarze 298: mp->form = sqlite3_column_int(s, 1);
1.50 schwarze 299: mp->bits = sqlite3_column_int64(s, 3);
1.26 schwarze 300: if (TYPE_Nd == outbit)
1.41 schwarze 301: mp->desc = mandoc_strdup((const char *)
1.26 schwarze 302: sqlite3_column_text(s, 0));
1.1 kristaps 303: ohash_insert(&htab, idx, mp);
304: }
305:
1.2 kristaps 306: if (SQLITE_DONE != c)
1.59 schwarze 307: warnx("%s", sqlite3_errmsg(db));
1.2 kristaps 308:
1.1 kristaps 309: sqlite3_finalize(s);
1.10 schwarze 310:
1.34 schwarze 311: c = sqlite3_prepare_v2(db,
1.35 schwarze 312: "SELECT sec, arch, name, pageid FROM mlinks "
313: "WHERE pageid=? ORDER BY sec, arch, name",
1.10 schwarze 314: -1, &s, NULL);
315: if (SQLITE_OK != c)
1.63 schwarze 316: errx((int)MANDOCLEVEL_SYSERR,
317: "%s", sqlite3_errmsg(db));
1.1 kristaps 318:
1.12 schwarze 319: c = sqlite3_prepare_v2(db,
1.35 schwarze 320: "SELECT bits, key, pageid FROM keys "
321: "WHERE pageid=? AND bits & ?",
1.12 schwarze 322: -1, &s2, NULL);
323: if (SQLITE_OK != c)
1.63 schwarze 324: errx((int)MANDOCLEVEL_SYSERR,
325: "%s", sqlite3_errmsg(db));
1.12 schwarze 326:
1.1 kristaps 327: for (mp = ohash_first(&htab, &idx);
328: NULL != mp;
329: mp = ohash_next(&htab, &idx)) {
330: if (cur + 1 > maxres) {
331: maxres += 1024;
1.36 schwarze 332: *res = mandoc_reallocarray(*res,
333: maxres, sizeof(struct manpage));
1.1 kristaps 334: }
1.10 schwarze 335: mpage = *res + cur;
1.47 schwarze 336: mpage->ipath = i;
1.50 schwarze 337: mpage->bits = mp->bits;
1.39 schwarze 338: mpage->sec = 10;
1.10 schwarze 339: mpage->form = mp->form;
1.53 schwarze 340: buildnames(search, mpage, db, s, mp->pageid,
1.19 schwarze 341: paths->paths[i], mp->form);
1.53 schwarze 342: if (mpage->names != NULL) {
343: mpage->output = TYPE_Nd & outbit ?
344: mp->desc : outbit ?
345: buildoutput(db, s2, mp->pageid, outbit) :
346: NULL;
347: cur++;
348: }
1.1 kristaps 349: free(mp);
350: }
1.10 schwarze 351:
352: sqlite3_finalize(s);
1.12 schwarze 353: sqlite3_finalize(s2);
1.10 schwarze 354: sqlite3_close(db);
1.1 kristaps 355: ohash_delete(&htab);
1.49 schwarze 356:
357: /*
358: * In man(1) mode, prefer matches in earlier trees
359: * over matches in later trees.
360: */
361:
362: if (cur && search->firstmatch)
363: break;
1.1 kristaps 364: }
1.39 schwarze 365: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.57 schwarze 366: if (chdir_status && getcwd_status && chdir(buf) == -1)
1.64 ! schwarze 367: warn("%s", buf);
1.1 kristaps 368: exprfree(e);
369: free(sql);
370: *sz = cur;
1.58 schwarze 371: return 1;
1.11 schwarze 372: }
373:
1.45 schwarze 374: void
375: mansearch_free(struct manpage *res, size_t sz)
376: {
377: size_t i;
378:
379: for (i = 0; i < sz; i++) {
380: free(res[i].file);
381: free(res[i].names);
382: free(res[i].output);
383: }
384: free(res);
385: }
386:
1.39 schwarze 387: static int
388: manpage_compare(const void *vp1, const void *vp2)
389: {
390: const struct manpage *mp1, *mp2;
391: int diff;
392:
393: mp1 = vp1;
394: mp2 = vp2;
1.58 schwarze 395: return (diff = mp2->bits - mp1->bits) ? diff :
396: (diff = mp1->sec - mp2->sec) ? diff :
397: strcasecmp(mp1->names, mp2->names);
1.39 schwarze 398: }
399:
1.17 schwarze 400: static void
1.53 schwarze 401: buildnames(const struct mansearch *search, struct manpage *mpage,
402: sqlite3 *db, sqlite3_stmt *s,
1.32 schwarze 403: uint64_t pageid, const char *path, int form)
1.11 schwarze 404: {
1.54 schwarze 405: glob_t globinfo;
406: char *firstname, *newnames, *prevsec, *prevarch;
1.19 schwarze 407: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.11 schwarze 408: size_t i;
1.54 schwarze 409: int c, globres;
1.11 schwarze 410:
1.25 schwarze 411: mpage->file = NULL;
1.17 schwarze 412: mpage->names = NULL;
1.54 schwarze 413: firstname = prevsec = prevarch = NULL;
1.11 schwarze 414: i = 1;
1.32 schwarze 415: SQL_BIND_INT64(db, s, i, pageid);
1.11 schwarze 416: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.17 schwarze 417:
1.22 schwarze 418: /* Decide whether we already have some names. */
1.17 schwarze 419:
420: if (NULL == mpage->names) {
1.11 schwarze 421: oldnames = "";
422: sep1 = "";
423: } else {
1.17 schwarze 424: oldnames = mpage->names;
1.11 schwarze 425: sep1 = ", ";
426: }
1.22 schwarze 427:
1.53 schwarze 428: /* Fetch the next name, rejecting sec/arch mismatches. */
1.22 schwarze 429:
1.41 schwarze 430: sec = (const char *)sqlite3_column_text(s, 0);
1.53 schwarze 431: if (search->sec != NULL && strcasecmp(sec, search->sec))
432: continue;
1.41 schwarze 433: arch = (const char *)sqlite3_column_text(s, 1);
1.53 schwarze 434: if (search->arch != NULL && *arch != '\0' &&
435: strcasecmp(arch, search->arch))
436: continue;
1.41 schwarze 437: name = (const char *)sqlite3_column_text(s, 2);
1.39 schwarze 438:
439: /* Remember the first section found. */
440:
441: if (9 < mpage->sec && '1' <= *sec && '9' >= *sec)
442: mpage->sec = (*sec - '1') + 1;
1.22 schwarze 443:
444: /* If the section changed, append the old one. */
445:
446: if (NULL != prevsec &&
447: (strcmp(sec, prevsec) ||
448: strcmp(arch, prevarch))) {
449: sep2 = '\0' == *prevarch ? "" : "/";
1.24 schwarze 450: mandoc_asprintf(&newnames, "%s(%s%s%s)",
451: oldnames, prevsec, sep2, prevarch);
1.22 schwarze 452: free(mpage->names);
453: oldnames = mpage->names = newnames;
454: free(prevsec);
455: free(prevarch);
456: prevsec = prevarch = NULL;
457: }
458:
459: /* Save the new section, to append it later. */
460:
461: if (NULL == prevsec) {
462: prevsec = mandoc_strdup(sec);
463: prevarch = mandoc_strdup(arch);
464: }
465:
466: /* Append the new name. */
467:
1.24 schwarze 468: mandoc_asprintf(&newnames, "%s%s%s",
469: oldnames, sep1, name);
1.17 schwarze 470: free(mpage->names);
471: mpage->names = newnames;
472:
473: /* Also save the first file name encountered. */
474:
1.51 schwarze 475: if (mpage->file != NULL)
1.17 schwarze 476: continue;
477:
1.48 schwarze 478: if (form & FORM_SRC) {
1.19 schwarze 479: sep1 = "man";
480: fsec = sec;
481: } else {
482: sep1 = "cat";
483: fsec = "0";
484: }
1.51 schwarze 485: sep2 = *arch == '\0' ? "" : "/";
486: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
487: path, sep1, sec, sep2, arch, name, fsec);
1.54 schwarze 488: if (access(mpage->file, R_OK) != -1)
489: continue;
490:
491: /* Handle unusual file name extensions. */
492:
493: if (firstname == NULL)
494: firstname = mpage->file;
495: else
496: free(mpage->file);
497: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.*",
498: path, sep1, sec, sep2, arch, name);
499: globres = glob(mpage->file, 0, NULL, &globinfo);
500: free(mpage->file);
501: mpage->file = globres ? NULL :
502: mandoc_strdup(*globinfo.gl_pathv);
503: globfree(&globinfo);
1.11 schwarze 504: }
1.51 schwarze 505: if (c != SQLITE_DONE)
1.59 schwarze 506: warnx("%s", sqlite3_errmsg(db));
1.11 schwarze 507: sqlite3_reset(s);
1.54 schwarze 508:
509: /* If none of the files is usable, use the first name. */
510:
511: if (mpage->file == NULL)
512: mpage->file = firstname;
513: else if (mpage->file != firstname)
514: free(firstname);
1.22 schwarze 515:
516: /* Append one final section to the names. */
517:
1.51 schwarze 518: if (prevsec != NULL) {
519: sep2 = *prevarch == '\0' ? "" : "/";
1.24 schwarze 520: mandoc_asprintf(&newnames, "%s(%s%s%s)",
521: mpage->names, prevsec, sep2, prevarch);
1.22 schwarze 522: free(mpage->names);
523: mpage->names = newnames;
524: free(prevsec);
525: free(prevarch);
526: }
1.12 schwarze 527: }
528:
529: static char *
1.32 schwarze 530: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.12 schwarze 531: {
532: char *output, *newoutput;
533: const char *oldoutput, *sep1, *data;
534: size_t i;
535: int c;
536:
537: output = NULL;
538: i = 1;
1.32 schwarze 539: SQL_BIND_INT64(db, s, i, pageid);
1.12 schwarze 540: SQL_BIND_INT64(db, s, i, outbit);
541: while (SQLITE_ROW == (c = sqlite3_step(s))) {
542: if (NULL == output) {
543: oldoutput = "";
544: sep1 = "";
545: } else {
546: oldoutput = output;
547: sep1 = " # ";
548: }
1.41 schwarze 549: data = (const char *)sqlite3_column_text(s, 1);
1.24 schwarze 550: mandoc_asprintf(&newoutput, "%s%s%s",
551: oldoutput, sep1, data);
1.12 schwarze 552: free(output);
553: output = newoutput;
554: }
555: if (SQLITE_DONE != c)
1.59 schwarze 556: warnx("%s", sqlite3_errmsg(db));
1.12 schwarze 557: sqlite3_reset(s);
1.58 schwarze 558: return output;
1.1 kristaps 559: }
560:
561: /*
1.7 schwarze 562: * Implement substring match as an application-defined SQL function.
563: * Using the SQL LIKE or GLOB operators instead would be a bad idea
564: * because that would require escaping metacharacters in the string
565: * being searched for.
566: */
567: static void
568: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
569: {
570:
571: assert(2 == argc);
572: sqlite3_result_int(context, NULL != strcasestr(
573: (const char *)sqlite3_value_text(argv[1]),
574: (const char *)sqlite3_value_text(argv[0])));
575: }
576:
577: /*
1.8 schwarze 578: * Implement regular expression match
579: * as an application-defined SQL function.
580: */
581: static void
582: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
583: {
584:
585: assert(2 == argc);
586: sqlite3_result_int(context, !regexec(
587: (regex_t *)sqlite3_value_blob(argv[0]),
588: (const char *)sqlite3_value_text(argv[1]),
589: 0, NULL, 0));
590: }
591:
1.13 schwarze 592: static void
593: sql_append(char **sql, size_t *sz, const char *newstr, int count)
594: {
595: size_t newsz;
596:
597: newsz = 1 < count ? (size_t)count : strlen(newstr);
598: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
599: if (1 < count)
600: memset(*sql + *sz, *newstr, (size_t)count);
601: else
602: memcpy(*sql + *sz, newstr, newsz);
603: *sz += newsz;
604: (*sql)[*sz] = '\0';
605: }
606:
1.8 schwarze 607: /*
1.1 kristaps 608: * Prepare the search SQL statement.
609: */
610: static char *
1.15 schwarze 611: sql_statement(const struct expr *e)
1.1 kristaps 612: {
613: char *sql;
614: size_t sz;
1.13 schwarze 615: int needop;
1.1 kristaps 616:
1.50 schwarze 617: sql = mandoc_strdup(e->equal ?
618: "SELECT desc, form, pageid, bits "
619: "FROM mpages NATURAL JOIN names WHERE " :
620: "SELECT desc, form, pageid, 0 FROM mpages WHERE ");
1.1 kristaps 621: sz = strlen(sql);
1.2 kristaps 622:
1.13 schwarze 623: for (needop = 0; NULL != e; e = e->next) {
624: if (e->and)
625: sql_append(&sql, &sz, " AND ", 1);
626: else if (needop)
627: sql_append(&sql, &sz, " OR ", 1);
628: if (e->open)
629: sql_append(&sql, &sz, "(", e->open);
1.26 schwarze 630: sql_append(&sql, &sz,
631: TYPE_Nd & e->bits
632: ? (NULL == e->substr
633: ? "desc REGEXP ?"
634: : "desc MATCH ?")
1.27 schwarze 635: : TYPE_Nm == e->bits
636: ? (NULL == e->substr
1.32 schwarze 637: ? "pageid IN (SELECT pageid FROM names "
1.27 schwarze 638: "WHERE name REGEXP ?)"
1.38 schwarze 639: : e->equal
1.50 schwarze 640: ? "name = ? "
1.32 schwarze 641: : "pageid IN (SELECT pageid FROM names "
1.27 schwarze 642: "WHERE name MATCH ?)")
1.26 schwarze 643: : (NULL == e->substr
1.32 schwarze 644: ? "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 645: "WHERE key REGEXP ? AND bits & ?)"
1.32 schwarze 646: : "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 647: "WHERE key MATCH ? AND bits & ?)"), 1);
1.13 schwarze 648: if (e->close)
649: sql_append(&sql, &sz, ")", e->close);
650: needop = 1;
1.1 kristaps 651: }
652:
1.58 schwarze 653: return sql;
1.1 kristaps 654: }
655:
656: /*
657: * Compile a set of string tokens into an expression.
658: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
659: * "(", "foo=bar", etc.).
660: */
661: static struct expr *
1.5 kristaps 662: exprcomp(const struct mansearch *search, int argc, char *argv[])
1.1 kristaps 663: {
1.27 schwarze 664: uint64_t mask;
1.13 schwarze 665: int i, toopen, logic, igncase, toclose;
1.27 schwarze 666: struct expr *first, *prev, *cur, *next;
1.1 kristaps 667:
668: first = cur = NULL;
1.53 schwarze 669: logic = igncase = toopen = toclose = 0;
1.1 kristaps 670:
671: for (i = 0; i < argc; i++) {
1.13 schwarze 672: if (0 == strcmp("(", argv[i])) {
673: if (igncase)
674: goto fail;
675: toopen++;
676: toclose++;
677: continue;
678: } else if (0 == strcmp(")", argv[i])) {
679: if (toopen || logic || igncase || NULL == cur)
680: goto fail;
681: cur->close++;
682: if (0 > --toclose)
683: goto fail;
684: continue;
685: } else if (0 == strcmp("-a", argv[i])) {
686: if (toopen || logic || igncase || NULL == cur)
687: goto fail;
688: logic = 1;
689: continue;
690: } else if (0 == strcmp("-o", argv[i])) {
691: if (toopen || logic || igncase || NULL == cur)
692: goto fail;
693: logic = 2;
694: continue;
695: } else if (0 == strcmp("-i", argv[i])) {
696: if (igncase)
697: goto fail;
698: igncase = 1;
699: continue;
1.1 kristaps 700: }
1.13 schwarze 701: next = exprterm(search, argv[i], !igncase);
702: if (NULL == next)
703: goto fail;
1.26 schwarze 704: if (NULL == first)
705: first = next;
706: else
1.1 kristaps 707: cur->next = next;
1.27 schwarze 708: prev = cur = next;
1.26 schwarze 709:
710: /*
711: * Searching for descriptions must be split out
712: * because they are stored in the mpages table,
713: * not in the keys table.
714: */
715:
1.27 schwarze 716: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
717: if (mask & cur->bits && ~mask & cur->bits) {
718: next = mandoc_calloc(1,
719: sizeof(struct expr));
720: memcpy(next, cur, sizeof(struct expr));
721: prev->open = 1;
722: cur->bits = mask;
723: cur->next = next;
724: cur = next;
725: cur->bits &= ~mask;
726: }
727: }
728: prev->and = (1 == logic);
729: prev->open += toopen;
730: if (cur != prev)
1.26 schwarze 731: cur->close = 1;
1.27 schwarze 732:
1.13 schwarze 733: toopen = logic = igncase = 0;
1.1 kristaps 734: }
1.53 schwarze 735: if ( ! (toopen || logic || igncase || toclose))
1.58 schwarze 736: return first;
1.15 schwarze 737:
1.13 schwarze 738: fail:
739: if (NULL != first)
740: exprfree(first);
1.58 schwarze 741: return NULL;
1.15 schwarze 742: }
743:
744: static struct expr *
1.8 schwarze 745: exprterm(const struct mansearch *search, char *buf, int cs)
1.1 kristaps 746: {
1.15 schwarze 747: char errbuf[BUFSIZ];
1.1 kristaps 748: struct expr *e;
1.38 schwarze 749: char *key, *val;
1.20 schwarze 750: uint64_t iterbit;
751: int i, irc;
1.1 kristaps 752:
753: if ('\0' == *buf)
1.58 schwarze 754: return NULL;
1.1 kristaps 755:
756: e = mandoc_calloc(1, sizeof(struct expr));
757:
1.45 schwarze 758: if (search->argmode == ARG_NAME) {
759: e->bits = TYPE_Nm;
1.8 schwarze 760: e->substr = buf;
1.38 schwarze 761: e->equal = 1;
1.58 schwarze 762: return e;
1.5 kristaps 763: }
764:
1.1 kristaps 765: /*
1.45 schwarze 766: * Separate macro keys from search string.
767: * If needed, request regular expression handling
768: * by setting e->substr to NULL.
1.1 kristaps 769: */
770:
1.45 schwarze 771: if (search->argmode == ARG_WORD) {
772: e->bits = TYPE_Nm;
773: e->substr = NULL;
1.61 schwarze 774: #if HAVE_REWB_BSD
1.45 schwarze 775: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
1.61 schwarze 776: #elif HAVE_REWB_SYSV
777: mandoc_asprintf(&val, "\\<%s\\>", buf);
778: #else
779: mandoc_asprintf(&val,
780: "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", buf);
781: #endif
1.46 schwarze 782: cs = 0;
1.45 schwarze 783: } else if ((val = strpbrk(buf, "=~")) == NULL) {
784: e->bits = TYPE_Nm | TYPE_Nd;
1.8 schwarze 785: e->substr = buf;
1.38 schwarze 786: } else {
787: if (val == buf)
1.45 schwarze 788: e->bits = TYPE_Nm | TYPE_Nd;
1.38 schwarze 789: if ('=' == *val)
790: e->substr = val + 1;
791: *val++ = '\0';
1.21 schwarze 792: if (NULL != strstr(buf, "arch"))
793: cs = 0;
1.38 schwarze 794: }
795:
796: /* Compile regular expressions. */
797:
798: if (NULL == e->substr) {
799: irc = regcomp(&e->regexp, val,
800: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.45 schwarze 801: if (search->argmode == ARG_WORD)
1.38 schwarze 802: free(val);
803: if (irc) {
1.15 schwarze 804: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
1.59 schwarze 805: warnx("regcomp: %s", errbuf);
1.8 schwarze 806: free(e);
1.58 schwarze 807: return NULL;
1.8 schwarze 808: }
1.38 schwarze 809: }
810:
811: if (e->bits)
1.58 schwarze 812: return e;
1.1 kristaps 813:
814: /*
815: * Parse out all possible fields.
816: * If the field doesn't resolve, bail.
817: */
818:
819: while (NULL != (key = strsep(&buf, ","))) {
820: if ('\0' == *key)
821: continue;
1.20 schwarze 822: for (i = 0, iterbit = 1;
823: i < mansearch_keymax;
824: i++, iterbit <<= 1) {
825: if (0 == strcasecmp(key,
826: mansearch_keynames[i])) {
827: e->bits |= iterbit;
828: break;
829: }
830: }
831: if (i == mansearch_keymax) {
832: if (strcasecmp(key, "any")) {
833: free(e);
1.58 schwarze 834: return NULL;
1.20 schwarze 835: }
836: e->bits |= ~0ULL;
1.1 kristaps 837: }
838: }
839:
1.58 schwarze 840: return e;
1.1 kristaps 841: }
842:
843: static void
844: exprfree(struct expr *p)
845: {
846: struct expr *pp;
847:
848: while (NULL != p) {
849: pp = p->next;
850: free(p);
851: p = pp;
852: }
853: }
CVSweb