Annotation of mandoc/mansearch.c, Revision 1.40
1.40 ! schwarze 1: /* $Id: mansearch.c,v 1.39 2014/07/24 20:30:45 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: #ifdef HAVE_CONFIG_H
19: #include "config.h"
20: #endif
21:
1.28 schwarze 22: #include <sys/mman.h>
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:
1.4 kristaps 35: #ifdef 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 */
1.1 kristaps 83: int form; /* 0 == catpage */
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,
126: PROT_READ | PROT_WRITE, MAP_ANON, -1, 0);
127:
128: if (MAP_FAILED == pagecache) {
129: perror("mmap");
130: pagecache = NULL;
131: return((int)MANDOCLEVEL_SYSERR);
132: }
133:
134: c = sqlite3_config(SQLITE_CONFIG_PAGECACHE,
135: pagecache, PC_PAGESIZE, PC_NUMPAGES);
136:
137: if (SQLITE_OK == c)
138: return((int)MANDOCLEVEL_OK);
139:
140: fprintf(stderr, "pagecache: %s\n", sqlite3_errstr(c));
141:
142: } else if (NULL == pagecache) {
143: fprintf(stderr, "pagecache missing\n");
144: return((int)MANDOCLEVEL_BADARG);
145: }
146:
147: if (-1 == munmap(pagecache, PC_PAGESIZE * PC_NUMPAGES)) {
148: perror("munmap");
149: pagecache = NULL;
150: return((int)MANDOCLEVEL_SYSERR);
151: }
152:
153: pagecache = NULL;
154: return((int)MANDOCLEVEL_OK);
155: }
1.1 kristaps 156:
157: int
1.5 kristaps 158: mansearch(const struct mansearch *search,
1.12 schwarze 159: const struct manpaths *paths,
160: int argc, char *argv[],
161: const char *outkey,
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;
197: if (NULL != outkey) {
1.20 schwarze 198: for (indexbit = 0, iterbit = 1;
199: indexbit < mansearch_keymax;
200: indexbit++, iterbit <<= 1) {
201: if (0 == strcasecmp(outkey,
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)
305: mp->desc = mandoc_strdup(
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;
1.39 schwarze 338: mpage->sec = 10;
1.10 schwarze 339: mpage->form = mp->form;
1.32 schwarze 340: buildnames(mpage, db, s, mp->pageid,
1.19 schwarze 341: paths->paths[i], mp->form);
1.26 schwarze 342: mpage->output = TYPE_Nd & outbit ?
343: mp->desc : outbit ?
1.32 schwarze 344: buildoutput(db, s2, mp->pageid, outbit) : NULL;
1.10 schwarze 345:
1.1 kristaps 346: free(mp);
347: cur++;
348: }
1.10 schwarze 349:
350: sqlite3_finalize(s);
1.12 schwarze 351: sqlite3_finalize(s2);
1.10 schwarze 352: sqlite3_close(db);
1.1 kristaps 353: ohash_delete(&htab);
354: }
1.39 schwarze 355: qsort(*res, cur, sizeof(struct manpage), manpage_compare);
1.2 kristaps 356: rc = 1;
1.1 kristaps 357: out:
1.30 schwarze 358: if (-1 != fd) {
359: if (-1 == fchdir(fd))
360: perror(buf);
361: close(fd);
362: }
1.1 kristaps 363: exprfree(e);
364: free(sql);
365: *sz = cur;
1.2 kristaps 366: return(rc);
1.11 schwarze 367: }
368:
1.39 schwarze 369: static int
370: manpage_compare(const void *vp1, const void *vp2)
371: {
372: const struct manpage *mp1, *mp2;
373: int diff;
374:
375: mp1 = vp1;
376: mp2 = vp2;
377: diff = mp1->sec - mp2->sec;
378: return(diff ? diff : strcasecmp(mp1->names, mp2->names));
379: }
380:
1.17 schwarze 381: static void
382: buildnames(struct manpage *mpage, sqlite3 *db, sqlite3_stmt *s,
1.32 schwarze 383: uint64_t pageid, const char *path, int form)
1.11 schwarze 384: {
1.22 schwarze 385: char *newnames, *prevsec, *prevarch;
1.19 schwarze 386: const char *oldnames, *sep1, *name, *sec, *sep2, *arch, *fsec;
1.11 schwarze 387: size_t i;
388: int c;
389:
1.25 schwarze 390: mpage->file = NULL;
1.17 schwarze 391: mpage->names = NULL;
1.22 schwarze 392: prevsec = prevarch = NULL;
1.11 schwarze 393: i = 1;
1.32 schwarze 394: SQL_BIND_INT64(db, s, i, pageid);
1.11 schwarze 395: while (SQLITE_ROW == (c = sqlite3_step(s))) {
1.17 schwarze 396:
1.22 schwarze 397: /* Decide whether we already have some names. */
1.17 schwarze 398:
399: if (NULL == mpage->names) {
1.11 schwarze 400: oldnames = "";
401: sep1 = "";
402: } else {
1.17 schwarze 403: oldnames = mpage->names;
1.11 schwarze 404: sep1 = ", ";
405: }
1.22 schwarze 406:
407: /* Fetch the next name. */
408:
1.19 schwarze 409: sec = sqlite3_column_text(s, 0);
410: arch = sqlite3_column_text(s, 1);
411: name = sqlite3_column_text(s, 2);
1.39 schwarze 412:
413: /* Remember the first section found. */
414:
415: if (9 < mpage->sec && '1' <= *sec && '9' >= *sec)
416: mpage->sec = (*sec - '1') + 1;
1.22 schwarze 417:
418: /* If the section changed, append the old one. */
419:
420: if (NULL != prevsec &&
421: (strcmp(sec, prevsec) ||
422: strcmp(arch, prevarch))) {
423: sep2 = '\0' == *prevarch ? "" : "/";
1.24 schwarze 424: mandoc_asprintf(&newnames, "%s(%s%s%s)",
425: oldnames, prevsec, sep2, prevarch);
1.22 schwarze 426: free(mpage->names);
427: oldnames = mpage->names = newnames;
428: free(prevsec);
429: free(prevarch);
430: prevsec = prevarch = NULL;
431: }
432:
433: /* Save the new section, to append it later. */
434:
435: if (NULL == prevsec) {
436: prevsec = mandoc_strdup(sec);
437: prevarch = mandoc_strdup(arch);
438: }
439:
440: /* Append the new name. */
441:
1.24 schwarze 442: mandoc_asprintf(&newnames, "%s%s%s",
443: oldnames, sep1, name);
1.17 schwarze 444: free(mpage->names);
445: mpage->names = newnames;
446:
447: /* Also save the first file name encountered. */
448:
449: if (NULL != mpage->file)
450: continue;
451:
1.19 schwarze 452: if (form) {
453: sep1 = "man";
454: fsec = sec;
455: } else {
456: sep1 = "cat";
457: fsec = "0";
458: }
1.22 schwarze 459: sep2 = '\0' == *arch ? "" : "/";
1.24 schwarze 460: mandoc_asprintf(&mpage->file, "%s/%s%s%s%s/%s.%s",
461: path, sep1, sec, sep2, arch, name, fsec);
1.11 schwarze 462: }
463: if (SQLITE_DONE != c)
464: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
465: sqlite3_reset(s);
1.22 schwarze 466:
467: /* Append one final section to the names. */
468:
469: if (NULL != prevsec) {
470: sep2 = '\0' == *prevarch ? "" : "/";
1.24 schwarze 471: mandoc_asprintf(&newnames, "%s(%s%s%s)",
472: mpage->names, prevsec, sep2, prevarch);
1.22 schwarze 473: free(mpage->names);
474: mpage->names = newnames;
475: free(prevsec);
476: free(prevarch);
477: }
1.12 schwarze 478: }
479:
480: static char *
1.32 schwarze 481: buildoutput(sqlite3 *db, sqlite3_stmt *s, uint64_t pageid, uint64_t outbit)
1.12 schwarze 482: {
483: char *output, *newoutput;
484: const char *oldoutput, *sep1, *data;
485: size_t i;
486: int c;
487:
488: output = NULL;
489: i = 1;
1.32 schwarze 490: SQL_BIND_INT64(db, s, i, pageid);
1.12 schwarze 491: SQL_BIND_INT64(db, s, i, outbit);
492: while (SQLITE_ROW == (c = sqlite3_step(s))) {
493: if (NULL == output) {
494: oldoutput = "";
495: sep1 = "";
496: } else {
497: oldoutput = output;
498: sep1 = " # ";
499: }
500: data = sqlite3_column_text(s, 1);
1.24 schwarze 501: mandoc_asprintf(&newoutput, "%s%s%s",
502: oldoutput, sep1, data);
1.12 schwarze 503: free(output);
504: output = newoutput;
505: }
506: if (SQLITE_DONE != c)
507: fprintf(stderr, "%s\n", sqlite3_errmsg(db));
508: sqlite3_reset(s);
509: return(output);
1.1 kristaps 510: }
511:
512: /*
1.7 schwarze 513: * Implement substring match as an application-defined SQL function.
514: * Using the SQL LIKE or GLOB operators instead would be a bad idea
515: * because that would require escaping metacharacters in the string
516: * being searched for.
517: */
518: static void
519: sql_match(sqlite3_context *context, int argc, sqlite3_value **argv)
520: {
521:
522: assert(2 == argc);
523: sqlite3_result_int(context, NULL != strcasestr(
524: (const char *)sqlite3_value_text(argv[1]),
525: (const char *)sqlite3_value_text(argv[0])));
526: }
527:
528: /*
1.8 schwarze 529: * Implement regular expression match
530: * as an application-defined SQL function.
531: */
532: static void
533: sql_regexp(sqlite3_context *context, int argc, sqlite3_value **argv)
534: {
535:
536: assert(2 == argc);
537: sqlite3_result_int(context, !regexec(
538: (regex_t *)sqlite3_value_blob(argv[0]),
539: (const char *)sqlite3_value_text(argv[1]),
540: 0, NULL, 0));
541: }
542:
1.13 schwarze 543: static void
544: sql_append(char **sql, size_t *sz, const char *newstr, int count)
545: {
546: size_t newsz;
547:
548: newsz = 1 < count ? (size_t)count : strlen(newstr);
549: *sql = mandoc_realloc(*sql, *sz + newsz + 1);
550: if (1 < count)
551: memset(*sql + *sz, *newstr, (size_t)count);
552: else
553: memcpy(*sql + *sz, newstr, newsz);
554: *sz += newsz;
555: (*sql)[*sz] = '\0';
556: }
557:
1.8 schwarze 558: /*
1.1 kristaps 559: * Prepare the search SQL statement.
560: */
561: static char *
1.15 schwarze 562: sql_statement(const struct expr *e)
1.1 kristaps 563: {
564: char *sql;
565: size_t sz;
1.13 schwarze 566: int needop;
1.1 kristaps 567:
1.35 schwarze 568: sql = mandoc_strdup(
569: "SELECT desc, form, pageid FROM mpages WHERE ");
1.1 kristaps 570: sz = strlen(sql);
1.2 kristaps 571:
1.13 schwarze 572: for (needop = 0; NULL != e; e = e->next) {
573: if (e->and)
574: sql_append(&sql, &sz, " AND ", 1);
575: else if (needop)
576: sql_append(&sql, &sz, " OR ", 1);
577: if (e->open)
578: sql_append(&sql, &sz, "(", e->open);
1.26 schwarze 579: sql_append(&sql, &sz,
580: TYPE_Nd & e->bits
581: ? (NULL == e->substr
582: ? "desc REGEXP ?"
583: : "desc MATCH ?")
1.27 schwarze 584: : TYPE_Nm == e->bits
585: ? (NULL == e->substr
1.32 schwarze 586: ? "pageid IN (SELECT pageid FROM names "
1.27 schwarze 587: "WHERE name REGEXP ?)"
1.38 schwarze 588: : e->equal
589: ? "pageid IN (SELECT pageid FROM names "
590: "WHERE name = ?)"
1.32 schwarze 591: : "pageid IN (SELECT pageid FROM names "
1.27 schwarze 592: "WHERE name MATCH ?)")
1.26 schwarze 593: : (NULL == e->substr
1.32 schwarze 594: ? "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 595: "WHERE key REGEXP ? AND bits & ?)"
1.32 schwarze 596: : "pageid IN (SELECT pageid FROM keys "
1.26 schwarze 597: "WHERE key MATCH ? AND bits & ?)"), 1);
1.13 schwarze 598: if (e->close)
599: sql_append(&sql, &sz, ")", e->close);
600: needop = 1;
1.1 kristaps 601: }
602:
603: return(sql);
604: }
605:
606: /*
607: * Compile a set of string tokens into an expression.
608: * Tokens in "argv" are assumed to be individual expression atoms (e.g.,
609: * "(", "foo=bar", etc.).
610: */
611: static struct expr *
1.5 kristaps 612: exprcomp(const struct mansearch *search, int argc, char *argv[])
1.1 kristaps 613: {
1.27 schwarze 614: uint64_t mask;
1.13 schwarze 615: int i, toopen, logic, igncase, toclose;
1.27 schwarze 616: struct expr *first, *prev, *cur, *next;
1.1 kristaps 617:
618: first = cur = NULL;
1.15 schwarze 619: logic = igncase = toclose = 0;
1.33 schwarze 620: toopen = NULL != search->sec || NULL != search->arch;
1.1 kristaps 621:
622: for (i = 0; i < argc; i++) {
1.13 schwarze 623: if (0 == strcmp("(", argv[i])) {
624: if (igncase)
625: goto fail;
626: toopen++;
627: toclose++;
628: continue;
629: } else if (0 == strcmp(")", argv[i])) {
630: if (toopen || logic || igncase || NULL == cur)
631: goto fail;
632: cur->close++;
633: if (0 > --toclose)
634: goto fail;
635: continue;
636: } else if (0 == strcmp("-a", argv[i])) {
637: if (toopen || logic || igncase || NULL == cur)
638: goto fail;
639: logic = 1;
640: continue;
641: } else if (0 == strcmp("-o", argv[i])) {
642: if (toopen || logic || igncase || NULL == cur)
643: goto fail;
644: logic = 2;
645: continue;
646: } else if (0 == strcmp("-i", argv[i])) {
647: if (igncase)
648: goto fail;
649: igncase = 1;
650: continue;
1.1 kristaps 651: }
1.13 schwarze 652: next = exprterm(search, argv[i], !igncase);
653: if (NULL == next)
654: goto fail;
1.26 schwarze 655: if (NULL == first)
656: first = next;
657: else
1.1 kristaps 658: cur->next = next;
1.27 schwarze 659: prev = cur = next;
1.26 schwarze 660:
661: /*
662: * Searching for descriptions must be split out
663: * because they are stored in the mpages table,
664: * not in the keys table.
665: */
666:
1.27 schwarze 667: for (mask = TYPE_Nm; mask <= TYPE_Nd; mask <<= 1) {
668: if (mask & cur->bits && ~mask & cur->bits) {
669: next = mandoc_calloc(1,
670: sizeof(struct expr));
671: memcpy(next, cur, sizeof(struct expr));
672: prev->open = 1;
673: cur->bits = mask;
674: cur->next = next;
675: cur = next;
676: cur->bits &= ~mask;
677: }
678: }
679: prev->and = (1 == logic);
680: prev->open += toopen;
681: if (cur != prev)
1.26 schwarze 682: cur->close = 1;
1.27 schwarze 683:
1.13 schwarze 684: toopen = logic = igncase = 0;
1.1 kristaps 685: }
1.15 schwarze 686: if (toopen || logic || igncase || toclose)
687: goto fail;
688:
1.33 schwarze 689: if (NULL != search->sec || NULL != search->arch)
690: cur->close++;
691: if (NULL != search->arch)
692: cur = exprspec(cur, TYPE_arch, search->arch, "^(%s|any)$");
693: if (NULL != search->sec)
694: exprspec(cur, TYPE_sec, search->sec, "^%s$");
1.15 schwarze 695:
696: return(first);
697:
1.13 schwarze 698: fail:
699: if (NULL != first)
700: exprfree(first);
701: return(NULL);
1.1 kristaps 702: }
703:
704: static struct expr *
1.15 schwarze 705: exprspec(struct expr *cur, uint64_t key, const char *value,
706: const char *format)
707: {
708: char errbuf[BUFSIZ];
709: char *cp;
710: int irc;
711:
1.24 schwarze 712: mandoc_asprintf(&cp, format, value);
1.15 schwarze 713: cur->next = mandoc_calloc(1, sizeof(struct expr));
714: cur = cur->next;
715: cur->and = 1;
716: cur->bits = key;
717: if (0 != (irc = regcomp(&cur->regexp, cp,
718: REG_EXTENDED | REG_NOSUB | REG_ICASE))) {
719: regerror(irc, &cur->regexp, errbuf, sizeof(errbuf));
720: fprintf(stderr, "regcomp: %s\n", errbuf);
721: cur->substr = value;
722: }
723: free(cp);
724: return(cur);
725: }
726:
727: static struct expr *
1.8 schwarze 728: exprterm(const struct mansearch *search, char *buf, int cs)
1.1 kristaps 729: {
1.15 schwarze 730: char errbuf[BUFSIZ];
1.1 kristaps 731: struct expr *e;
1.38 schwarze 732: char *key, *val;
1.20 schwarze 733: uint64_t iterbit;
734: int i, irc;
1.1 kristaps 735:
736: if ('\0' == *buf)
737: return(NULL);
738:
739: e = mandoc_calloc(1, sizeof(struct expr));
740:
1.38 schwarze 741: if (MANSEARCH_MAN & search->flags) {
742: e->bits = search->deftype;
1.8 schwarze 743: e->substr = buf;
1.38 schwarze 744: e->equal = 1;
1.5 kristaps 745: return(e);
746: }
747:
1.1 kristaps 748: /*
1.38 schwarze 749: * Look for an '=' or '~' operator,
750: * unless forced to some fixed macro keys.
1.1 kristaps 751: */
752:
1.38 schwarze 753: if (MANSEARCH_WHATIS & search->flags)
754: val = NULL;
755: else
756: val = strpbrk(buf, "=~");
757:
758: if (NULL == val) {
759: e->bits = search->deftype;
1.8 schwarze 760: e->substr = buf;
1.1 kristaps 761:
1.38 schwarze 762: /*
763: * Found an operator.
764: * Regexp search is requested by !e->substr.
765: */
766:
767: } else {
768: if (val == buf)
769: e->bits = search->deftype;
770: if ('=' == *val)
771: e->substr = val + 1;
772: *val++ = '\0';
1.21 schwarze 773: if (NULL != strstr(buf, "arch"))
774: cs = 0;
1.38 schwarze 775: }
776:
777: /* Compile regular expressions. */
778:
779: if (MANSEARCH_WHATIS & search->flags) {
780: e->substr = NULL;
781: mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", buf);
782: }
783:
784: if (NULL == e->substr) {
785: irc = regcomp(&e->regexp, val,
786: REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE));
787: if (MANSEARCH_WHATIS & search->flags)
788: free(val);
789: if (irc) {
1.15 schwarze 790: regerror(irc, &e->regexp, errbuf, sizeof(errbuf));
791: fprintf(stderr, "regcomp: %s\n", errbuf);
1.8 schwarze 792: free(e);
793: return(NULL);
794: }
1.38 schwarze 795: }
796:
797: if (e->bits)
798: return(e);
1.1 kristaps 799:
800: /*
801: * Parse out all possible fields.
802: * If the field doesn't resolve, bail.
803: */
804:
805: while (NULL != (key = strsep(&buf, ","))) {
806: if ('\0' == *key)
807: continue;
1.20 schwarze 808: for (i = 0, iterbit = 1;
809: i < mansearch_keymax;
810: i++, iterbit <<= 1) {
811: if (0 == strcasecmp(key,
812: mansearch_keynames[i])) {
813: e->bits |= iterbit;
814: break;
815: }
816: }
817: if (i == mansearch_keymax) {
818: if (strcasecmp(key, "any")) {
819: free(e);
820: return(NULL);
821: }
822: e->bits |= ~0ULL;
1.1 kristaps 823: }
824: }
825:
826: return(e);
827: }
828:
829: static void
830: exprfree(struct expr *p)
831: {
832: struct expr *pp;
833:
834: while (NULL != p) {
835: pp = p->next;
836: free(p);
837: p = pp;
838: }
839: }
840:
841: static void *
1.37 schwarze 842: hash_calloc(size_t nmemb, size_t sz, void *arg)
1.1 kristaps 843: {
844:
1.37 schwarze 845: return(mandoc_calloc(nmemb, sz));
1.1 kristaps 846: }
847:
848: static void *
849: hash_alloc(size_t sz, void *arg)
850: {
851:
852: return(mandoc_malloc(sz));
853: }
854:
855: static void
1.37 schwarze 856: hash_free(void *p, void *arg)
1.1 kristaps 857: {
858:
859: free(p);
860: }
CVSweb