Annotation of mandoc/mansearch.c, Revision 1.59
1.59 ! schwarze 1: /* $Id: mansearch.c,v 1.58 2015/10/06 18:32:19 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>
1.59 ! schwarze 24: #include <err.h>
1.52 schwarze 25: #include <errno.h>
1.1 kristaps 26: #include <fcntl.h>
27: #include <getopt.h>
1.54 schwarze 28: #include <glob.h>
1.6 schwarze 29: #include <limits.h>
1.8 schwarze 30: #include <regex.h>
1.1 kristaps 31: #include <stdio.h>
32: #include <stdint.h>
33: #include <stddef.h>
34: #include <stdlib.h>
35: #include <string.h>
36: #include <unistd.h>
37:
1.44 schwarze 38: #if HAVE_OHASH
1.1 kristaps 39: #include <ohash.h>
1.4 kristaps 40: #else
41: #include "compat_ohash.h"
42: #endif
1.1 kristaps 43: #include <sqlite3.h>
1.40 schwarze 44: #ifndef SQLITE_DETERMINISTIC
45: #define SQLITE_DETERMINISTIC 0
46: #endif
1.1 kristaps 47:
48: #include "mandoc.h"
1.23 schwarze 49: #include "mandoc_aux.h"
1.56 schwarze 50: #include "manconf.h"
1.1 kristaps 51: #include "mansearch.h"
52:
1.20 schwarze 53: extern int mansearch_keymax;
54: extern const char *const mansearch_keynames[];
55:
1.3 kristaps 56: #define SQL_BIND_TEXT(_db, _s, _i, _v) \
1.8 schwarze 57: do { if (SQLITE_OK != sqlite3_bind_text \
1.2 kristaps 58: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.59 ! schwarze 59: warnx("%s", sqlite3_errmsg((_db))); \
1.8 schwarze 60: } while (0)
1.3 kristaps 61: #define SQL_BIND_INT64(_db, _s, _i, _v) \
1.8 schwarze 62: do { if (SQLITE_OK != sqlite3_bind_int64 \
1.2 kristaps 63: ((_s), (_i)++, (_v))) \
1.59 ! schwarze 64: warnx("%s", sqlite3_errmsg((_db))); \
1.8 schwarze 65: } while (0)
66: #define SQL_BIND_BLOB(_db, _s, _i, _v) \
67: do { if (SQLITE_OK != sqlite3_bind_blob \
68: ((_s), (_i)++, (&_v), sizeof(_v), SQLITE_STATIC)) \
1.59 ! schwarze 69: warnx("%s", sqlite3_errmsg((_db))); \
1.8 schwarze 70: } while (0)
1.2 kristaps 71:
1.1 kristaps 72: struct expr {
1.38 schwarze 73: regex_t regexp; /* compiled regexp, if applicable */
74: const char *substr; /* to search for, if applicable */
75: struct expr *next; /* next in sequence */
1.34 schwarze 76: uint64_t bits; /* type-mask */
1.38 schwarze 77: int equal; /* equality, not subsring match */
1.13 schwarze 78: int open; /* opening parentheses before */
79: int and; /* logical AND before */
80: int close; /* closing parentheses after */
1.1 kristaps 81: };
82:
83: struct match {
1.32 schwarze 84: uint64_t pageid; /* identifier in database */
1.50 schwarze 85: uint64_t bits; /* name type mask */
1.26 schwarze 86: char *desc; /* manual page description */
1.48 schwarze 87: int form; /* bit field: formatted, zipped? */
1.1 kristaps 88: };
89:
1.53 schwarze 90: static void buildnames(const struct mansearch *,
91: struct manpage *, sqlite3 *,
1.19 schwarze 92: sqlite3_stmt *, uint64_t,
93: const char *, int form);
1.12 schwarze 94: static char *buildoutput(sqlite3 *, sqlite3_stmt *,
95: uint64_t, uint64_t);
1.1 kristaps 96: static void *hash_alloc(size_t, void *);
1.37 schwarze 97: static void hash_free(void *, void *);
98: static void *hash_calloc(size_t, size_t, void *);
1.34 schwarze 99: static struct expr *exprcomp(const struct mansearch *,
1.5 kristaps 100: int, char *[]);
1.1 kristaps 101: static void exprfree(struct expr *);
1.8 schwarze 102: static struct expr *exprterm(const struct mansearch *, char *, int);
1.39 schwarze 103: static int manpage_compare(const void *, const void *);
1.13 schwarze 104: static void sql_append(char **sql, size_t *sz,
105: const char *newstr, int count);
1.7 schwarze 106: static void sql_match(sqlite3_context *context,
107: int argc, sqlite3_value **argv);
1.8 schwarze 108: static void sql_regexp(sqlite3_context *context,
109: int argc, sqlite3_value **argv);
1.15 schwarze 110: static char *sql_statement(const struct expr *);
1.28 schwarze 111:
1.34 schwarze 112:
1.28 schwarze 113: int
114: mansearch_setup(int start)
115: {
116: static void *pagecache;
117: int c;
118:
119: #define PC_PAGESIZE 1280
120: #define PC_NUMPAGES 256
121:
122: if (start) {
123: if (NULL != pagecache) {
1.59 ! schwarze 124: warnx("pagecache already enabled");
1.58 schwarze 125: return (int)MANDOCLEVEL_BADARG;
1.28 schwarze 126: }
127:
128: pagecache = mmap(NULL, PC_PAGESIZE * PC_NUMPAGES,
1.42 schwarze 129: PROT_READ | PROT_WRITE,
130: MAP_SHARED | MAP_ANON, -1, 0);
1.28 schwarze 131:
132: if (MAP_FAILED == pagecache) {
133: perror("mmap");
134: pagecache = NULL;
1.58 schwarze 135: return (int)MANDOCLEVEL_SYSERR;
1.28 schwarze 136: }
137:
138: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
139: pagecache, PC_PAGESIZE, PC_NUMPAGES);
140:
141: if (SQLITE_OK == c)
1.58 schwarze 142: return (int)MANDOCLEVEL_OK;
1.28 schwarze 143:
1.59 ! schwarze 144: warnx("pagecache: %s", sqlite3_errstr(c));
1.28 schwarze 145:
146: } else if (NULL == pagecache) {
1.59 ! schwarze 147: warnx("pagecache missing");
1.58 schwarze 148: return (int)MANDOCLEVEL_BADARG;
1.28 schwarze 149: }
150:
151: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
152: perror("munmap");
153: pagecache = NULL;
1.58 schwarze 154: return (int)MANDOCLEVEL_SYSERR;
1.28 schwarze 155: }
156:
157: pagecache = NULL;
1.58 schwarze 158: return (int)MANDOCLEVEL_OK;
1.28 schwarze 159: }
1.1 kristaps 160:
161: int
1.5 kristaps 162: mansearch(const struct mansearch *search,
1.12 schwarze 163: const struct manpaths *paths,
164: int argc, char *argv[],
1.1 kristaps 165: struct manpage **res, size_t *sz)
166: {
1.32 schwarze 167: int64_t pageid;
1.20 schwarze 168: uint64_t outbit, iterbit;
1.6 schwarze 169: char buf[PATH_MAX];
1.11 schwarze 170: char *sql;
1.10 schwarze 171: struct manpage *mpage;
1.1 kristaps 172: struct expr *e, *ep;
173: sqlite3 *db;
1.12 schwarze 174: sqlite3_stmt *s, *s2;
1.1 kristaps 175: struct match *mp;
176: struct ohash_info info;
177: struct ohash htab;
178: unsigned int idx;
179: size_t i, j, cur, maxres;
1.57 schwarze 180: int c, chdir_status, getcwd_status, indexbit;
181:
182: if (argc == 0 || (e = exprcomp(search, argc, argv)) == NULL) {
183: *sz = 0;
1.58 schwarze 184: return 0;
1.57 schwarze 185: }
1.1 kristaps 186:
1.37 schwarze 187: info.calloc = hash_calloc;
1.1 kristaps 188: info.alloc = hash_alloc;
1.37 schwarze 189: info.free = hash_free;
1.32 schwarze 190: info.key_offset = offsetof(struct match, pageid);
1.1 kristaps 191:
1.57 schwarze 192: cur = maxres = 0;
1.1 kristaps 193: *res = NULL;
194:
1.45 schwarze 195: if (NULL != search->outkey) {
1.55 schwarze 196: outbit = TYPE_Nd;
1.20 schwarze 197: for (indexbit = 0, iterbit = 1;
198: indexbit < mansearch_keymax;
199: indexbit++, iterbit <<= 1) {
1.45 schwarze 200: if (0 == strcasecmp(search->outkey,
1.20 schwarze 201: mansearch_keynames[indexbit])) {
202: outbit = iterbit;
1.12 schwarze 203: break;
204: }
205: }
1.55 schwarze 206: } else
207: outbit = 0;
1.12 schwarze 208:
1.1 kristaps 209: /*
1.57 schwarze 210: * Remember the original working directory, if possible.
211: * This will be needed if the second or a later directory
212: * is given as a relative path.
213: * Do not error out if the current directory is not
214: * searchable: Maybe it won't be needed after all.
1.1 kristaps 215: */
216:
1.57 schwarze 217: if (getcwd(buf, PATH_MAX) == NULL) {
218: getcwd_status = 0;
219: (void)strlcpy(buf, strerror(errno), sizeof(buf));
220: } else
221: getcwd_status = 1;
1.1 kristaps 222:
1.15 schwarze 223: sql = sql_statement(e);
1.1 kristaps 224:
225: /*
226: * Loop over the directories (containing databases) for us to
227: * search.
228: * Don't let missing/bad databases/directories phase us.
229: * In each, try to open the resident database and, if it opens,
230: * scan it for our match expression.
231: */
232:
1.57 schwarze 233: chdir_status = 0;
1.1 kristaps 234: for (i = 0; i < paths->sz; i++) {
1.57 schwarze 235: if (chdir_status && paths->paths[i][0] != '/') {
236: if ( ! getcwd_status) {
1.59 ! schwarze 237: warnx("%s: getcwd: %s", paths->paths[i], buf);
1.57 schwarze 238: continue;
239: } else if (chdir(buf) == -1) {
240: perror(buf);
241: continue;
242: }
243: }
244: if (chdir(paths->paths[i]) == -1) {
1.1 kristaps 245: perror(paths->paths[i]);
246: continue;
1.34 schwarze 247: }
1.57 schwarze 248: chdir_status = 1;
1.1 kristaps 249:
1.34 schwarze 250: c = sqlite3_open_v2(MANDOC_DB, &db,
251: SQLITE_OPEN_READONLY, NULL);
1.1 kristaps 252:
1.2 kristaps 253: if (SQLITE_OK != c) {
1.59 ! schwarze 254: warn("%s/%s", paths->paths[i], MANDOC_DB);
1.1 kristaps 255: sqlite3_close(db);
256: continue;
257: }
258:
1.8 schwarze 259: /*
260: * Define the SQL functions for substring
261: * and regular expression matching.
262: */
1.7 schwarze 263:
264: c = sqlite3_create_function(db, "match", 2,
1.31 schwarze 265: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
266: NULL, sql_match, NULL, NULL);
1.8 schwarze 267: assert(SQLITE_OK == c);
268: c = sqlite3_create_function(db, "regexp", 2,
1.31 schwarze 269: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
270: NULL, sql_regexp, NULL, NULL);
1.8 schwarze 271: assert(SQLITE_OK == c);
1.7 schwarze 272:
1.1 kristaps 273: j = 1;
1.2 kristaps 274: c = sqlite3_prepare_v2(db, sql, -1, &s, NULL);
275: if (SQLITE_OK != c)
1.59 ! schwarze 276: warnx("%s", sqlite3_errmsg(db));
1.1 kristaps 277:
278: for (ep = e; NULL != ep; ep = ep->next) {
1.8 schwarze 279: if (NULL == ep->substr) {
280: SQL_BIND_BLOB(db, s, j, ep->regexp);
281: } else
282: SQL_BIND_TEXT(db, s, j, ep->substr);
1.27 schwarze 283: if (0 == ((TYPE_Nd | TYPE_Nm) & ep->bits))
1.26 schwarze 284: SQL_BIND_INT64(db, s, j, ep->bits);
1.1 kristaps 285: }
286:
287: memset(&htab, 0, sizeof(struct ohash));
288: ohash_init(&htab, 4, &info);
289:
290: /*
291: * Hash each entry on its [unique] document identifier.
292: * This is a uint64_t.
293: * Instead of using a hash function, simply convert the
294: * uint64_t to a uint32_t, the hash value's type.
295: * This gives good performance and preserves the
296: * distribution of buckets in the table.
297: */
1.2 kristaps 298: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.32 schwarze 299: pageid = sqlite3_column_int64(s, 2);
1.34 schwarze 300: idx = ohash_lookup_memory(&htab,
301: (char *)&pageid, sizeof(uint64_t),
302: (uint32_t)pageid);
1.1 kristaps 303:
304: if (NULL != ohash_find(&htab, idx))
305: continue;
306:
307: mp = mandoc_calloc(1, sizeof(struct match));
1.32 schwarze 308: mp->pageid = pageid;
1.26 schwarze 309: mp->form = sqlite3_column_int(s, 1);
1.50 schwarze 310: mp->bits = sqlite3_column_int64(s, 3);
1.26 schwarze 311: if (TYPE_Nd == outbit)
1.41 schwarze 312: mp->desc = mandoc_strdup((const char *)
1.26 schwarze 313: sqlite3_column_text(s, 0));
1.1 kristaps 314: ohash_insert(&htab, idx, mp);
315: }
316:
1.2 kristaps 317: if (SQLITE_DONE != c)
1.59 ! schwarze 318: warnx("%s", sqlite3_errmsg(db));
1.2 kristaps 319:
1.1 kristaps 320: sqlite3_finalize(s);
1.10 schwarze 321:
1.34 schwarze 322: c = sqlite3_prepare_v2(db,
1.35 schwarze 323: "SELECT sec, arch, name, pageid FROM mlinks "
324: "WHERE pageid=? ORDER BY sec, arch, name",
1.10 schwarze 325: -1, &s, NULL);
326: if (SQLITE_OK != c)
1.59 ! schwarze 327: warnx("%s", sqlite3_errmsg(db));
1.1 kristaps 328:
1.12 schwarze 329: c = sqlite3_prepare_v2(db,
1.35 schwarze 330: "SELECT bits, key, pageid FROM keys "
331: "WHERE pageid=? AND bits & ?",
1.12 schwarze 332: -1, &s2, NULL);
333: if (SQLITE_OK != c)
1.59 ! schwarze 334: warnx("%s", sqlite3_errmsg(db));
1.12 schwarze 335:
1.1 kristaps 336: for (mp = ohash_first(&htab, &idx);
337: NULL != mp;
338: mp = ohash_next(&htab, &idx)) {
339: if (cur + 1 > maxres) {
340: maxres += 1024;
1.36 schwarze 341: *res = mandoc_reallocarray(*res,
342: maxres, sizeof(struct manpage));
1.1 kristaps 343: }
1.10 schwarze 344: mpage = *res + cur;
1.47 schwarze 345: mpage->ipath = i;
1.50 schwarze 346: mpage->bits = mp->bits;
1.39 schwarze 347: mpage->sec = 10;
1.10 schwarze 348: mpage->form = mp->form;
1.53 schwarze 349: buildnames(search, mpage, db, s, mp->pageid,
1.19 schwarze 350: paths->paths[i], mp->form);
1.53 schwarze 351: if (mpage->names != NULL) {
352: mpage->output = TYPE_Nd & outbit ?
353: mp->desc : outbit ?
354: buildoutput(db, s2, mp->pageid, outbit) :
355: NULL;
356: cur++;
357: }
1.1 kristaps 358: free(mp);
359: }
1.10 schwarze 360:
361: sqlite3_finalize(s);
1.12 schwarze 362: sqlite3_finalize(s2);
1.10 schwarze 363: sqlite3_close(db);
1.1 kristaps 364: ohash_delete(&htab);
1.49 schwarze 365:
366: /*
367: * In man(1) mode, prefer matches in earlier trees
368: * over matches in later trees.
369: */
370:
371: if (cur && search->firstmatch)
372: break;
1.1 kristaps 373: }
1.39 schwarze 374: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.57 schwarze 375: if (chdir_status && getcwd_status && chdir(buf) == -1)
376: perror(buf);
1.1 kristaps 377: exprfree(e);
378: free(sql);
379: *sz = cur;
1.58 schwarze 380: return 1;
1.11 schwarze 381: }
382:
1.45 schwarze 383: void
384: mansearch_free(struct manpage *res, size_t sz)
385: {
386: size_t i;
387:
388: for (i = 0; i < sz; i++) {
389: free(res[i].file);
390: free(res[i].names);
391: free(res[i].output);
392: }
393: free(res);
394: }
395:
1.39 schwarze 396: static int
397: manpage_compare(const void *vp1, const void *vp2)
398: {
399: const struct manpage *mp1, *mp2;
400: int diff;
401:
402: mp1 = vp1;
403: mp2 = vp2;
1.58 schwarze 404: return (diff = mp2->bits - mp1->bits) ? diff :
405: (diff = mp1->sec - mp2->sec) ? diff :
406: strcasecmp(mp1->names, mp2->names);
1.39 schwarze 407: }
408:
1.17 schwarze 409: static void
1.53 schwarze 410: buildnames(const struct mansearch *search, struct manpage *mpage,
411: sqlite3 *db, sqlite3_stmt *s,
1.32 schwarze 412: uint64_t pageid, const char *path, int form)
1.11 schwarze 413: {
1.54 schwarze 414: glob_t globinfo;
415: char *firstname, *newnames, *prevsec, *prevarch;
1.19 schwarze 416: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.11 schwarze 417: size_t i;
1.54 schwarze 418: int c, globres;
1.11 schwarze 419:
1.25 schwarze 420: mpage->file = NULL;
1.17 schwarze 421: mpage->names = NULL;
1.54 schwarze 422: firstname = prevsec = prevarch = NULL;
1.11 schwarze 423: i = 1;
1.32 schwarze 424: SQL_BIND_INT64(db, s, i, pageid);
1.11 schwarze 425: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.17 schwarze 426:
1.22 schwarze 427: /* Decide whether we already have some names. */
1.17 schwarze 428:
429: if (NULL == mpage->names) {
1.11 schwarze 430: oldnames = "";
431: sep1 = "";
432: } else {
1.17 schwarze 433: oldnames = mpage->names;
1.11 schwarze 434: sep1 = ", ";
435: }
1.22 schwarze 436:
1.53 schwarze 437: /* Fetch the next name, rejecting sec/arch mismatches. */
1.22 schwarze 438:
1.41 schwarze 439: sec = (const char *)sqlite3_column_text(s, 0);
1.53 schwarze 440: if (search->sec != NULL && strcasecmp(sec, search->sec))
441: continue;
1.41 schwarze 442: arch = (const char *)sqlite3_column_text(s, 1);
1.53 schwarze 443: if (search->arch != NULL && *arch != '\0' &&
444: strcasecmp(arch, search->arch))
445: continue;
1.41 schwarze 446: name = (const char *)sqlite3_column_text(s, 2);
1.39 schwarze 447:
448: /* Remember the first section found. */
449:
450: if (9 < mpage->sec && '1' <= *sec && '9' >= *sec)
451: mpage->sec = (*sec - '1') + 1;
1.22 schwarze 452:
453: /* If the section changed, append the old one. */
454:
455: if (NULL != prevsec &&
456: (strcmp(sec, prevsec) ||
457: strcmp(arch, prevarch))) {
458: sep2 = '\0' == *prevarch ? "" : "/";
1.24 schwarze 459: mandoc_asprintf(&newnames, "%s(%s%s%s)",
460: oldnames, prevsec, sep2, prevarch);
1.22 schwarze 461: free(mpage->names);
462: oldnames = mpage->names = newnames;
463: free(prevsec);
464: free(prevarch);
465: prevsec = prevarch = NULL;
466: }
467:
468: /* Save the new section, to append it later. */
469:
470: if (NULL == prevsec) {
471: prevsec = mandoc_strdup(sec);
472: prevarch = mandoc_strdup(arch);
473: }
474:
475: /* Append the new name. */
476:
1.24 schwarze 477: mandoc_asprintf(&newnames, "%s%s%s",
478: oldnames, sep1, name);
1.17 schwarze 479: free(mpage->names);
480: mpage->names = newnames;
481:
482: /* Also save the first file name encountered. */
483:
1.51 schwarze 484: if (mpage->file != NULL)
1.17 schwarze 485: continue;
486:
1.48 schwarze 487: if (form & FORM_SRC) {
1.19 schwarze 488: sep1 = "man";
489: fsec = sec;
490: } else {
491: sep1 = "cat";
492: fsec = "0";
493: }
1.51 schwarze 494: sep2 = *arch == '\0' ? "" : "/";
495: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
496: path, sep1, sec, sep2, arch, name, fsec);
1.54 schwarze 497: if (access(mpage->file, R_OK) != -1)
498: continue;
499:
500: /* Handle unusual file name extensions. */
501:
502: if (firstname == NULL)
503: firstname = mpage->file;
504: else
505: free(mpage->file);
506: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.*",
507: path, sep1, sec, sep2, arch, name);
508: globres = glob(mpage->file, 0, NULL, &globinfo);
509: free(mpage->file);
510: mpage->file = globres ? NULL :
511: mandoc_strdup(*globinfo.gl_pathv);
512: globfree(&globinfo);
1.11 schwarze 513: }
1.51 schwarze 514: if (c != SQLITE_DONE)
1.59 ! schwarze 515: warnx("%s", sqlite3_errmsg(db));
1.11 schwarze 516: sqlite3_reset(s);
1.54 schwarze 517:
518: /* If none of the files is usable, use the first name. */
519:
520: if (mpage->file == NULL)
521: mpage->file = firstname;
522: else if (mpage->file != firstname)
523: free(firstname);
1.22 schwarze 524:
525: /* Append one final section to the names. */
526:
1.51 schwarze 527: if (prevsec != NULL) {
528: sep2 = *prevarch == '\0' ? "" : "/";
1.24 schwarze 529: mandoc_asprintf(&newnames, "%s(%s%s%s)",
530: mpage->names, prevsec, sep2, prevarch);
1.22 schwarze 531: free(mpage->names);
532: mpage->names = newnames;
533: free(prevsec);
534: free(prevarch);
535: }
1.12 schwarze 536: }
537:
538: static char *
1.32 schwarze 539: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.12 schwarze 540: {
541: char *output, *newoutput;
542: const char *oldoutput, *sep1, *data;
543: size_t i;
544: int c;
545:
546: output = NULL;
547: i = 1;
1.32 schwarze 548: SQL_BIND_INT64(db, s, i, pageid);
1.12 schwarze 549: SQL_BIND_INT64(db, s, i, outbit);
550: while (SQLITE_ROW == (c = sqlite3_step(s))) {
551: if (NULL == output) {
552: oldoutput = "";
553: sep1 = "";
554: } else {
555: oldoutput = output;
556: sep1 = " # ";
557: }
1.41 schwarze 558: data = (const char *)sqlite3_column_text(s, 1);
1.24 schwarze 559: mandoc_asprintf(&newoutput, "%s%s%s",
560: oldoutput, sep1, data);
1.12 schwarze 561: free(output);
562: output = newoutput;
563: }
564: if (SQLITE_DONE != c)
1.59 ! schwarze 565: warnx("%s", sqlite3_errmsg(db));
1.12 schwarze 566: sqlite3_reset(s);
1.58 schwarze 567: return output;
1.1 kristaps 568: }
569:
570: /*
1.7 schwarze 571: * Implement substring match as an application-defined SQL function.
572: * Using the SQL LIKE or GLOB operators instead would be a bad idea
573: * because that would require escaping metacharacters in the string
574: * being searched for.
575: */
576: static void
577: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
578: {
579:
580: assert(2 == argc);
581: sqlite3_result_int(context, NULL != strcasestr(
582: (const char *)sqlite3_value_text(argv[1]),
583: (const char *)sqlite3_value_text(argv[0])));
584: }
585:
586: /*
1.8 schwarze 587: * Implement regular expression match
588: * as an application-defined SQL function.
589: */
590: static void
591: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
592: {
593:
594: assert(2 == argc);
595: sqlite3_result_int(context, !regexec(
596: (regex_t *)sqlite3_value_blob(argv[0]),
597: (const char *)sqlite3_value_text(argv[1]),
598: 0, NULL, 0));
599: }
600:
1.13 schwarze 601: static void
602: sql_append(char **sql, size_t *sz, const char *newstr, int count)
603: {
604: size_t newsz;
605:
606: newsz = 1 < count ? (size_t)count : strlen(newstr);
607: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
608: if (1 < count)
609: memset(*sql + *sz, *newstr, (size_t)count);
610: else
611: memcpy(*sql + *sz, newstr, newsz);
612: *sz += newsz;
613: (*sql)[*sz] = '\0';
614: }
615:
1.8 schwarze 616: /*
1.1 kristaps 617: * Prepare the search SQL statement.
618: */
619: static char *
1.15 schwarze 620: sql_statement(const struct expr *e)
1.1 kristaps 621: {
622: char *sql;
623: size_t sz;
1.13 schwarze 624: int needop;
1.1 kristaps 625:
1.50 schwarze 626: sql = mandoc_strdup(e->equal ?
627: "SELECT desc, form, pageid, bits "
628: "FROM mpages NATURAL JOIN names WHERE " :
629: "SELECT desc, form, pageid, 0 FROM mpages WHERE ");
1.1 kristaps 630: sz = strlen(sql);
1.2 kristaps 631:
1.13 schwarze 632: for (needop = 0; NULL != e; e = e->next) {
633: if (e->and)
634: sql_append(&sql, &sz, " AND ", 1);
635: else if (needop)
636: sql_append(&sql, &sz, " OR ", 1);
637: if (e->open)
638: sql_append(&sql, &sz, "(", e->open);
1.26 schwarze 639: sql_append(&sql, &sz,
640: TYPE_Nd & e->bits
641: ? (NULL == e->substr
642: ? "desc REGEXP ?"
643: : "desc MATCH ?")
1.27 schwarze 644: : TYPE_Nm == e->bits
645: ? (NULL == e->substr
1.32 schwarze 646: ? "pageid IN (SELECT pageid FROM names "
1.27 schwarze 647: "WHERE name REGEXP ?)"
1.38 schwarze 648: : e->equal
1.50 schwarze 649: ? "name = ? "
1.32 schwarze 650: : "pageid IN (SELECT pageid FROM names "
1.27 schwarze 651: "WHERE name MATCH ?)")
1.26 schwarze 652: : (NULL == e->substr
1.32 schwarze 653: ? "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 654: "WHERE key REGEXP ? AND bits & ?)"
1.32 schwarze 655: : "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 656: "WHERE key MATCH ? AND bits & ?)"), 1);
1.13 schwarze 657: if (e->close)
658: sql_append(&sql, &sz, ")", e->close);
659: needop = 1;
1.1 kristaps 660: }
661:
1.58 schwarze 662: return sql;
1.1 kristaps 663: }
664:
665: /*
666: * Compile a set of string tokens into an expression.
667: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
668: * "(", "foo=bar", etc.).
669: */
670: static struct expr *
1.5 kristaps 671: exprcomp(const struct mansearch *search, int argc, char *argv[])
1.1 kristaps 672: {
1.27 schwarze 673: uint64_t mask;
1.13 schwarze 674: int i, toopen, logic, igncase, toclose;
1.27 schwarze 675: struct expr *first, *prev, *cur, *next;
1.1 kristaps 676:
677: first = cur = NULL;
1.53 schwarze 678: logic = igncase = toopen = toclose = 0;
1.1 kristaps 679:
680: for (i = 0; i < argc; i++) {
1.13 schwarze 681: if (0 == strcmp("(", argv[i])) {
682: if (igncase)
683: goto fail;
684: toopen++;
685: toclose++;
686: continue;
687: } else if (0 == strcmp(")", argv[i])) {
688: if (toopen || logic || igncase || NULL == cur)
689: goto fail;
690: cur->close++;
691: if (0 > --toclose)
692: goto fail;
693: continue;
694: } else if (0 == strcmp("-a", argv[i])) {
695: if (toopen || logic || igncase || NULL == cur)
696: goto fail;
697: logic = 1;
698: continue;
699: } else if (0 == strcmp("-o", argv[i])) {
700: if (toopen || logic || igncase || NULL == cur)
701: goto fail;
702: logic = 2;
703: continue;
704: } else if (0 == strcmp("-i", argv[i])) {
705: if (igncase)
706: goto fail;
707: igncase = 1;
708: continue;
1.1 kristaps 709: }
1.13 schwarze 710: next = exprterm(search, argv[i], !igncase);
711: if (NULL == next)
712: goto fail;
1.26 schwarze 713: if (NULL == first)
714: first = next;
715: else
1.1 kristaps 716: cur->next = next;
1.27 schwarze 717: prev = cur = next;
1.26 schwarze 718:
719: /*
720: * Searching for descriptions must be split out
721: * because they are stored in the mpages table,
722: * not in the keys table.
723: */
724:
1.27 schwarze 725: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
726: if (mask & cur->bits && ~mask & cur->bits) {
727: next = mandoc_calloc(1,
728: sizeof(struct expr));
729: memcpy(next, cur, sizeof(struct expr));
730: prev->open = 1;
731: cur->bits = mask;
732: cur->next = next;
733: cur = next;
734: cur->bits &= ~mask;
735: }
736: }
737: prev->and = (1 == logic);
738: prev->open += toopen;
739: if (cur != prev)
1.26 schwarze 740: cur->close = 1;
1.27 schwarze 741:
1.13 schwarze 742: toopen = logic = igncase = 0;
1.1 kristaps 743: }
1.53 schwarze 744: if ( ! (toopen || logic || igncase || toclose))
1.58 schwarze 745: return first;
1.15 schwarze 746:
1.13 schwarze 747: fail:
748: if (NULL != first)
749: exprfree(first);
1.58 schwarze 750: return NULL;
1.15 schwarze 751: }
752:
753: static struct expr *
1.8 schwarze 754: exprterm(const struct mansearch *search, char *buf, int cs)
1.1 kristaps 755: {
1.15 schwarze 756: char errbuf[BUFSIZ];
1.1 kristaps 757: struct expr *e;
1.38 schwarze 758: char *key, *val;
1.20 schwarze 759: uint64_t iterbit;
760: int i, irc;
1.1 kristaps 761:
762: if ('\0' == *buf)
1.58 schwarze 763: return NULL;
1.1 kristaps 764:
765: e = mandoc_calloc(1, sizeof(struct expr));
766:
1.45 schwarze 767: if (search->argmode == ARG_NAME) {
768: e->bits = TYPE_Nm;
1.8 schwarze 769: e->substr = buf;
1.38 schwarze 770: e->equal = 1;
1.58 schwarze 771: return e;
1.5 kristaps 772: }
773:
1.1 kristaps 774: /*
1.45 schwarze 775: * Separate macro keys from search string.
776: * If needed, request regular expression handling
777: * by setting e->substr to NULL.
1.1 kristaps 778: */
779:
1.45 schwarze 780: if (search->argmode == ARG_WORD) {
781: e->bits = TYPE_Nm;
782: e->substr = NULL;
783: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
1.46 schwarze 784: cs = 0;
1.45 schwarze 785: } else if ((val = strpbrk(buf, "=~")) == NULL) {
786: e->bits = TYPE_Nm | TYPE_Nd;
1.8 schwarze 787: e->substr = buf;
1.38 schwarze 788: } else {
789: if (val == buf)
1.45 schwarze 790: e->bits = TYPE_Nm | TYPE_Nd;
1.38 schwarze 791: if ('=' == *val)
792: e->substr = val + 1;
793: *val++ = '\0';
1.21 schwarze 794: if (NULL != strstr(buf, "arch"))
795: cs = 0;
1.38 schwarze 796: }
797:
798: /* Compile regular expressions. */
799:
800: if (NULL == e->substr) {
801: irc = regcomp(&e->regexp, val,
802: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.45 schwarze 803: if (search->argmode == ARG_WORD)
1.38 schwarze 804: free(val);
805: if (irc) {
1.15 schwarze 806: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
1.59 ! schwarze 807: warnx("regcomp: %s", errbuf);
1.8 schwarze 808: free(e);
1.58 schwarze 809: return NULL;
1.8 schwarze 810: }
1.38 schwarze 811: }
812:
813: if (e->bits)
1.58 schwarze 814: return e;
1.1 kristaps 815:
816: /*
817: * Parse out all possible fields.
818: * If the field doesn't resolve, bail.
819: */
820:
821: while (NULL != (key = strsep(&buf, ","))) {
822: if ('\0' == *key)
823: continue;
1.20 schwarze 824: for (i = 0, iterbit = 1;
825: i < mansearch_keymax;
826: i++, iterbit <<= 1) {
827: if (0 == strcasecmp(key,
828: mansearch_keynames[i])) {
829: e->bits |= iterbit;
830: break;
831: }
832: }
833: if (i == mansearch_keymax) {
834: if (strcasecmp(key, "any")) {
835: free(e);
1.58 schwarze 836: return NULL;
1.20 schwarze 837: }
838: e->bits |= ~0ULL;
1.1 kristaps 839: }
840: }
841:
1.58 schwarze 842: return e;
1.1 kristaps 843: }
844:
845: static void
846: exprfree(struct expr *p)
847: {
848: struct expr *pp;
849:
850: while (NULL != p) {
851: pp = p->next;
852: free(p);
853: p = pp;
854: }
855: }
856:
857: static void *
1.37 schwarze 858: hash_calloc(size_t nmemb, size_t sz, void *arg)
1.1 kristaps 859: {
860:
1.58 schwarze 861: return mandoc_calloc(nmemb, sz);
1.1 kristaps 862: }
863:
864: static void *
865: hash_alloc(size_t sz, void *arg)
866: {
867:
1.58 schwarze 868: return mandoc_malloc(sz);
1.1 kristaps 869: }
870:
871: static void
1.37 schwarze 872: hash_free(void *p, void *arg)
1.1 kristaps 873: {
874:
875: free(p);
876: }
CVSweb