Annotation of mandoc/mansearch.c, Revision 1.55
1.55 ! schwarze 1: /* $Id: mansearch.c,v 1.54 2015/02/27 16:02:10 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: *
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: */
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>
1.52 schwarze 24: #include <errno.h>
1.1 kristaps 25: #include <fcntl.h>
26: #include <getopt.h>
1.54 schwarze 27: #include <glob.h>
1.6 schwarze 28: #include <limits.h>
1.8 schwarze 29: #include <regex.h>
1.1 kristaps 30: #include <stdio.h>
31: #include <stdint.h>
32: #include <stddef.h>
33: #include <stdlib.h>
34: #include <string.h>
35: #include <unistd.h>
36:
1.44 schwarze 37: #if HAVE_OHASH
1.1 kristaps 38: #include <ohash.h>
1.4 kristaps 39: #else
40: #include "compat_ohash.h"
41: #endif
1.1 kristaps 42: #include <sqlite3.h>
1.40 schwarze 43: #ifndef SQLITE_DETERMINISTIC
44: #define SQLITE_DETERMINISTIC 0
45: #endif
1.1 kristaps 46:
47: #include "mandoc.h"
1.23 schwarze 48: #include "mandoc_aux.h"
1.1 kristaps 49: #include "manpath.h"
50: #include "mansearch.h"
51:
1.20 schwarze 52: extern int mansearch_keymax;
53: extern const char *const mansearch_keynames[];
54:
1.3 kristaps 55: #define SQL_BIND_TEXT(_db, _s, _i, _v) \
1.8 schwarze 56: do { if (SQLITE_OK != sqlite3_bind_text \
1.2 kristaps 57: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.8 schwarze 58: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
59: } while (0)
1.3 kristaps 60: #define SQL_BIND_INT64(_db, _s, _i, _v) \
1.8 schwarze 61: do { if (SQLITE_OK != sqlite3_bind_int64 \
1.2 kristaps 62: ((_s), (_i)++, (_v))) \
1.8 schwarze 63: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
64: } while (0)
65: #define SQL_BIND_BLOB(_db, _s, _i, _v) \
66: do { if (SQLITE_OK != sqlite3_bind_blob \
67: ((_s), (_i)++, (&_v), sizeof(_v), SQLITE_STATIC)) \
68: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
69: } while (0)
1.2 kristaps 70:
1.1 kristaps 71: struct expr {
1.38 schwarze 72: regex_t regexp; /* compiled regexp, if applicable */
73: const char *substr; /* to search for, if applicable */
74: struct expr *next; /* next in sequence */
1.34 schwarze 75: uint64_t bits; /* type-mask */
1.38 schwarze 76: int equal; /* equality, not subsring match */
1.13 schwarze 77: int open; /* opening parentheses before */
78: int and; /* logical AND before */
79: int close; /* closing parentheses after */
1.1 kristaps 80: };
81:
82: struct match {
1.32 schwarze 83: uint64_t pageid; /* identifier in database */
1.50 schwarze 84: uint64_t bits; /* name type mask */
1.26 schwarze 85: char *desc; /* manual page description */
1.48 schwarze 86: int form; /* bit field: formatted, zipped? */
1.1 kristaps 87: };
88:
1.53 schwarze 89: static void buildnames(const struct mansearch *,
90: struct manpage *, sqlite3 *,
1.19 schwarze 91: sqlite3_stmt *, uint64_t,
92: const char *, int form);
1.12 schwarze 93: static char *buildoutput(sqlite3 *, sqlite3_stmt *,
94: uint64_t, uint64_t);
1.1 kristaps 95: static void *hash_alloc(size_t, void *);
1.37 schwarze 96: static void hash_free(void *, void *);
97: static void *hash_calloc(size_t, size_t, void *);
1.34 schwarze 98: static struct expr *exprcomp(const struct mansearch *,
1.5 kristaps 99: int, char *[]);
1.1 kristaps 100: static void exprfree(struct expr *);
1.8 schwarze 101: static struct expr *exprterm(const struct mansearch *, char *, int);
1.39 schwarze 102: static int manpage_compare(const void *, const void *);
1.13 schwarze 103: static void sql_append(char **sql, size_t *sz,
104: const char *newstr, int count);
1.7 schwarze 105: static void sql_match(sqlite3_context *context,
106: int argc, sqlite3_value **argv);
1.8 schwarze 107: static void sql_regexp(sqlite3_context *context,
108: int argc, sqlite3_value **argv);
1.15 schwarze 109: static char *sql_statement(const struct expr *);
1.28 schwarze 110:
1.34 schwarze 111:
1.28 schwarze 112: int
113: mansearch_setup(int start)
114: {
115: static void *pagecache;
116: int c;
117:
118: #define PC_PAGESIZE 1280
119: #define PC_NUMPAGES 256
120:
121: if (start) {
122: if (NULL != pagecache) {
123: fprintf(stderr, "pagecache already enabled\n");
124: return((int)MANDOCLEVEL_BADARG);
125: }
126:
127: pagecache = mmap(NULL, PC_PAGESIZE * PC_NUMPAGES,
1.42 schwarze 128: PROT_READ | PROT_WRITE,
129: MAP_SHARED | MAP_ANON, -1, 0);
1.28 schwarze 130:
131: if (MAP_FAILED == pagecache) {
132: perror("mmap");
133: pagecache = NULL;
134: return((int)MANDOCLEVEL_SYSERR);
135: }
136:
137: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
138: pagecache, PC_PAGESIZE, PC_NUMPAGES);
139:
140: if (SQLITE_OK == c)
141: return((int)MANDOCLEVEL_OK);
142:
143: fprintf(stderr, "pagecache: %s\n", sqlite3_errstr(c));
144:
145: } else if (NULL == pagecache) {
146: fprintf(stderr, "pagecache missing\n");
147: return((int)MANDOCLEVEL_BADARG);
148: }
149:
150: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
151: perror("munmap");
152: pagecache = NULL;
153: return((int)MANDOCLEVEL_SYSERR);
154: }
155:
156: pagecache = NULL;
157: return((int)MANDOCLEVEL_OK);
158: }
1.1 kristaps 159:
160: int
1.5 kristaps 161: mansearch(const struct mansearch *search,
1.12 schwarze 162: const struct manpaths *paths,
163: int argc, char *argv[],
1.1 kristaps 164: struct manpage **res, size_t *sz)
165: {
1.20 schwarze 166: int fd, rc, c, indexbit;
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;
180:
1.37 schwarze 181: info.calloc = hash_calloc;
1.1 kristaps 182: info.alloc = hash_alloc;
1.37 schwarze 183: info.free = hash_free;
1.32 schwarze 184: info.key_offset = offsetof(struct match, pageid);
1.1 kristaps 185:
1.2 kristaps 186: *sz = cur = maxres = 0;
1.1 kristaps 187: sql = NULL;
188: *res = NULL;
189: fd = -1;
190: e = NULL;
1.2 kristaps 191: rc = 0;
1.1 kristaps 192:
193: if (0 == argc)
194: goto out;
1.5 kristaps 195: if (NULL == (e = exprcomp(search, argc, argv)))
1.1 kristaps 196: goto out;
197:
1.45 schwarze 198: if (NULL != search->outkey) {
1.55 ! schwarze 199: outbit = TYPE_Nd;
1.20 schwarze 200: for (indexbit = 0, iterbit = 1;
201: indexbit < mansearch_keymax;
202: indexbit++, iterbit <<= 1) {
1.45 schwarze 203: if (0 == strcasecmp(search->outkey,
1.20 schwarze 204: mansearch_keynames[indexbit])) {
205: outbit = iterbit;
1.12 schwarze 206: break;
207: }
208: }
1.55 ! schwarze 209: } else
! 210: outbit = 0;
1.12 schwarze 211:
1.1 kristaps 212: /*
213: * Save a descriptor to the current working directory.
214: * Since pathnames in the "paths" variable might be relative,
215: * and we'll be chdir()ing into them, we need to keep a handle
216: * on our current directory from which to start the chdir().
217: */
218:
1.6 schwarze 219: if (NULL == getcwd(buf, PATH_MAX)) {
1.30 schwarze 220: perror("getcwd");
1.1 kristaps 221: goto out;
222: } else if (-1 == (fd = open(buf, O_RDONLY, 0))) {
223: perror(buf);
224: goto out;
225: }
226:
1.15 schwarze 227: sql = sql_statement(e);
1.1 kristaps 228:
229: /*
230: * Loop over the directories (containing databases) for us to
231: * search.
232: * Don't let missing/bad databases/directories phase us.
233: * In each, try to open the resident database and, if it opens,
234: * scan it for our match expression.
235: */
236:
237: for (i = 0; i < paths->sz; i++) {
238: if (-1 == fchdir(fd)) {
239: perror(buf);
240: free(*res);
241: break;
242: } else if (-1 == chdir(paths->paths[i])) {
243: perror(paths->paths[i]);
244: continue;
1.34 schwarze 245: }
1.1 kristaps 246:
1.34 schwarze 247: c = sqlite3_open_v2(MANDOC_DB, &db,
248: SQLITE_OPEN_READONLY, NULL);
1.1 kristaps 249:
1.2 kristaps 250: if (SQLITE_OK != c) {
1.52 schwarze 251: fprintf(stderr, "%s/%s: %s\n",
252: paths->paths[i], MANDOC_DB, strerror(errno));
1.1 kristaps 253: sqlite3_close(db);
254: continue;
255: }
256:
1.8 schwarze 257: /*
258: * Define the SQL functions for substring
259: * and regular expression matching.
260: */
1.7 schwarze 261:
262: c = sqlite3_create_function(db, "match", 2,
1.31 schwarze 263: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
264: NULL, sql_match, NULL, NULL);
1.8 schwarze 265: assert(SQLITE_OK == c);
266: c = sqlite3_create_function(db, "regexp", 2,
1.31 schwarze 267: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
268: NULL, sql_regexp, NULL, NULL);
1.8 schwarze 269: assert(SQLITE_OK == c);
1.7 schwarze 270:
1.1 kristaps 271: j = 1;
1.2 kristaps 272: c = sqlite3_prepare_v2(db, sql, -1, &s, NULL);
273: if (SQLITE_OK != c)
274: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
1.1 kristaps 275:
276: for (ep = e; NULL != ep; ep = ep->next) {
1.8 schwarze 277: if (NULL == ep->substr) {
278: SQL_BIND_BLOB(db, s, j, ep->regexp);
279: } else
280: SQL_BIND_TEXT(db, s, j, ep->substr);
1.27 schwarze 281: if (0 == ((TYPE_Nd | TYPE_Nm) & ep->bits))
1.26 schwarze 282: SQL_BIND_INT64(db, s, j, ep->bits);
1.1 kristaps 283: }
284:
285: memset(&htab, 0, sizeof(struct ohash));
286: ohash_init(&htab, 4, &info);
287:
288: /*
289: * Hash each entry on its [unique] document identifier.
290: * This is a uint64_t.
291: * Instead of using a hash function, simply convert the
292: * uint64_t to a uint32_t, the hash value's type.
293: * This gives good performance and preserves the
294: * distribution of buckets in the table.
295: */
1.2 kristaps 296: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.32 schwarze 297: pageid = sqlite3_column_int64(s, 2);
1.34 schwarze 298: idx = ohash_lookup_memory(&htab,
299: (char *)&pageid, sizeof(uint64_t),
300: (uint32_t)pageid);
1.1 kristaps 301:
302: if (NULL != ohash_find(&htab, idx))
303: continue;
304:
305: mp = mandoc_calloc(1, sizeof(struct match));
1.32 schwarze 306: mp->pageid = pageid;
1.26 schwarze 307: mp->form = sqlite3_column_int(s, 1);
1.50 schwarze 308: mp->bits = sqlite3_column_int64(s, 3);
1.26 schwarze 309: if (TYPE_Nd == outbit)
1.41 schwarze 310: mp->desc = mandoc_strdup((const char *)
1.26 schwarze 311: sqlite3_column_text(s, 0));
1.1 kristaps 312: ohash_insert(&htab, idx, mp);
313: }
314:
1.2 kristaps 315: if (SQLITE_DONE != c)
316: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
317:
1.1 kristaps 318: sqlite3_finalize(s);
1.10 schwarze 319:
1.34 schwarze 320: c = sqlite3_prepare_v2(db,
1.35 schwarze 321: "SELECT sec, arch, name, pageid FROM mlinks "
322: "WHERE pageid=? ORDER BY sec, arch, name",
1.10 schwarze 323: -1, &s, NULL);
324: if (SQLITE_OK != c)
325: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
1.1 kristaps 326:
1.12 schwarze 327: c = sqlite3_prepare_v2(db,
1.35 schwarze 328: "SELECT bits, key, pageid FROM keys "
329: "WHERE pageid=? AND bits & ?",
1.12 schwarze 330: -1, &s2, NULL);
331: if (SQLITE_OK != c)
332: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
333:
1.1 kristaps 334: for (mp = ohash_first(&htab, &idx);
335: NULL != mp;
336: mp = ohash_next(&htab, &idx)) {
337: if (cur + 1 > maxres) {
338: maxres += 1024;
1.36 schwarze 339: *res = mandoc_reallocarray(*res,
340: maxres, sizeof(struct manpage));
1.1 kristaps 341: }
1.10 schwarze 342: mpage = *res + cur;
1.47 schwarze 343: mpage->ipath = i;
1.50 schwarze 344: mpage->bits = mp->bits;
1.39 schwarze 345: mpage->sec = 10;
1.10 schwarze 346: mpage->form = mp->form;
1.53 schwarze 347: buildnames(search, mpage, db, s, mp->pageid,
1.19 schwarze 348: paths->paths[i], mp->form);
1.53 schwarze 349: if (mpage->names != NULL) {
350: mpage->output = TYPE_Nd & outbit ?
351: mp->desc : outbit ?
352: buildoutput(db, s2, mp->pageid, outbit) :
353: NULL;
354: cur++;
355: }
1.1 kristaps 356: free(mp);
357: }
1.10 schwarze 358:
359: sqlite3_finalize(s);
1.12 schwarze 360: sqlite3_finalize(s2);
1.10 schwarze 361: sqlite3_close(db);
1.1 kristaps 362: ohash_delete(&htab);
1.49 schwarze 363:
364: /*
365: * In man(1) mode, prefer matches in earlier trees
366: * over matches in later trees.
367: */
368:
369: if (cur && search->firstmatch)
370: break;
1.1 kristaps 371: }
1.39 schwarze 372: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.2 kristaps 373: rc = 1;
1.1 kristaps 374: out:
1.30 schwarze 375: if (-1 != fd) {
376: if (-1 == fchdir(fd))
377: perror(buf);
378: close(fd);
379: }
1.1 kristaps 380: exprfree(e);
381: free(sql);
382: *sz = cur;
1.2 kristaps 383: return(rc);
1.11 schwarze 384: }
385:
1.45 schwarze 386: void
387: mansearch_free(struct manpage *res, size_t sz)
388: {
389: size_t i;
390:
391: for (i = 0; i < sz; i++) {
392: free(res[i].file);
393: free(res[i].names);
394: free(res[i].output);
395: }
396: free(res);
397: }
398:
1.39 schwarze 399: static int
400: manpage_compare(const void *vp1, const void *vp2)
401: {
402: const struct manpage *mp1, *mp2;
403: int diff;
404:
405: mp1 = vp1;
406: mp2 = vp2;
1.50 schwarze 407: return( (diff = mp2->bits - mp1->bits) ? diff :
408: (diff = mp1->sec - mp2->sec) ? diff :
409: strcasecmp(mp1->names, mp2->names));
1.39 schwarze 410: }
411:
1.17 schwarze 412: static void
1.53 schwarze 413: buildnames(const struct mansearch *search, struct manpage *mpage,
414: sqlite3 *db, sqlite3_stmt *s,
1.32 schwarze 415: uint64_t pageid, const char *path, int form)
1.11 schwarze 416: {
1.54 schwarze 417: glob_t globinfo;
418: char *firstname, *newnames, *prevsec, *prevarch;
1.19 schwarze 419: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.11 schwarze 420: size_t i;
1.54 schwarze 421: int c, globres;
1.11 schwarze 422:
1.25 schwarze 423: mpage->file = NULL;
1.17 schwarze 424: mpage->names = NULL;
1.54 schwarze 425: firstname = prevsec = prevarch = NULL;
1.11 schwarze 426: i = 1;
1.32 schwarze 427: SQL_BIND_INT64(db, s, i, pageid);
1.11 schwarze 428: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.17 schwarze 429:
1.22 schwarze 430: /* Decide whether we already have some names. */
1.17 schwarze 431:
432: if (NULL == mpage->names) {
1.11 schwarze 433: oldnames = "";
434: sep1 = "";
435: } else {
1.17 schwarze 436: oldnames = mpage->names;
1.11 schwarze 437: sep1 = ", ";
438: }
1.22 schwarze 439:
1.53 schwarze 440: /* Fetch the next name, rejecting sec/arch mismatches. */
1.22 schwarze 441:
1.41 schwarze 442: sec = (const char *)sqlite3_column_text(s, 0);
1.53 schwarze 443: if (search->sec != NULL && strcasecmp(sec, search->sec))
444: continue;
1.41 schwarze 445: arch = (const char *)sqlite3_column_text(s, 1);
1.53 schwarze 446: if (search->arch != NULL && *arch != '\0' &&
447: strcasecmp(arch, search->arch))
448: continue;
1.41 schwarze 449: name = (const char *)sqlite3_column_text(s, 2);
1.39 schwarze 450:
451: /* Remember the first section found. */
452:
453: if (9 < mpage->sec && '1' <= *sec && '9' >= *sec)
454: mpage->sec = (*sec - '1') + 1;
1.22 schwarze 455:
456: /* If the section changed, append the old one. */
457:
458: if (NULL != prevsec &&
459: (strcmp(sec, prevsec) ||
460: strcmp(arch, prevarch))) {
461: sep2 = '\0' == *prevarch ? "" : "/";
1.24 schwarze 462: mandoc_asprintf(&newnames, "%s(%s%s%s)",
463: oldnames, prevsec, sep2, prevarch);
1.22 schwarze 464: free(mpage->names);
465: oldnames = mpage->names = newnames;
466: free(prevsec);
467: free(prevarch);
468: prevsec = prevarch = NULL;
469: }
470:
471: /* Save the new section, to append it later. */
472:
473: if (NULL == prevsec) {
474: prevsec = mandoc_strdup(sec);
475: prevarch = mandoc_strdup(arch);
476: }
477:
478: /* Append the new name. */
479:
1.24 schwarze 480: mandoc_asprintf(&newnames, "%s%s%s",
481: oldnames, sep1, name);
1.17 schwarze 482: free(mpage->names);
483: mpage->names = newnames;
484:
485: /* Also save the first file name encountered. */
486:
1.51 schwarze 487: if (mpage->file != NULL)
1.17 schwarze 488: continue;
489:
1.48 schwarze 490: if (form & FORM_SRC) {
1.19 schwarze 491: sep1 = "man";
492: fsec = sec;
493: } else {
494: sep1 = "cat";
495: fsec = "0";
496: }
1.51 schwarze 497: sep2 = *arch == '\0' ? "" : "/";
498: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
499: path, sep1, sec, sep2, arch, name, fsec);
1.54 schwarze 500: if (access(mpage->file, R_OK) != -1)
501: continue;
502:
503: /* Handle unusual file name extensions. */
504:
505: if (firstname == NULL)
506: firstname = mpage->file;
507: else
508: free(mpage->file);
509: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.*",
510: path, sep1, sec, sep2, arch, name);
511: globres = glob(mpage->file, 0, NULL, &globinfo);
512: free(mpage->file);
513: mpage->file = globres ? NULL :
514: mandoc_strdup(*globinfo.gl_pathv);
515: globfree(&globinfo);
1.11 schwarze 516: }
1.51 schwarze 517: if (c != SQLITE_DONE)
1.11 schwarze 518: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
519: sqlite3_reset(s);
1.54 schwarze 520:
521: /* If none of the files is usable, use the first name. */
522:
523: if (mpage->file == NULL)
524: mpage->file = firstname;
525: else if (mpage->file != firstname)
526: free(firstname);
1.22 schwarze 527:
528: /* Append one final section to the names. */
529:
1.51 schwarze 530: if (prevsec != NULL) {
531: sep2 = *prevarch == '\0' ? "" : "/";
1.24 schwarze 532: mandoc_asprintf(&newnames, "%s(%s%s%s)",
533: mpage->names, prevsec, sep2, prevarch);
1.22 schwarze 534: free(mpage->names);
535: mpage->names = newnames;
536: free(prevsec);
537: free(prevarch);
538: }
1.12 schwarze 539: }
540:
541: static char *
1.32 schwarze 542: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.12 schwarze 543: {
544: char *output, *newoutput;
545: const char *oldoutput, *sep1, *data;
546: size_t i;
547: int c;
548:
549: output = NULL;
550: i = 1;
1.32 schwarze 551: SQL_BIND_INT64(db, s, i, pageid);
1.12 schwarze 552: SQL_BIND_INT64(db, s, i, outbit);
553: while (SQLITE_ROW == (c = sqlite3_step(s))) {
554: if (NULL == output) {
555: oldoutput = "";
556: sep1 = "";
557: } else {
558: oldoutput = output;
559: sep1 = " # ";
560: }
1.41 schwarze 561: data = (const char *)sqlite3_column_text(s, 1);
1.24 schwarze 562: mandoc_asprintf(&newoutput, "%s%s%s",
563: oldoutput, sep1, data);
1.12 schwarze 564: free(output);
565: output = newoutput;
566: }
567: if (SQLITE_DONE != c)
568: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
569: sqlite3_reset(s);
570: return(output);
1.1 kristaps 571: }
572:
573: /*
1.7 schwarze 574: * Implement substring match as an application-defined SQL function.
575: * Using the SQL LIKE or GLOB operators instead would be a bad idea
576: * because that would require escaping metacharacters in the string
577: * being searched for.
578: */
579: static void
580: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
581: {
582:
583: assert(2 == argc);
584: sqlite3_result_int(context, NULL != strcasestr(
585: (const char *)sqlite3_value_text(argv[1]),
586: (const char *)sqlite3_value_text(argv[0])));
587: }
588:
589: /*
1.8 schwarze 590: * Implement regular expression match
591: * as an application-defined SQL function.
592: */
593: static void
594: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
595: {
596:
597: assert(2 == argc);
598: sqlite3_result_int(context, !regexec(
599: (regex_t *)sqlite3_value_blob(argv[0]),
600: (const char *)sqlite3_value_text(argv[1]),
601: 0, NULL, 0));
602: }
603:
1.13 schwarze 604: static void
605: sql_append(char **sql, size_t *sz, const char *newstr, int count)
606: {
607: size_t newsz;
608:
609: newsz = 1 < count ? (size_t)count : strlen(newstr);
610: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
611: if (1 < count)
612: memset(*sql + *sz, *newstr, (size_t)count);
613: else
614: memcpy(*sql + *sz, newstr, newsz);
615: *sz += newsz;
616: (*sql)[*sz] = '\0';
617: }
618:
1.8 schwarze 619: /*
1.1 kristaps 620: * Prepare the search SQL statement.
621: */
622: static char *
1.15 schwarze 623: sql_statement(const struct expr *e)
1.1 kristaps 624: {
625: char *sql;
626: size_t sz;
1.13 schwarze 627: int needop;
1.1 kristaps 628:
1.50 schwarze 629: sql = mandoc_strdup(e->equal ?
630: "SELECT desc, form, pageid, bits "
631: "FROM mpages NATURAL JOIN names WHERE " :
632: "SELECT desc, form, pageid, 0 FROM mpages WHERE ");
1.1 kristaps 633: sz = strlen(sql);
1.2 kristaps 634:
1.13 schwarze 635: for (needop = 0; NULL != e; e = e->next) {
636: if (e->and)
637: sql_append(&sql, &sz, " AND ", 1);
638: else if (needop)
639: sql_append(&sql, &sz, " OR ", 1);
640: if (e->open)
641: sql_append(&sql, &sz, "(", e->open);
1.26 schwarze 642: sql_append(&sql, &sz,
643: TYPE_Nd & e->bits
644: ? (NULL == e->substr
645: ? "desc REGEXP ?"
646: : "desc MATCH ?")
1.27 schwarze 647: : TYPE_Nm == e->bits
648: ? (NULL == e->substr
1.32 schwarze 649: ? "pageid IN (SELECT pageid FROM names "
1.27 schwarze 650: "WHERE name REGEXP ?)"
1.38 schwarze 651: : e->equal
1.50 schwarze 652: ? "name = ? "
1.32 schwarze 653: : "pageid IN (SELECT pageid FROM names "
1.27 schwarze 654: "WHERE name MATCH ?)")
1.26 schwarze 655: : (NULL == e->substr
1.32 schwarze 656: ? "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 657: "WHERE key REGEXP ? AND bits & ?)"
1.32 schwarze 658: : "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 659: "WHERE key MATCH ? AND bits & ?)"), 1);
1.13 schwarze 660: if (e->close)
661: sql_append(&sql, &sz, ")", e->close);
662: needop = 1;
1.1 kristaps 663: }
664:
665: return(sql);
666: }
667:
668: /*
669: * Compile a set of string tokens into an expression.
670: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
671: * "(", "foo=bar", etc.).
672: */
673: static struct expr *
1.5 kristaps 674: exprcomp(const struct mansearch *search, int argc, char *argv[])
1.1 kristaps 675: {
1.27 schwarze 676: uint64_t mask;
1.13 schwarze 677: int i, toopen, logic, igncase, toclose;
1.27 schwarze 678: struct expr *first, *prev, *cur, *next;
1.1 kristaps 679:
680: first = cur = NULL;
1.53 schwarze 681: logic = igncase = toopen = toclose = 0;
1.1 kristaps 682:
683: for (i = 0; i < argc; i++) {
1.13 schwarze 684: if (0 == strcmp("(", argv[i])) {
685: if (igncase)
686: goto fail;
687: toopen++;
688: toclose++;
689: continue;
690: } else if (0 == strcmp(")", argv[i])) {
691: if (toopen || logic || igncase || NULL == cur)
692: goto fail;
693: cur->close++;
694: if (0 > --toclose)
695: goto fail;
696: continue;
697: } else if (0 == strcmp("-a", argv[i])) {
698: if (toopen || logic || igncase || NULL == cur)
699: goto fail;
700: logic = 1;
701: continue;
702: } else if (0 == strcmp("-o", argv[i])) {
703: if (toopen || logic || igncase || NULL == cur)
704: goto fail;
705: logic = 2;
706: continue;
707: } else if (0 == strcmp("-i", argv[i])) {
708: if (igncase)
709: goto fail;
710: igncase = 1;
711: continue;
1.1 kristaps 712: }
1.13 schwarze 713: next = exprterm(search, argv[i], !igncase);
714: if (NULL == next)
715: goto fail;
1.26 schwarze 716: if (NULL == first)
717: first = next;
718: else
1.1 kristaps 719: cur->next = next;
1.27 schwarze 720: prev = cur = next;
1.26 schwarze 721:
722: /*
723: * Searching for descriptions must be split out
724: * because they are stored in the mpages table,
725: * not in the keys table.
726: */
727:
1.27 schwarze 728: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
729: if (mask & cur->bits && ~mask & cur->bits) {
730: next = mandoc_calloc(1,
731: sizeof(struct expr));
732: memcpy(next, cur, sizeof(struct expr));
733: prev->open = 1;
734: cur->bits = mask;
735: cur->next = next;
736: cur = next;
737: cur->bits &= ~mask;
738: }
739: }
740: prev->and = (1 == logic);
741: prev->open += toopen;
742: if (cur != prev)
1.26 schwarze 743: cur->close = 1;
1.27 schwarze 744:
1.13 schwarze 745: toopen = logic = igncase = 0;
1.1 kristaps 746: }
1.53 schwarze 747: if ( ! (toopen || logic || igncase || toclose))
748: return(first);
1.15 schwarze 749:
1.13 schwarze 750: fail:
751: if (NULL != first)
752: exprfree(first);
753: return(NULL);
1.15 schwarze 754: }
755:
756: static struct expr *
1.8 schwarze 757: exprterm(const struct mansearch *search, char *buf, int cs)
1.1 kristaps 758: {
1.15 schwarze 759: char errbuf[BUFSIZ];
1.1 kristaps 760: struct expr *e;
1.38 schwarze 761: char *key, *val;
1.20 schwarze 762: uint64_t iterbit;
763: int i, irc;
1.1 kristaps 764:
765: if ('\0' == *buf)
766: return(NULL);
767:
768: e = mandoc_calloc(1, sizeof(struct expr));
769:
1.45 schwarze 770: if (search->argmode == ARG_NAME) {
771: e->bits = TYPE_Nm;
1.8 schwarze 772: e->substr = buf;
1.38 schwarze 773: e->equal = 1;
1.5 kristaps 774: return(e);
775: }
776:
1.1 kristaps 777: /*
1.45 schwarze 778: * Separate macro keys from search string.
779: * If needed, request regular expression handling
780: * by setting e->substr to NULL.
1.1 kristaps 781: */
782:
1.45 schwarze 783: if (search->argmode == ARG_WORD) {
784: e->bits = TYPE_Nm;
785: e->substr = NULL;
786: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
1.46 schwarze 787: cs = 0;
1.45 schwarze 788: } else if ((val = strpbrk(buf, "=~")) == NULL) {
789: e->bits = TYPE_Nm | TYPE_Nd;
1.8 schwarze 790: e->substr = buf;
1.38 schwarze 791: } else {
792: if (val == buf)
1.45 schwarze 793: e->bits = TYPE_Nm | TYPE_Nd;
1.38 schwarze 794: if ('=' == *val)
795: e->substr = val + 1;
796: *val++ = '\0';
1.21 schwarze 797: if (NULL != strstr(buf, "arch"))
798: cs = 0;
1.38 schwarze 799: }
800:
801: /* Compile regular expressions. */
802:
803: if (NULL == e->substr) {
804: irc = regcomp(&e->regexp, val,
805: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.45 schwarze 806: if (search->argmode == ARG_WORD)
1.38 schwarze 807: free(val);
808: if (irc) {
1.15 schwarze 809: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
810: fprintf(stderr, "regcomp: %s\n", errbuf);
1.8 schwarze 811: free(e);
812: return(NULL);
813: }
1.38 schwarze 814: }
815:
816: if (e->bits)
817: return(e);
1.1 kristaps 818:
819: /*
820: * Parse out all possible fields.
821: * If the field doesn't resolve, bail.
822: */
823:
824: while (NULL != (key = strsep(&buf, ","))) {
825: if ('\0' == *key)
826: continue;
1.20 schwarze 827: for (i = 0, iterbit = 1;
828: i < mansearch_keymax;
829: i++, iterbit <<= 1) {
830: if (0 == strcasecmp(key,
831: mansearch_keynames[i])) {
832: e->bits |= iterbit;
833: break;
834: }
835: }
836: if (i == mansearch_keymax) {
837: if (strcasecmp(key, "any")) {
838: free(e);
839: return(NULL);
840: }
841: e->bits |= ~0ULL;
1.1 kristaps 842: }
843: }
844:
845: return(e);
846: }
847:
848: static void
849: exprfree(struct expr *p)
850: {
851: struct expr *pp;
852:
853: while (NULL != p) {
854: pp = p->next;
855: free(p);
856: p = pp;
857: }
858: }
859:
860: static void *
1.37 schwarze 861: hash_calloc(size_t nmemb, size_t sz, void *arg)
1.1 kristaps 862: {
863:
1.37 schwarze 864: return(mandoc_calloc(nmemb, sz));
1.1 kristaps 865: }
866:
867: static void *
868: hash_alloc(size_t sz, void *arg)
869: {
870:
871: return(mandoc_malloc(sz));
872: }
873:
874: static void
1.37 schwarze 875: hash_free(void *p, void *arg)
1.1 kristaps 876: {
877:
878: free(p);
879: }
CVSweb