Annotation of mandoc/mansearch.c, Revision 1.49
1.49 ! schwarze 1: /* $Id: mansearch.c,v 1.48 2014/09/03 18:09:14 schwarze Exp $ */
1.1 kristaps 2: /*
3: * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv>
1.13 schwarze 4: * Copyright (c) 2013, 2014 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>
24: #include <fcntl.h>
25: #include <getopt.h>
1.6 schwarze 26: #include <limits.h>
1.8 schwarze 27: #include <regex.h>
1.1 kristaps 28: #include <stdio.h>
29: #include <stdint.h>
30: #include <stddef.h>
31: #include <stdlib.h>
32: #include <string.h>
33: #include <unistd.h>
34:
1.44 schwarze 35: #if HAVE_OHASH
1.1 kristaps 36: #include <ohash.h>
1.4 kristaps 37: #else
38: #include "compat_ohash.h"
39: #endif
1.1 kristaps 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"
1.1 kristaps 47: #include "manpath.h"
48: #include "mansearch.h"
49:
1.20 schwarze 50: extern int mansearch_keymax;
51: extern const char *const mansearch_keynames[];
52:
1.3 kristaps 53: #define SQL_BIND_TEXT(_db, _s, _i, _v) \
1.8 schwarze 54: do { if (SQLITE_OK != sqlite3_bind_text \
1.2 kristaps 55: ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
1.8 schwarze 56: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
57: } while (0)
1.3 kristaps 58: #define SQL_BIND_INT64(_db, _s, _i, _v) \
1.8 schwarze 59: do { if (SQLITE_OK != sqlite3_bind_int64 \
1.2 kristaps 60: ((_s), (_i)++, (_v))) \
1.8 schwarze 61: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
62: } while (0)
63: #define SQL_BIND_BLOB(_db, _s, _i, _v) \
64: do { if (SQLITE_OK != sqlite3_bind_blob \
65: ((_s), (_i)++, (&_v), sizeof(_v), SQLITE_STATIC)) \
66: fprintf(stderr, "%s\n", sqlite3_errmsg((_db))); \
67: } while (0)
1.2 kristaps 68:
1.1 kristaps 69: struct expr {
1.38 schwarze 70: regex_t regexp; /* compiled regexp, if applicable */
71: const char *substr; /* to search for, if applicable */
72: struct expr *next; /* next in sequence */
1.34 schwarze 73: uint64_t bits; /* type-mask */
1.38 schwarze 74: int equal; /* equality, not subsring match */
1.13 schwarze 75: int open; /* opening parentheses before */
76: int and; /* logical AND before */
77: int close; /* closing parentheses after */
1.1 kristaps 78: };
79:
80: struct match {
1.32 schwarze 81: uint64_t pageid; /* identifier in database */
1.26 schwarze 82: char *desc; /* manual page description */
1.48 schwarze 83: int form; /* bit field: formatted, zipped? */
1.1 kristaps 84: };
85:
1.17 schwarze 86: static void buildnames(struct manpage *, sqlite3 *,
1.19 schwarze 87: sqlite3_stmt *, uint64_t,
88: const char *, int form);
1.12 schwarze 89: static char *buildoutput(sqlite3 *, sqlite3_stmt *,
90: uint64_t, uint64_t);
1.1 kristaps 91: static void *hash_alloc(size_t, void *);
1.37 schwarze 92: static void hash_free(void *, void *);
93: static void *hash_calloc(size_t, size_t, void *);
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 *);
1.15 schwarze 97: static struct expr *exprspec(struct expr *, uint64_t,
98: const char *, const char *);
1.8 schwarze 99: static struct expr *exprterm(const struct mansearch *, char *, int);
1.39 schwarze 100: static int manpage_compare(const void *, const void *);
1.13 schwarze 101: static void sql_append(char **sql, size_t *sz,
102: const char *newstr, int count);
1.7 schwarze 103: static void sql_match(sqlite3_context *context,
104: int argc, sqlite3_value **argv);
1.8 schwarze 105: static void sql_regexp(sqlite3_context *context,
106: int argc, sqlite3_value **argv);
1.15 schwarze 107: static char *sql_statement(const struct expr *);
1.28 schwarze 108:
1.34 schwarze 109:
1.28 schwarze 110: int
111: mansearch_setup(int start)
112: {
113: static void *pagecache;
114: int c;
115:
116: #define PC_PAGESIZE 1280
117: #define PC_NUMPAGES 256
118:
119: if (start) {
120: if (NULL != pagecache) {
121: fprintf(stderr, "pagecache already enabled\n");
122: return((int)MANDOCLEVEL_BADARG);
123: }
124:
125: pagecache = mmap(NULL, PC_PAGESIZE * PC_NUMPAGES,
1.42 schwarze 126: PROT_READ | PROT_WRITE,
127: MAP_SHARED | MAP_ANON, -1, 0);
1.28 schwarze 128:
129: if (MAP_FAILED == pagecache) {
130: perror("mmap");
131: pagecache = NULL;
132: return((int)MANDOCLEVEL_SYSERR);
133: }
134:
135: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
136: pagecache, PC_PAGESIZE, PC_NUMPAGES);
137:
138: if (SQLITE_OK == c)
139: return((int)MANDOCLEVEL_OK);
140:
141: fprintf(stderr, "pagecache: %s\n", sqlite3_errstr(c));
142:
143: } else if (NULL == pagecache) {
144: fprintf(stderr, "pagecache missing\n");
145: return((int)MANDOCLEVEL_BADARG);
146: }
147:
148: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
149: perror("munmap");
150: pagecache = NULL;
151: return((int)MANDOCLEVEL_SYSERR);
152: }
153:
154: pagecache = NULL;
155: return((int)MANDOCLEVEL_OK);
156: }
1.1 kristaps 157:
158: int
1.5 kristaps 159: mansearch(const struct mansearch *search,
1.12 schwarze 160: const struct manpaths *paths,
161: int argc, char *argv[],
1.1 kristaps 162: struct manpage **res, size_t *sz)
163: {
1.20 schwarze 164: int fd, rc, c, indexbit;
1.32 schwarze 165: int64_t pageid;
1.20 schwarze 166: uint64_t outbit, iterbit;
1.6 schwarze 167: char buf[PATH_MAX];
1.11 schwarze 168: char *sql;
1.10 schwarze 169: struct manpage *mpage;
1.1 kristaps 170: struct expr *e, *ep;
171: sqlite3 *db;
1.12 schwarze 172: sqlite3_stmt *s, *s2;
1.1 kristaps 173: struct match *mp;
174: struct ohash_info info;
175: struct ohash htab;
176: unsigned int idx;
177: size_t i, j, cur, maxres;
178:
1.37 schwarze 179: info.calloc = hash_calloc;
1.1 kristaps 180: info.alloc = hash_alloc;
1.37 schwarze 181: info.free = hash_free;
1.32 schwarze 182: info.key_offset = offsetof(struct match, pageid);
1.1 kristaps 183:
1.2 kristaps 184: *sz = cur = maxres = 0;
1.1 kristaps 185: sql = NULL;
186: *res = NULL;
187: fd = -1;
188: e = NULL;
1.2 kristaps 189: rc = 0;
1.1 kristaps 190:
191: if (0 == argc)
192: goto out;
1.5 kristaps 193: if (NULL == (e = exprcomp(search, argc, argv)))
1.1 kristaps 194: goto out;
195:
1.12 schwarze 196: outbit = 0;
1.45 schwarze 197: if (NULL != search->outkey) {
1.20 schwarze 198: for (indexbit = 0, iterbit = 1;
199: indexbit < mansearch_keymax;
200: indexbit++, iterbit <<= 1) {
1.45 schwarze 201: if (0 == strcasecmp(search->outkey,
1.20 schwarze 202: mansearch_keynames[indexbit])) {
203: outbit = iterbit;
1.12 schwarze 204: break;
205: }
206: }
207: }
208:
1.1 kristaps 209: /*
210: * Save a descriptor to the current working directory.
211: * Since pathnames in the "paths" variable might be relative,
212: * and we'll be chdir()ing into them, we need to keep a handle
213: * on our current directory from which to start the chdir().
214: */
215:
1.6 schwarze 216: if (NULL == getcwd(buf, PATH_MAX)) {
1.30 schwarze 217: perror("getcwd");
1.1 kristaps 218: goto out;
219: } else if (-1 == (fd = open(buf, O_RDONLY, 0))) {
220: perror(buf);
221: goto out;
222: }
223:
1.15 schwarze 224: sql = sql_statement(e);
1.1 kristaps 225:
226: /*
227: * Loop over the directories (containing databases) for us to
228: * search.
229: * Don't let missing/bad databases/directories phase us.
230: * In each, try to open the resident database and, if it opens,
231: * scan it for our match expression.
232: */
233:
234: for (i = 0; i < paths->sz; i++) {
235: if (-1 == fchdir(fd)) {
236: perror(buf);
237: free(*res);
238: break;
239: } else if (-1 == chdir(paths->paths[i])) {
240: perror(paths->paths[i]);
241: continue;
1.34 schwarze 242: }
1.1 kristaps 243:
1.34 schwarze 244: c = sqlite3_open_v2(MANDOC_DB, &db,
245: SQLITE_OPEN_READONLY, NULL);
1.1 kristaps 246:
1.2 kristaps 247: if (SQLITE_OK != c) {
1.1 kristaps 248: perror(MANDOC_DB);
249: sqlite3_close(db);
250: continue;
251: }
252:
1.8 schwarze 253: /*
254: * Define the SQL functions for substring
255: * and regular expression matching.
256: */
1.7 schwarze 257:
258: c = sqlite3_create_function(db, "match", 2,
1.31 schwarze 259: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
260: NULL, sql_match, NULL, NULL);
1.8 schwarze 261: assert(SQLITE_OK == c);
262: c = sqlite3_create_function(db, "regexp", 2,
1.31 schwarze 263: SQLITE_UTF8 | SQLITE_DETERMINISTIC,
264: NULL, sql_regexp, NULL, NULL);
1.8 schwarze 265: assert(SQLITE_OK == c);
1.7 schwarze 266:
1.1 kristaps 267: j = 1;
1.2 kristaps 268: c = sqlite3_prepare_v2(db, sql, -1, &s, NULL);
269: if (SQLITE_OK != c)
270: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
1.1 kristaps 271:
272: for (ep = e; NULL != ep; ep = ep->next) {
1.8 schwarze 273: if (NULL == ep->substr) {
274: SQL_BIND_BLOB(db, s, j, ep->regexp);
275: } else
276: SQL_BIND_TEXT(db, s, j, ep->substr);
1.27 schwarze 277: if (0 == ((TYPE_Nd | TYPE_Nm) & ep->bits))
1.26 schwarze 278: SQL_BIND_INT64(db, s, j, ep->bits);
1.1 kristaps 279: }
280:
281: memset(&htab, 0, sizeof(struct ohash));
282: ohash_init(&htab, 4, &info);
283:
284: /*
285: * Hash each entry on its [unique] document identifier.
286: * This is a uint64_t.
287: * Instead of using a hash function, simply convert the
288: * uint64_t to a uint32_t, the hash value's type.
289: * This gives good performance and preserves the
290: * distribution of buckets in the table.
291: */
1.2 kristaps 292: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.32 schwarze 293: pageid = sqlite3_column_int64(s, 2);
1.34 schwarze 294: idx = ohash_lookup_memory(&htab,
295: (char *)&pageid, sizeof(uint64_t),
296: (uint32_t)pageid);
1.1 kristaps 297:
298: if (NULL != ohash_find(&htab, idx))
299: continue;
300:
301: mp = mandoc_calloc(1, sizeof(struct match));
1.32 schwarze 302: mp->pageid = pageid;
1.26 schwarze 303: mp->form = sqlite3_column_int(s, 1);
304: if (TYPE_Nd == outbit)
1.41 schwarze 305: mp->desc = mandoc_strdup((const char *)
1.26 schwarze 306: sqlite3_column_text(s, 0));
1.1 kristaps 307: ohash_insert(&htab, idx, mp);
308: }
309:
1.2 kristaps 310: if (SQLITE_DONE != c)
311: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
312:
1.1 kristaps 313: sqlite3_finalize(s);
1.10 schwarze 314:
1.34 schwarze 315: c = sqlite3_prepare_v2(db,
1.35 schwarze 316: "SELECT sec, arch, name, pageid FROM mlinks "
317: "WHERE pageid=? ORDER BY sec, arch, name",
1.10 schwarze 318: -1, &s, NULL);
319: if (SQLITE_OK != c)
320: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
1.1 kristaps 321:
1.12 schwarze 322: c = sqlite3_prepare_v2(db,
1.35 schwarze 323: "SELECT bits, key, pageid FROM keys "
324: "WHERE pageid=? AND bits & ?",
1.12 schwarze 325: -1, &s2, NULL);
326: if (SQLITE_OK != c)
327: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
328:
1.1 kristaps 329: for (mp = ohash_first(&htab, &idx);
330: NULL != mp;
331: mp = ohash_next(&htab, &idx)) {
332: if (cur + 1 > maxres) {
333: maxres += 1024;
1.36 schwarze 334: *res = mandoc_reallocarray(*res,
335: maxres, sizeof(struct manpage));
1.1 kristaps 336: }
1.10 schwarze 337: mpage = *res + cur;
1.47 schwarze 338: mpage->ipath = i;
1.39 schwarze 339: mpage->sec = 10;
1.10 schwarze 340: mpage->form = mp->form;
1.32 schwarze 341: buildnames(mpage, db, s, mp->pageid,
1.19 schwarze 342: paths->paths[i], mp->form);
1.26 schwarze 343: mpage->output = TYPE_Nd & outbit ?
344: mp->desc : outbit ?
1.32 schwarze 345: buildoutput(db, s2, mp->pageid, outbit) : NULL;
1.10 schwarze 346:
1.1 kristaps 347: free(mp);
348: cur++;
349: }
1.10 schwarze 350:
351: sqlite3_finalize(s);
1.12 schwarze 352: sqlite3_finalize(s2);
1.10 schwarze 353: sqlite3_close(db);
1.1 kristaps 354: ohash_delete(&htab);
1.49 ! schwarze 355:
! 356: /*
! 357: * In man(1) mode, prefer matches in earlier trees
! 358: * over matches in later trees.
! 359: */
! 360:
! 361: if (cur && search->firstmatch)
! 362: break;
1.1 kristaps 363: }
1.39 schwarze 364: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.2 kristaps 365: rc = 1;
1.1 kristaps 366: out:
1.30 schwarze 367: if (-1 != fd) {
368: if (-1 == fchdir(fd))
369: perror(buf);
370: close(fd);
371: }
1.1 kristaps 372: exprfree(e);
373: free(sql);
374: *sz = cur;
1.2 kristaps 375: return(rc);
1.11 schwarze 376: }
377:
1.45 schwarze 378: void
379: mansearch_free(struct manpage *res, size_t sz)
380: {
381: size_t i;
382:
383: for (i = 0; i < sz; i++) {
384: free(res[i].file);
385: free(res[i].names);
386: free(res[i].output);
387: }
388: free(res);
389: }
390:
1.39 schwarze 391: static int
392: manpage_compare(const void *vp1, const void *vp2)
393: {
394: const struct manpage *mp1, *mp2;
395: int diff;
396:
397: mp1 = vp1;
398: mp2 = vp2;
399: diff = mp1->sec - mp2->sec;
400: return(diff ? diff : strcasecmp(mp1->names, mp2->names));
401: }
402:
1.17 schwarze 403: static void
404: buildnames(struct manpage *mpage, sqlite3 *db, sqlite3_stmt *s,
1.32 schwarze 405: uint64_t pageid, const char *path, int form)
1.11 schwarze 406: {
1.22 schwarze 407: char *newnames, *prevsec, *prevarch;
1.19 schwarze 408: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.48 schwarze 409: const char *gzip;
1.11 schwarze 410: size_t i;
411: int c;
412:
1.25 schwarze 413: mpage->file = NULL;
1.17 schwarze 414: mpage->names = NULL;
1.22 schwarze 415: prevsec = prevarch = NULL;
1.11 schwarze 416: i = 1;
1.32 schwarze 417: SQL_BIND_INT64(db, s, i, pageid);
1.11 schwarze 418: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.17 schwarze 419:
1.22 schwarze 420: /* Decide whether we already have some names. */
1.17 schwarze 421:
422: if (NULL == mpage->names) {
1.11 schwarze 423: oldnames = "";
424: sep1 = "";
425: } else {
1.17 schwarze 426: oldnames = mpage->names;
1.11 schwarze 427: sep1 = ", ";
428: }
1.22 schwarze 429:
430: /* Fetch the next name. */
431:
1.41 schwarze 432: sec = (const char *)sqlite3_column_text(s, 0);
433: arch = (const char *)sqlite3_column_text(s, 1);
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:
472: if (NULL != mpage->file)
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: }
1.48 schwarze 482: if (form & FORM_GZ)
483: gzip = ".gz";
484: else
485: gzip = "";
1.22 schwarze 486: sep2 = '\0' == *arch ? "" : "/";
1.48 schwarze 487: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s%s",
488: path, sep1, sec, sep2, arch, name, fsec, gzip);
1.11 schwarze 489: }
490: if (SQLITE_DONE != c)
491: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
492: sqlite3_reset(s);
1.22 schwarze 493:
494: /* Append one final section to the names. */
495:
496: if (NULL != prevsec) {
497: sep2 = '\0' == *prevarch ? "" : "/";
1.24 schwarze 498: mandoc_asprintf(&newnames, "%s(%s%s%s)",
499: mpage->names, prevsec, sep2, prevarch);
1.22 schwarze 500: free(mpage->names);
501: mpage->names = newnames;
502: free(prevsec);
503: free(prevarch);
504: }
1.12 schwarze 505: }
506:
507: static char *
1.32 schwarze 508: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.12 schwarze 509: {
510: char *output, *newoutput;
511: const char *oldoutput, *sep1, *data;
512: size_t i;
513: int c;
514:
515: output = NULL;
516: i = 1;
1.32 schwarze 517: SQL_BIND_INT64(db, s, i, pageid);
1.12 schwarze 518: SQL_BIND_INT64(db, s, i, outbit);
519: while (SQLITE_ROW == (c = sqlite3_step(s))) {
520: if (NULL == output) {
521: oldoutput = "";
522: sep1 = "";
523: } else {
524: oldoutput = output;
525: sep1 = " # ";
526: }
1.41 schwarze 527: data = (const char *)sqlite3_column_text(s, 1);
1.24 schwarze 528: mandoc_asprintf(&newoutput, "%s%s%s",
529: oldoutput, sep1, data);
1.12 schwarze 530: free(output);
531: output = newoutput;
532: }
533: if (SQLITE_DONE != c)
534: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
535: sqlite3_reset(s);
536: return(output);
1.1 kristaps 537: }
538:
539: /*
1.7 schwarze 540: * Implement substring match as an application-defined SQL function.
541: * Using the SQL LIKE or GLOB operators instead would be a bad idea
542: * because that would require escaping metacharacters in the string
543: * being searched for.
544: */
545: static void
546: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
547: {
548:
549: assert(2 == argc);
550: sqlite3_result_int(context, NULL != strcasestr(
551: (const char *)sqlite3_value_text(argv[1]),
552: (const char *)sqlite3_value_text(argv[0])));
553: }
554:
555: /*
1.8 schwarze 556: * Implement regular expression match
557: * as an application-defined SQL function.
558: */
559: static void
560: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
561: {
562:
563: assert(2 == argc);
564: sqlite3_result_int(context, !regexec(
565: (regex_t *)sqlite3_value_blob(argv[0]),
566: (const char *)sqlite3_value_text(argv[1]),
567: 0, NULL, 0));
568: }
569:
1.13 schwarze 570: static void
571: sql_append(char **sql, size_t *sz, const char *newstr, int count)
572: {
573: size_t newsz;
574:
575: newsz = 1 < count ? (size_t)count : strlen(newstr);
576: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
577: if (1 < count)
578: memset(*sql + *sz, *newstr, (size_t)count);
579: else
580: memcpy(*sql + *sz, newstr, newsz);
581: *sz += newsz;
582: (*sql)[*sz] = '\0';
583: }
584:
1.8 schwarze 585: /*
1.1 kristaps 586: * Prepare the search SQL statement.
587: */
588: static char *
1.15 schwarze 589: sql_statement(const struct expr *e)
1.1 kristaps 590: {
591: char *sql;
592: size_t sz;
1.13 schwarze 593: int needop;
1.1 kristaps 594:
1.35 schwarze 595: sql = mandoc_strdup(
596: "SELECT desc, form, pageid FROM mpages WHERE ");
1.1 kristaps 597: sz = strlen(sql);
1.2 kristaps 598:
1.13 schwarze 599: for (needop = 0; NULL != e; e = e->next) {
600: if (e->and)
601: sql_append(&sql, &sz, " AND ", 1);
602: else if (needop)
603: sql_append(&sql, &sz, " OR ", 1);
604: if (e->open)
605: sql_append(&sql, &sz, "(", e->open);
1.26 schwarze 606: sql_append(&sql, &sz,
607: TYPE_Nd & e->bits
608: ? (NULL == e->substr
609: ? "desc REGEXP ?"
610: : "desc MATCH ?")
1.27 schwarze 611: : TYPE_Nm == e->bits
612: ? (NULL == e->substr
1.32 schwarze 613: ? "pageid IN (SELECT pageid FROM names "
1.27 schwarze 614: "WHERE name REGEXP ?)"
1.38 schwarze 615: : e->equal
616: ? "pageid IN (SELECT pageid FROM names "
617: "WHERE name = ?)"
1.32 schwarze 618: : "pageid IN (SELECT pageid FROM names "
1.27 schwarze 619: "WHERE name MATCH ?)")
1.26 schwarze 620: : (NULL == e->substr
1.32 schwarze 621: ? "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 622: "WHERE key REGEXP ? AND bits & ?)"
1.32 schwarze 623: : "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 624: "WHERE key MATCH ? AND bits & ?)"), 1);
1.13 schwarze 625: if (e->close)
626: sql_append(&sql, &sz, ")", e->close);
627: needop = 1;
1.1 kristaps 628: }
629:
630: return(sql);
631: }
632:
633: /*
634: * Compile a set of string tokens into an expression.
635: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
636: * "(", "foo=bar", etc.).
637: */
638: static struct expr *
1.5 kristaps 639: exprcomp(const struct mansearch *search, int argc, char *argv[])
1.1 kristaps 640: {
1.27 schwarze 641: uint64_t mask;
1.13 schwarze 642: int i, toopen, logic, igncase, toclose;
1.27 schwarze 643: struct expr *first, *prev, *cur, *next;
1.1 kristaps 644:
645: first = cur = NULL;
1.15 schwarze 646: logic = igncase = toclose = 0;
1.33 schwarze 647: toopen = NULL != search->sec || NULL != search->arch;
1.1 kristaps 648:
649: for (i = 0; i < argc; i++) {
1.13 schwarze 650: if (0 == strcmp("(", argv[i])) {
651: if (igncase)
652: goto fail;
653: toopen++;
654: toclose++;
655: continue;
656: } else if (0 == strcmp(")", argv[i])) {
657: if (toopen || logic || igncase || NULL == cur)
658: goto fail;
659: cur->close++;
660: if (0 > --toclose)
661: goto fail;
662: continue;
663: } else if (0 == strcmp("-a", argv[i])) {
664: if (toopen || logic || igncase || NULL == cur)
665: goto fail;
666: logic = 1;
667: continue;
668: } else if (0 == strcmp("-o", argv[i])) {
669: if (toopen || logic || igncase || NULL == cur)
670: goto fail;
671: logic = 2;
672: continue;
673: } else if (0 == strcmp("-i", argv[i])) {
674: if (igncase)
675: goto fail;
676: igncase = 1;
677: continue;
1.1 kristaps 678: }
1.13 schwarze 679: next = exprterm(search, argv[i], !igncase);
680: if (NULL == next)
681: goto fail;
1.26 schwarze 682: if (NULL == first)
683: first = next;
684: else
1.1 kristaps 685: cur->next = next;
1.27 schwarze 686: prev = cur = next;
1.26 schwarze 687:
688: /*
689: * Searching for descriptions must be split out
690: * because they are stored in the mpages table,
691: * not in the keys table.
692: */
693:
1.27 schwarze 694: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
695: if (mask & cur->bits && ~mask & cur->bits) {
696: next = mandoc_calloc(1,
697: sizeof(struct expr));
698: memcpy(next, cur, sizeof(struct expr));
699: prev->open = 1;
700: cur->bits = mask;
701: cur->next = next;
702: cur = next;
703: cur->bits &= ~mask;
704: }
705: }
706: prev->and = (1 == logic);
707: prev->open += toopen;
708: if (cur != prev)
1.26 schwarze 709: cur->close = 1;
1.27 schwarze 710:
1.13 schwarze 711: toopen = logic = igncase = 0;
1.1 kristaps 712: }
1.15 schwarze 713: if (toopen || logic || igncase || toclose)
714: goto fail;
715:
1.33 schwarze 716: if (NULL != search->sec || NULL != search->arch)
717: cur->close++;
718: if (NULL != search->arch)
719: cur = exprspec(cur, TYPE_arch, search->arch, "^(%s|any)$");
720: if (NULL != search->sec)
721: exprspec(cur, TYPE_sec, search->sec, "^%s$");
1.15 schwarze 722:
723: return(first);
724:
1.13 schwarze 725: fail:
726: if (NULL != first)
727: exprfree(first);
728: return(NULL);
1.1 kristaps 729: }
730:
731: static struct expr *
1.15 schwarze 732: exprspec(struct expr *cur, uint64_t key, const char *value,
733: const char *format)
734: {
735: char errbuf[BUFSIZ];
736: char *cp;
737: int irc;
738:
1.24 schwarze 739: mandoc_asprintf(&cp, format, value);
1.15 schwarze 740: cur->next = mandoc_calloc(1, sizeof(struct expr));
741: cur = cur->next;
742: cur->and = 1;
743: cur->bits = key;
744: if (0 != (irc = regcomp(&cur->regexp, cp,
745: REG_EXTENDED | REG_NOSUB | REG_ICASE))) {
746: regerror(irc, &cur->regexp, errbuf, sizeof(errbuf));
747: fprintf(stderr, "regcomp: %s\n", errbuf);
748: cur->substr = value;
749: }
750: free(cp);
751: return(cur);
752: }
753:
754: static struct expr *
1.8 schwarze 755: exprterm(const struct mansearch *search, char *buf, int cs)
1.1 kristaps 756: {
1.15 schwarze 757: char errbuf[BUFSIZ];
1.1 kristaps 758: struct expr *e;
1.38 schwarze 759: char *key, *val;
1.20 schwarze 760: uint64_t iterbit;
761: int i, irc;
1.1 kristaps 762:
763: if ('\0' == *buf)
764: return(NULL);
765:
766: e = mandoc_calloc(1, sizeof(struct expr));
767:
1.45 schwarze 768: if (search->argmode == ARG_NAME) {
769: e->bits = TYPE_Nm;
1.8 schwarze 770: e->substr = buf;
1.38 schwarze 771: e->equal = 1;
1.5 kristaps 772: return(e);
773: }
774:
1.1 kristaps 775: /*
1.45 schwarze 776: * Separate macro keys from search string.
777: * If needed, request regular expression handling
778: * by setting e->substr to NULL.
1.1 kristaps 779: */
780:
1.45 schwarze 781: if (search->argmode == ARG_WORD) {
782: e->bits = TYPE_Nm;
783: e->substr = NULL;
784: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
1.46 schwarze 785: cs = 0;
1.45 schwarze 786: } else if ((val = strpbrk(buf, "=~")) == NULL) {
787: e->bits = TYPE_Nm | TYPE_Nd;
1.8 schwarze 788: e->substr = buf;
1.38 schwarze 789: } else {
790: if (val == buf)
1.45 schwarze 791: e->bits = TYPE_Nm | TYPE_Nd;
1.38 schwarze 792: if ('=' == *val)
793: e->substr = val + 1;
794: *val++ = '\0';
1.21 schwarze 795: if (NULL != strstr(buf, "arch"))
796: cs = 0;
1.38 schwarze 797: }
798:
799: /* Compile regular expressions. */
800:
801: if (NULL == e->substr) {
802: irc = regcomp(&e->regexp, val,
803: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
1.45 schwarze 804: if (search->argmode == ARG_WORD)
1.38 schwarze 805: free(val);
806: if (irc) {
1.15 schwarze 807: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
808: fprintf(stderr, "regcomp: %s\n", errbuf);
1.8 schwarze 809: free(e);
810: return(NULL);
811: }
1.38 schwarze 812: }
813:
814: if (e->bits)
815: return(e);
1.1 kristaps 816:
817: /*
818: * Parse out all possible fields.
819: * If the field doesn't resolve, bail.
820: */
821:
822: while (NULL != (key = strsep(&buf, ","))) {
823: if ('\0' == *key)
824: continue;
1.20 schwarze 825: for (i = 0, iterbit = 1;
826: i < mansearch_keymax;
827: i++, iterbit <<= 1) {
828: if (0 == strcasecmp(key,
829: mansearch_keynames[i])) {
830: e->bits |= iterbit;
831: break;
832: }
833: }
834: if (i == mansearch_keymax) {
835: if (strcasecmp(key, "any")) {
836: free(e);
837: return(NULL);
838: }
839: e->bits |= ~0ULL;
1.1 kristaps 840: }
841: }
842:
843: return(e);
844: }
845:
846: static void
847: exprfree(struct expr *p)
848: {
849: struct expr *pp;
850:
851: while (NULL != p) {
852: pp = p->next;
853: free(p);
854: p = pp;
855: }
856: }
857:
858: static void *
1.37 schwarze 859: hash_calloc(size_t nmemb, size_t sz, void *arg)
1.1 kristaps 860: {
861:
1.37 schwarze 862: return(mandoc_calloc(nmemb, sz));
1.1 kristaps 863: }
864:
865: static void *
866: hash_alloc(size_t sz, void *arg)
867: {
868:
869: return(mandoc_malloc(sz));
870: }
871:
872: static void
1.37 schwarze 873: hash_free(void *p, void *arg)
1.1 kristaps 874: {
875:
876: free(p);
877: }
CVSweb